THE LIBRARY 
 
 OF 
 
 THE UNIVERSITY 
 
 OF CALIFORNIA 
 
 RIVERSIDE 
 
 GIFT OF 
 
 U.S. Naval Ordnance Laboratory 
 
 through 
 Dr. E.L.Harrington Estate
 
 of Cbicago 
 
 A REDETERMINATION OF 
 THE COEFFICIENT OF VISCOSITY OF AIR 
 
 A DISSERTATION 
 
 SUBMITTED TO THE FACULTY OF THE OGDEN SCHOOL OF GRADUATE 
 
 SCIENCE IN CANDIDACY FOR THE DEGREE OF 
 
 DOCTOR OF PHILOSOPHY 
 
 (DEPARTMENT OF PHYSICS) 
 
 BY 
 
 ERTLE LESLIE HARRINGTON 
 
 A Private Edition 
 
 Distributed by 
 The University of Chicago Libraries 
 
 Reprinted from THE PHYSICAL REVIEW, N. S., Vol. VIII, No. 6, 
 December, 1916
 
 A REDETERMINATION OF THE ABSOLUTE VALUE OF THE 
 COEFFICIENT OF VISCOSITY OF AIR.
 
 [Reprinted from the PHYSICAL REVIEW, N.S.. Vol. VIII, No. 6, December, 1916.] 
 
 A REDETERM I NATION OF THE ABSOLUTE VALUE OF THE 
 COEFFICIENT OF VISCOSITY OF AIR. 
 
 BY ERTLE LESLIE HARRINGTON. 
 
 IN view of certain results on the "Coefficients of Slip" obtained by him 
 from his study of the laws of fall of small spheres in gases, Prof. 
 Millikan suggested to me that I attempt to check these results by making 
 some measurements on the coefficients of viscosity at very low pressures 
 with the constant deflection apparatus designed by himself and Dr. 
 Gilchrist. In the course of this work the method was so perfected as to 
 make it capable of a precision apparently unapproached heretofore in 
 measurements on the viscosity of gases. 
 
 Since the knowledge of the exact value of this coefficient for air is of 
 such fundamental importance in many lines of physical research it 
 seemed worth while to turn aside for the sake of attempting again to 
 fix its value with all the precision possible. This was especially needed 
 since some question 1 has been raised recently relative to the reliability 
 of the estimate made by Prof. Millikan as to the most probable value of 
 this constant. From determinations made under his direction by Gil- 
 christ 2 by the constant deflection method, and by Rapp 3 by an improved 
 capillary tube method, taken in connection with reliable determinations 
 made elsewhere and by other methods, Prof. Millikan published 4 as the 
 most probable value of this constant, ?;, at 23 C., .0001824, and estimated 
 its uncertainty at not more than one tenth per cent. 
 
 In connection with another problem, Vogel, 5 in 1914, was led to make 
 and publish a summary of the results of nearly all the determinations of 
 this constant that had ever been made, but he arrived at a value about 
 .8 per cent, higher than the above. In this, however, he includes deter- 
 minations which certainly involve gross error and which are at least very 
 far from what is now known to be the approximate value of rj. In fact, 
 he lists values with a total range of nearly n per cent., even in the same 
 method, while observers with present laboratory methods and facilities, 
 
 1 A. Gille, Ann. der Phys., 48, p. 799, 1915. 
 
 2 PHYS. REV., I., p. 124, 1913. 
 
 3 PHYS. REV., 2, p. 363, 1913. 
 
 4 Ann. der Phys., 41, p. 759, 1913. 
 
 6 Ann. der Phys., 43, p. 1235, 1914.
 
 VOL. VIII.l 
 No. 6. 
 
 COEFFICIENT OF VISCOSITY OF AIR. 
 
 739 
 
 using any method whatsoever, would admit no greater error than a 
 fraction of a per cent., hence the inclusion of such widely variant values 
 of 77 renders more doubtful the validity of any mean so obtained. 
 
 The determination by Gilchrist appears also in this summary by 
 Vogel, and although he gives it relatively more weight than any other 
 single determination by any method, he nevertheless expresses the feeling 
 that the method and theory had not as yet received full development. 
 Since the publication of the Gilchrist article the method has been used 
 by Timiriazeff, 1 but his experimental arrangements were such as to make 
 it better suited to relative than to absolute determinations. It therefore 
 seemed important to subject the method itself to as critical a study as 
 possible while making the new determination of 77. 
 
 APPARATUS. 
 
 The apparatus is the same as that used by Gilchrist save for 
 pension and for certain other features which 
 it was necessary to modify in order to adapt 
 it to running in vacuum. The diagram shows 
 the general arrangement of the apparatus. 
 The outer, O, of two concentric brass cylin- 
 ders is made to rotate with constant velocity 
 about the inner, 7, which is suitably hung by 
 an elastic suspension. The viscosity of the 
 air produces a drag upon the inner cylinder 
 causing it to be deflected from its position of 
 rest to such an angle that the restoring couple 
 of the suspension brought into play exactly 
 counteracts the drag of the air, the angle of 
 deflection being measured by the usual mirror 
 telescope and scale method. 
 
 The cylinder frame, F, is mounted upon a 
 heavy steel plate, P, accurately machined and 
 having a raised rim in order to provide a 
 mercury seal for the large glass jar, /, which 
 covers the whole. To one side of the plate is 
 drilled a hole to provide pump connections, 
 and in the bottom is screwed a steel pipe, Q, 
 enclosing the driving shaft, and of such length 
 as to serve as a barometer column, and the 
 lower ends of pipe, rod, and gearing were suit- 
 ably constructed to dip into a mercury cup, 
 
 1 Ann. der Phys., 40, p. 971, 1913. 
 
 the sus- 
 
 Fig. 1.
 
 74 ERTLE LESLIE HARRINGTON. FSECOND 
 
 [SERIES. 
 
 C, through which the motion must be transmitted. The cylinders are 
 carried by a heavy brass frame, F, provided with leveling screws and levels. 
 The outer cylinder is supported on a slightly conical bearing, B, at the bot- 
 tom, and at the top rotates on the tube, T, as an axle. The tube, T, at 
 the same time supports the suspension head, A, and is rigidly held by 
 the main frame. To eliminate end effects the inner cylinder is provided 
 at either end with a guard cylinder, G, of the same radius, and rigidly 
 held at a small distance from it and in perfect alignment with it by three 
 brass posts which connect the two guard cylinders and suitable end disks. 
 The upper part of this guard system and the suspension tube above 
 mentioned are constructed as one piece so that the upper guard cylinder 
 is held accurately centered with respect to the upper part of the rotating 
 cylinder. The lower part rests on a conical bearing which is a part of, 
 and accurately centered with respect to the bottom of the outer cylinder. 
 The suspension head, A, is so constructed as to permit considerable 
 translatory motion in any direction, as well as rotatory, thus making 
 possible accurate adjustment. The above apparatus was constructed 
 with great care and precision by Wm. Gaertner & Co., and to them 
 much credit is therefore due for the success of the experiment. Such 
 accuracy in construction made it possible to have the guard cylinders 
 very close (.025 cm.) to the suspended cylinder, and thereby completely 
 eliminate the need of end effect corrections. 
 
 A chronograph, K, provided with extra weights serves the double 
 purpose of driving the apparatus, and also of leaving a permanent record 
 of the speed of rotation. It was so arranged that the cylinder could be 
 instantly thrown out of gear and the chronograph used independently 
 of the other apparatus, this being done each time the period of vibration 
 of the inner cylinder was determined. 
 
 A Beckmann thermometer, calibrated with a standard Baudin ther- 
 mometer, and reading directly to .01 degree was hung beside the outer 
 cylinder. A high-grade Centigrade thermometer graduated to .1 degree 
 was hung beside it to serve as a check. The room itself was one of the 
 constant temperature rooms of the Ryerson laboratory; a basement 
 room, with no windows, lined on. all sides, top and bottom, with a 5-in. 
 layer of cork, and provided with heavy inner and outer doors. The 
 room was efficiently heated by a new style electric heater (furnished by 
 the Lee Electric Radiator Co.) used in connection with a sensitive thermo- 
 static control, and the air was constantly stirred by a large fan. The 
 temperature control was such that during any run the temperature 
 variation recorded by the Beckmann thermometer was never more than 
 a few hundred ths of a degree, and often no change at all was detected.
 
 VOL. VIII.1 
 No. 6. J 
 
 COEFFICIENT OF VISCOSITY OF AIR. 
 
 741 
 
 The air in the apparatus was kept dry by an enclosed dish of phos- 
 phorus pentoxide. 
 
 CRITICAL STUDY OF APPARATUS AND DETERMINATION OF DIMENSIONS. 
 
