]\gt ; VI .
On the pear-shaped Figure o.f Equilibrinm of a Rotating Mass of Liquid .
G. H. DARWIN , F.R.S. , Plumian Professor Fellow of Trinity College , Ccrmbridge .
fficeived October 21 , \mdash ; Read November 21 , 1901 .
INTRODUCTION .
THIS is the sequel to a previous paper on ( Ellipsoidal Harmonic Analysis\ldquo ; ( ' Phil. Trans , vol. 197 , pp. 461-557 ) .
I here make use of the methods of that paper , and for brevity shall refer to it as " " Harmonics The sections 1 to 4 are preparatory , and might have been included in " " Harmonics but seem more appropriate here .
Section 5 is an independent investigation of so much of M. POINCARE'S celebrated memoir on rotating as relates to the immediate object in view .
It is not necessary to say more here , since I give a short summary in the last section . .
The rmonics of the Third Degree .
It was remarked in S 7 of " " Harmonics\ldquo ; that this group of harmonics may be determined in rigorous algebraic form ; I now proceed to do this .
The requisite formulae are contained in S 6 of " " Harmonics and I take the several harmonics in succession .
Throughout this section we have , of course , ; type OEC , and , with The equation for is If we write the proper solution of the quadratic equation is .
'Acta Mathematica , ' vol. 7 , 1885 .
305 .
) 17.3.1902 .
Also , with Then since , we find .
( 1 ) .
; type OOC , and , with The equation for is or If we write , the proper solution of the quadratic equation is Also , and , with Then since , we find ( 2 ) .
, type OOS and , with The equation for is A comparison with the last case shows that we have only to change the sign of ; accordingly if , we have , , with 1 .
OF EQUILIBRIUM OF A OTATING MASS OF LIQUID .
On substitution we find ( 3 ) ; type OEC , , with The equation for is We have already defined as , and find the proper solution of the quadratic equation to be .
Also , with With the known values of and of , we find ( 4 ) .
A comparison with ( 1 ) for shows that the last factors in each only differ in the sign of ; type OES , .
Since , we have at once ( 5 ) .
; type OOC , , with The equation for is , 2 { 3 , 3 } We have already defined , and find for the proper solution .
Also , and , with Whence on substitution ( 6 ) .
; OOS , , with The equation for is We may derive the result from the last case by introducing , aIJ.dehaIJging the sign of , so that , , with Whence on substitution ( 7 ) .
The forms of the corresponding functions of are the same , except that and replace the corresponding factors .
I have not determined the cosine-and sine-functions , because they may be written down at once from the results already obtained .
The three roots of the fundamental cubic are , and Hence we have only to replace by this last function in the seven formulae ( 1 ) in order to obtain functions proportional to the seven cosine- and sine-functions .
If the definition of the latter functions is to agree with that given in " " Harmonics the factors must be determined appropriately , the questio1l as to the value of the factor will not arise here .
OF TING MASS OF LIQUID .
S2 .
Change of Notation .
It will be convenient , with a view to future work , to change the notation , and I desire to adopt a notation which shall not only agree in the main with that used in " " Harmonics but shall also facilitate reference to a previous paper on JACOBI'S ellipsoid ( ' Roy .
Soc. Proc. , ' vol. 41 , pp. 319-336 ) .
I write It may be noted that what I here write wafi denoted by in " " Harmonics rice ver I have in general the current -ordin ates , and the ellipsoid of reference , so that the squares of the semi-axes are I now propose to write as the squares of three semi-axes of the ellipsoid of reference Comparing these two we see that , and For the current co-ordinates I retain and write The three roots of the fundamental cubic are therefore .
The rectangular co-ordinates are therefore now expressible as follows:\mdash ; in COS COS COS , .
( 8 ) .
VOL. CXCVIII.\mdash ; A. 2 SSOR G. H. DARWIN ON THE SHAPED F These give At the surface , and we have In the formulae for the third harmonics , in every case but one , and in two out of the five harmonics of the second degree , there occurs a factor of the form \mdash ; constant ) ; in each such case I write that constant in the form , and Thus will have a different value for each harmonic .
