]\gt ; On Kinetic bility .
By HORACE LAMB , F.RS .
} ( Received November 21 , \mdash ; Read December 12 , \mdash ; Revised December 19 , 1907 .
) 1 .
The object of this paper is to illustrate the theory of kinetic stability , so far as such a theory can be said to exist , by a few simple examples .
As the theory itself appears to be by no means widely known , some preliminary recapitulation seems advisable .
The difficulty of framing a definition of kinetic stability which shall be comprehensive and at the same time conform to natural prepossessions has been recognised .
* Thus , according to one definition which has been proposed , the vertical fall of a particle under gravity would be unstable ; according to another the revolution of a particle in a circular orbit about a centre of force varying inversel .
as the cube of the distance would be reckoned as stable , although the slightest disturbance would cause the particle either to fall ultimately.into the centre , or to recede to infinity , after describing in either case a spiral path with an infinite number of convolutions .
There are , however , certain restricted classes of cases where a natural definition of stability is possible and the corresponding criterion can be formulated .
Suppose , in the first ] , that we have a dynamical system which is the seat of cylic motions whose momenfa ( in the generalised are constant .
Apart from the cyclic motions the configuration depends on a certain number of " " palpable\ldquo ; -ordinates , , , and an " " equilibrium\ldquo ; configuration is one in which these co-ordinates can remain constant when the system is left to itself .
Such an equilibrium configuration is said to be stable when the .
extreme variations of these co-ordinates , consequent on an arbitrary disturbance , are confined within limits which diminish indeiinitely with the energy of the disturbance .
Any alrangement of frictionless gyrostats gives a system of this kind ; on a larger scale we have the problem of the free rotation of a liquid mass under its own gravitation .
In a second class of cases we have ( again ) certain co-ordinates whose values do not affect the kinetic or the potential energy , and the corre- *Cf .
F. Klein .
A. Sommerfeld , ' Ueber die Theory des Kreisels , ' Leipzig , 1898 , p. 342 .
Cf .
Thomson and Tait , ' Natural Philosophy , ' S319 , example ( G ) ; Lamb , 'Hydromics , ' 1906 , SS140 , 141 .
On Kinetic Stability .
sponding vdocities are now supposed to be maintained constant by the application of suitable forces .
* We have then to investigate the stability ( in the same sense as before ) of an " " equilibrium\ldquo ; configuration in which the remaining co-ordinates , , have constant values .
As an example , the system may be attached to a rigid body which rotates with constant speed .
The theory of the stability of an ocean a rotating globe also comes under this elass .
It has been customary , in treatises on dyna mics , to discuss all such questions by the classical method of " " small oscillabions If the variations of the co-ordinates , be arded as infinitely small , the solution of the equations of disturbed motion is obtained in the form , ( 1 ) the values of being determined by an algebraical or ( in the case of an infinite number of degrees of freedom ) a transcendental equation .
If these values of are found to be all real and ative , the undisturbed configuration is reckoned as stable , whilst if any of them are positive or complex , it is accounted as unstable .
As familiar instances of problems discussed from this standpoint , we have the stability of the conical pendulum , of the steady precessional motion of a top , and so on .
The general theol.y of the method , including the condivions of stability ( in this sense ) , has been investigated by Routh .
M. PoincareS has , however , insisted on the fact that this method may , from a practical point of view , be misleading as to the ultimate behaviour of the system .
If deviations from the equilibrium configuration be resisted ( as in practice they always are ) by forces of a viscous character affecting the co-ordinates , , , then in the case of absolute ( statical ) equilibrium the usual criterion of stability , , that the potential energy must be a minimum , is not affected .
But in such cases of 1-inetic equilibrium as have been referred to , it may happen that the effect of the viscous forces is gradually to increase the deviation , even the equilibrium configul.ation is ( i.e. , from the\ldquo ; classical\ldquo ; standpoint ) thoroughly stable .
