]\gt ; PHILOSOPTIICAL TRANSACT[ONS .
I. On the Instability of the Pear-shaped Figure ofof Rotating Mass of Liquid .
J. H. , F.R.S. Received March 21 , \mdash ; Read Apri16 , 1916 .
1 .
THE main result obtained in this paper is sufficiently indicated by its title .
In a recent paper*I showed that the stability of the pear-shaped figure could only be decided on after the figure itself had been calculated as far as terms involving the cube of the parameter , which is used to measure the deviatio of the pear-shaped figure from the critical Jacobian ellipsoid .
In the present papel these third-order terms are calculated , and the pear-shaped is litoly s to be A statement of the results obtained , and discussion of their bearing on the wider question of which this problem forms a part , ill be found at the end of the paper ( SS23-27 ) .
2 .
The discussion has to begin with a determination of the potential of a distorted ellipsoid , carried as far as the third order of the small quantities involved .
With a view to shortening very lengthy computations , it is convenient to arrange the algebraic solution in a form solnewhat different from that previously given .
The tion now given can readily be extended to any order of small quantities , and appears to lead to the most concise series of computations for terlns of all degrees above the Potential of Distorted Ellipsoid to thoe of Higher Orders .
3 .
As before , the undisturbed ellipsoid ( which will ultimately be supposed to be ) critical Jacobian ) is taken to be the surface in the nily of surfaces , ( 1 ) while the disturbed ellipsoid whose potential we require ( which will ultimately be " " On Potential of Ellipsoidal Bodies , and the Figures of Equilibrium of Rotating Liquid Masses ' Phil. Trans , vol. 215 , p. 27 .
VOL. CCXVII .
July ll , J. H. JLANS ON THE INSTABILITY OF THE PEAlt-SHAPEr ) supposed ) be the ) figure ) is taken to ) the surface in the more general family , . . . . .
( 2 ) Here ( is a function of , and , different values for representing different distortions of the fundamental ellipsoid .
If are external at ( or their analytical continuations as explained in the previous per , S3 ) , we have seen that must be of the form , ( 3 ) , . . . . .
( 4 ) where .
( 5 ) In these integrations , are treated as constants while is the value of at the point as determined by equation ( 2 ) .
Furthermore , ( l ) must not be selected at random ; equations ( 3 ) and ( 4 ) will only give the true potentials if is chosen so as to satisfy . . .
( 6 ) This equation , as before , is most conveniently solved by a solution , . . . . .
( 7 ) in which must satisf.y the equations , ( 8 ) .
( 9 ) 4 .
We attack the second equation first .
As in the previous paper , let us introduce new co-ordinates defined by \amp ; c. , and the equation is found to reduce to ( cf. equation ( 37 ) of the previous paper ) OF EQUILLBRIUM OF A ROTATING MASS OF LIQUID .
.3 this equation we may put , or , since , we may put of conrse it ) be legitimate to coefficients of these quantities .
The equation educes to , ( 10 ) and this be readily solved in powers of the parameter on assuming a series of the form ( 11 ) Equating coefficients of the different powers of we obtain , and so on .
To satisfy the first of these equations , must be a function of only , say P. To satisfy the remaining equations , write , \amp ; c. , so that , \amp ; c. if is written for , the equation for is of which the solution is , . . . . .
( 12 ) where is a function of only .
Proceeding in the same way , we find . . .
where is another function of only , and so 011 .
5 .
We now return to equation ( 8 ) .
By a transformation given in the previous paper , it is found that this is satisfied by taking when , . . . . .
( 14 ) and .
( 15 ) where may be any function of and , which vanishes when and when .
For equation ( 15 ) we may try provisionally a solution . . . .
( 16 ) where are quantities satisfying .
( 17 ) .
( 18 ) \amp ; c. ( 19 ) After a good deal of simplification , the left-hand member of equation ( 15 ) is found to reduce to .
( 20 ) The quantity is so far undetermined .
Let it be given by , ( 21 ) so thal the quantity in the square bracket on the right of equation ( 20 ) becomes When , this vanishes through the factor .
It will vanish when through the last factor if we make , when ( 23 ) If last condition is satisfied , expression ( 22 ) satisfies completely the conditions which have to be satiSfied by in equation ( 15 ) .
HeIlce the value of given by FIGURE OF EQUILIBRIUM OF A ROTATING OF LIQUID .
.5 equation ( 16 ) will be a solution of equation ( 15 ) .
, vanish when ) will also vanish when , so that equation ( 14 ) will also ) satisfied .