 Inasmuch as one of the objects of the experiment was to study the 
 
 possibilities of the method as well as to find the absolute value of 17, 
 
 considerable time was spent on this phase of the work, and w enever 
 
 Fig. 2. 
 
 possible various methods of measurement were used for the purpose of 
 cross checking. 
 
 i. The Inner Cylinder. Three methods were used to measure the 
 diameter of the inner cylinder. In order to insure getting different 
 diameters and to entitle each different one to equal weight in the computa- 
 tion of the mean a large number of points evenly distributed over the 
 surface of the cylinder were systematically numbered and the measure- 
 ments were taken at these points. The first method was to place the 
 cylinder vertically on the bed of a dividing engine having mounted at one
 
 742 ERTLE LESLIE HARRINGTON. 
 
 side with its axes perpendicular to the direction of motion a short focus 
 telescope provided with cross hairs. The distance between tangent 
 lines was then determined from the screw readings. The second method 
 was to adjust the jaws of a large-size vernier caliper to the diameter of 
 the cylinder and then determine the perpendicular distance between the 
 locked jaws by placing the same on the bed of the dividing engine, and 
 measuring by the usual method. The third method was to use a 
 micrometer screw caliper of sufficiently large size to measure directly 
 the diameter. The last two methods were by far the more convenient 
 and yielded as accurate results. These two results differed by only I 
 part in 13,000, and their mean agreed to Gilchrist's value to within I 
 part in 6,000, which is a liberal estimate of the possible error in this 
 dimension. The length was determined by cathetometer methods, and 
 the value found to agree exactly with that given by Gilchrist, and is 
 probably correct to within I part in 6,000. 
 
 2. The Outer Cylinder. The most satisfactory method of measuring 
 the diameter of the outer cylinder was found to be by filling it with dis- 
 tilled water observing the temperature and depth, and weighing the water 
 used. Results thus obtained differed by only I part in 15,000, and the 
 mean differed from that by Gilchrist by exactly the same amount, 
 so the error is perhaps no more than I part in 8,000. 
 
 The accuracy in the inner surface was studied in this way : The inner 
 cylinder, the lower guard ring, and the posts were removed, leaving the 
 upper part of the guard cylinder system, which, as above described, form 
 the axle for the outer cylinder. A heavy rod with one end hollowed 
 out to fit the conical bearing in the bottom of the outer cylinder was 
 inserted in the opening and brought to rest on the bearing. This rod 
 carried a lever arm suitably curved to rest against the wall of the cylinder, 
 capable of being adjusted to different heights, and bearing at the fulcrum 
 a mirror. A suitable telescope with vertical scale placed at a distance 
 of nearly three meters made it possible to detect the slightest deviation 
 from constancy in the radius or perfection in symmetry. In doing this 
 the lever was held lightly against the wall by a weight, and the cylinder 
 slowly rotated at a constant speed. From the readings of the observer 
 at the telescope and a measurement of all distances concerned, any 
 variations could be quantitatively determined. It is important to note 
 that this method not only tests the accuracy of the inner surface but 
 tests also the symmetry of this surface about the same bearing that is to 
 carry and hold in position the inner cylinder system. The results of this 
 test were very satisfactory, inasmuch as the variations were of the order 
 of i part in 2,500 and of such nature as to be self-counteracting in their
 
 No L 6 VIIL ] COEFFICIENT OF VISCOSITY OF AIR. 743 
 
 effect on the final results, which were therefore affected probably less 
 than i part in 5,000. 
 
 3. Moment of Inertia of the Inner Cylinder. This was found as usual 
 by determining the period of vibration of the cylinder alone, and then 
 when a known moment of inertia was added. The cylinder was sus- 
 pended by piano wire and the cylindrical surface made plumb and 
 symmetrical about the axis by suitable adjustment of the three support 
 bars which connect the cylinder with the suspension clamp, and by use 
 of small weights fastened to the upper supporting vane of the cylinder. 
 The heavy, double support of the clamp from which the cylinder was 
 suspended rested on a stone bench, and although the room was apparently 
 free from air currents the cylinder was surrounded by a much larger one 
 in order to insure entire absence of them. The passage through the zero 
 position of the cylinder was indicated by the flash of a light into a tele- 
 scope at the opposite side of the room, and an electric key enabled the 
 operator to accurately register this time on the chronograph. The time 
 divisions were marked by impulses from the standard laboratory clock 
 and involved no appreciable error whatever, and the divisions could be 
 read to one one hundredth of a second. By taking the period from a 
 rather long run thus measured, the individual runs varied from the mean 
 for any suspension by an amount of the order of I part in 10,000, which 
 may be considered the probable error from this source. The added 
 known inertia consisted of a bar and ring, accurately machined, and 
 placed on the cylinder symmetrically. The dimensions of the ring and 
 bar were determined on the dividing engine, and their weights by the 
 analytical balance. The inertia of the combined ring and bar was 
 computed by the usual formula derived for such bodies, and the result 
 agreed to I part in 8,000 with that obtained by Gilchrist, this ratio 
 probably representing the accuracy of this determination. No faulty 
 adjustment to symmetry about the axis of suspension could have been 
 constant to the different series of runs, since the ring and bar were 
 frequently removed and replaced. 
 
 Having the two periods and the one known inertia, the moment of 
 inertia of the inner cylinder was computed from the usual formula, 
 
 _ 
 
 r 2 / 
 
 The average departure of the various results from the mean for the 
 moment of inertia of the inner cylinder obtained from the different 
 suspensions used was I part in 5,000 and the mean differed from that 
 obtained by Gilchrist by exactly the same amount.
 
 744 ERTLE LESLIE HARRINGTON. 
 
 DIMENSIONS. 
 
 Dimensions that are not involved in the computation of the results are 
 here included for descriptive purposes, but only approximate values for 
 such are given. 
 
 Vertical distance from base of chronograph to extreme top of cover 185 cm. 
 
 Diameter and height of glass jar, respectively 28 and 62 cm. 
 
 Length of outer cylinder 46 
 
 Length of suspension 23.5 
 
 Length of each guard cylinder 10 
 
 Length of inner cylinder 24.88 
 
 Distance between guard cylinders 24.93 
 
 Weight of inner cylinder 321 g. 
 
 Radius of outer cylinder 6.06317 cm. 
 
 Radius of inner cylinder 5.341 16 
 
 Moment of inertia of inner cylinder (experimentally determined) . . 7,617.3 
 
 ADJUSTMENT OF INSTRUMENT. 
 
 Before being placed in the instrument the inner cylinder was again 
 tested for symmetry and perpendicularity with the same methods and 
 precautions used preliminary to the determination of its moment of 
 inertia. In this three plumb bobs instead of one were used for the 
 purpose of expediting matters and permitting simultaneous observations 
 in three directions without in any way disturbing the cylinder or plumb 
 bobs. 
 
 The lower end of the outer cylinder being in place, the inner cylinder 
 and the guard cylinder system were put into position, fastened rigidly 
 by the top brace, and after suspending the inner cylinder the suspension 
 head screws and the leveling screws of the base of the instrument were so 
 manipulated as to bring the surface of the guard cylinders into alignment 
 with the inner cylinder as perfectly as possible without the use of a 
 telescope. The outer cylinder was then put into position, the suspension 
 tube again clamped, and the deflection of the inner cylinder noted as the 
 outer cylinder was rotated at the speed to be used later in determinations. 
 The suspension head was then rotated until the zero position was brought 
 as much to one side as the deflection position was to the other side of 
 the telescope which was in the middle of, and perpendicular to the 
 scale. Three small support screws in the base of the instrument were 
 then brought just to touch the bottom of the outer cylinder in order to 
 hold it in its exact position during the final adjustment. The outer 
 cylinder was loosened from its bottom and removed and the top brace 
 replaced and fastened firmly in position. The final, and more careful 
 adjustment of the guard cylinder system to alignment with the inner 
 cylinder was made by means of a short focus telescope. With a suitable
 
 JJ L - 6 y iIL ] COEFFICIENT OF VISCOSITY OF AIR. 745 
 
 clamp system attached to the frame the outer cylinder was replaced 
 without at any stage of the process inclining or otherwise disturbing the 
 inner cylinder. The preliminary adjustment to zero position made un- 
 necessary the turning of either the guard cylinder system or the suspen- 
 sion head after this final adjustment, and thereby eliminated any error 
 that might have come from a lack of perfect coincidence of the suspension 
 with the axis of the suspension head collar. The use of the support screws 
 for the base of the outer cylinder holds it, and therefore the lower end of 
 the inner cylinder system, in precisely the position they occupy during 
 a run. Keeping the guard cylinder system in place while replacing the 
 outer cylinder eliminates the question as to whether it returns to the 
 exact position it had during adjustment. With these precautions it 
 would seem unlikely that any appreciable error could arise from faulty 
 adjustment, and the results later given involving results obtained before 
 and after dissembling and readjusting furnish convincing evidence of 
 the absence of such error. 
 