A distinction is drawn between " " ordi1ln \ldquo ; or " " temporary\ldquo ; stability , i.e. , stability as by the method of small oscillations , and ' secular\ldquo ; or ' permanent\ldquo ; stability , , stability when regard is had to possible viscous forces the co-ordinates Thomson and Tait , S319 , example .
HydrodyIlamics , ' 1906 , SS202 , 203 , 204 .
Stability of Motion , ' 1877 ; 'Advanced Rigid Dynamics , 6th 1905 , chap .
vi .
S " " Sir l'Equilibre Mass Fluide animee d'un Mouvement de .
Rotation , \ldquo ; ' Acta Math 1885 , vol. 7 , p. 269 .
above referred to , and in particular to the question of stability of equilibrium relative to a rigid body which is maintained in constant rotation about a fixed axis .
2 .
The trivial character of the first example may be excused on the ground that it shows almost intuitively the necessity for some qualification to the doctrine of ' ordinary\ldquo ; stability .
We consider a particle movable on the inner surface of a spherical bowl which rotates with constant angular velocity about the vertical diameter .
If the bowl be smooth the equilibrium of the particle when in the lowest position is " " ordinarily\ldquo ; stable , since the rotation of the bowl is quite irrelevant .
But if we admit the existence of friction , , between the particle and the bowl , the lowest position is " " permanently\ldquo ; stable only so long as , where is the radius .
This results immediately from the consideration of the formula for the kinetic potential , , ( 2 ) where is the mass of the particle , and is its angular distance from the lowest point .
When the above value of is exceeded , the only permanently stable position is that in which ' ( 3 ) See Poincare , .
cit. , or the author 's ' Hydrodynamics , ' .
cit. latest edition ( 1905 ) of Routh 's ' Advanced Rigid Dynamics ' contains no reference to the matter .
1907 .
] On Kinetic Stability .
when the particle rotates with the bowl like the bob of a conical pendulum .
To examine in detail the initial stage when the particle is slightly disturbed from its lowest position we may ( for mathematical convenience ) adopt the hypothesis of a frictional force vary ing as the relative velocity .
If we employ horizontal rectangular axes Ox , Oy passing through the lowest point , and rotating with the bowl , we have , when are small , ( 4 ) where is the frictional coefficient .
These equations may be combined into where .
If we assume , ( 6 ) we find , if the square of be neglected .
be Cartesian co-ordinates referred to fixed axes throughO , the complete solution is , ( 8 ) where .
( 9 ) If this be put in real form we perceive that the motion is made up of two superposed circular vibrations , in opposite directions , of period ; moreover that , if is positive , so that that circular vibration whose sense agrees with continually increases in amplitude .
The particle works its way outwards in an ever } spiral path , approximating to the stable position of relative equilibrium indicated by ( 3 ) .
3 .
The next illustration is of a more practical character , and admits of being realised with considerable exactness .
A pendulum symmetrical about a longitudinal axis hangs by a Hooke 's joint from a vertical spindle which is made to rotate with a constant angular velocity .
The pendulum used by the writer FIG. 1 .
was constructed originally without any reference to the ( 10 ) .
where denote the principal monlents of inertia of the pendulum at the centre of the joint .
Hence , if , we have provided denote the distance of the centre of gravity from the joint .
This expression ceases to be a minimum for , if lfgh ' and the only stable positions are then those in which the makes an angle with the vertical , iven by To examine the motion about the vertical position we neglect , in ( 10 ) , terms in and of higher order than the second .
Thus const .
( 14 ) .
Hence 's equations give As in the case of ( 4 ) , we find that these satisfied by 1907 .
] On Kinetic Stability .
provided or ( 19 ) The vertical position is therefore " " ordinarily\ldquo ; stable , whatever the value of It .
is evident that are the rectangular co-ordinates , relative to rotating axes , of a point on the axis of the pendulum .
