]\gt ; [ 27 ] II .
On the Potential of Etlipsoidal Bodies , and Figures of Equilibrium of Rotating Liquid Masses .
By J. H. JEANS , M.A. , F.R. S. Received , \mdash ; Read June 25 , 1914 .
BY an ellipsoidal body is meant , in the present paper , any homogeneous body which can be arrived at by continuous distortion of an ellipsoid .
If is the equation of the ellipsoid from which we start , and is a parameter , the distortion of the ellipsoid may be supposed to proceed by increasing from the value upwards , and the final figure may be taken to be For very small distortions the first two terms will adequately represent the distorted , and as we pass to higher orders the remaining terms will enter successively .
The potential problem , to some extent interesting in itself , derives its chief importance from its application to the determination of the possible figures of equilibrium of a rotating mass of liquid .
POINCARE , *using his ingenious method of double layers , has shown how the potential of an ellipsoidal body can be carried as far as the second-order terms when the distortion is small , but ve*no indication of how it is possible to carry it further , and indeed his method is one which hardly seems susceptible of being developed further than he himself has developed it .
It is clear , however , that progress with the problem of rotating liquids can only be made when a method is available for writing down the potential of an ellipsoidal body distorted as far as we please .
I believe the method explained in the present paper will be found capable of giving the potential of a body distorted to any extent , although ( for reasons which will be explained later ) I have not in the present paper carried the calculations further than second-order terms .
The theory of figures of equilibrium of rotating masses of liquid is at present in an unsatisfactory state .
It has been shown by Lord KELVIN that the Jacobian ellipsoid is stable at the point at which it coalesces with the Maclaurin spheroid , and it has been shown by POINCARE to remain stable up to the point at which * ' Sir la Stabilite ' de l'Equilibre des Figures Pyriformes affecte'ss par un Mass uide e\ldquo ; ' Phil. , vol. 198 , p. 333 .
[ Published , 1915 .
the series of Jacobian ellipsoids coalesces with the Poincare series of pear-shaped figures .
After this point the series of Jacobian ellipsoids must , in accordance with doctrine of exchange of stabilities at a point of bifurcation , lose its stability , but the question of how it loses its stability is in a state of doubt .
DABWIN believed he had proved the Poincare series to be initially stable , IAPOUNOFF has maintained that this series is initially unstable .
The importance of this question to theories of cosmogony is , of course , great , although perhaps liable to be overrated .
A caution of POINCARE'S may be borne in mind : " " Quelle que soit l'hypothese [ stability or instability ] que doive triompher un jour , je tiens a mettre tout de suite en guard contre les consequences cosmogoniques qu'on pourrait en tirer .
Les masses de la nature no sont pas homogenes , et si on reconnaissait que les figures pyriformes sont instables , il pourrait neanmoins arriver qu'une mass heterogene fut susceptible de prendre un form d'equilibre analogue aux figures pyriformes , et qui serait stable .
Le contraire pourrait d'ailleurs arriver e'galement.\ldquo ; The present investigation was tarted primarily in the hope of setting this question of stability at rest .
I realised that to make a new series of computations on the subject could be of little value , for whatever the result , there would have been two opinionS on the one side to one on the other .
Moreover , DARWIN has stated clearly that he does not think the divergence of opinion between M. LIAPOUNOFF and himself is one to be settled by new computationsS : " " I feel a conviction that the source of our disagreement will be found in some matter of .
principle I had hoped that it might be found possible to discuss the problem by a purely algebraical method , involving neither laborious computations nor intricate physical arguments , and that if such a discussion did not give a convincing and satisfying answer to the question in hand , at least it might reveal the source of disagreement between DARWIN and LIAPOUNOFF .
The result arrived at is one which , as will readily be understood from its nature , is only put forward with the utmost diffidence , but it is one from which I can find absolute]y no escape .
It is that underlying the whole question there is a complication , unsuspected equally by POINCARE , DARWIN , and ( iu so far as I can read his writings ) TNOFF , which renders nugatory the work of all these investigators on the stability of the pear-shaped figure .
If my method is sound , it appears , as will be explained later , that it is impossible to draw any inference as to the stability of the pear from computations carried only as far as the second order of small quantities .
The ' ' The Stability of the Pear-shaped Figure of Equilibrium of a Rotating NIass of Liquid ' Phil. Trans 200 ( 1902 ) , p. 251 ; also papers in 'Phil .
Trans 208 , .
] , and ' Proc. Roy .
Soc. , 82 ( 1909 ) , p. 188 , all combined in one paper in ' Coll. Scientific Papers , ' vol. 3 , p. 317 .
" " Sir un Problem de Tchebychef ' Memoires de l'Academie de St. Petersbourg , ' xvii .
, 3 ( 1905 ) .
Loc .
cit. , p. 335 .
S 'Coll .
Scientific Papers , ' 3 , p. 392 .
THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
materials for an answer to the question are to be found only through the thirdorder terms .
* Fortunately the method of the present paper admits of extension to the computation of third-order terms , and so it ought to , and I hope will be quite feasible to decide as to the stability or instability of the pear , a question reserved for a subsequent paper .
The reader who is mterested in the main conclusions of the paper rather than in details of theory , method , or calculations , may care to pass directly to S35 .
GENERAL THEORY OF POTENTIAL OF ELLIPSOIDAL BODIES .
2 .
We proceed to develop a method for writing down the potentials of certain homogeneous solids ; in particular of ellipsoids and distorted ellipsoids .
We are for the present concerned solely with potential-theory\mdash ; the discussion of rotating liquids does not enter before S19 .
As will soon be evident , the problem in potential theory amounts to the following : to write the equation of the boundary of a homogeneous solid in such a form , that the potential at the boundary is of the form , where is a function containing the same algebraic terlLb as , but having in general different coefficients .
If this can be done , it only remains to equate to , and we have at once , on equating coefficients , a series of equations which will determine the possible figures of equilibrium for a liquid mass rotating with angular velocity 3 .
Let be the boundary of any homogeneous solid of density Assuming it to be possible , let be a function of position satisfying at all points of space and coinciding with the potential of the solid at all points inside the solid , and let similarly be a function of position satisfying at all points of space , except possibly the origin or other infinitesimal region inside solid , and coinciding with the potential of the solid at all points outside the solid .
Then must be equal to at the boundary of the solid , and we must also have at the boundary , where denotes differentiation along the normal to the surface .
* Since writing this paper , I have been surprised to find that this conclusion is quite clearly implied in a paper which I published in 1902 , " " On the Equilibrinm of Rotating Liquid Cylinders 'Phil .
Trans , 200 , p. 67 .
See below , S36 .
I have not examined in any detail the conditions that this may be possible , because the resuIt of the paper proves that it is possible in the cases which of importance .
SimilarIy I ha ve not examined in detail the difficulties which might arise at the origin or at infinity , ) ecausc in the final result they do not arise .