It has been already remarked that for most purposes it is immaterial by what constants the several functions are multiplied .
Although it would be easy to determine the constant in each case so as to make the function agree with its value as defined in Harmonics yet I shall not take that course , and shall omit factors as being in most cases redundant .
For the sake of completeness I will give the first and second harmonics in the new notation , as well as the third .
Since the harmonics of the first degree are expressed by $ it is clear that in the new notation ( 9 ) .
It appears from S 12 of ' ' Harmonics\ldquo ; that where , and In accordance with the notation suggested above , let EQUILIBRIUM OF A ROTATIN MASS OF LIQUII ) .
and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and and both cases Hence and and , and .Then substituting for , we find ( 10 ) , where , and , with upper sign f the first and the lower sign for the second .
It from ( 19 ) itnd ( 20 ) , S 7 , of " " Harmonics\ldquo ; that , ( 11 ) , nnd from ( 21 ) and ( 22 ) that ( 12 ) .
Lastly , from ( 25 ) and ( 26 ) ( 13 ) .
Turning to the rmonics of the third degree , we found in the cases where the type is and 2 .
put , we find , with the above alternative form for and and , ( 14 ) .
and Again in the two cases where the type is OOC we found and \mdash ; Putting we find , and Thelefore , with above alternative folnl for and , and \mdash ; , ( 15 ) .
and In the two cases where the type is OOS we found OF OF A LI(JUII ) .
Puttmg we find , and Therefore , with the above alternatiye form for and and , ( 16 ) .
and The seventh of these harmonics , which is of type , stands by itself We had This gives in the new notation ( 17 ) .
The formulae ( 9 ) to ( 17 ) give the tifteen sets of three functions constituting the fifteen harlnonic functions of the first three degrees .
It would be lthough somewhat tedious , to find the coefficient by which each function is be so that its definition may agree with that of the .
Expressions for }Solid Harmonics in Co-ordinates .
The three roots of the original cubic equation were , and in the new notation the three roots of .310 G. H. DARWIN ON THE ) FIGUR Hence it follows that we have the identity si .
Putting .
This expression , together with those for in ( 8 ) , enables us to write down the results at once .
As before , I drop the several factors as being redundant for most purposes .
From ( 9 ) ( 18 ) .
( 19 ) , where , and so that , ( 12 ) , and ( 13 ) ( 20 ) .
From ( 14 ) ( 21 ) , where , and so that From ( 15 ) and ( 22 ) , OF FQUILIBRIUh[ OF A ROTATING MASS OF LIQUID .
where , and so that From ( 16 ) and ?
' ( 23 ) , where and so that .
Lastly , from ( 1.7 ) , . . . .
( 24 ) .
It is easy to verify that each of these expressions satisfies LAPLACE'S equation . .
The Expression for the -functions in Elliptic In this paper I drop the factors o and which were found to be necessary when the -functions were expressed in series .
We make the following definition:\mdash ; and a similar formula holds for Qi .
It is clear that ?
may be multiplied by any constant factor without changing the result ; hence we may use the forms which have been found in SS 2 , 3 .
The notation must now be changed .
We have and Therefore , when is adopted as the limits are to , and the sign of the whole is changed .
But and Therefore In accordance with the uSage in elliptic integrals , I write under the integral sign , or outside the integral .
shall also for brevity write under the integral , or outside the integral .
We hove then I apply this formula successively to the several functions , as given in ( 9 ) to ( 17 ) , and introduce the abridged notation just defined , but I do not reiterate the special meanings to be attached to ] ) symbol in each case .
Since , we have ( dropping the now unnecessary suffix ) , and , and ( 25 ) .
and and OF EQUILIBRIUM OF A MASS OF LIQUID .
All these integrals are expressible in terms of the elliptic integrals , II It will , however , be found that in fact the coefficient of vanishes in every case .
The cases of and are very simple , and we have , It is possible by direct differentiation to verify the following results , although the verification will be found pretty tedious .
\mdash ; These are all the integrals needed for the harmonics of the second degree .