For the corre- sponding co-ordinates relative to fixed axes we have , ( 20 ) where .
( 21 ) The motion is therefore made up of two superposed circular vibrations of different periods , the more rapid vibration being the one whose direction of revolution agrees with that of the spindle .
To investigate the of permanent stability we introduce into the left-hand members of ( 16 ) terms to represent the viscous forces at the joint .
The modified equations are satisfied by , ( 22 ) provided .
( 23 ) If be the two values of given by ( 19 ) , this may be written , ( 24 ) the roots of which are , if we neglect the square of When the two values of have opposite , and the real parts of are both ative .
The ver.tical position is then permanently , as well as " " ordinarily\ldquo ; stable .
But if both values of are negative , and if be the smaller in absolute magnitude , the real part of will be positive , and that of yative .
If we pass to fixed axes , writing as before , ( 26 ) we find that the periods of the two circular vibrations are to a first approximation unaffected by a small degree of friction , but that the amplitude of one of these vibrations , viz. , the one whose direction of revolubion agrees with that of the spindle , increases exponentially with the time , whilst the amplitude of the other sinks asymptotically to zero .
These points 174 Prof H. Lamb .
[ Nov. 21 , are illustrated in a striking manner by the apparatus referred to.* Sub- * stantially the same experiment can be made in a simpler form by means of a heavy metal ball hanging by a stout string from a hook at the lower end of the spindle .
If due precautions be taken to check the violent evolutions which the ball is sometimes apt in the first instance to perform , the torsion of the soon brings the latter into a state of steady rotation about a vertical diameter , with practically the angular velocity of the spindle .
When the steady state has been attained the ball may be left to itself , with the string vertical .
The friction of motion relative to the spindle is in this form of the experiment very slight , and a close observation may soon detect the tendency to a circular vibration of continually increasing amplitude in the direction of revolution of the spindle , some time may elapse before this becomes really conspicuous .
The final result is , however , unmistakable .
4 .
The question is not seriously modified by a slight amount of deyiation from the theoretical conditions , , in the problem of S 3 , by a slight defect of alignment between the axis of rotation of the spindle and the centre of the joint .
The configuration of relative equilibrium about which the observed oscillations take place is only slightly altered , except in the case cf approximate agreement between the imposed period of rotation and what would be the natural period of vibration in the absence of rotation .
The effects of a want of perfect alignment in S 3 can be studied in their simplest form if we the moment of inertia ( C ) about the axis of the pendulum .
The case is then that of a particle suspended from the lower surface of a horizontal disc , which is made to rotate about a vertical axis .
If be the length of the string , and the distance of the point of suspension from the axis of rotation , the inclination of the string to the vertical in a position of relative equilibrium is given b.y , ( 27 ) where .
If this has three solutions , for two of which is negative ; in one of these , moreover , is numerically greater , and in the other numerically less , It may be worth while to give roughly the dimensions .
The steel rod shown in fig. 1 had a length of 36 in .
and a thickness of .
The diameter of the iron disc which could be fixed in valious positions along the rod was 7 in .
and its thickness*in .
The spiudle was driven from a small electromotor , by means of the small pulley shown , at speeds ranging up to about 25 revolutions per second .
a typical experiment the ball was 3 in .
in diameter , and was suspended by a string 33 in .
long ; and the speed was about 7 revolutions per second .
The circular vibration took about 18 minutes to attain an amplitude of 1 inch .
1907 .
] On Kinetic Stability .
than .
These three positions are shown in fig. 3 .
The position is found to be both " " ordinarily\ldquo ; and " " manently \ldquo ; stable , whilst the position III is on either reckoning unstable .
Case II is " " permanently\ldquo ; unstable , but the question of " " ordinary\ldquo ; stability is less simple .
For sufficiently great values of the equilibrium may become unstable from this point of view , but there is no difficulty in the conditions so that FIG. there may be " " ordinary\ldquo ; staDility with " " permanent\ldquo ; instability .