We are searching for , and ultimately find , a certain of the potential cquationls , and after the solution has been obtained it is easy to verify directly that it eally is a solution , and that it involves no complications either at infinity or at the centre of the solid .
MR. J. H. JEANS ON THE POTENTIAL OF ELLIPSOIDAL BODIES , AND Introduce a new function , defined by . . . . . .
( 1 ) at all points of space , then we must have . . . . .
( 2 ) at all points of space , and , at the boundary , and These last two conditions are equivalent to .
( 3 ) at the boundary , together with one further condition .
Equations ( 3 ) require that shall have a constant value all over the boundary ; the further condition is that constant value shall be zero .
4 .
Let be the equation of a family of surfaces obtained by varying the parameter , and such that the boundary of the solid is the surface .
The surfaces of this family will divide up the solid into a series of thin shells .
There will be a contribution from each shell to and also to .
Thus may be regarded as the sum of a number of contributions , one from each shell .
Let the thicknesses of the separate thin shells be determined by small increments in , say , , starting from the boundary .
Then we may write , . . . . .
( 4 ) where represents the contribution to from the shell , and so on .
Similarly . . . . . .
( 5 ) Suppose that , and are bsing evaluated at a point on the shell at which the value of is .
Then if is any shell inside the shell , the contributions to and from the shell will be the same ; we have .
Hence from equations ( 1 ) , ( 4 ) , and ( 5 ) , , ( 6 ) or expressed as an integral , .
( 7 ) 5 .
This form for satisfies automatically the last of the conditions of S3 , namely , that shall vanish at the boundary .
We proceed to determine so as to satisfy the remaining conditions which are expressed by equations ( 2 ) and ( 3 ) .
THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
Let be used to denote geometrical diflerentiation with respect to , differentiation in space and constant\mdash ; and let be used to denote algebraic differentiation , i.e. , partial differentiation with respect to keeping , and constant .
Then \amp ; c. , . . . . .
( 8 ) and have ( 9 ) If the poin is the boundary we have , so from equation ( 7 ) , , and it appears that equations ( S ) will all be satisfied if , . . . . .
( 10 ) at all points on the boundary We notice from a comparison of equatiolB ( 6 ) and ( 7 ) that , and the right hand of course vanishes when is on the shell .
Thus we must have , at all points , . . . . . . . . . .
( 11 ) identically , provided that has the value appropriate to the point .
This condition of course imposes more restriction on the value of than does equation ( 10 ) .
Equation ( 10 ) was adequate to ensure that the boundary condition ( 3 ) should be satisfied , but the remark just made shows that for ( 3 ) and ( 2 ) both to be satisfied\mdash ; i. e. , for to give the true value of , equation ( 11 ) must necessarily be satisfied .
We shall now assume that equation ( 11 ) is satisfied , and proceed to satisfy the remaining condition expressed by equation ( 2 ) , namely , . . . . . .
( 12 ) In virtue of equation ( 11 ) , equation ( 9 ) reduces to awe whence on further differentiation ( cf. equation ( 8 ) ) , so that equation ( 12 ) becomes ( 13 ) From equations ( 8 ) and ( 11 ) , so that equation ( 13 ) may be written in either of the equivalent forms : , ( 14 ) ( 15 ) in which stands for ) , and for Thus if satisfies either equation ( 14 ) or ( 15 ) , and also equation ( 11 ) , then , as given by equation(7 ) , satisfies all the conditions which have been seen to be necessary , and will therefore give the true value of * Suppose there , if possible , two solutions to the same problem , say and Since is determined when the problem is fixed , we must have so that , or where . . . . .
( i ) for all values of .
Thus , if is such as to satisfy ( i ) we may add a term to and still obtain a solution of the same problem .
A ) ecial case in which ( i ) is satisfied is when where is function which vanishes identically for the value of appropriate to the point THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID IsfASSES .
6 .
As a matter of convenience , neither loss nor gain of generality , we shall write , . . . . .
( 16 ) in which is any function of a quite eneral function of and .
Then , from equation ( 11 ) , we must . . . . .
( 17 ) identically at all points , when has the value appropriate to the point .
We accordingly have , \amp ; c. , so that equations ( 14 ) and ( 15 ) reduce to ( 18 ) .
( 19 ) Moreover the mily of cons .
may now be supposed to be determined equation ( 17 ) , the boundary will be given by . . . .
( 20 ) , to sum up , if and are such that either equation ( I8 ) or ( 1 is satisfied , then the potential of the homogeneous solid of density whose boundary is determined by equation ( 20 ) , will be given by the value of being evaluated at the point , and being determined from the equation 7 .
The boundary is of course fixed by the solid whose potential is required , but we are left with a certain amount of choice as to the disposition of the surfaces and is any function of whatever .
Replacing by , we find that if is a soIution then will also be a solution of the same ) roblem , .
CCXV.\mdash ; A. .34 MR. J. H. JEANS ON THE POTENTIAL OF ELLIPSOIDAL BODIES , AND cons .
We shall now limit this amount of freedom by assuming that the region at infinity is made to coincide with the surface Consider a new function defined at any point of space , by ( 22 ) ( 23 ) by equation ( 18 ) .
, if ( as will be the case in all our applications ) the value of the limit when of the term in square brackets is zero , we shall have . . . . . .
( 24 ) At finity V must vanish , so that at infinity , by equation ( 21 ) , V : Thus V ' .
vanishes at infinity , and satisfies at all points of space ; whence ( except for a.possible singularity at the origin , which will be found not to cause trouble ) we must have , so that is the internal potential .
Knowing and we find immediately by equation ( 1 ) , and have ( 25 ) ( 26 ) To recapitulate , the condition that these equations shall give the true values of the potential are ( i ) that shall be finite at the origin , ( ii ) that shall vanish at infinity .
If these conditions are satisfied , as they will be without trouble in all our applications , then equations ( 25 ) and ( 26 ) will give the potentials .
8 .
If is the equation of the boundary , the potential at the boundary will be and therefore will contain exactly the same terms in as the equation of the THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
boundary , but with different coefficients .
The method is therefore exactly suited for the determination of figures of equilibrium of rotating fluid ( cf. S2 ) .
As a method of determining potentials , the procedure is indirect in the sense that we cannot pass by any direct series of processes from the equation of the boundary of the solid , as expressed by equation ( 20 ) , to the general function .
We must first search for solutions of equations ( 18 ) or ( 19 ) , and then examine } problem is solved .
An obvious case to examine first is that in which is an algebraic function of the second degree .
In this case is a function of only , so that the equation for can be satisfied .
if the last term in equation ( 18 ) or ( 19 ) is a function of only .
EXAMPLES 0F GENERAL THEORY .
I. A Sphere .
9 .
A quite trivial example may perhaps be taken first , namely that of the sphere .