It follows that equations ( 16 ) to ( 19 ) , with ( 23 ) , cont a complotc solution of equation ( 8 ) .
6 .
These equations cill solved in powers of the parameter .
Lot us for expansions in the form , and for ?
, \amp ; c. , expansions , , The coefficients in the expansions of are of course not dent of those in the expansion ( 11 ) already assumed for We find easily enough the rela.tions , \amp ; c. The value of ( equation ( 21 ) ) is found to be On equating coefficients of powers of in equations ( 17 1J ) , we , . . .
( 24 ) , \amp ; c. , ( 25 ) , ( 26 ) , \amp ; c. . . . . .
( 27 ) MIR .
J. H. JLANS ON TfIE I OF ) are an inite llumbe of sets of equations , of which we shall only the addition sets above .
To simplify equations as much as , let us limit ourselves to the type of distortion which leads to the -shaped series of figures equilibrium .
For this , as we saw in the previous papel , ?
is of degree 3 in , so that must by equation ( 24 ) be of degree unity , eqnation ( 25 ) ?
must vanish .
Similarly ?
is of degree 4 , so that is of is of , and .
Again will be of degree 5 , of degree 3 , ?
of unity , ; and so on .
The set of equations for now reduces to ( 28 ) . . . . .
( 29 ) 7 .
Let us now introduce the operator , already used in the previous paper ( S14 ) , defined by . .
, . . . . .
( 30 ) By differentiation with respect to , we have so that is the same as transformed into co-ordinates .
We can now solve equations ( 24 ) to ( 29 ) at once by transforming into co-ordinates .
Equation ( 24 ) becol1les in co-ordinates and since is a lnctio of only , this has the integral . . . . . . . . .
( 31 ) No stant of integration must be added , for vanishes when and must also be made to vanish when ( cf. equations ( 23 ) ) .
Thus equation ( 31 ) ives the true of we also seen a ) that ?
value of is accordingly ( 32 ) FIGURE OF EQUILIBRIUM OF A ROTATING MASS OF From SS6 and 4 , we now as the value of . . . . . . . . . .
( 33 ) The .
of can next be found from equation ( 26 ) .
The right-hand embel reduces to , and the equation , expressed in co-ordinates , becomes of which the solution is .
, .
( 34 ) and similarly equation ( 27 ) leads to . . . . . .
( 35 ) These values , and are with those obtained in the earlier paper , although obtained a slightly different method .
We 11 proceed to the third order .terms .
On substituting for its value as obtained in Sr we find , in place of equation ( 28 ) , which , after a good deal of simplification , .
( 36 ) Introducing the various values which havs been obtained for we find , after simplification , that .
( 37 ) where , \amp ; .
, ( ) , \amp ; .
, \amp ; c. Relation ( 37 ) can accordingly be put in the form ( PQ ) quation ( , after transformation to co-ordinates and integration with respect to , gives ( PQ)DR . .
( 38 ) Let us now put , ( 39 ) so that will be a function ) only .
We then have , ( 40 ) whence , on operating with .
( 41 ) Equation ) , transformed to -ordinates , now becomes OF QUILIBRIUM O A OF LIQUID .
( i ) giving on integration , 8 .
solution is now complete as as third ] ) sed in a more form .
We found the value of , . . . .
( 42 ) vlJere Z is formed of terms volving the function has its value given by The term in ) gives rise to a in this in turn to a term in the fun ction prev i ) ) , which solution is derived .
Using the value for has just been obtained , readily find that , in co-ordinates , - . . . . .
( 43 ) We found , }lowever see to 1 .
32 ( .
previous ) ) , that given potential , the value is .
] ctioI of the potential problem , then it was , ( 44 ) will a tion of the same probleln , provided that is any function of , , and when when relation ( 2 ) ) , and u is any function of whatever .
with these conditions we take and the new solution ( 44 ) becomes VOL. CCXVII.\mdash ; A. In this solution , already colltains a term involving , namely that given by ression ( 43 ) .
Combining this with the term in ession ( , we find that the new solution can be put in the form ( I ) , ( 46 ) where is the old solution with the terms in omitted .
The last term in expression ( , being proportional to order small quantities .
Thus in a solution as far as only , this term be omitted , and will be a solution .
In other words the term ) entirely from equation ( 42 ) , and the remaining terms will still give an accurate solution for Omitting this -term , we obtain for the tbird order terms , ( PQ)R - . . . . .
( 47 ) This completes the solution of the general potential problem .
Potential of the Pear-shaped Figure .
9 .
Collecting the results obtained in SS3-8 , we have found that as far as terms of the third order of small quantities , a value of which satisfies the necessary difFerential equation ( 6 ) is . . . . . .