In the case of the first we have Thus the coefficient of vanishes and the results are and In the first of these and VOL. CXCVIII.\mdash ; A. 2 The following integrals may also be vel.ified by differentiation : ( 26 ) , ( 27 ) , ( 28 ) , ( 29 ) .
Now in ( 26 ) we have to put ; in ( 27 ) ; and in ( 28 ) Introducing these values , and taking the integrals between the limits and , we find : and ( 30 ) .
and .
( 31 ) .
and ( 32 ) .
OF EQUILIBRIUM OF A ROTATING IsIASS OF LIQUID .
( 3.3 ) .
In ( 30 ) In ( 31 ) In ( 32 ) S5 .
Bifurcation of Jacobi 's Ellipsoid .
If a mass of liquid be rotating like a rigid body about an , with uniform angular velocity , the determination of the figure of equilibrium may be treated as a statical problem , if the mass be subjected to a potential The energy lost in , the concentration of a body from a condition of infinite dispersion is equal to the potential of the body in its final configuration at the position of each molecule , multiplied by the mass of the molecule and summed throughout the body .
In the proposed system , as rendered a statical one , it is necessary to add to the gravitation potential before lnaking the summation .
If denotes the moment of inertia of the body about , this latter portion of the sum is , and is therefore the kinetic energy of the system .
If denote any pair of molecules and the distance between them , and the energy lost , we have If the system had been considered as a dynamical one , the ession f. the energy of the system , say , would have resembled that for , but the former of these terms would have presented itself with a negative sign .
It is clear that the variation of , when the moment of momentum is kept constant , is equal and opposite to the variation of the same function when the angular velocity is kept constant .
The condition for a figure of equilibrium is that shall be stationary for constant moment of momentu , or stationary for constant , in both cases subject to the .
condition of constancy of volume .
The variations in question lead to identical results , and I shall proceed from the variation of If the internal potential of an ellipsoid of mass and semi-axes is Hence . .
Now if denote the principal moments of of the ellipsoid about and similar formulae hold for the two other axes .
Therefore But since is a homogeneous function of degree\mdash ; l in , the sum of the three differential terms is equal to .
Hence this expression is equal to Since we have If be varied , whilst and are constant , it is stationary if @ Eliminating S , Sb , we have the well-known conditions for JACOBI'S ellipsoid .
( 34 ) ' OF EQUILIBRIUM OF A ROTATING MASS OF LIQUID .
If we add together the first two of these , and ayail ourselvss of the property that is homogeneous of degree , we easily prove that the stationary value of is Since the potential of the ellipsoid must satisfy equation Also By means of these and two out of the three equations ( 34 ) , we eliminate the differentials of , and writing for the density find ( 35 ) .
I do not happen to have seen this form for the angular velocity of JACOBI'S ellipsoid in any book .
It is easy also to the stationary value of may be written We may now express the potential , say , of the system entirely in terms and \mdash ; for ' , We thus verify that is constant over the surface of the ellipsoid .
Let denote the value of gravity at the surface .
Then if be an element of the outward normal , Since where Now change the notation and write .
Then ) Now and so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , so that , and ( 36 ) We may note in passing that the condition for a Jacobian ellipsoid ( the last equation of ( 34 ) ) is reducible to the form On examining the series of functions given in ( 25 ) , we see that it may be written This agrees with M. POINCARE\ldquo ; S equation ( 1 ) on p. 341 of his memoir .
We will now suppose that the body , instead of being an ellipsoid , is an ellipsoidal harmonic deformation of an ellipsoid , which is itself a figure of equilibrium for rotation The addition to will consist of three parts ; first that due to the mutual OF EQUILIBRIUM OF A ROTATING MASS OF LIQUID .
energy of the layer of deformation ; secondly that due to the ellipsoid and the layer ; thirdly that due to the change in the moment of inertia .
If a subscript denotes integration throughout the space occupied by the layer , the potential of the ellipsoid , and an element of volume , If denotes the thickness of the layer standing on the element , the first of these terms is The value of throughout the layer is equal to , where is the constant value of over the surface of the ellipsoid , and is the distance measured along the normal to the element of volume .