It is seen without trouble that any way of forming the function will lead to a solution , provided that this function involves only through , provided the surfaces are taken to be concentric spheres .
For instance , we may take then equation ( 18 ) reduces to , of which the solution is found to be The potentials are now given by ' where is written for , and II .
An Ellipsoid .
10 .
It is readily seen that a solution of equation ( 19 ) can be obtained by taking .
( 27 ) .
J. .
JEANS ON THE POTENTIAL OF ELLIPSOIDA BODIES , AND The equation reduces to , ( 28 ) of which the solution is readily found to be .
( 29 ) These values are found to satisfy the conditions of S7 , so that the potentials are given by III .
Distorted Ellipsoid .
11 .
an ellipsoid distorted in any , and without any limitation ( at present ) as to th distortion small , assume where and have the same meanings as before , being given by equations ( 27 ) and ( 29 ) , and , is any general function of , and .
The boundary of the distorted ellipsoid is of course Equation ( 19 ) becomes in have been replaced by now that there is no danger of fusion , and ) , are itten f ) .
If we further put THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
the equation becomes But from equation ( 28 ) , so that on subtraction we find as the equation to be satisfied by .
( 30 ) The equation is too complicated to be attacked directly , but can be effectively broken up by assuming a solution in which and are general functions of , and , while is given by equation ( 27 ) .
On substituting this value for , equation ( 30 ) reduces , after considerable simplification , to and this will be satisfied if we satisfy separately the equations , ( 31 ) .
( 32 ) On substituting for and , and writing ABC , equation ( 31 ) becomes in which Thus equation ( 31 ) is equivalent to be satisfied for all values of , so that we must have ( as is clear on putting in the equation ) when Tt will be relnembered that the boundary of the distorted figure is by and it is now clear that ) reduces to .
Thus the generality of the boundary must be involved in the generality of , and provided is kept general , we shall obtain a general solution of the problem , even if we take the simplest possible value for .
The most general way of satisfying equation ( 83 ) is to take ( 34 ) where may be any function of , and which vanishes for or , but the simplest } of satisfying the equation is to take .
( 35 ) In each of these equations the sign of identity is used in place of the sign of equality , because in the integrand of equation ( 33 ) the value of is not the same as the value of in the upper limit of the integral , which is determined b.y the values of 12 .
To shorten the algebra we may change to a new set of variables connected with the old variables by the relations Differentiation with respect to the new variable is given by where , since , we have , and so new old .
( 36 ) THE FIGURES OF I OF ROTATING LIQUID MASSES .
The old symbol has been to mean ' which in the new variables becomes ' and this will be denoted by Equations ( 32 ) and ( 35 ) in the new co-ordinates , reduce to , ( 37 ) .
( 38 ) We are assuming that when has the value appropriate to the values of , so that in the first equation may be replaced by \mdash ; but in the second equation this may not be done .
13 .
It is , for the purposes of the present paper , to suppose the to start from the undistorted ellipsoid , and to proceed in powers of a ameter e Thus we assume aod in the equation ( 37 ) since ) , it is clear is be a small quantity of the smallness of .
As far as , equation ( 37 ) reduces to giving .
Equation ( 38 ) then gives ; equation ( 87 ) taken as far as will then give , and ( 38 ) will give taken as far as ' will give , ( 38 ) will give and so on .
As far as only , equation ( 37 ) reduces to , of which the solution is where is the most general functio1l of , and .
At the reduces to or to , so that the generality of the function enables us to deal with the most general small displacement possible .
At present we shall consider only solutions for which is algebraic and of not greater thau 3 in .
For these solutions equation ( 38 ) shows that will be of degree not greater than 1 in , so that , and the equation reduces to ' or , ( 39 ) since must vanish when Remembering that , and that , the terms in in ( 37 ) now give giving on integration 4 ?
, ( 40 ) in which is again theInost general function possible of ( enabling us to carry on the distortion to the second order in any way we please ) , and the lower limit of the integral is taken to be zero simply as a lmatter of convenience .
The addition of a perfectly general function would be equivalent to the superposition of a perfectly general distortion ( proportional to ) on to the distortion already under consideration .
The real object of the present analysis is to be found in its ultimate application to the problem of the rotating fluid , and to solve this problem , it will be nd that need contain no terms of degree higher than 4 in , this being also the degree of the other terms in ) .
Hence in what follows it will be supposed that contains no terms of degree higher than 4 in A value is obtainable from equation ( 38 ) , but there are , as has been seen , many possible forms for , and the most convenient is , in point of fact , obtained by going back to equation ( 34 ) , which in co-ordinates is ( 41 ) where may be function of , and which vanishes ( to the power of we are now concerned with ) both for and Let two new functions ?
) and be introduced , defined by , ( 42 ) .
( 43 ) THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
Clearly since is of degree 4 in will be of degree 2 , so that will be a function of only .
The term equation ( 43 ) is therefore zero in the present instance , but is inserted to maintain symmetry .
We now have so that after simplification , .
Since vanishes for the appropriate value of will vanish for both and provided is made to vanish when .
Thus will satisfy the condition to be satisfied by in equation , and a solution of this equation will be . . . . . .
( 44 ) Since must vanish when ( S11 ) it appears that both and must vanish separately when .
On transforming ( 41 ) and ( 42 ) to co-ordinates ( cf. equation ( 36 ) ) and integrating , we obtain as the values of and which vanish when ; .
( 45 ) 14 .
Let us introduce a differential operator , defined by ' ( 46 ) noticing that , as a function of is purely a multiplier .
We have and when CCXV .
The value of already obtained ( equation ( 39 ) ) is , . . . . . .
( 47 ) and the value of ( equation ( 40 ) ) is . . . . . .
( 48 ) Hence from equations(45 ) . . . . . . . . .
( 49 ) 15 .
This completes the solution as far as the second order of small quantities .
We shall not attempt to evaluate and , as the problems discussed in the present paper require a solution as far as only .
As far as , the value of has been seen to be given by . . . . .
( 50 ) and the potentials can now be found directly from the formula ( S11 ) As in S7 , examine a function V ' : defined by then .
( 51 ) Now the value is by S10 , equal to , while from equation ( 31 ) THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
we have Inspection of the values obtained shows that the limit of when is zero , so that equation ( 51 ) reduces to , and since is equal to the true value of at infinity , and is finite at the origin , V ' : must be the true value of the internal potential .
Thus the potentials are given by , ( 52 ) ( 53 ) in which all the quantities must be transformed into -ordinates before integration .
When reduces to by equation ( 40 ) , or to while , also and all vanish when , so that ( cf. equation ( 50 ) ) and the boundary of the distorted ellipsoid is .
( 54 ) 16 .
Before proceeding further it will be convenient to examine in detail the first order solutions which can be obtained from the foregoing analysis , classifying them according to the degree of the algebraic function , and , for brevity , omitting the continual multiplier .