 TRIAL RUNS AND A STUDY OF FACTORS AFFECTING RESULTS. 
 As will appear later, the value of rj is computed from the time of rota- 
 tion, the period of vibration of the inner cylinder, the deflection, the 
 distance to the scale, and the temperature observations. The method of 
 taking these observations was this: The zero position was read when the 
 cylinder was at rest and checked by causing the cylinder to make small 
 vibrations (about I cm. as seen on the scale) and determining the zero 
 position as is done in using balances. This was done as a precaution 
 against any error due to sticking, although experience showed this step 
 really unnecessary. The outer cylinder was then put into rotation, the 
 speed at first being modified by a brake attachment to the chronograph 
 which could be operated from the position of the telescope in order to 
 bring the inner cylinder quickly to near rest in its deflected position. 
 As soon as a steady state was attained the stylus was lowered on to the 
 waxed paper of the chronograph drum and the rotation continued until 
 the stylus had traveled the length of the drum, which meant, with the 
 speed used, an interval of about 13 min. In practice the inner cylinder 
 was allowed to vibrate through two or three millimeters as seen on the 
 scale, readings being taken at short intervals; this plan giving a large 
 number of independent readings, eliminating the barely possible source 
 of error due to sticking, and affecting a great saving of time that would 
 otherwise be required to bring the cylinder to absolute rest, inasmuch as 
 the period of vibration was very long, and the damping very small. 
 At frequent intervals during this period the temperature was read on
 
 746 ERTLE LESLIE HARRINGTON. [SEMES! 
 
 the Beckmann thermometer. At the close of the run the cylinder was 
 slowly let back and its zero position again checked. Now the outer 
 cylinder was given a rotation sufficient to set the inner cylinder in vibra- 
 tion, and then thrown out of gear, thus allowing the chronograph to be 
 used merely as such, whereupon the period of vibration was taken over a 
 period of about 45 min. To do this an electric key at the position of 
 the telescope enabled the operator to make accurate chronograph record 
 of the passage through the zero position, the first five and the last five 
 being recorded in order to provide means of cross checking and thus 
 insure absence of appreciable error from this source. The periods thus 
 obtained were probably accurate to within I part in 8,000. 
 
 The scale on which the deflections were read was a carefully selected 
 straight meter stick tested with a standard metric steel scale. The 
 magnification of the telescope was such that .1 mm. could be read easily 
 and since the deflections were of the order of 600 mm. the readings were 
 probably correct to within I part in 6,000. A steel tape was used to set 
 the ends of the scale equidistant fiom the mirror and to determine 
 its perpendicular distance from the mirror. The error in this was prob- 
 ably not greater than I part in 6,oob. 
 
 Irregularities in the speed of the chronograph might offer a source of 
 error, not on account of uncertainty as to what the speed is, for that is 
 obtainable directly from the record, but on account of the resulting 
 unsteadiness of the deflection. Fortunately it was found that the chrono- 
 graph drove the apparatus at a very constant speed, though of course 
 not perfectly so, owing perhaps to the irregularities in the friction, or a 
 lack of perfection in the gearing. However, a small auxiliary weight, 
 at the side of the operator, brought into series with the main weight by 
 two pulleys, enabled him after some experience to almost completely 
 neutralize any such irregularities, thereby reducing the error from this 
 source to probably I part in 5,000. 
 
 The high consistency in the early trials in the results for any suspension 
 indicated a satisfactory control in all the above sources of error, but it 
 was found that variations in the suspension produced rather great 
 variations in the results. In view of the experience of Gilchrist the 
 bifilar form of suspension was tried at first, but on account of the rather 
 great weight (321 g.) of the inner cylinder, and its small distance (.25 
 mm.) from the guard cylinders which permitted no sag, it was not possible 
 to use silk fibers which permit, perhaps, the closest approach to the true 
 bifilar type. Metallic ribbons were therefore used, but after two months 
 of experience with them they were wholly discarded, since it was not 
 found possible to entirely eliminate the effect of a change in the ribbon
 
 No"6 VIIL ] COEFFICIENT OF VISCOSITY OF AIR. 747 
 
 or even of a mere change in the separation of the strands or in the manner 
 of clamping, upon the results obtained. The trouble no doubt lies par- 
 tially in the rather great discrepancy between such a bifilar and those 
 ordinarily treated theoretically since the deflection brings into play not 
 only a restoring couple due to the slight raise in the weight, but also a 
 restoring couple due to the twist in the strands themselves. Such a 
 suspension is therefore a sort of hybrid between the bifilar and the 
 ordinary elastic unifilar suspension. The greatest error, however, prob- 
 ably comes from the fact that the ribbons have widths of the same order 
 of magnitude as the separation of the strands which makes it quite likely 
 that as deflection occurs the two edges of either strand may assume 
 varying portions of the load, thereby causing a virtual change in the 
 distance between the strands. If the strands be clamped at both ends 
 it is unlikely that the load is equalized between the two strands, and if 
 instead the strand be simply looped about a pin at one end in order to 
 permit constant equalization of the load there is the possibility of error 
 due to a rolling of the strands about the pin as the deflection occurs. 
 Unrolled phosphor brojize wire in place of the ribbon was even less 
 satisfactory owing to the residual coil and the consequent drift, nor was 
 either found satisfactory later as a unifilar suspension. In fact the 
 writer was led to conclude that little dependence could be placed on phos- 
 phor bronze where precise results are expected. Quartz fibers were tried, 
 but it was not found possible to obtain fibers coarse enough to support 
 the rather great weight and at the same time fine enough to give sufficient 
 deflections. The smallest piano steel wire obtainable was much too stiff, 
 but it was found possible by reducing the size of the smallest obtained, 
 by very carefully rubbing with fine emery paper, to secure a sufficiently 
 large deflection and yet retain adequate tensile strength. After the 
 adoption of this plan no further suspension troubles were experienced. 
 A number of such were made, and in all cases there was a good zero 
 return and a satisfactory absence of drift. Moreover, different suspen- 
 sions, though differing greatly in stiffness, yielded entirely concordant 
 results. Later some samples of the uncoiled stock from which the hair 
 springs of watches are made were furnished by the Elgin Watch Company 
 and found to have the proper range of stiffness. 
 
 As will be seen later the torsion constant of the suspension is expressed 
 in terms of the inertia of the inner cylinder and the period of vibration. 
 From the observed damping, and the theoretical relation between the 
 magnitude of damping and the effect on the period as given, for example, 
 by Helmholtz, it was calculated that the total effect due to the damping 
 factors would be of the order of I part in 10,000, or quite inappreciable.
 
 748 ERTLE LESLIE HARRINGTON. 
 
 Moreover, by experiment no effect on the period could be observed when 
 the outer cylinder was removed and the guard cylinders were separated 
 many times as far from the vibrating cylinder. But, although such 
 factors as viscosity involved above do not appreciably affect the period, 
 the fact must not be overlooked that the vibrating cylinder does carry 
 the air with it, and the moment of inertia of this air must be taken into 
 account. This point was first called to my attention by Dr. Lunn. 
 The periods taken in vacua were actually found to be i part in 750 less 
 than the periods taken at ordinary pressures, so the data given make due 
 allowance for this effect. 
 
 COMPUTATION OF RESULTS. 
 
 For the calculation of the results the well-known and very simple 
 formula 1 was used in this form : 
 
 - a 2 ) 
 
 where 77 is the coefficient of viscosity, I the moment of inertia of the inner 
 cylinder, a and b the radii of the inner and outer cylinders respectively, 
 I the length of the inner cylinder, <f> the angular deflection of the inner 
 cylinder, T the period of vibration of the inner cylinder, and o> the 
 constant angular velocity of the outer cylinder. If the period of rotation 
 of the outer cylinder, /, be substituted for 27r/w, and a constant, K, for 
 the product [I(b 2 a 2 )]/(2a 2 & 2 /) (having here a numerical value of 
 1.20188) the formula becomes: 
 
 tK<f> tK s 
 
 77 = or i, = tan l , 
 
 where 5 is the deflection as read on a straight scale, and d the distance 
 of the scale from the mirror. The latter form was used in all computa- 
 tion. A development of the above formula involving a more general 
 treatment which considers the coefficient of slip will be given in a paper, 
 following this, which will consider the problem of viscosity at the low 
 pressures where the effect of slip becomes appreciable. In apparatus 
 of the dimensions here used the correction for slip at ordinary pressures 
 amounts to about 2 parts in 100,000. 
 
 All determinations were made at temperatures in the neighborhood of 
 23 C., and the values for 77 reduced to that temperature by the use of 
 the formula suggested by Prof. Millikan (1. c.), 
 
 1723 = 779 + .000000493(23 - 0), 
 where 779 is the value of the viscosity coefficient obtained at C. This 
 
 1 See Poynting and Thompson, Properties of Matter, p. 213.
 
 VOL. VIII.l 
 No. 6. 
 
 COEFFICIENT OF VISCOSITY OF AIR. 
 
 749 
 
 simple formula, found satisfactory through the range mentioned by him 
 must certainly hold in the small ranges here involved, which are, with 
 but two exceptions, less than one degree. 
 
 RESULTS. 
 
 The following data show the results of thirty-one determinations of 
 77 made at various times during about three months, and involve the use 
 of six different suspensions. Suspensions C and D were really the same 
 suspension under different physical conditions. In most cases during 
 the series of runs for a given suspension the apparatus was taken apart 
 and readjusted in order to make sure there was no error of adjustment. 
 
 Sus. 
 
 /(Cm.). 
 
 j. 
 
 e. 
 
 t. 
 
 T. 
 
 1>H X 1C*. 
 