This was illustrated by an experiment in which the excentricity lvas purposely made appreciable .
The metal ball referred to was suspended by a stout string about 3 feet long from a point 1 inch out from the centre of the disc .
If the ball be carefully steadied in the central position before left to itself , its subsequent demeanour differs in no essential way from what is observed when the suspension is made as nearly axial as possible .
5 .
The next example is one in which the number of degrees of freedom is nite .
For a reason to be given it is hardly a practical one , but it may serve to illustrate the limitations to which the application of the tbeory is subject .
We consider a cylindrical shaft rotating in fixed beariugs placed at isolated points , and the question is at what speed the straight form becomes unstable .
If the circumferential of the shaft be Prof. H. Lamb .
[ Nov. 2 small compared with the elastic wave-velocities of the material , the ular momentum about the axis may be ignored .
Under this condition it is obvious that the straight form is " " ordinarily\ldquo ; stable , the fact of the rotation being irrelevant .
To investigate the " " permanent\ldquo ; stability , consider , for definiteness , a length between two bearings , B. If the axis of be taken along the length of the shaft , and if denote the lateral deviation , we have , by the usual theory of flexure , , ( 29 ) , ( 30 ) where is the cross-section , is the radius of gyration about a diameter , denotes Young 's modulus , and is the density of the material .
Hence , , ( 31 ) where .
( 32 ) The integrated terms vanish if each end be either free , or 1nelely supported , or fixed also in direction .
It is known from the ordinary theory of transversal that any arbitrary function which is subject to the given terminal conditions can be expanded , for , in a series of normal functions , .
( 33 ) Here , satisfy the differential equations , ( 34 ) .and the proper terminal conditions , , , the roots of a certain transcendental equation , arranged in ascending order of magnitude .
If we substitute from ( 33 ) in ( 31 ) , and omit terms which vanish in consequence of the orthogonal property of different conjugate functions , we find .
( 35 ) *See Rayleigh , ' Theory of Sound , ' chap .
viii .
Thus , if the shaft be merely supported at the ends , the equation is ; if it be fixed in direction at one end and free at the other , we have 1907 .
] On Kinetic Stability .
The frequencies of the various modes of natural vibration of the shaft are determined by the relation . .
( 36 ) Hence is a minimum , in the straight condition , or the equilibrium is permanently stable , only so long as , i.e. , so long as the period of rotation is greater than that of the ravest mode of transverse vibration .
The incipient of the instability might be studied as in the previous problems .
The motion can be analysed into circular vibrations , and it appears that the amplitude oi one at least of these , having the salve direction of revolution as the shaft , should increase exponentially with the time , provided exceed the smallest root of the transcendental equation which determines We conclude that a truly symmetrical shaft , rotating accurately about axis , in rigidly fixed , with any speed exceeding that of the ravest mode of transverse vibration , would be rendered viscous ; cting the motion , such as are , in fact , present owing to the , internal friction of the substance .
The instability , indeed , take time to develop itself , but the result would be inevitable .
The fact that shafts can be , and are , safely driven at speeds the critical limit thus indicated* must be ascribed to the operation of dissipative forces ( so far fnored ) affecting the absolute as well as the relative vibrabions .
The seat of such forces is probably to be found in a of the bearings .
For similar reason the " " permanent\ldquo ; instability illustrated by the experiments of SS 2 , 3 above might be wholly masked if the resistance , of the air much greater than it actually is , or if the whole apparatus were immersed in } a viscous liquid .
* The observed " " whirling\ldquo ; of shafts at a series of critical speeds is due to of absolute symmetry , and is to be regarded as a forced oscillation of exaggerated amplitude , due to approximate synchronism .
( See Dunkerley , 'Phil .
, vol. 185 , Stodola , ' Die Dampfturbinen , ' Berlin , 1904 , p. 157 .
) .
LXXX.\mdash ; A.