Hence Since is constant and the total mass of the layer is zero , this is equal to It follows that The axes of the ellipsoid have been chosen so as to make our original stationary , and the further condition to be satisfied is that SE shall be stationary .
Let us suppose that , which expression shall be deemed to include any one of the other types of harmonic .
Then it is shown in ( 51 ) of " " Harmonics\ldquo ; that the potential of this layer at the surface of the ellipsoid is .
Since the mass of an element is , we have With the value of found in ( 36 ) .320 SSOR G. H. ON THE PEAR-SHAPED Hence [ $ In order that the new figure may be one of equilibrium , this expression must oe stationary for variations of .
It follows that we must either have , which leads back to JACOBI'S ellipsoid , or else This last condition is what M. calls the vanishing of a coefficient of stability .
* It shows that if and satisfy not only the condition for the Jacobian ellipsoid , namely , ) , but also this equation , we have arrived at a figure which belongs at the same time to two series , and there is a bifurcation at this point .
The form of the figure is found by attributing to any arbitrary but small S6 .
The Properties of the Successive Coefficients of bility .
Correspondin to each harmonic deformation of the ellipsoid , there is a coefficient of stability of one of the two forms 1 \mdash ; These coefficients may be written , or according to an easily intelligible notation , The soid is defined by , and the question arises as to the possibility of the vanishing of the several 's as gradually ishes from infinity , that is to say , as the ellipsoid lengthens .
An harmonic of the first order merely denotes a shift of the centre of inertia along one of the three axes ; one of the second order denotes a change of ellipticity of the ellipsoid .
Since we must keep the centre of inertia at the origin , and since the ellipticity is determined by the consideration that the ellipsoid is a these harmonics need not be considere , and we may begin with those of the third order .
I shall not attempt to follow M. POINCARE in his master , discussion of the properties of the coefficients of stability , but will merely restate in my own notation the principal conclusions at which he has arrived .
'Acta Math vol. 7 , 1885 , p. 321 .
The factors and ( or , if is the degree of the harmonic ) which occur in his form of the condition are included in my functions .
Sections 10 and 12 of his memoir .
I to thank him for saving me from making a ious m in this portion of my work .
OF EQUILIBRIUM OF A OTATING MASS OF LIQUID .
1st .
The equation is not satisfied by any value of between 1 and infinity , if or is divisible by .
It appears from the forms of the functions as given in S4 of " " Harmonics\ldquo ; that the functions are so divisible .
These functions appertain to the types EES , OOC , OES , EOC , and therefore the ellipsoid cannot bifurcate into deformations of these types .
2nd .
The equation has no solution if is divisible by .
We again see from S 4 of ' ' Harmonics\ldquo ; that is so divisible if it is of types OOS , EOS .
Hence the ellipsoid cannot bifurcate into these types .
The only types remaining are EEC , OEC .
3rd .
The equation has no solution if any of the roots of outside the limits to \mdash ; 1 .
The only of the types EEC , OEC which has all its roots inside the limits to \mdash ; 1 is the zonal harmonic for which O. Hence the ellipsoid can only bifurcate into a zonal harmonic .
4th .
The equation must have a solution between 1 and infinity for all values of It follows from these four propositions that the Jacobian ellipsoid is stable for all deformations except the zonal ones , and that as it lengthens it must at successive stages biful.cate into each and all the zonal deformations .
5th .
As the ellipsoid lengthens , the first coefficient of stability to vanish is that of the third zonal harmonic .
This stage is the end of the stability of the Jacobian ellipsoids , and there is almost certainly exchange of stability with the pear-shaped figure defined by this harmonic .
6th .
It has not been rigorously proved that there is only one solution of the equation even in the case where , but M. ' believes that this is almost certainly the 7th .
The functions have always the same sign as increases from to infinity , provided that.v and are both greater than zero , and greater than 2 .
The seventh of the preceding propositions renders it easy to determine the relative magnitudes of all the 's belonging to a single degree In what follows I may take the symbols $ , as including also , Q. VOL. CXCVIII .