Solution is .
Solution is , .
A physical interpretation of these first few solutions can readily be found .
For the undisturbed ellipsoid of axes , and origin at , ( 55 ) and the special ellipsoid which has been under consideration has been that for which .
We can change the centre and axes of the ellipsoid contemplated in equation ( 55 ) by varying , and .
If we change by an amount in equation ( 55 ) , the change in is given by so that may be regarded as replaced by where .
Thus the solution represents a change from to ; physically it represents a change in the scale of the ellipsoid .
Similarly , if in equation ( 55 ) is changed by by , and by , we find that , so that , and the solution is seen to represent a motion of the centre of the ellipsoid .
Similarly , if we put and in equation ( 55 ) so that and differentiate logarithmically with respect to , we obtain whence Clearly , then , the first three terms in the solution represent a distortion of the original ellipsoid produced by a change in the lengths of the axes , and it is easily seen that the complete solution represents a change of this kind combined with a small rotation of the axes . .
There are ten terms in the general cubic function of .
For the present purpose it is convenient to regard this general cubic function as made up of a term , and the sum of three expressions such as .
For the solution given by , we have , so that and THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
It will be shown later that an ellipsoid distorted in this way cannot possibly be a figure of equilibrium for a rotating fluid ( S20 ) .
For the solution . . .
( 56 ) we have and the solution is .
( 57 ) It will be shown later that this distortion leads to the Darwin-Poincare ' series of pear-shaped figures of equilibrium of a rotating fluid . .
The analysis of SS13 , 14 was co1lfined to the case in which was supposed of degree not greater than 3 in .
But if , in the solution finally obtained in S15 , we take , which involves taking also , we are left with a solution ( cf. equation ( 50 ) ) , in which , and do not vanish on account of the occurrence of the arbitrary function .
And since has been supposed of the fourth degree , this solution gives us the solution of degree to the first order , the parameter replacing the former .
Thus the solution of degree 4 is general function of degree 4 in ( the old ) This solution is not discussed in detail in the present paper , but is classified here with the other solutions for the sake of teness .
An ellipsoid distorted in accordance with this solution would give rise to a series of dumb-bell shaped figures , which would be figures of equilibrium for a rotating liquid .
They would be unstable for a homogeneous mass , but the corresponding figures might conceivably be stable for a heterogeneous mass ( cf. POINCARE'S remark quoted in S1 of the present paper ) .
17 .
One point of interest must be mentioned here in connection with the potentials derived from these solutions .
In the potentials arising from the solution of degree , the internal or boundary potential will be of the form , where do not involve , or .
Since this must be a solution of LAPLACE'S equation , must vanish , and the potential must be expressible in the form .
All the other potentials may be similarly treated .
Making use of this simplification , we arrive at the following scheme for the contributions to the internal or boundary potentials of the various solutions up to .
Only typical terms are taken ; represents the value of at the boundary , represents the contribution from the typical term to the boundary or internal potential . .
( ) ( ii ) .
( i ) ( ii ) ( iii ) 18 .
In any physical application of this method , and in particular in its application to the discussion of rotating masses of liquid , it will be important to know what changes are produced by the distortion upon the mass ( or density ) , the position of the centre of gravity , and the moments of inertia of the body .
These changes are given at once by a study of the limiting form of the external potential at infinity .
The potential at finity of any mass whatever , taken as far as terms of order ' has the limiting form where is the mass of the whole body , , are the co-ordinates of the centre of gravity , and , , are products of inertia defined JII \amp ; c. The moment of inertia about the axis of is , and so on .
THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
For the solution of degree , the limit at infinity of the contribution to the potential is showing that the distortion involves a change of mass , accompanied of course with a change in the inertia terms .
For the solution ) of order so that this distortion represents a motion of the centre of gravity by an amount Solutions of degree will clearly involve changes in the moments and products of inertia .
limiting potentials are found to be as follows : ( i ) ' ( ii ) The first solution does not involve a change in mass , whilst the second does ; both distortions affect the inertia .
For the solutions of degree , the limiting values are as follows : ( i ) This distortion changes neither mass , centre of gravity , nor inertia .
( ) \mdash ; ( iii ) abc abc These two latter distortions move the centre of gravity , but do not affect the mass or inertia .
It is clear , without detailed examination , that the distortions represented by solutions of degree cannot affect the centre of gravity , but may affect the mass and inertia .
* FIGURE OF EQUILIBRIUM 0F ROTATING MASSES .
Jacobian Ellipsoids .
19 .
The condition that a single figure shall be a figure of equilibrium for a rotation about the axis of is that the centre of gravity shall lie on the axis of , and that shall have a constant value over the boundary In searching for a series of figures of equilibrium we must add a further condition of constancy of mass .
For the undisturbed ellipsoid , with the notation of S10 , .
( 58 ) For this to be constant over the surface , it must be identical with , where is a constant .
If we put the equations obtained by comparing coefficients are , ( 59 ) , ( 60 ) .
( 61 ) Ellipsoids with Distortions of the First Order .
20 .
We proceed to consider which of the distorted ellipsoids can give rise to possible figures of equilibrium .
The solutions of degrees , 1 , 2 lead to nothing except new ellipsoids , so that the inclusion of these distortions could only represent a step along the already known series of Jacobian ellipsoids or Maclaurin spheroids .
Consider next an ellipsoid distorted by the addition of a solution of the type ( i ) of degree 3 ( S18 ) , say .
This distortion , as we have seen ( S18 ) , does not affect the total mass or the position of the centre of gravity .
The boundary of the distorted ellipsoid is ; THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID IsIASSES .
while the additional term which has to be inserted on the right of equation ( 58 ) is Hence the additional equation which has to be satisfied , in addition to equations ( 59 ) to ( 61 ) , is . . . . . .
( 62 ) Eliminating from this equation and ( 61 ) we obtain , as an equation which must be satisfied if the distorted ellipsoid is to be a figure of equilibrium , .
( 63 ) This obviously cannot be satisfied , for the integrand is positive for all values of .
We conclude that the distortion now under consideration cannot possibly give rise to a figure of equilibrium .
21 .
There remain nine terms for consideration in the general cubic function .
Inspectio will show , or it will soon become apparent as we proceed with the analysis , that these fall into three groups , as in S16 , and that the three terms of any one group just suffice to give a possible figure of equilibrium when combined with a term to restore the centre of gravity to its position on the axis of rotation .
We shall accordingly consider a distortion in which the cubic terms are those already written down in equation ( 56 ) .
These terms are seen ( S18 ) to move the centre of gravity parallel to the axis of , and to correct this we shall add a ( cf. S16 ) , to Thus , for the distorted ellipsoid now under consideration , the boundary will be .
( 64 ) As far as terms in , the value of the potential at infinity is ( cf. S18 ) so that for the centre of gravity to remain at the origin we must have .
