]\gt ; III .
Experiments on f ?
uid Viscosity .
A. MALLOCK .
by Lord KELVIN , P.R. Received July 26 , \mdash ; Read November 21 , 1895 .
THESE experiments consisted of the measurement of the moment transmitted , by fluid visoosity , across the annular space between two concentric cylinders , one of which revolves while the other is stationary .
The fluid used in these experiments was water .
Three distinct sets of conditions were tried , .\mdash ; ( 1 ) Out cylinder revolving , inner cylinder stationary .
( 2 ) Inner , , , , outer ( 3 ) Repetition of series ( 1 ) with an annuIus of different width .
It had been intended to have had a repetition of series ( 2 ) with a larger annuIus , but the motion in the fluid in series ( 2 ) was so thoroughly unstable that this was not done .
The object of the experiments was chiefly to examine the limits between which the motion of the fluid in the annulus was stable , and the manner in which stability broke down .
For obtaining the actual value of the coefficient of viscosity , other methods , such as the fiow through capillary tubes , would be more suitable .
Fig. 1 is a tion of the cylinders as arranged for the experiment of series ( 1 ) .
The inner cylinder A is suspended by a torsion wire , attached by a gymbal rin to the top of the stem B. This stem is guided by two rings of balls , held by the gunmetal castillgC , the rings being adjustable in their own planes by means of four set screws at each ring , thus allowing the axes of the suspended and revolving cylinders to be made coincident .
* The stem carries a divided circle which is read by the telescope T. is the outer cylinder , carried on the axis F. This cylinder was driven at constant speeds by a small electromotor connected with a governor which euts off the current when , and not before , the desired speed is attained .
Surrounding Though this adjustment was made with ease in the main experiments , preliminary experiments with the arces of the cylinders parallel , bnt separated by known intervals , had shown that an error in oentreing the cylinders , if mall , produoed a of the second order onIy in the moment transmitted by the fluid .
NDCCCXCVI .
10.3.96 the cylinder is another , , and the space between and , and the interior of are kept filled with water in which thermometers placed , and the temperature 04 the fluid in the annulus betweeu and A is taken as the mean of the temperatures in A and the outer annuIus .
The cylinder A has its floor about half-an-inch up from the lower edge , and during the experiments the space under this floor is filled with air .
Thus the fluid in the annuIus only touches the cylindrical surface of A. The axis of the outel cylinde is hollow , and a rod , , passing through a watertight joint at its upper end , carries a short piece of cylinder , of the same diameter as A. The arm at the lower end of , where it projects beyond , is held fixed , thus keeping the short cylinder stationary whilst revolves .
The space between and is filled with mercury .
The object of this arrangement is to cause the lower surface of the water in the annulus to be in contact with a surface whose velocity at any point is nearly the same as its own .
Of course , in the mercury , in consequence of its being in contact with the floor of bhe cylinder , there is a certain amount of circulation set up , which causes an outward radial drift in the mercury in contact with the floor and a return curtent flowing inwards in the mercury surface in contact with the water , but the radial velocity is slow compared with the circumferential , and the water in the annulus between and A is very nearly in the same condition as it would be if and A were infinitely Fi gs .
( 2 ) and ( 3 ) show the arrangement used in the second and third series of experiments respectively .
Attached to reading telescope was a small camera-lucida , by means of which the image of the divided circle could be seen on the surface of a cylinder , driven by clock- work and carrying continuous paper , P. On this paper an electric clock recorded seconds , and another pen and electromagnet in connection with a contact-maker on the axis marked every revolution of the cylinder E. A third pen , worked by a hand lever , was used to follow the motion of the image of the divided circle seen on the paper through the camera-lucida .
In making an experiment the usual course was to bring the wire of the telescope to the zero of the divided circle when everything was at rest , then to start the paper cylinder and mark the position of the zero on the paper .
The experimental cylinder was then set in motion , and when the suspended cylinder had reached its approximately stationary position ( about four minutes after the revolving cylinder was started ) the mark on the divided circle , which was nearest to the centre of the paper , was followed by the third pen for a minute or more , after which the zero mark was again made on the paper , and the number of the division ( on the circle ) followed by the pen was writlen on .
In some experiments , however , the motions of the suspended cylinder were followed from the moment the revolving cylinder was started .
The diagrams so made gave a permanent rocord of the result of each experiment , which could be analyzed at leisure : Table I. ( p. 47 ) gives the dimensions and constants of the apparatus .