Now as , If we express the 's in terms of integrals this becomes he scventh proposition sholvs that and re greater than zero , and is reater than the elements of the same .
Hence the question Therefore we ] to arrange all the in descending order of magnitude , and shall thereby obtain the non-zonal 's in ascending order .
I wish first to show that these coefficients may to a great extent be sorted by considering the inequality . .
Suppose , if possible , that whereas , for the ellipsoids defined by , yet Then there must be some value of for which for values of greater than It is almost obvious that there is no on .
value of which renders this equation but consider for example the case of Now .
It we substitute this in the equation we find .
This can only be satisfied by , and hence the hypothesis is negatived .
Similarly the assumption of other values of .
and leads to an impossibility .
Thus we may consider the functions in place of the functions .
OF OF A OTATING MASS OF LIQUID .
Con .
cider the inequality , for , 2 . .
, If the inequality is determined for any value of , it is detel.mined for all values .
Now when is very large ' Hence our inequality becomes .
This inequality is of the same kind for all values of .
Now involves factor and involves .
Puttin efore the left-hand side involves and the right .
It follows that unless is equal to the -hand side is greater than the right ; but is necessarily equal to at greatest .
Therefore .
Hence 's with smaller are less than those with eater s It remains to discriminate between the two sorts of -functions which occur in ellipsoidal harmonic analysis ; that is to say we must determine Since the of " " Hal.monics\ldquo ; is equal to in the present notation , when cTnd are small we have by the mulae of that paper . . .
, . .
When is very great and very small , so it suffices to determine the inequality ; and this may be considered for any value of greater than unity .
By taking very large and very small the inequality becomes or But , hence the first sign holds true and whence Thus it follows that for order . . . .
The order of magnitude of these coefficients is thel.efol.e completely determined .
As of the ectness of this result it may be mentioned that I find that when When and ( anothe Jacobian ellipsoid ) the numbers run 130 , 224 , 460 , 465 , 604 , 614 .
We see that for the hal.monics of higher order the ellipsoid is more stable than it those of lower order less stable .
S7 .
The critical Jacobian EllipsoId .
From a nulnbe of I saw reason to believe that the critical ellipsoid would be found within the egion cised between and , and and If we write where the amplitudes of and are and their moduli , the existence of the Jacobian ellipsoid is determined by The coefficient of stability is The formulae for computing are given in S4 .
The values of and are from LEGENDRE'S tables .
*Sec .
Soc. ' vol. 41 , p. 323 , where the formula is reduced to form conve1lient for computation .
OF LQUILIBRIUM A ROTATING OF LIQUID .
Now I find ; ; ; ; By interpolation we get the results:\mdash ; The Jacobian ellipsoid is given by The of the coefficient of stability is given by In these equations the minute of arc is the unit .
them I find With these values I find that the three axes , where are 5 , The last place of decimals in these is certainly doubtful .
The formula for is given in ( 35 ) .
Now , Then since In this formula , must correspond with values interpolated amongst those used in obtaining the solution .
From this I find In the paper on the Jacobian ellipsoid referred to above the moment of nlomentum is tabulated by means of , where the moment of momentum is .
The nula for is given in ( 25 ) of that paper , and , modified to suit the present notation , is For the critical soid I find The following table gives the numerical values a number of Jacobian ellipsoids , with the initial one and terminating just beyond instability .
The last ) rives the corresponding values for the critical ellipsoid .
JACOBI'S Ellipsoid I have been criticised with respect to my paper on ellipsoid , from which these results are extracted , by M. S. KRUGER ( Nieuw Archief voor Wiskunde , Tweede Reeks , Derde Deel and ' Ellipsoidale tsvormen , ' \amp ; c. , Thesis for Degree of Doctor , Leiden , J. W. van Leeuwen , Hoogewoerd 89 , 1896 ) , because I wrote it in ignorance of certain previous work , especially of a paper by PLANA .
But I cannot but congratulate myself on my since it appears that PLANA a nnmber of numerit , al results which were wholly vrong .