( 48 ) At the boundary , this value of reduces to , . . . . .
( 49 ) and since are entirely at our disposal this is capable of representing the most general displacement possible , as far as the third order of small quantities .
We have , however , to save the printing of additional terms , already assumed that is of a degree not higher than the third i and of a degree not higher than the fourth .
Subject to these limitations , the potential of the ellipsoid deformed in any way can be obtained by lnserting the value ( 48 ) for into equations ( 3 ) and ( 4 ) .
FIGUPffi OF EQUILIBRIUM OF A ROTATING MASS OF LIQUID .
10 .
We shall now limit ourselves to the distortion which was found , in the earlier paper , to lead to the pear-shaped figure of equilibrium .
For this particular distortion we found ( cf. equations ( 93 ) and ( 94 ) of the earlier paper ) , , . . . . .
( 50 ) .
( 51 ) We shall now assume for the value , ( 52 ) this foun adequate to satisfy all the conditions for , a figure of of a rotating liquid mass .
11 .
The values of ?
and of were calculated in the form paper ( see equations ( 121 ) , ( 122 ) and ( 132 In the same way the value of ?
, as given by equation ( 47 ) is found to be [ ?
AB AB BC , . . . . . . . . . .
( 53 ) stands , and is given by . . .
( 54 ) 12 .
The potential of the distorted ellipsoid , . . . . .
vrittcI its as terms in The terms in of course , have been calculated in the previous paper .
The terms in in the potential the boundary , are ( cf. equation ( 3 ) ) , , ( 56 ) in which has the value just given in equation ( 53 ) .
OF A ROTATING LIQUII ) .
Let us put .
( 57 ) of the various fficicnts in this expression are found to ) as follows:\mdash ; AB AL $ BC .
( 58 ) AC 1-1 MR. J. H. JANS ON THE INSTABILITy OF THE PEAR-SHAPED value of be written down from symmetry . .
( 60 ) The value of can be written down from symmetry .
: - BC .
( 61 ) FIGURE OF EQUILIBRIUM OF A ROTATIN MASS OF ) .
The values of the four coefficients are most conveniently expressed in , ( G3 ) , ( 64 ) , ( 65 ) where are quantities which do not depelld on , 8 , being given by the followin equations:\mdash ; [ . . . . . . . . .
( 66 ) BC . . . . . . . . . . . . .
The value of } vritte down sylnnletry ; that of ich is of the in present 13 . .
potential which we have bad in hand .
It naturally be feared some lnisbake have been made either in principle or iu detail , and it lllust ) remembered that even one mistake the answer to the whole .
I have , therefore , here and , taken the nt ost care to check the accuracy of my work in every way .
The following ) , I , show no error need be feared in the solution which has just been obtained .
value of , if obtained tely , ought to satisfy , and the term must entirely the terms dependent of in .
Thus the terms in are lnultiplied by and ought separately to be spherical harmonics .
It was verified in the paper that the terms in ) were in actual lact of this form .
The terms in .
will ) harnonic if the quantity on the right of equation ( 57 ) is harmonic , and the conditions for this are expressed by the equations , . . . . . .
( 68 ) , . . . . . .
( 69 ) , . . . .
( 70 ) . . . . .
( 71 ) I inserted the values just obtained for , in these equations , and have verified that they are all satisfied .
( The necessary transforlnations of the various rals are tedious , but involve no special dificulties .
) It follows that the solution obtained accurately the potential of some solid of uniform density .
By the method explained in the earlier sections of the paper , it is easy to ) roblem f ) ackwards and to verify that the equation of the boundary of the solid in question is obtained by putting the equation .
Thus we verify that our solutio1l accurately the potential of solid of boundary . . . . . . . . . .
( 72 ) ditions that be on of for a Jiotating 14 .
order that the equation shall be one of equilibriun for a rotating liquid , the potential at the dary plus must , as in equation ( 98 ) the former paper , be identical with constant .
FIGUlffi OF A ATING MASS OF LIQUID .
Let us ourselves to the terms in } .
Using the value of given by equatio1l ( 56 ) and that assulned in equation ( 52 ) , we find we must have we ; ; , .
( 73 ) ; ; , ( 74 ) ; ; , ( 75 ) .
( 76 ) , in addition ) those fouled for the -order terms in the paper , the condition that the third-order shall be a possible figure of equil i 15 .
On .
the of which been obtained S12 into the six equations ( 73 ) and ( 74 ) , obtain of six equations from which it is possible to determine the six $ , .
The solution is actually effected in S17 below .