( 65 ) VOL. CCXV.\mdash ; A. Collecting the terms in as calculated in S18 , we find .
( 66 ) For brevity in printin , introduce the following notation .
Let ; so that , for instance , , ( 67 ) And , for the problem immediately in hand , write further ' and as before put , then equation ( 66 ) becomes .
( 68 ) THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID The distorted ellipsoid will be a possible figure of equilibrium , as far as the first order of small quantities , provided the right-hand member of equation ( 68 ) is identical with .
( 69 ) Equating coefficients , we obtain , ( 70 ) , ( 71 ) , ( 72 ) , ( 73 ) , ( 74 ) and from these equations , together with equation ( 65 ) , we must eliminate or obtain the coefficients .
If we put ( 75 ) then equations ( 71 ) to ( 73 ) reduce to , ( 76 ) , ( 77 ) , ( 78 ) whilst on substitution for from equation ( 65 ) and from ( 70 ) , ( 74 ) reduces to . . . . .
( 79 ) We have now to deal with four equations ( 76 ) , ( 77 ) , ( 78 ) , and ( 79 ) , and have to examine whether , and how , these can be satisfied by values of , and Since there must , from general physical principles , be an equation of some sort for points of bifurcation ( whether capable of being satisfied by real values or not ) , we are led to suspect that these four equations are not really independent .
Equation ( 79 ) was obtained by the elimination of from two equations ( 65 ) and ( 74 ) , each of expres sed in effect the condition that the centre of gravity of the lnass should be at the ; in fact , equatio1l ( 65 ) was only a short way ivin the value of , which would in any case have been given by equation ( 74 ) .
We therefore expect that equation ( 79 ) , derived from ( 65 ) and ( 74 ) , will prove only to be an identity of which the truth is involved in the three other tions ( 76 ) to ( 78 ) .
And , as a matter of procedure which is entirely at our cboice , we shall elect first to solve equations ( 76 ) to ( 78 ) , and then to verify the of ( 79 ) .
The elimination of from equations ( 76 ) , ( 77 ) , and ( 78 ) gives a determinant which on expansion reduces to - , ( 80 ) and this is accordingly the equation giving points of bifurcation on the Jacobian series of ellipsoids .
22 .
Two identities of importance are the following : , ( 81 ) . .
( 82 ) From equation ( 70 ) , . . . . .
( 83 ) , ( 84 ) , ( 85 ) so that on addition , by the use of ( 81 ) , , ( 86 ) giving , and hence , and , in terms of , and .
We further have ( 87 ) , ( 88 ) THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
and is given by equation ( 80 ) as soon as the values of and known .
Substituting for , from equations ( 83 ) to ( 85 ) , we obtain Finally we bay .
( 89 ) , .
With this material it ought to be possible to find the points of bifurcation from equation ( 80 ) .
As , however , DARWIN'S results are available , it will be sufficient to make use of his results and merely verify that his quantities satisfy equation ( 80 ) , as they are in point of fact found to do .
23 .
DARWIN'S values , calculated for an ellipsoid such that , are whence , by equations ( 86 ) and ( 89 ) , With the use of these values , equations ( 76 ) , ( 77 ) and ( 78 ) become . . . . .
( 90 ) . . . . .
( 91 ) . . . . .
( 92 ) The values of , are , of course , indeterminate to within a common multiplier .
The simplest set of values , obtained by cross multiplication of the coefficients of equations ( 90 ) and ( 91 ) , is If we substitute these values in equation ( 92 ) , we find The fact that the error occurs only in the seventh place of decimals adequately verifies DARWIN'S calculations , but the tendency of small errors to accumulate in computation is forcibly illustrated by the circumstance that in the above equation the final error is as much as one-six-hundredth part of the whole value of the middle term .
With the values .
obtained for , I find verifying that equation ( 79 ) is satisfied , again as far as the sixth place of decimals .
24 .
With a view to subsequent computations , it is convenient to take a standard set of values such that .
These values are found to be and with these we have , by equation ( 65 ) , These numerical values substituted in equation ( 64 ) will give the equation of POINCARE'S pear-shaped figure as far as small terms of the first order .
The Pear-shaped Figure Calculated to the Second Order .
25 .
The question as to whether the pear-shaped figure is stable depends upon the change effected by the distortion upon the angular momentum of the ellipsoid .
But ( cf. S18 ) the first-order distortion so far considered can be easily seen to produce no effect at all upon the angular momentum of the figure .
It is therefore necessary to proceed to terms of a higher order , and we now consider terms of the second order .
The first-order terms have been found to be given by , . . . . .
( 93 ) with ( cf. equation ( 47 ) ) .
The value of will be given by equation ( 48 ) , in which is to be assigned the value ( 93 ) , and will be taken to be given by . . .
( 94 ) It has to be shown that this value for makes it possible for the figure distorted to the second order in this way to be a figure of equilibrium .
THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
It will be noticed that with the value we have assumed for , the value of becomes a function of , of degree 4 involving only even powers of ; its value is . . .
( 95 ) The values of and of are easily seen to be similar in form , so that all the second-order terms in ) are of this form ( cf. equation ( 50 ) ) .
Multiplying by and integrating from to , we obtain as the terms of the second order in the potential an expression of the form .
If this figure can be a figure of equilibrium at all , it will be for a rotation differing only by a second-order quantity from that of the original ellipsoid .
Let us suppose that for it ; then at the boundary , as far as -degree terms in , the same as before } . . . . . .
( 96 ) At the boundary , , ( 97 ) so that for the figure to be one of equilibrium , the right-hand member of equation ( 96 ) must be identical with .
( 98 ) MR. J. H. JEANS ON THE fIAL OF ELLIPSOIDAL BODIES , AND Upon equating coefficients of the terms of degree 1 and 3 in , we obtain exactly the same equations as were obtained before when the terms in were omitted .
Thus no alteration is produced in the cubic terms until the solution is carried as far as The equations obtained by comparison of the coefficients of , and , and two similar equations .
In these the terms in and the terms independent of must be equal separately .
The latter terms again give equations ( 70 ) , while the terms in give . . . . .
( 99 ) ( 100 ) ( 101 ) Finally , upoIl equating terms of the fourth degree , we obtain the six equations : ; ; ( 102 ) to ( 104 ) ; ; ( 105 ) to ( 107 ) We have seen that equations ( 70 ) to ( 73 ) must still be true , so that will be the same as before .
26 .
At present there are eleven quantities to be determined or eliminated , namely , , and The equations giving these quantities can be simplified in the following way .
By an argument already used in S17 , it appears that the terms in on the right hand of equation ( 96 ) must be harmonic .
Thus we must have . . . . . . . . . .
( 108 ) . . .
, . .