MR. A. MALLOCK ON EXPERIMENTS ON FLmD VISCOSITY .
The Table II .
( pp. 48-52 ) contains the resuIts of all the experiments ( except a few made with the medium cylinder ) as obtained directly from the analysis of the diagrams taken with the camera-lucida .
hese results were afterwards corrected for the variation of the depth of the fluid in contact with the suspended cylinder , dne to the curvature of the free surface , and furtt ) corrected for the difference , if any , of this depth from inches ( a value which the dimensions of the cylinder rendered convenient ) .
Most of the quantitative results are summarized in diagrams 6 , and 10 .
In Diagrams 5 and 6 , and also in Diagrams 4 , 8 , and 9 , the abscissa is the number of revoIutions per minute of the revolving cylinder and ordinates are the readings of the divided circle on A. In Diagrams 5 and 6 , in addition to the number of revolutions per minute , the velocity ( in feet per second ) is given , which two parallel planes must have with reference to one another , if their distance apart is equal to the width of the annuIus , and the fluid between them is distorted at the same rate as the fluid at the surface of the cylinder A. In Diagrams 4 , 5 , and 6 , it will be seen that up to a certain velocity of the cylinders and , the ordinates increase almost exactly as the velocity .
Above this velocity , and until a considerably higher velocity is reached , it appears that the ordinates may have any value between certain limits .
At the higher velocities the readings obtained for the ordinates again become very precise and constant , but lie on a curve instead of a straight line .
This curve , if continued in the direction of decreasing veIocities , will be found to form the upper limit of the value for the ordinates in that part of the diagram where the readings are fluctuating , the lower limit being the continuation of the straight line which precedes it .
The curves of Diagram 7 record the motion of the cylinder , from the moment of the starting of until the fluid in the annuIus has assumed the steady motion appropriate to the velocity at which is being driven ; and they show that , whether the velocity of is large or small , provided it does not exceed 150 feet per minute about , the steady motion state is reached in approximately the same time .
The curves of Diagram 7 , as immediately obtained ffom the camera-lucida tracings , are partly obscured by the natural oscillation of the cylinder on its torsion wire , but these oscillations have been eliminated , and the curves show the zero about which the oscillations are taking place , at each instant .
The curves on Diagram 8 are given to show that both the mercury and the short cylinder have a great effect in inoreasing the range through which the transmitted moment varies as the first power of the velocity .
Diagram 9 contains the results of the expeliments of series ( 2 ) , in which the outer cylinder was suspended by the torsion wire while the inner cylinder ( A ) revolved . .
A. MALLOCK ON EXPERIMENTS ON FLUID VISCOSITY .
a few experiments were made with this arrangement , as the motion of the fluid was eddying and unstable , even at very low velocities .
Diagram 10 gives the coefficient of viscosity in terms of temperature , as deduced from the experiment of series ( 1 ) and ( 3 ) .
With these results , for the sake of comparison , I have drawn the curve representing the true value of the coefficient { taken from Professor EVERETT'S What the origin of the high values found for the coefficient by my experiments is , I am not at present in a position to explain , but frona the fact of the moment transmitted by the fluid being ectly proportional to the I do not think it can be put down to eddy-makmg in the ordinary sense of the word , or , if due to formation of isolated eddies , the magnitude of and rate at which such eddies are formed 1nust be such as to keep the total transmitted ( at any rate very nearly ) proportional to the velocity .
As on this point the curves of Diagram 11 have been introduced .
These curves are actual tracings by the camera-lucida of the angular motion of the divided circle on cylinder A during portions of four experiments .
It will be seen that , for low velocities curves of the revolving cylinder , the suspended cylinder A remains nearly stationary , the chief movement being a slight harmonic oscitlation of about fifteen seconds ' period ( the period of the suspended cylinder on the torsion wire ) , but that this movement is more marked and more irregular ill ( than in When the speed approaches the limits of stability ( curve c ) the disturbance becomes very large , showing that at these speeds the motion of the fluid in the annulus is at times , but not always , irregular .
At much higher speed ( curve ) , when the motion is thoroughly unstable , and the fluid , as it were , saturated with eddies , the curve , on the whok , is again fairly straight , but the nature of the irregularities indicates that the forces : at work are large .
hen the velocity approached that at which instabiht , was liable to occur , it was interesting to notice how small a disturbance of the system was sufficient to change the entire .
character of the motion .