 A 
 
 200.0 
 
 40.21 
 
 23.97 
 
 30.059 
 
 140.70 
 
 1,823.7 
 
 A 
 
 200.0 
 
 40.26 
 
 24.45 
 
 30.020 
 
 140.61 
 
 1,823.5 
 
 A 
 
 200.0 
 
 40.06 
 
 23.06 
 
 30.018 
 
 140.62 
 
 1,820.9 
 
 A 
 
 200.0 
 
 40.05 
 
 23.31 
 
 30.085 
 
 140.62 
 
 1,823.3 
 
 A 
 
 199.7 
 
 40.025 
 
 23.18 
 
 30.000 
 
 140.58 
 
 1,821.4 
 
 B 
 
 200.2 
 
 35.81 
 
 23.61 
 
 30.014 
 
 132.81 
 
 1,821.2 
 
 B 
 
 200.2 
 
 35.665 
 
 22.88 
 
 30.114 
 
 132.79 
 
 1,824.1 
 
 B 
 
 200.2 
 
 35.73 
 
 23.14 
 
 30.031 
 
 132.79 
 
 1,821.0 
 
 B 
 
 200.2 
 
 35.837 
 
 23.43 
 
 29.971 
 
 132.80 
 
 1,821.2 
 
 B 
 
 200.2 
 
 35.80 
 
 23.46 
 
 30.032 
 
 132.80 
 
 1,822.8 
 
 C 
 
 200.8 
 
 60.51 
 
 22.88 
 
 30.090 
 
 172.13 
 
 1,825.9 
 
 C 
 
 200.8 
 
 60.67 
 
 23.055 
 
 29.940 
 
 172.14 
 
 1,820.5 
 
 C 
 
 200.8 
 
 61.064 
 
 22.93 
 
 29.840 
 
 172.21 
 
 1,825.1 
 
 C 
 
 200.8 
 
 60.624 
 
 23.165 
 
 30.014 
 
 172.18 
 
 1,822.2 
 
 C 
 
 200.8 
 
 60.604 
 
 22.84 
 
 29.962 
 
 172.13 
 
 1,821.1 
 
 c 
 
 200.8 
 
 60.68 
 
 23.09 
 
 29.936 
 
 172.13 
 
 1,820.8 
 
 C 
 
 200.8 
 
 60.73 
 
 22.91 
 
 29.940 
 
 172.13 
 
 1,823.1 
 
 C 
 
 200.8 
 
 60.77 
 
 23.19 
 
 29.907 
 
 172.14 
 
 1,821.0 
 
 C 
 
 200.8 
 
 60.75 
 
 22.90 
 
 29.895 
 
 172.14 
 
 1,820.9 
 
 C 
 
 200.6 
 
 60.905 
 
 24.10 
 
 29.912 
 
 172.13 
 
 1,822.7 
 
 C 
 
 200.6 
 
 60.54 
 
 23.00 
 
 29.982 
 
 172.10 
 
 1,822.0 
 
 D 
 
 200.7 
 
 - 61.154 
 
 22.97 
 
 30.014 
 
 172.98 
 
 1,822.9 
 
 D 
 
 200.7 
 
 61.216 
 
 22.99 
 
 29.954 
 
 172.95 
 
 1,821.6 
 
 D 
 
 200.7 
 
 61.07 
 
 23.22 
 
 30.075 
 
 172.96 
 
 1,823.2 
 
 E 
 
 200.85 
 
 51.66 
 
 23.27 
 
 30.178 
 
 159.39 
 
 1,824.8 
 
 E 
 
 200.85 
 
 51.168 
 
 23.28 
 
 29.888 
 
 159.39 
 
 1,824.7 
 
 F 
 
 200.9 
 
 63.34 
 
 23.16 
 
 29.946 
 
 175.62 
 
 1,823.8 
 
 F 
 
 200.9 
 
 62.988 
 
 23.26 
 
 30.087 
 
 175.60 
 
 1,822.3 
 
 F 
 
 200.9 
 
 62.957 
 
 23.11 
 
 30.103 
 
 175.56 
 
 1,824.0 
 
 F 
 
 200.9 
 
 62.974 
 
 23.10 
 
 30.066 
 
 175.58 
 
 1,821.9 
 
 F 
 
 200.9 
 
 63.093 
 
 23.19 
 
 30.008 
 
 175.55 
 
 1,821.9 
 
 Mean value of r, at 23 C. = 1822.6 X 10^.
 
 750 ERTLE LESLIE HARRINGTON. 
 
 Moreover, the runs show such variations in the various factors involved 
 as to make each determination an independent one, and no determination 
 made with satisfactory control in the manipulation of all steps was 
 discarded. 
 
 The strongest evidence of the advantages of this method for the deter- 
 mination of 77 lies in the remarkable consistency of the results here shown. 
 If from the individual deviations from the above mean one computes 
 the probable error by the usual least square method the result is found 
 to be .19 or i part in 9,600. Moreover, it should be emphasized that this 
 really includes every source of error except those involved in the instru- 
 ment constant, K. The probable error for each of the various determina- 
 tions involved in this constant has been given above in detail and if the 
 probable error in this constant be computed by the same method as 
 above it is found to be 1.9 parts in 5,000. The combined or total error 
 would be by this method of calculation only I part in 2,500 or .04 per 
 cent. Moreover, if the means for the different suspensions be compared 
 it is found that the maximum variation from the above mean is less than 
 .03 per cent, with the one exception of suspension , a watch-spring 
 suspension the use of which was discontinued on account of its tendency 
 to drift. Considering these two striking results, and making any reason- 
 able allowance for any unreliability in the least square method of com- 
 puting errors or of the estimates made of any individual probable error 
 it seems entirely justifiable to claim that the above mean is correct to 
 within less than .1 per cent, of the true value of t\ at the temperature 
 considered. 
 
 A comparison of the consistency of these results with the lack of con- 
 sistency of the results obtained by any other method, more especially 
 by the capillary tube method, shows the marked superiority of this 
 method. The uncertainties of the capillary tube method need not be 
 considered here inasmuch as they are well known and have been dis- 
 cussed by Prof. Millikan, 1 by Fisher, 2 Vogel, 3 and others. Only a few 
 points of contrast need be mentioned; the capillary tube method is 
 based on incomplete theory since it does not consider the effect of the 
 radial component of the velocity and other uncertainties due to the 
 increase in the volume of the gas as it passes along the tube, the general 
 end effects in addition to the question of the effect of irregularities in the 
 bore upon the stream lines, the difficulties in getting the exact pressures, 
 and above all, the impossibility of getting perfect tubes of the small radii 
 usually employed, and the great difficulty of subjecting any tube selected 
 to accurate examination as to uniformity, circularity, and even as to 
 
 1 L. c. * PHYS. REV., 29, p. 147, 1909. * L. c.
 
 No L 6 ] COEFFICIENT OF VISCOSITY OF AIR. 751 
 
 the value of the radius itself, which enters, it should be recalled, in the 
 fourth power. The wide variation in the results obtained by this method 
 by different observers, and even by the same observer with different 
 tubes, is ample proof that these uncertainties exist. On the other hand, 
 with the method described above the theory is complete, there are no 
 end corrections involved, no expansion takes place, every step in the 
 construction of the cylinders is subject to control, the dimensions are so 
 great that they may be determined with great accuracy, and every portion 
 of the apparatus is subject to minute study for irregularities, and even 
 should such exist, their effect would be far less serious than in the case 
 of the other method. As to consistency with other observers we are 
 essentially limited to the value obtained by Gilchrist who first used the 
 method and made the claim that his result contained not more than .2 
 per cent, error. The value here obtained differs from his by less than 
 that amount. That his values fluctuate through a greater range than 
 the range here obtained is without doubt due principally to the suspension 
 troubles mentioned above which were here so largely eliminated, for 
 every determination of his of any instrument constant that could at this 
 time be checked was most critically examined, and no one of them found 
 to differ by more than i part in 6,000 from the value here given. 
 
 It is interesting to note that Rapp, 1 who perhaps came more nearly 
 completely eliminating the errors in the capillary tube method than 
 anyone else, and who used a large number of tubes, is practically identical 
 with the result here obtained, as is also that obtained by Hogg, 2 who used 
 an oscillation method. The result obtained by Grindley and Gibson, 3 
 using still a different plan differs only by about .03 per cent. More 
 significant still is the fact that the value published by Prof. Millikan as 
 correct to within .1 per cent, is less than .08 per cent, above the value 
 here obtained, and as his value was based on perhaps the most accurate 
 determinations by five different methods, it would seem that the above 
 claim that the result here obtained is within .1 per cent, of the true value 
 is well founded, since the above mentioned values all lie well within 
 this limit. 
 
 In conclusion the writer wishes to acknowledge his gratitude to Prof. 
 Millikan, who suggested the problem and maintained such constant and 
 helpful interest in the research during its progress, and to Prof. Michelson, 
 the head of the department, for various helpful suggestions. 
 RYERSON PHYSICAL LABORATORY, 
 UNIVERSITY OF CHICAGO, 
 June, 1916. 
 
 iL. c. 
 
 2 Am. Acad. Proc., 40, 18, p. 611, 1905. 
 
 * Proc. Roy. Soc., A, 80, p. 114, 1908.
 
 [Reprinted from THE PHYSICAL REVIEW, Vol. 21, No. 3, March, 1923.] 
 
 A DETERMINATION BY THE CONSTANT DEFLECTION 
 
 METHOD OF THE VALUE OF THE COEFFICIENT OF 
 
 SLIP FOR ROUGH AND FOR SMOOTH 
 
 SURFACES IN AIR. 
 
 BY LELAND JOHNSON STACY. 
 
 ABSTRACT. 
 