These are , of course , not new equations to be satisfied ; they reduce to identities when the calculated values are inserted for , \amp ; c. THE FIGURE OF EQUILIBRIUM OF ROTATING LIQUID 'MASSES .
With the help of ( 111 ) , equations ( 99 ) to ( 101 ) may be replaced by , ( 112 ) , ( 113 ) , ( 114 ) while from equations ( 102 ) to ( 110 ) we obtaiu , ( 115 ) .
( 116 ) .
( 117 ) These three equations with equations ( 102 ) to ( 104 ) , namely , , ( 118 ) , ( 119 ) ( 120 ) may be used to replace the group ( 102 ) to ( 107 ) .
27 .
We proceed to evaluate these quantities in detail .
The values of ?
and from equation ( 47 ) are , . . . . .
( 121 ) , ( 122 ) and the value of is already given by equation ( 95 ) .
For convenience in computation we shall combine all the terms in which are independent of in the first two lines , with the similar terms in the last line ; we accordingly write ? . . . . . .
( 123 ) VOL. CCXV.\mdash ; A. 1 MR. J. H. JEANS ON THE POTENTIAL OF ELLIPSOIDAL BODIES , AND The value of is.now given by ( 124 ) }yhence , by eqnation ( 45 ) , , . . . . .
( 125 ) where the 's are constants , to be chosen so as to make vanish when The value of is derived from that of by replacing by and omitting all other terms .
Thus we have , again from equation ( 45 ) , , ( 126 ) in which is a new constant , chosen so as to make vanish when On collecting terms , we obtain .
( 127 ) where ( 128 ) THE OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
( 129 ) ( 130 ) ( 131 ) If the value of is taken to be we have as the value of , . . . . .
( 132 ) and we have already supposed ( S25 ) that .
( 133 ) Hence , by comparison , we obtain as the values of the coefficients ( 134 ) ( 135 ) ( 136 ) ( 137 ) ( 138 ) ( 139 ) ( 140 ) ( 141 ) ( 142 ) .
( 143 ) of Integrals .
28 .
It is clear that before these coefficients can be evaluated , certain integrals must be calculated of the types , where the notation is that of S21 ( equation ( 67 ) ) .
values of have already been evaluated in S23 , as also of and .
We also have These 10 integrals will form an adequate basis from which to calculate all the THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
others .
The required integrals can all be obtained by successive applications of formulae of the following types , which can be verified without trouble : ; \amp ; .
; , \amp ; .
abc .
A great number of the integrals can be calculated in two or more ways , and owing to this circumstance it is possible to provide very complete checks on the accuracy of calculation .
Two complications are worthy of mention .
In the first place one must consider the ordinary cumulative error of all prolonged computation .
Since is nearly equal to , any error present will be increased when we evaluate a new quantity of the type ; consequently where a quantity can be evaluated in two ways , the one in which division by is not involved has been taken to be the true value , and the one derived by division by , has been used merely as a check , and has generally been found to differ in the sixth or seventh place ( or near the end of the computations even in the fifth place ) from the other values .
Secondly , if the 10 integrals used as base were known with perfect accuracy , the checks ought to be satisfied fully except for the error in the last one or two figures .
But , as has been indicated in S23 , the 10 integrals are not themselves perfectly selfconsistent , so that different methods of computation will lead to a difference of the final results comparable with the errors in the basic integrals .
The following table gives the values I have selected as the best for the various integrals required .
I have not it necessary to record the checks or estimate the probable errors here , as a much more searching test of the accuracy of the whole computation can be provided at a later stage .
Evaluation of thoe Coefficients , .
29 .
It will be noticed that the coefficients , are linear in , , , , so that the various contributions may be calculated separately and independently .
ibutions fTerrns in As regards the contributions from these coefficients , we may take ( cf. equations ( 125 ) and ( 126 ) ) ; ; THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
whence we obtain as the contributions to ( cf. equations ( 128 ) to ( 131 ) , It now appears that , 7 contribute nothing to the values of , , ( cf. equations ( 134 ) to ( 139 ) , p. 60 .
) Their contributions to , and 4 are as follows : so that the contributions Since this part of the potential should be harmonic , we ought to have ( cf. equation ( 111 ) ) .
This is clearly the case , in virtue of the identity I have verified that these identities are satisfied by the values in the table opposite , and the contributions are found to be , Contributions from Terms in 30 .
As regards these terms , we may take ( cf. equations ( 125 ) and ( 126 ) ) , and the contributions to , are found to be ( cf. equations ( 128 ) to ( 131 ) ) , We find as the contribution to , ( equation ( 140 ) , p. 60 ) \amp ; c. , from which it is easily verified that , as it ought to be .
In virtue of this relation it is only necessary to evaluate two of the contributions , but I have calculated all three directly from the table on p. 61 , so as to obtain a check on the amount of error involved from the causes mentioned in S28 , as well as a check on the accuracy of my own computations .
The values I find are From these figures the value of would be given by This check gives an idea of the amount of error involved .
It illustrates the tendency of the errors to accumulate in terms where 's and 's are plentiful , and to be absent THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
where 's are present .
This is a consequence of the method of procedure explained in S28 .
In calculating the contributions to , , I have ( unwisely ) taken the variables to be , , instead of ' .
In virtue of relations ( 108 ) to ( 110 ) , only three contributions need have been calculated , but I have calculated five out of the six quite independently , so as to have two independent checks on the computations .
For the contribution to , I find .
The other five contributions can be similarly written down .
It then can be verified algebraically that the three relations ( 108 ) to ( 110 ) are satisfied ; this provides a check on the method and the algebra , but it hardly seems worth it here , as the numerical checks to be given later provide simultaneously checks on the method , the algebra , and the computations .
By independent computations I find for the contributions to , the following : From these I have not calculated independently , but a check is provided by comparison of the values of deduced from the above figures by the use of equations ( 109 ) and ( 110 ) VOL. CCXV.\mdash ; A. separately .
For these values I find the following , the upper value being that derived from equation ( 109 ) : Contribution in , , 31 .
For the computation of these contributions it is convenient to take the standard set of value obtained in S24 , namely \mdash ; These give the values Then , as ards the terms in , , \amp ; c. , including , we find ( cf. equations ( 125 ) and ( 126 ) ) giving on substitution of numerical values , as regards terms in , only THE FIGURES EQUILIBRIUM OF ROTATING LIQUID MASSES .
The values of originating from these terms are 1 have evaluated independently the three contributions to and find for them , and there are corresponding values for the contributions to and The numerical values found by independent computations are As a check on the computations , the sum of these contributions ought to be zero ; we find The six contributions to , have also all been calculated separately .
The values obtained are 0'0271162 , as regards the checks which ought to be satisfied by these values , I find Solution of the Equation .
32 .
We proceed to form and solve the final equations .
It is convenient to deal with the equations in first .
By comparison of coefficients in equation ( 123 ) , we obtain .