A slight blow on the support which carried the apparatus , or a retardation for a few moments of the rotation of the outer cylinder , was almost sure to produce the effect .
( Note that reducing the velocity of had the effect of increasing for the time the moment acting on A. ) The unstable motion so produced , however , was not necessarily permanent , and 1 have seen the stable form of motion change to the eddyin one and back again many times in auccession at irregular intervals , and for no apparent reason , when the speed near the hmit of stability .
Th. appearance of the surface of the fluid in the annulus was as certain a criterion of the of the motion as the torsion produced on the inner cylinder .
As long as the motion was stable , the surface remained as smooth as gl-ass ; the beginning MR. A. MALLOCK ON EXPERIlfENTS ON FLUID scosmr .
of the instability being marked by the appearance of small dimples and which , when the high velocities were reached , covered the ole surface .
The experiments of series ( 3 ) , where the width of the annulus was about half-aninch , give a nearer approach to the true value of the coefficient of viscosity than , where the annulus had a width of an inch .
In some former experiments of mine , described in 'Proc .
Roy .
Soc December , 1888 , where the annulus was little more than inch wide , the approximation to the true value was very close .
* The effect of temperature in altering the critical velocity was not as marked as I had expected it to be .
From Professor OSBORNE REYNOLDS ' experiments I had supposed that the critical veJocity would be proportional directly to the viscosity , but Diagrams 6 and 7 show that in this form of experiment , at any rate , this is not the case .
At a temperature of C. the viscosity of water is only about a of what it is at C. , but , at lormer temperature , instability begins at a speed only 11 or 12 per cent. less than at the latter .
If we deduce the coefficient of fluid friction from the experiments at the higher speeds of series and ( 3 ) it will be found that the formula which best represents the curves is , and coefficient of friction is .
per square foot of area at 10 feet per second instead of 23 lb. ( FROUDE ) and .
( UNWIN ) .
And both Mr. FROUDE and Professor found the frictional resistance increase as the power of the velocity for smooth metal surfaces such as se used in my experiments .
It would seem this , even when water in the annulus is in the completely eddying condition , the character of the motion cannot be the same as that in the neighbourhood of Mr. FROUDE'S plane or Professor UNWIN'S disc .
The case is quite different , however , in the experiments of series ( 2 ) .
Here the motion seems essentially unstable at all speeds , and from such experiments no value of the coefficient of viscosity can be deduced , but the coefficient of friction which they *Since the above was written it has been pointed out to me , by G. STOKmS , that the formula which I used for computing the coefficient of viscosity , from these experiments and also from those of the present series , was incorrect .
The values of have , therefore , been recomputed for all the experiments ( including those of from the formula where , and are the radii of the external and internal oylindeoe the tangential force acting on the surface of the internal cylinder , the area of the surface of the internal oylinder in contact with the water , and the velocity of the surfaoe of the external oylindsr .
The resuIts of the 1888 experiments are indicated by the spots , near the curve , on diagram 10 ) , but they are close an approximation to the true value of to allow of a rate curve being drawn through them .
MR. A. MALLOCK ON EXPERIMENTS ON VISCOSITY .
indicate is a little ore than2 lb. per square foot at ten feet per second ( practicaIIy the same as Mr. FROUDE'S and Professor UNWlN'S ) .
The exponent also of the velocity is not far from , but the velocities used in this series were so low that I should not attach much importance to the numerical determinations except as showing that the motion is really different in character from that which takes place when the inner cylinder is stationary and the outer one revolves .
[ Note added July 10th , 1895.\mdash ; Since writing the above , I have , at Sir G. G. STOKES ' suggestion , added the set of experiments shown in Diagram 12 .
These experiments show that the moment transmitted is directly proportional to the depth of the fluid in the annulus measured from the lower edge of the suspended cylinder .
They give therefore further and independent evidence that the mercury floor and short cylinder do really supplyboundary conditions such as must exist at any cross-section of an infinitely]ong annulus , if particles having the same radu on either side of the cross-section are to be without relative motion .
] TABLE I.\mdash ; Dimension and Constants of Apparatus .
The letters refer to figs. ( 1 ) , ( 2 ) , and ( 3 ) .
Radius of Cylinder Height of cylinders 'Inches .
Centims .
3 .
9.943 3 .
8.687 24.5 Number of divisions on circle , 400 .
TANGENTIAL Force at Surface of Cylinder A required to turn it through one division of the circle .
With torsion wire I. II .
, grain .
gram .