A knowledge of that paper would no doubt have caused me much further trouble .
paper gives a number of solutions of the problem which I believe to be correct .
Unfortunately the methods of the paper are clumsy , and there are several mistakes .
The formula for .
used in this present paper , is much better than that given there .
The complicated formula on p. 325 is susceptible of reduction to a simple form , for on substituting for its form ( i ) we have simply where S5 The linal numerical result , howeve , nearly right , for I now , I .
The is the same the used here .
The formula the top of 1 ) .
326 which is reproduced as on 1 ) .
is , I thiuk , illusory , for if in the OF EQUILIBRIUM OF A ROTATING MASS OF LIQUID .
In order to determine the question as to whether or not it is possible that should have another solution than that found in the next section , I have computed the value of this coefficient for the Jacobian ellipsoid , and find it to be .
From the manner in which the numbers in the computation present themselves , it is obvious that for more ated ellipsoids will always remain negative , and will become nulnerically greater .
I have therefore not thought it necessary to seek for an algebraic proof that there is no second root of the equation .
Very long Jacobian ellipsoids tend to figures of revolution , and the coefficients of stability tend to assume the fornoe forms of these functions are well known , and I think that fair approximations to the incidences of the successive figures of bifurcation might be derived from the vanishing of this expression .
For example have not , however , attempted to solve the equation found by equating these two expressions to one another .
Even when and ( the critical Jacobian ) this rough approximation makes the coefficient of stability very small , but it is to be admitted that and differ very sensibly from and , although in such a way that the errors compensate one another .
first term we put S ( as is clearly allowable in approximation ) the term with coefficient or disappears .
This shows that it was necessary to proceed in the approximation as far in order to obtain a result .
The methods of approximation adopted on pp. 326-7 are of doubtful propriety , but will , I think , lead to nearly correct results .
There is , however , a mistake towards the bottom of p. 327 lvhich runs on to the end .
M. KRUGER correctly points out that the second line of formula 24 ) p. 329 should rum Lastly , on p. 33 lino 13 , for , read ; and p. 336 , line 7 , for , read ; and for , read .328 .
H. S 8 .
The pear-shaped Figure of Equihbrium .
By ( 21 ) the normal displacement Sn for the third zonal harmonic deformation may be written subject condition The expression has been arranged so that when , we have Hence and are the normal displacements the stalk and bhmt end of the respectively .
In the section , this may be written The nodal points are given by In the section , since , it may be written The nodal points are given by The section is obviously another nodal line for all sections .
By means of these it is easy to compute the normal displacements from the surface of the critical Jacobian .
The figure opposite showing the three sections , is drawn from these formulae , the dotted line being the critical .
Tacobian and the firm line the pear .
The scale of the normal displacements is , of course , arbitrary .
Comparison with M. , sketch shows that ths figure is coltsiderably longer than he supposed .
In this first approximation the positions of the nodal lines are independent of the magnitude of , and they lie so near the ends that it is impossible to construct an exaggerated figure , for if we do so the blunt end acquires a dimple , which is It might have been hoped that such an exaggeration would afford us some idea of the mode of development of the pear .
OF EQUILIBRIUM OF A ROTATINT MASS OF LIQUID .
029 VOL. CXCVIII.\mdash ; A. 2 PltO G. H ON THE FIGURE M. SCHWARZSCHILD has remarked* that it is not absolutely certain that the principle of exchange of stability holds with reference to this figure , and that we cannot feel absolutely certain that the pear is stable unless we can prove that the moment of momentum is greater than in the critical Jacobian .
With reference to this objection , M. POINCARE writes to me as follows:\mdash ; " " Faisons croitre le moment de rotation , que j'appellerai M. Deux hypotheses sont possibles .
" " Ou bien pour ( the moment of momentum of the Jacobian ) , nous aurons un seule figure , stable , a savoir l'ellipsoide de JACOBI , et pour trois figures , un instable , l'ellipsoide , et deux stables ( d'ailleurs egales entre else ) , les deux figures pyriformes .