 Coefficient of slip for rough and smooth surfaces in air, determined by the 
 constant deflection method. The apparent coefficient of viscosity measured 
 by this exceptionally precise method does not come out constant for the lower 
 pressures unless correction is made for the slip at the surfaces. The relation 
 7p'(i + kp) = rj = constant, between rj p ', the apparent viscosity coefficient, 
 and f p , the coefficient of slip, enables f p to be determined from measurements of 
 r) p f at pressures of 0.2 mm or lower. The apparatus included a vacuum- 
 tight chamber inside which a cylinder of radius 5.341 cm was suspended by a 
 steel wire concentric with a cylinder of radius 6.063 cm which could be rotated 
 at a constant slow rate so as to cause a steady deflection of the inner cylinder. 
 The accuracy of this method of measuring rj p ' is so great that the values of 
 frs = fp/76 all lie within 4 per cent of the mean. The chief difficulty 
 was in keeping the air pure because of the gradual evolution of gas, probably 
 hydrogen, inside the apparatus; but by taking observations only shortly 
 after evacuation this effect was avoided. For brass surfaces, f, reduced to 
 23 and 76 cm, came out 66.15 X io~ 7 which is practically the theoretical 
 minimum deduced by Millikan for a completely diffusing surface, 65.9 X io~ 7 . 
 For surfaces coated with shellac, the coefficient was found to be 97 X io~ 7 for a 
 fresh surface, in agreement with the value obtained by Lee from droplet 
 measurements, but it decreased steadily with time, presumably because of a 
 roughening due to oxidation, falling in two months to within 3 per cent of the 
 theoretical minimum. The early part of this work was done in collaboration 
 with E. L. Harrington. 
 
 Coefficient of viscosity of air at o.i mm is the same as at atmospheric pres- 
 sure when correction is made for the slip effect. The constancy of the values 
 obtained for f provides new evidence that the coefficient is independent of 
 the pressure. 
 
 T 
 
 I. HISTORICAL DEVELOPMENT OF THE IDEA OF SLIP. 
 
 HE theory that the coefficient of viscosity of a gas should be inde- 
 pendent of the pressure was first deduced by Maxwell * from a 
 consideration of the internal friction of molecules assumed to be rigid 
 spheres. He first deduced the relation 
 
 (i) 77 = pel, 
 
 where 17 = coefficient of viscosity; p = density of the gas; c = mean 
 1 Phil. Mag., 1860, Vol. 19, p. 31.
 
 240 LELAND JOHNSON STACY. 
 
 , molecular velocity; / = mean free path of gas molecule. Since the 
 density is directly proportional to the pressure, while the mean free path 
 is inversely proportional to the same quantity, the product pi, and there- 
 fore 17, should be independent of the pressure. 1 In a later paper, 2 Maxwell 
 reported some experimental tests of his theory for pressures ranging 
 from 30 in. to 0.5 in. of mercury. His apparatus consisted of a torsion 
 pendulum of three plane-parallel plates suspended by an elastic fiber 
 between four fixed plane-parallel plates. The whole system was enclosed 
 in an airtight vessel and the logarithmic decrement of the oscillations 
 was observed for different pressures. He found no observable change in the 
 decrement, which is a measure of the viscosity, within the pressure range 
 studied. 
 
 The same relation (i) was derived later by O. E. Meyer. 3 By an 
 experimental arrangement similar to Maxwell's he checked the theoretical 
 deductions for the same range of pressures. Results at pressures below 
 1/60 atmosphere showed a falling off of the viscosity coefficient, which 
 he later ascribed to the fact that, in the theory of the experimental 
 method, the external friction (e) has been considered infinitely large in 
 comparison with the internal friction or viscosity. Thus he introduced 
 into the Kinetic Theory of Gases the slip coefficient f = rj/e, which 
 Helmholtz 4 had previously defined for liquids. 
 
 II. DETERMINATIONS OF THE COEFFICIENT OF SLIP IN AIR. 
 
 The first experimental determination of the value of the coefficient of 
 slip was made by Kundt and Warburg 5 who used the capillary tube 
 method. Their results were: 
 
 Tube No. j Pressure. Temperature. 
 
 I 
 
 33.8 mm 
 
 15 c 
 
 .00017 
 
 76 X io~ 7 
 
 2 
 
 39-0 ' 
 
 
 .00016 
 
 82 X io~ 7 
 
 2 
 
 33-8 " 
 
 
 .00018 
 
 80 X io~ 7 
 
 
 
 
 Mean 
 
 79 X io~ 7 
 
 _. 
 
 
 
 
 
 The mean of their values at 15 C and 76 cm is about 79 X io~ 7 . 
 Millikan in a preceding article has reduced this to 23 C, getting f 76 at 
 23 C = 82 X io- 7 . 
 
 1 Using Stokes' value Vij/p = .116 for air, Maxwell made the first calculation of the 
 mean free path of a gas molecule. 
 
 2 Phil. Trans., 1866, Vol. 156, p. 249. 
 
 3 Pogg. Ann., 1865, Vol. 125, p. 177. 
 
 4 Wiener Sitzung., 1860, Vol. 40, p. 607. 
 6 Pogg. Ann., 1876, Vol. 159, p. 399.
 
 COEFFICIENT OF SLIP IN AIR. 241 
 
 In the determination of the elementary electrical charge by the falling 
 drop method Millikan * found that very small droplets of oil did not 
 follow Stokes' equation 
 
 (2) v = 2 ~^( ~ P)- 
 
 From his observations he made an empirical correction of the above 
 equation, writing the corrected law in the form 
 
 (3) . : -?<.-. 
 
 where Al was determined from the curve of his observations. Millikan 
 pointed out that the correction of Stokes' Law for Slip z gave 
 
 which, for small values of f/o, reduces to 
 
 (5) *=-^(<r-p)(i + r/a). 
 
 9 1 
 
 Thus, the Al determined by Millikan was really the coefficient of slip 
 for oil and air. His value for 23 C was f 76 = 77 X io~ 7 . In a later 
 determination 3 a more detailed study of the failure of Stokes' Law for 
 small oil drops gave the value 76 at 23 C = 82.2 X io~ 7 . 
 
 For small drops of shellac, Lee 4 observed Al = fr 6 at 76 cm and 23 C 
 to be 100 X io~ 7 . 
 
 These experimental results led Millikan to a theoretical study of slip 
 for different boundary conditions. He concluded that when no gas 
 molecules were regularly reflected after impact upon the walls, the 
 maximum of external friction, and hence the minimum of slip, would 
 result. For a mechanically rough surface, which would cause such 
 diffuse reflection of all gas molecules, he calculated the minimum value 
 of f at 23 C and 76 cm to be 65.9 X lO" 7 . 
 
 III. THEORY FOR SLIP DETERMINATIONS BY THE CONSTANT DEFLECTION 
 
 METHOD. 
 
 The theory of the Constant Deflection Method of determining viscosity 
 coefficients gives 
 
 1 PHYS. REV., 1911, 32, 382. 
 
 2 Bassett, Hydrodynamics, Vol. II., p. 271. 
 J PHYS. REV., 1913, Vol. II., p. 139. 
 
 4 PHYS. REV., 1914, IV., 420.
 
 242 LELAND JOHNSON STACY. 
 
 I = moment of inertia of suspended cylinder ; <f> angular displacement 
 of suspended cylinder; a = radius of suspended cylinder; / = length of 
 suspended cylinder; T = period of oscillation of suspended cylinder; 
 co = angular velocity of outer cylinder; b = radius of outer cylinder, 
 when it is assumed that there is no slip at the surfaces of the cylinders. 
 At ordinary pressures, this assumption involves an error too small to be 
 observed by this method. For the case where slip becomes appreciable 
 the complete equation as developed by Millikan is 
 
 (7) u = -. 
 
 a 2 \/ , 6 3 + a 3 \ 
 
 2 J\ ^ P ab 3 -a s b) 
 
 The slip term at atmospheric pressure may be calculated for this appa- 
 ratus since a = 5.3412 cm, b = 6.0632 cm, and f 76 = 66 X io~ 7 (from 
 theory). This gives a value for the term 
 
 b 3 + a 3 
 
 2{ P -=~ - = .00002 
 ab 3 a 3 b 
 
 which is quite negligible, since the experimental error is about .1 per cent. 
 Hereafter the equation (6) will be written 
 
 and tip denned as the apparent viscosity coefficient at the pressure p. 
 The value 1776' will be taken as the true value of the viscosity coefficient. 
 Equation (7) may now be written 
 
 and solving for f p 
 
 (10) fp = (-^-- 
 
 Thus to determine the slip coefficient by this method it is necessary, 
 first, to determine the coefficient of viscosity (77 = T7 76 ') and then ij p at 
 some other pressure. Application of equation (10) gives f p , and f 76 is 
 calculated from the formula f 76 = fp(/76). 17' and 77 p must, of course, 
 be observed at the same temperature or reduced to the same conditions 
 by the proper formula. 
 
 The investigation of slip coefficients by this method was undertaken, 
 at Prof. Millikan's suggestion, by E. L. Harrington x to determine 
 whether the slip depended on the nature of the surface as the lack of 
 
 1 PHYS. REV., 1916, VIII., p. 738.
 
 COEFFICIENT OF SLIP IN AIR. 
 
 agreement between the values for oil and shellac drops had indicated. 
 Harrington devoted his time chiefly, however, to the improvement of 
 the precision of the method of determining the viscosity coefficient, 
 leaving the writer to carry on the slip determinations. The writer 
 assisted Harrington for a short time in the first determinations of the 
 slip coefficient. Some of Harrington's results will be in- 
 cluded in this paper. 
 