( 144 ) ( 145 ) ( 146 ) ( 147 ) - ( 148 ) ( 149 ) ( 150 ) ( 151 ) ( 152 ) ( 153 ) These coefficients can be checked by comparing the sums of the coefficients in the two values of ( equation ( 123 ) ) ; i.e. , taking .
For the difference of the sums which ought to vanish , I find .
Using the values just obtained , Collecting the various contributions to , equations ( 102 ) to ( 104 ) now become THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
while equations ( 115 ) to ( 117 ) give The solution of this set of six equation is .
( 154 ) No check is needed on the accuracy of these values beyond the fact that they satisfy the equations .
On substituting directly into the equations , I find that they are all satisfied accurately to the last place of decimals .
The corresponding yalues of ( see opposite ) are , .
and the values of , , calculated directly from equations ( 102 ) to ( 107 ) are ( 156 ) These ought to sat.isfy the checks afforded by equations ( 108 ) to ( 110 ) .
In point of fact , I find 33 .
The first use which must be made of these numbers is to determine the contributions to , evaluated on p. 64 .
We have by separate computation , 1672652 , Hence collecting all contributions to , we obtain as their total values From the values already obtained for , we can transform equations ( 112 ) to ( 114 ) into the following :\mdash ; On substituting the values of , and , these reduce to The solution of these equations is found to be ( 157 ) ( 58 ) ( 159 ) and this solution has been verified by direct insertion into all the equations .
34 .
This completes the solution of all the equations , and the determination of all the coefficients except THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
On collecting the values of , from pp. 62 , 63 , and 66 , and inserting the numerical values already obtained , we find - To evaluate we have to examine the form assumed by the potential at infinity .
The additional terms in the potential , as far as terms in , produced by the distortion are readily found to be , and if is the additional mass produced by the distortion , this must be identical with Hence , if is determined by the condition of constancy of mass , we have giving ( cf. equation ( 153 Discussion of the Figure .
35 .
The boundary of the pear-shaped figure , as far as the second order of small quantities has been found to be ( 160 ) In this equation all the coefficients have been determined ; the coefficients .
and have been found to involve defined in S25 by the equation while the remainder are pure numbers . .
J. H. ON THE NTIAL OF ELLIPSOIDAL BODIES , AND Let us put and similarly for , and , and let us put ; then the equation may be put in the form .
( 161 ) For any values whatever of and , provided only that they are sufficiently small , equation ( 161 ) will give a figure of equilibrium .
If we put , but retain , the equation becomes which is an ellipsoid of semi-axes , given by , or , numerically , It is at once clear that as varies this ellipsoid coincides with the various Jacobian ellipsoids near to the standard ellipsoid .
If we put but retain in equation ( 161 ) we obtain equation ( 160 ) with ; i.e. we obtain a series of figures of equilibrium all having the same angular velocity as the standard Jacobian ellipsoid from which they are derived .
The two series of figures obtained by putting and in equation ( 161 ) may be represented by the two intersecting straight lines POP ' , QOQ ' in fig. 1 , the point of bifurcation bein of course represented by the point O. The general figure of equilibrium represented in equation ( 161 ) is , however , arrived at by assigning values to both and , these values being limited by the condition only that and shall be so small that and shall be negligible .
Thus the figures of equilibrium given by equation ( 161 ) are represented by all the points inside certain rectangle ABCD surrounding the point O. They do not fall into two linear series , as it seems to be tacitly assumed by DARWIN and POINCARE that they will do .
36 .
The circumstance that the two linear series lose their identity and become merged indistinguishab]y into a general area seems to be predicted as a direct consequence of THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
POINCARE'S general analysis , coupled with the linearity of the equations \amp ; c. ) which lead to figures of equilibrium .
For the vanishing of the Hessian in the notation of POINCARE ' * ) which expresses the condition that a poin of bifurcation should exist , expresses also the condition that the two linear series should be merged with an area of linear series as they approach the point of bifurcation .
There is , of course , no question that in the neighbourhood of the point two linear series do actually exist , such as may be represented by the lines POP ' , ROR ' in fig. 1 ; this is abundantly proved by POINCARE'S general argument .
What is now maintained is that an expansion as far as only , does not suffice to reveal the Fig. 1 .
directions in which these linear series start out from the point of bifurcation .
So long as our vision is limited to interior of the rectallgle ABCD in fig. 1 , we can know nothing of the direction in which the line OR starts out from O. And the whole difficulty is merely one introduced by the artificial method of expansion in powers of a parameter ; as soon as this artificial method is abandoned the rectangle ABCD shrinks to an infinitesimal size , and the curves POP ' and ROR ' become merely two lines intersecting in the point without any complications .
An exactly " " Sir l'equilibre d'une mass animee d'un mouvement de rotation ' Acta Math , p. 259 .
Cf .
footnote to p. 74 . .
cit. , S2 .
VOL. CCXV.\mdash ; A. analogous situation arises in considering the directions in which lines of force start out from a point of equilibrium in an electrostatic field .
* 37 .
A more interesting illustration of the difficulty will be found in an investigation of the figures of equilibrium of rotating liquid cylinders which I published in 1902 .
In this paper bhe equation of the cross-section of a of equilibrium corresponding to a rotation is supposed to be expressed in the .
( 162 ) where are functions of ars complex co-ordinates given by , and are ordinary Cartesian co-ordinates measured from the axis of rotation .
The quantity is a parameter , analogous to the of the present paper , measuring distance from the point of bifurcation at which the pear-shaped figure and the elliptic cylinder coalesce .
At this point of bifurcation , , so that , of which the value is shown to be .
If V is the potential of an electrostatic field , the equation of a line of force will be ' where are direction -cosines .
lines of force will meet in a point of equilibrium ( jusl as two linear series meet in a point of bifurcation ) , and the condition for this is ii .
Let point equilibrium satisfying , then , if is a small quantity .
of the first order , point will , as varies from zero upwards , trace out a line passing through / ? .
The condition that this shall be a line of force is , as far as first powers of and this is satisfied ( analytically ) because of equations for all values of .
Thus , as far as first powers of , there are as many lines of force through the point of equilibrium as there are values of ; an infinite nnmber .
But on going as far as it becomes clear that there are only true lines of force ough this point .
The condition that a point of equilibrium shall exist is also the condition that , if the approximations are not carried far enough , there shall be the confusion of an infinite number of lines appearing to satisfy the condition for a line of force , and the analogous statement is true for points of bifurcation and linear series of figures of equilibrium .
" " On the of Robating Liquid Cylinders ' Phil. Trans , 200 , p. 67 . .
cit. , equation ( 71 ) , p. 86 , THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
The calculation of presents no difficulty , and equation ( 16.2 ) as far as only is shown to be the equation of a pear-shaped figure .