" " Ou bien pour , nous aurons trois figures d'equilibre , deux pyriformes instables , un stable , l'ellipsoide , et pour un seule figure instable , l'ellipsoide\mdash ; auquel cas la mass fluide devrait se dissoudre par un cataclysme subit .
" " Il a donc a verifier si pour les figures pyriformes , It seems very improbable that the latter can be the case ; but this opinion is not a proof Since is stationary for the initial pear , a small change in the angular velocity will certainly produce a great change in the figure of the pear .
If this investigation has , in fact , its counterpart in the genesis of satellites and planets , it seems clear that the birth of a new body , although not cataclysmal , is rapid .
S9 .
Summary .
It is possible by the methods explained in my previous paper on " " Harmonics\ldquo ; to form rigorous expressions for the ellipsoidal harmonics of the third degree .
Accordingly in S1 I proceed to form those functions .
In S2 the notation is changed with a view to convenience in subsequent work , and for the sake of completeness the harmonics of the first and second degrees are also given .
In S 3 the corresponding solid harmonics are expressed in rectangular co-ordinates .
In S 4 I find the -functions , the harmonic functions of the second kind , and express the results in terms of the elliptic integrals and .
It appears that both the P- and -functions of the third degree of harmonics occur in three pairs which have the same algebraic forms , and that in each pair one of them only differs from the other in the value of a certain parameter .
There is , lastly , a seventh function which stands by itself ; this last corresponds to the solid harmonic In S5 the equations for JACOBI'S ellipsoid are determined by the consideration that the energy must be stationary , and the superficial value of gravity is found in terms of the appropriate and -functions .
I then proceed to find the additional terms * ' Die Poincaresche Theory des Gleichgewichts Almalen der K. Sternwarte , Munchen , ' Bd .
III .
OF EQUILIBRIUM OF A ROTATING MASS OF LIQUID .
in the energy when the mass of fluid is subject to an ellipsoidal harmonic deformation .
This section is a paraphrase of M. PoINCARA'S work , but the notation and manner of presentation are somewhat different .
The additional terms in the energy are shown to involve a certain coefficient , which is called by POINCARE a coefficient of stability .
It is clear that when any coefficient vanishes we are at a point of bifurcation , and the particular Jacobian ellipsoid for which it vanishes is also a member of another series of figures of equilibrium .
In S 6 the principal properties of these coefficients , as established by M. POINCARE , are enumerated .
He has shown that the ellipsoid can bifurcate only into figures defined by zonal harmonics ; that it must do so for all degrees , and that the first bifurcation occurs with the third zonal harmonic .
The order of lnagnitude of the coefficients of the seyeral orders and of the same degree is determined .
A numerical result seems to indicate that as the ellipsoid lengthens , it becomes more stable as regards deformations of the third degree and of higher orders , and less stable as regards the lower orders of the same degree .
In S 7 the numerical solutio of the vanishing of the coefficient corresponding to the third zonal harmonic is found , and it is shown that the critical ellipsoid has its three axes proportional to 65066 , 81498 , , and that the square of the angular velocity is given by .
A short table is also given showing the march of the axes of the Jacobian ellipsoids from their beginning on to instability at this critical stage .
The nature of the formula for the third zonal coefficient of stability seems to show that it can only vanish point which it appears that M. POINCARE ' found himself unable to prove rigorously .
A suggestion is made for the approximate determination of the bifurcations into the successive zonal deformations , but no numerical results are given .
In S 8 the nature of the pear-shaped figure is determined numerically , and the reader may refer to the figure above , where it is delineated .
It will be seen to be longer than was shown in M. POINCARE\ldquo ; S conjectural sketch .
If , as M. POINCARt suggests , the bifurcation into the pear-shaped body leads onward stably and continuously to a planet attended by a satellite , the bifurcation into the fourth zonal harmonic probably leads unstably to a planet with a satellite on each side , that into the fifth to a planet with two satellites on one side and one on the other , and so on .
The pear-shaped bodies are almost certainly stable , but a rigorous and conclusive proof is wanting until the angular velocity and moment of momentum corresponding to a given pear are determined .
To do this further approximation is needed .