 IV. ADAPTATION OF THE APPARATUS FOR SLIP DETER- 
 MINATIONS. 
 
 The Constant Deflection Apparatus consists essen- 
 tially of two concentric brass cylinders, the inner, /, 
 being suspended on an elastic suspension, s, so that, 
 when the outer cylinder, O, is driven at a constant speed 
 by a clock driving mechanism, K, a constant torque due 
 to viscosity will cause a constant deflection of the sus- 
 pended cylinder from its equilibrium position. To elim- 
 inate end effects the inner cylinder is suspended between 
 two guard rings, G, of the same diameter and less than 
 .3 mm from it. A small mirror mounted at the base of 
 the suspension wire makes it possible to observe the de- 
 
 Fig, i. 
 
 flection by a telescope and scale. The period of rotation (t = 27r/o>) of 
 the outer cylinder was determined by a chronograph attached to the 
 driving mechanism. 
 
 The constants of the apparatus as determined by Harrington were used. 
 
 Moment of inertia of inner cylinder / = 7617.3 
 
 Radius of inner cylinder a = 5.3410 cm 
 
 Length of inner cylinder / = 24.88 cm 
 
 Radius of outer cylinder b = 6.0632 cm 
 
 For work at low pressures, the cylinders were set upon a steel plate, 
 P, about 30 cm in diameter. A large-mouthed glass bottle, /, about 
 28 cm in diameter and 62 cm high was found and the mouth ground 
 plane to fit tightly a rubber gasket laid on the steel base-plate. An 
 opening was ground in the bottom of the bottle for the suspension head, 
 T, which projected some 25 cm above the large bottle. A glass tube, A, 
 sealed at the top was fitted by a ground glass joint into the bottom of the 
 large bottle thus closing the apparatus. A plate glass observing window 
 was sealed over a small hole in the side of the suspension cover to prevent 
 distortion of the image seen in the observing mirror. A discharge tube 
 for spectroscopic work was also sealed into the side of the suspension cover. 
 The driving shaft for the rotating cylinder was led through an iron pipe,
 
 244 LELAND JOHNSON STACY. 
 
 Q, sealed into the steel base-plate. The lower end of this pipe was 
 immersed in a vessel of mercury, C, thus sealing it from the outside. 
 A rim around the edge of the base-plate made it possible to use a mercury 
 seal at this joint and, the bottom of the bottle being somewhat concave, 
 a mercury seal was also used at the upper joint. Thus the apparatus 
 was enclosed tightly enough to permit evacuation to about .001 mm 
 pressure. 
 
 A Gaede mercury pump backed by a rotary oil pump made it possible 
 to reduce the pressure within the apparatus to .1 mm in a little over 
 two hours. The volume of the apparatus was about 150 liters. Pres- 
 sures were read by a McLeod gauge calibrated to read .0001 mm. Tem- 
 peratures within the apparatus were read from a Beckmann thermometer 
 which had been calibrated by comparison with a standard Baudin 
 instrument. The observing telescope and scale were mounted about 
 200 cm distant from the suspended mirror and the deflection could be 
 read to .1 mm. The steel suspension wire used throughout this work 
 gave an observed scale deflection of 62 to 64 cm when the period of rota- 
 tion of the.outer cylinder was about 30 sec. 
 
 V. EXPERIMENTAL METHODS AND ELIMINATION OF ERRORS. 
 
 The determination of the viscosity coefficient was made from the 
 formula 
 
 / -.-,- t , S < 2?T - S 
 
 rj p = K tan" 1 where / = ; $ = tan" 1 > 
 T 2 2# co 2a 
 
 The period T and the scale distance d being determined beforehand, it 
 was only necessary to observe 5 and / so that r) p ' could be calculated. 
 Temperatures and pressures were read as already described. 
 
 Having determined f]^'(= >/), it was necessary to determine -rj p ' at 
 some low pressure and calculate fre from the equations 
 
 *:,\tt.!t .*: 
 
 For atmospheric pressure the observed scale deflection 5 was about 63 
 cm with an error of .3 mm. At 0.12 mm pressure and t = 30 sec. the 
 scale deflection was found to be about 56 cm. Thus a difference of 7 cm 
 with an error of .3 mm in reading would give an apparent error of not 
 more than I part in 200. The gauge reading to .0001 mm, this error 
 should be only i part in 1,200. Errors in observing all the other factors 
 were much less than these, so they may be neglected. It was found,
 
 COEFFICIENT OF SLIP IN AIR. 245 
 
 however, that the principal source of inaccuracy was due to change of 
 the pressure during observations. This rise of pressure was considerable 
 in the first few hours after evacuation was stopped, but reached a steady 
 value after 24 hours or so. It was found by experiment that during the 
 lo-minute interval necessary for one complete observation the pressure 
 change was from .001 to .003 mm. After one or two days the pressure 
 rose less than that in 24 hours. This was explained by supposing the 
 increase in pressure to be due to gases released from the glass and metal 
 surfaces. By taking pressure readings before and after each observation 
 a fairly accurate mean value was obtained. The temperature was also 
 read at frequent intervals and the mean value used. The apparatus 
 being set up in a constant temperature room, the observations were made 
 within a very narrow range (22 to 24 C). Variations from 23 C were 
 corrected for by Millikan's formula (Ann. der Physik, 1913, Vol. 41, p. 
 
 759)- 
 
 -no = >?23 - .000000493(23 - 0). 
 
 In the early work on slip determinations, a series of observations was 
 made after a single evacuation. The values of fre calculated from these 
 observations gave an initial value of about 70 X io~ 7 but rose steadily 
 until a value of 200 X io~ 7 was found about two weeks later. The 
 pressure change was from .1238 to .1448 mm during this interval. Since 
 f 7 e should be independent of the pressure, this indicates that the increase 
 in pressure must be due to the presence of some gas of a lower viscosity 
 than air. This suggested that hydrogen (viscosity about one half that 
 of air) was being released from occlusion by the metal parts of the 
 apparatus. Spectroscopic examination of the discharge tube showed a 
 definite increase in the intensity of the hydrogen lines when the appa- 
 ratus was allowed to stand several days at a low pressure. 
 
 Admission of air to full atmospheric pressure to flush out the apparatus 
 and a second evacuation gave the same result, viz., a low value during 
 the first two or three hours after the pumps were stopped, then a steady 
 rise in the value of the slip constant fre. To eliminate this variation it 
 was found advisable to admit air immediately after a set of observations 
 was completed and to evacuate only a short time before readings were 
 to be taken. Thus the time during which the "hydrogen effect" might 
 be present was so short that it did not affect the results appreciably. 
 It was found that the values of f 76 obtained within three or four hours 
 after an evacuation were quite consistent. Observations were usually 
 made within an hour after evacuation. After many evacuations this 
 "hydrogen effect" was less marked but was always present. By taking 
 observations shortly after evacuation, it was avoided.
 
 246 
 
 LELAND JOHNSON STACY. 
 
 VI. TABLE OF OBSERVATIONS AND CALCULATED DATA FOR BRASS 
 SURFACES IN AIR. 
 
 Results on Brass Surfaces in Air. 
 d = 200.5 cm; T = 175.48; 11 = 1822.6 X io~ 7 . 
 
 5 (cm). 
 
 * (sec.). 
 
 e ( C). 
 
 7?p X I0 7 . 
 
 r P x io". 
 
 p (mm). 
 
 fn x io 7 . 
 
 56.28. . . . 
 56.56. . . . 
 
 . 30.134 
 30.040 
 
 23.04 
 23-05 
 
 1643.0 
 
 1646.0 
 
 3884 
 3811 
 
 1303 
 .1328 
 
 66.6 
 66.6 
 
 56.71---. 
 56.93 -.. 
 
 30.034 
 29.989 
 
 22.76 
 22.81 
 
 1649.4 3701 
 
 l6 53.2 3616 
 
 .1361 
 .1388 
 
 66.4 
 66.0 
 
 57-23.... 
 55-95.... 
 
 56.76. . . . 
 56.63.... 
 
 29-500 
 30.267 
 
 29.992 
 30.124 
 
 22.67 
 22.67 
 
 22.72 
 22.72 
 
 1634.7 4044 
 1640.2 3912 
 
 1646.0 3726 
 1646.2 3721 
 
 .1238 
 .1266 
 
 .1340 
 .1362 
 
 65-9 
 65-2 
 
 65-7 
 66.7 
 
 55-77.... 
 56.25.... 
 
 30.187 
 29-975 
 
 22.81 
 22.82 
 
 1628.2 4166 
 1630.6 4 IO 7 
 
 .1200 65.8 
 .1209 65.3 
 
 57-25 -.. 
 57-02.... 
 
 29.720 
 29.880 
 
 23-70 
 23.66 
 
 1645.0 
 1647.1 
 
 3856 
 
 3797 
 
 1253 
 .1268 
 
 63-6 
 63-4 
 
 57-10.... 
 56.68.... 
 
 29-795 
 30.032 
 
 22.81 
 22.81 
 
 1644.8 
 1645-9 
 
 3764 
 3738 
 
 .1287 
 -I3II 
 
 63-8 
 64-5 
 
 56.71--.. 
 56.86. . . . 
 
 30.134 
 30.079 
 
 22.80 
 22.83 
 
 1652.9 
 1653.8 
 
 3570 
 3556 
 
 .1361 
 1367 
 
 63-9 
 64.0 
 
 57-20. . . . 
 57-27.-.. 
 