On calculating its value is found to be of the form ( cf. S22 of the " " Cylinders\ldquo ; paper ) , where is analogous to the of the present paper ; to be exact the rotation for any value of is supposed given by Again , then , as far as there is a doubly-infinite series of figures of equilibrium , not two singly-infinite series .
In this earlier and simpler investigation , it was an easy matter to carry the computations to the third , fourth , and fifth orders of small quantities .
It was found that the equations giving for a figure of equilibrium could not be satisfied so long as was kept indeterminate , they could only satisfied for one special value of namely .
After having determined the value of in this way , it was possible to investigate the stability , and the pear-shaped cylinder was found in point of fact to be stable .
What is important in connection with the present paper is that it not possible to determine the stability of the pear until after its figure had been determined to the third order of small quantities .
* 38 .
The work of POINCARE can hardly be compared in detail with the investigation of the present paper , because POINCARE tacitly assumes the whole poin at issue ; namely , that it is possible to determine the beginning of the pear-shaped series from an investigation of figures of equilibrium going only as far as second-order terms .
The work of DARWIN admits of detailed comparison , because DARWIN'S work claims actually to have effected the determination which I am compelled to believe , after my investigation , to be impossible .
It will be clear that any extra condition in addition to the conditions that the figure should be one of equilibrium will provide an additiorlal equation which will reduce the doubly-infinite series inside the rectangle ABCD down to a singly-infinite series represented by a straight line .
For instance , the condition that the angular velocity shall remain constant would reduce the rectangle to the straight line QOQ ' ; the condition that the angular momentum should remain constant would reduce it to * The previous ation on cylinders and the present one on three-dimensional bodies follow widely different methods ; the present paper is in no sense a translation of the former from two into three dimensions .
The two papers were written at an interval of twelve years , and I had hardly referred to the former paper in writing the present one until after I had erlcountered the difficulty of not being able to determine the stability from the second-order figure ; I then discovered that precisely same situation had arisen in my former investigation .
some curve at present undeterminsd .
Hence , if the present paper is sound , we should anticipate that DARWIN obtained a linear series instead of a rectangle of configurations by the introduction of some adventitious condition not necessary to equilibrium .
39 .
DARWIN supposes his pear defined as far as the second order by the equation .
( 163 ) where gives a measure of the surface displacement , is a parameter analogous to our ( although not numerically the same ) , and the quantities are coefficients which must vary as we pass along the linear series , but are constants , as is also , for any single figure of equilibrium .
The quantities are ellipsoidal harmonics .
DARWIN supposes to be a small quantity of the first order , and the coefficients to be small quantities of the second order .
The energy of the distorted ellipsoid will differ from that of the original ellipsoid from which the displacement is measured by a small quantity which will involve and , where is the increase in the value of for the distorted figure .
The first order displacement of the ellipsoid will be represented by the first terms , . . . . .
( 164 ) of equation ( 163 ) , and this is supposed to be a possible figure of equilibrium with .
The corresponding value of ] ] be of the form , . . . . .
( 165 ) and the coefficient must therefore vamsh if the original ellipsoid was one corresponding to a point of bifurcation .
If we do not at present assume that , the value of arising from the secondorder displacement ( 163 ) will be of the form ( 166 ) in which a single term in has been taken as being typical of all the terms in the coefficients .
The condition that the displacement ( 163 ) together with an increase in shall give rise to a figure of equilibrium is that shall be stationary for all variations of and ; it is expressed by the equations ( 167 ) Expression ( 166 ) is the same in form as that given by DARWIN ( ' Coll. Works , ' vol. 4 , p. 349 ) , except that he omits the term from ; and equations ( 167 ) are THE FIGURES OF EQUILIBRIUM OF ROTATING LIQUID MASSES .
the same in form as those from which DARWIN obtains the conditions of equilibrium .
The equation , written out in full , becomes .
( 168 ) If second-order terms are neglected equation ( 168 ) reduces to , . . . . .
( 169 ) and if is taken equal to zero , equation ( 168 ) reduces to .
( 170 ) DARWIN'S method is in effect to replace equation ( 168 ) by equations ( 169 ) and ( 170 ) .
This , in my opinion , introduces one limitation too many on the values of the riables , and so reduces the doubly-infinite series of figures of second order to two singlyinfinite series .
Equation ( 169 ) must undoubtedly be true if second-order terms are neglected , but we may take ( 171 ) where is a quantity of the second order entirely at our disposal , and still satisfy all the conditions necessary for equilibrium .
I think it will be found that DARWIN'S procedure in effect introduces the spurious condition .
DARWIN'S equations are of course sufficient to ensure equilibrium ; what is maintained is that they are not necessary and so do not disclose all possible figures of equilibrium .
In this way I believe it will be found that DARWIN has limited himself to one linear series of figures instead of the doubly-infinite series represented inside the rectangle ABCD in fig. 1 .
If this series should happen to run on continuously at the edge of the rectangle with the true series ROR ' in fig. 1 , then of course DARWIN'S investigation would stand .
But no reason suggests itself , and certainly DARWIN ( not foreseeing the complication of the doubly-infinite series ) gives no reason , why this should be the case .
For some value of he- two series will run on continuously , but there is no assignable reason why this value should be What , then , will happen if we try to carry DARWIN'S approximation on to thirdorder terms by his method ?
I think it will be found that his series comes to a dead end before the third-order terms are arrived at , precisely as if it ran on to the boundary of the rectangle ABCD in fig. 1 , and could get no further .
If the displace- meant goes as far as third-order terms , must go as far as sixth-order terms , and will contain terms of orders 2 , 4 , and 6 , say The equations necessary for equilibrium are \amp ; ( 172 ) while the equations provided by DARWIN'S methods would be and it will readily be seen that there are equations than can be satisfied by the variables at our disposal .
The conclusions of this section are put forward with the utmost diffidence , but to the present writer they seem inevitable .
40 .
Assuming that no glaring error has been made , the present investigation seems to indicate that so far the knowledge we have as to the stability of the pear-shaped figure amounts to absolutely nothing .
The required knowledge can only be obtained by carrying the figure of the pear to a still higher approximation .
In the parallel investigation on cylinders it was found that the stability could be examined as soon as the figure was determined as far as third-order terms , and doubtless the same will prove to be the case in the present problem .
Fortunately the method of the present paper is one which lends itself to indefinite extension , limited only by labour of computation , so that it ought to be possible to proceed to third-order terms and determine the stability of the pear , and if the pear then proves to be stable , to pro- ceed to higher orders and so examine the series of pear-shaped forms .
In the previous investigation on cylindrical figures it was found that an expansion as far as fifth-order terms gave a good approximation to the pear-shaped figure up to a stage where it was obviously just about to divide into two separate masses .
It is only in the hope that I shall be able to carry the present investigation further , that I have ventured to put forward the present somewhat lengthy piece of work which has so far led only to such negative and disappointing conclusions .