 29.929 
 29.922 
 
 22.82 
 22.79 
 
 1655-1 3520 
 1656.7 3480 
 
 .1405 
 .1428 
 
 65-1 
 65-4 
 
 57-15.-.. 
 57-05-... 
 
 30.000 
 30.053 
 
 22.80 
 22.73 
 
 1657.6 3459 
 1657-7 3451 
 
 .1445 65.8 
 
 .1459 66.2 
 
 57.32-... 
 57-30. . . . 
 
 29.922 
 29.930 
 
 22.79 
 22.77 
 
 1656.2 
 1656.5 
 
 3437 
 3444 
 
 .1467 
 1475 
 
 66.3 
 66.8 
 
 56.51-... 
 
 29.952 
 
 22.87 
 
 1640.5 3836 
 
 .1368 69.0 
 
 55.77.... 
 55.65..-. 
 
 29.917 
 30.000 
 
 22.84 
 22.78 
 
 1614.0 
 1614.6 
 
 4469 
 4436 
 
 .1178 69.3 
 .1166 68.1 
 
 56.47.... 
 
 29.873 
 
 22.90 
 
 1631.2 4053 
 
 .1300 69.3 
 
 56.53- 
 
 30.037 
 
 22.42 
 
 1641.9 3733 
 
 .1402 68.9 
 
 56.73-.. - 
 
 30.043 
 
 23-01 
 
 1651.4 3405 
 
 .1508 67.6 
 
 56.62... . 
 
 29-915 
 
 22.73 
 
 1637.0 3886 
 
 .1343 68.7 
 
 56.68.... 
 
 30.102 
 
 22.84 
 
 1649.8 3591 
 
 .1349 63.7 
 
 56.46. . . . 
 56.76. . . . 
 
 30.100 
 30.018 
 
 22.81 
 22.79 
 
 1647.1 3646 
 1647.4 3630 
 
 .1378 66.1 
 .1424 68.0 
 
 56.68.... 
 57.00. . . . 
 
 29.972 
 29.860 
 
 22.87 
 22.99 
 
 1642.6 
 1645.6 
 
 3762 
 3710 
 
 .1308 64.7 
 .1338 65.3 
 
 57-03.... 
 
 30.158 
 
 23-33 
 
 1662.9 
 
 3338 
 
 1558 
 
 68.3 
 
 56.34- 
 
 30.080 
 
 23-46 
 
 1638.8 
 
 3915 
 
 .1258 
 
 64.8 
 
 Mean . . . 
 
 
 
 
 
 
 66.15
 
 COEFFICIENT OF SLIP IN AIR. 247 
 
 The observations were made within a pressure range of .1 to .18 mm 
 since at lower pressures the error due to change of pressure during an 
 observation was large, while at pressures above .18 mm the deflection 
 5 differed so little from the deflection at atmospheric pressure that 
 the error of observing this difference became considerable. 
 
 In general, the experimental conditions were as follows: 
 
 Temperature 22 C to 24 C 
 Pressure .1000 mm to .1800 mm 
 Period (t) 30 sec. (.2) 
 Period (r) - 175.5 sec. 
 
 Scale deflection, for p = 76 cm -630 mm 1 
 
 " p = o.i 8 mm - 580 mm > approximately 
 " p = o.io mm - 550 mm J 
 
 Observations were usually made in pairs within an hour after evacua- 
 tion. 
 
 The mean value (66.15 X io~ 7 ) is very close to Millikan's theoretical 
 minimum value (65.9 X io 7 ) but is considerably lower than any of the 
 values found by other methods for oil or for glass surfaces. 
 
 VII. DETERMINATION OF SLIP FOR SHELLAC SURFACES IN AIR. 
 
 The cylinders were next coated with a thin layer of shellac, dried by 
 an air blast and replaced in position. From the weight of the cylinders 
 before and after the shellac was applied and the surface area of the 
 cylinders, the average thickness of the shellac film was calculated. This 
 was found to be .01 mm and made only a small correction in the values 
 of the constants a and b. 
 
 After the determination of the viscosity coefficient at atmospheric 
 pressure, the pressure was reduced and observations made as before. 
 At first a high value of fre = 97 X io~ 7 was found while the shellac was 
 fresh. Later, the values of fa dropped steadily toward the minimum 
 value. This result is shown in the two sets of data here given ; the first 
 set of observations being due to Harrington and the second to the writer. 
 In two other trials, accidental experimental difficulties prevented the 
 writer from making observations while the shellac was fresh but values 
 between 80 and 90 X io~ 7 were found several days after application 
 of the shellac. The fall in the observed value of the slip coefficient 
 indicates a change in the surface due probably to oxidation of the 
 shellac which seems to produce a rough surface.
 
 248 
 
 LELAND JOHNSON STACY. 
 
 Data on Fresh Shellac Surfaces in Air. 
 
 d = 200.6, T = 176.00. 
 E. L. Harrington: Fresh Shellac, August i, ipi6. 
 
 Date. 
 
 S. 
 
 t. 
 
 e. 
 
 ij/ X io 7 . 
 
 TP X IO B . 
 
 p (mm). 
 
 ft. X io 7 . 
 
 Aug. 9 
 
 55-93 
 
 29.719 
 
 23-38 
 
 1602.8 
 
 4911 
 
 .1474 
 
 95-2 
 
 
 55-72 
 
 29.727 
 
 23-44 
 
 1597-3 
 
 5059 
 
 .1478 
 
 98.5 
 
 Aug. 10, A.M. . 
 
 53-42 
 
 29-715 
 
 23.76 
 
 I53L3 
 
 6842 
 
 .1048 
 
 94-3 
 
 
 52-55 
 
 29.960 
 
 23-86 
 
 I5I9-4 
 
 7188 
 
 .1058 
 
 IOO.O 
 
 Aug. 10, P.M.. 
 
 53-99 
 
 29.936 
 
 23.67 
 
 1559-0 
 
 6080 
 
 .1125 
 
 90.0 
 
 
 53-19 
 
 30.263 
 
 23.60 
 
 1552.9 
 
 6240 
 
 1133 
 
 93-0 
 
 Aug. ii, A.M.. 
 
 53-43 
 
 30.227 
 
 23.46 
 
 1556.6 
 
 6121 
 
 .1107 
 
 89.0 
 
 
 52.96 
 
 30.440 
 
 23-50 
 
 1553-9 
 
 6204 .1112 
 
 90.8 
 
 Aug. ii, P.M.. 
 
 55-19 
 
 30.100 
 
 23-13 
 
 1602.0 
 
 4886 
 
 .1288 
 
 82.8 
 
 
 55-H 
 
 30.100 
 
 23-13 
 
 1600.5 
 
 4920 
 
 .1298 
 
 84.1 
 
 Aug. 12 
 
 56-94 
 
 29.827 
 
 23.64 
 
 1637.0 
 
 4099 
 
 .1438 
 
 77-6 
 
 
 56.75 
 
 29.917 
 
 23.70 
 
 1636.5 
 
 4108 
 
 .1438 
 
 77-7 
 
 Aug. 30 
 
 57-28 
 
 29.780 
 
 23.10 
 
 1653-1 
 
 3538 
 
 .1506 
 
 70.1 
 
 L. J. Stacy: Fresh Shellac, March 7, 1917. 
 
 Mar. 9, A.M.. 
 
 56.04 
 
 29.789 
 
 23-36 
 
 1578.8 
 
 5588 
 
 .1330 
 
 97-7 
 
 
 55-92 
 
 30.010 
 
 23.40 
 
 1582.5 
 
 5381 
 
 1352 
 
 
 Mar. 9, P.M.. 
 
 57-66 
 
 29.800 
 
 23-14 
 
 1619.7 
 
 4580 
 
 .1470 
 
 88.6 
 
 
 57-26 
 
 30.080 
 
 23.19 
 
 1623.7 
 
 4486 
 
 .1482 
 
 87.5 
 
 Mar. io, A.M.. 
 
 58.15 
 
 29-723 
 
 23-55 
 
 1628.2 
 
 4414 
 
 .1418 
 
 82.3 
 
 
 57-76 
 
 29.983 
 
 23-59 
 
 1632.4 
 
 4318 
 
 .1441 
 
 81.9 
 
 Mar. io, P.M.. 
 
 57.56 
 
 30.164 
 
 23-53 
 
 1636.7 
 
 4206 
 
 1493 
 
 82.6 
 
 
 58.12 
 
 29.896 
 
 23-50 
 
 1637.7 
 
 4179 
 
 1543 
 
 84.8 
 
 
 57-30 
 
 30.322 
 
 23.24 
 
 1637-9 
 
 4152 
 
 1545 
 
 84.4 
 
 Mar. ii 
 
 56.83 
 
 29.985 
 
 22.87 
 
 1 606.6 
 
 4876 
 
 I3I3 
 
 84.2 
 
 
 56.06 
 
 30-432 
 
 22.86 
 
 1608.7 
 
 4820 
 
 1323 
 
 83-9 
 
 Mar. 12 
 
 57.38 
 
 30.334 
 
 23.12 
 
 1640.8 
 
 4066 
 
 .1487 
 
 79-5 
 
 
 57-77 
 
 30.188 
 
 23.21 
 
 1643.8 
 
 3999 
 
 1495 
 
 78.7 
 
 Mar. 13 
 
 57-75 
 
 30-144 
 
 23-52 
 
 1640.9 
 
 4102 .1443 
 
 77-9 
 
 
 58-05 
 
 30.031 
 
 23-52 
 
 1643.1 
 
 4048 .1473 
 
 78.4 
 
 Mar. 16 
 
 58-16 
 
 30.169 
 
 23.07 
 
 1653-7 
 
 3747 
 
 .1518 
 
 74.8 
 
 Mar. 17 
 
 57-42 
 
 30.084 
 
 22.85 
 
 1628.4 
 
 4242 
 
 .1332 
 
 74-3 
 
 The following results were obtained using the same experimental 
 method as with brass surfaces. The shellac surfaces were about two 
 months old when the first readings were taken. 
 
 The results of these experiments furnish independent evidence of the 
 fact that the viscosity of a gas is independent of the pressure, since foe 
 turns out to be a constant for a given surface in air. The value of fre 
 for rough surfaces checks Millikan's theoretically deduced values within 
 the limit of error of the experiment. The variation in slip for different 
 surfaces has been checked and the result for fresh shellac, f 76 = 96.8X io~ 7 , 
 is very close to the result Lee obtained, fre = 100 X io~ 7 , from the 
 correction of Stokes' Law for falling shellac drops. 
 
 In conclusion the writer wishes to express his indebtedness to Professor 
 R. A. Millikan who suggested the problem and directed the experimental
 
 COEFFICIENT OF SLIP IN AIR. 
 
 249 
 
 Data for Old Shellac Surfaces in Air. 
 d = 200.6 cm; T = 176.00 sec. 
 
 5. 
 
 t. 
 
 e. 
 
 r, p ' X I0 7 . 
 
 f, X 10 
 
 P (mm). 
 
 fte X io 7 . 
 
 55-39 .... 
 
 30.027 
 
 23.67 
 
 1605.1 
 
 4833 
 
 .1088 
 
 69.2 
 
 57.49. . . . 
 57.06 
 
 29.767 
 30.076 
 
 24.28 
 24-34 
 
 1648.5 
 1653-3 
 
 3875 
 3767 
 
 .1338 
 .1370 
 
 68.2 
 67.9 
 
 5749. 
 
 30.023 
 
 23.11 
 
 1661.5 
 
 3443 
 
 1535 69.5 
 
 57.02.... 
 57.20. . . . 
 
 30.058 
 30.123 
 
 22.76 
 22.87 
 
 1653.8 
 1662.5 
 
 3639 
 3431 
 
 .1444 
 .1478 
 
 69.1 
 66.7 
 
 57.38- 
 57-28.... 
 
 30.023 
 30.074 
 
 22.90 
 22.95 
 
 1661.7 
 I662.I 
 
 3447 
 3444 
 
 .1488 67.5 
 .1493 67.7 
 
 57.64- - 
 
 57-72. . . . 
 
 29.885 
 29.930 
 
 22.92 
 22.97 
 
 1659.2 
 1664.4 
 
 3507 
 3392 
 
 .1478 68.2 
 .1495 66.7 
 
 57-44.... 
 57-60.... 
 
 30.144 
 30.050 
 
 23.08 
 23.11 
 
 1668.9 
 1668.2 
 
 3299 
 3317 
 
 1573 68.3 
 .1585 69.2 
 
 56.75 .. 
 56.90. . . . 
 
 30.111 
 30.112 
 
 23.24 
 23.28 
 
 1647.3 
 1651.1 
 
 3830 
 3744 
 
 1351 
 .1373 
 
 68.1 
 67.6 
 
 57-I3-- 
 
 30.067 
 
 23.21 
 
 1656.0 
 
 3613 
 
 1437 68.3 
 
 55.19..-. 
 55.23...- 
 
 57.15.... 
 57.27.... 
 
 30.014 
 30.087 
 
 30.210 
 30.271 
 
 23.06 
 23.10 
 
 23-57 
 23.66 
 
 1597-2 
 1602.3 
 
 1664.0 
 1670.8 
 
 5032 
 4910 
 
 3464 
 33i6 
 
 1057 
 .1076 
 
 .1498 
 1533 
 
 70.0 
 
 69-5 
 
 68.3 
 66.9 
 
 56.91..-. 
 57.I7-... 
 
 30.211 
 30.174 
 
 23-21 
 
 23-31 
 
 1657.2 
 1662.6 
 
 3585 
 3469 
 
 .1411 66.6 
 .1445 66.0 
 
 57.36. . . . 
 57.37...- 
 
 29.963 
 30.019 
 
 22.55 
 22.58 
 
 1650.5 
 1652.8 
 
 3823 
 3793 
 
 1353 
 .1380 
 
 68.1 
 68.6 
 
 56.74. . . . 
 
 30.300 
 
 22.51 
 
 1644.5 
 
 3780 
 
 .1288 
 
 64-1 
 
 57.33..-. 
 57.25...- 
 
 30.161 
 30.160 
 
 22.87 
 
 22.86 
 
 1655.3 
 1652.6 
 
 3742 
 3806 
 
 1356 
 .1364 
 
 66.8 
 68.3 
 
 58.38.-.- 
 57.23..-. 
 
 29.508 
 30-152 
 
 21.94 
 
 22.12 
 
 1647-3 
 1655.5 
 
 3835 
 
 3777 
 
 1363 
 
 -1373 
 
 68.8 
 68.2 
 
 Mean . . . 
 
 
 
 
 
 67.7 
 
 This result is 2\ per cent higher than the value found for brass surfaces in air. 
 
 work ; to the other members of the Physics Department of the University 
 of Chicago for their interest and assistance throughout the investigation ; 
 and, in particular, to Dr. E. L. Harrington with whom the author 
 worked in the first determination of slip by this method. 
 RYERSON PHYSICAL LABORATORY, 
 UNIVERSITY OF CHICAGO, 
 September 22, I922. 1 
 
 *The work described in this paper was completed in February 1919.
 
 
 Pftys 
 DATE 
 
 ! ' ca ' Sciences 
 DUE 
 
 M&raiy 
 
 Phys.Sci. 
 > QC189 Harrington, E. 
 
 
 37J 
 
 
 
 H37 
 A redetermination of the 
 
 
 
 
 
 coefficient of viscosity 
 
 
 
 
 
 of air. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 1 1 III! LI 1 II II 1 II 1 
 
 
 
 
 
 lillH 
 
 
 
 
 
 3 W*JC^ee524221 
 
 
 
 
 
 Gty 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 Phys . Sci 
 
 
 
 
 
 QC189 Harrington, E 
 - H37 
 
 
 
 
 
 A redetermination of the 
 
 GAYLORD 
 
 
 
 PR.NTEO.NU.S.A. 
 
 coefficient of viscosity 
 of air 
 
 
 
 
 
 
 
 
 
 
 3ATEtOANED| ISSUED TO } DATE DUE 
 
 Physical Sciences Library 
 
 University of California 
 
 Riverside
 
 TIP 
 
 eprinted from NATURE, Vol. 136, page 682, October 26, 1935.) 
 
 Viscosity of Air and the Electronic Charge 
 ' 
 
 THE greatest uncertainty in determining the 
 electronic charge e by the oil drop method of 
 Millikan is introduced by the uncertainty in the 
 assumed value of the coefficient of viscosity of air, TJ. 
 The value adopted by Millikan in 1917 
 
 T] 23 = (1822-6l-2) x 10- 7 
 
 is probably too low, and its accuracy overestimated, 
 as is pointed out by Shiba 1 . 
 
 Considering the fundamental importance of the 
 constant e, I have undertaken a new determination 
 of v), using the rotating cylinder method also em- 
 ployed by Millikan and his co-workers 2 > 3 : An inner 
 cylinder of electron metal, suspended vertically by 
 a fine phosphor-bronze wire between two guard 
 cylinders of equal diameter is deviated from its 
 equilibrium position through an angle 9 by a con- 
 centric outer cylinder, rotating with constant velocity, 
 Y) being calculated from the equation 
 
 / (ft 2 - a 2 ). 9 * 
 73 = 2 a*6* Z T 2 ' where 
 a = the radius of the inner cylinder = 2-81767 cm. 
 
 at 20 ; 
 
 6 = the radius of the outer cylinder = 3-26628 cm. 
 at 20 or = 3-18328 cm. at 20 (two different 
 cylinders) ; 
 I = the length of the inner cylinder = 9-9981 cm. 
 
 at 20; 
 t = the time of revolution of the outer cylinder 
 
 (20-150 sec.); 
 
 T = the period of oscillation of the suspended system 
 
 (53 128 sec., using different suspensions) ; 
 
 / = the moment of inertia of the suspended system 
 
 about the line of suspension = 423-22 gm.cm 2 . 
 
 The mean value of 51 determinations of TJ for dry 
 air at temperatures between 18-9 and 20-9 is 
 
 7] 20 = (1820-0 3-0) x 10- 7 corresponding to 
 7] 23 = (1834-8 3-0) x 10- 7 . 
 
 From this we get 
 
 /1834-8\ 3 / 2 
 e = (ig22^6J x 4 ' 770 x 10- 10 = (4-818 0-012) x 
 
 the uncertainty stated being due only to the viscosity, 
 other sources of error not being considered here. 
 I am, therefore, of the opinion that the discrepancy