Tf we substitute the values of , il1( obtained in S12 ( equations into the three equations ( 75 ) , we obtain three equations can be vrittel in the form : ( 77 ) , ( 78 ) .
( 79 ) It be seen do not occur on equations .
VOL. CCXVII .
These equations are not , as might at first be thought , a system of three simple equations determining .
They will be found to be of the type known as " " \ldquo ; that is to say , equations which are inconsistent unless the coefficients satisfy a certain relation , and such that , when this is satisfied , the equations lave an infinite } ) of solutions .
Let us , for brevity , write equations in the form : .
( 80 ) Let us use also the abbreviated notation of the previous paper ( p. 50 ) , such that Then , by simple transformations of the integrals , we obtain ( 81 ) With these values of the coefficients , it will be found that equations ( 71 ) of the previous paper reduce to , ( 82 ) so that , . . .
( 83 ) is a solution of atiol 1 , when detailed inspection of the equations , it could been foreseen that this would necessarily be the case .
For our general solution , . . . . . .
( 84 ) See ' HOBSON'S Plane Trigonometry , ' S73 , or IVOLSTENHOLNE , 'Proc .
London Math. Soc vol. 4 .
FIGURL OF EQUILIBRIUM OF A ROTATING MASS OF LIQUII ) .
must cover all possible figures of as far hird ordel such figure is , however , known to be ( 85 ) and the corresponding solution for is that expressed by From equations ( 82 ) , it follows at once that we must have .
( 86 ) Indeed , it is now clear that this is precisely the e which determined the existence of the point of bifurcation on the series of Jacobian ellipsoids .
* If this were the relation between the coefficients in equations ( 80 ) , these equations could have no solution other than .
Let us , however , multiply the three equations ( 80 ) by the three minors of in the determinant of equation ( 86 ) , and add .
We obtain a relation of the type , ( 87 ) and it is clear that these equations can now have a solution in which are not all infinite if we have .
( 88 ) It is only when this relation is satisfied that it is possible to continue our linear series of equilibrium configurations beyond the second order terms .
When it is not satisfied , our third order solution ( 84 ) lapses back into the , namely , which is virtually the first order solution with replacing parameter .
16 .
The relation ( 88 ) can be expressed in a much simpler form .
Independently of the values of and , we have already seen ( equation ( 71 ) ) that we must have . . . .
( 89 ) * As to the relation of this equation to the general theory , see S36 of my } ) rovious 1 ) } ) ON TH1 THI ) by equ in to , we obtain With the help of these relations equation ( 88 ) reduces quite simply to .
( 90 ) In this equation the coefficicnts of and , depend only on semi-axes of the ollipsoid , and so are fully known .
quantities however depend the ) rder coofficients , ) were calculated in previous paper , but could not be fully deterlnined , since they were found to depend ( measured the change in angular velocity This it was found impossible to evaluate ) long as the investigation was confined to second-order terms .
It now appears that equation ( 90 ) is in effect all equation .
The equation is linear in so it gives only one value for When ?
has this value we on the true linear series , but has any other value our solution , when we to extend it to ternls , tes into a solution of the of ( 85 ) , with which no ress c be llade .
Our plan , then , is to evaluate the tel'lns which ( ccnr in equation ( 90 ) so obtain the value of On inserting this into , values of were obtained in the previous papc we complete th as second-order terms , and then proceed to the stability criterion .
Numerical 17 .
It is at once apparent atio of , , iven by equations ) , can be made to depend on integrals of the occurred in vious paper , namely defined by IGURE OF OF A MASS LIQUII Tn ( which in pape , * have calculated values dditioI of third-order terlns .
, , The first task is the cvaluatio of , .
These quantities are connected by the three relations ( 68 ) , so that three of the six quantities need have been calculated .
I , however , calculated four coefficients , namely , and quite independently , the object being to obtain a check on , accuracy of computations .
In this way a check is obtained at step of the computations .
* The following { true lues of integrals which were incorrectly given in the previous In each case the error was one of printing only , did not entcr into the coinpntations .
The whole table given in the previous was recomputed before extending it for pnrposes of the prcsent ) aper .
values obtained , ( 91 ) , ( 92 ) , ( 93 ) . . .
( 94 ) Two values of can now be deduced from equations ( 69 ) and ( 70 ) respectively .
These are found to be The reement of these values provides a check on the computations of the coefficients , , and of the integrals from which they have been calculated .
In virtue of relations ( 73 ) , equations ( 68 ) become , ( 95 ) , ( 96 ) , ( 97 ) while relations ( 73 ) and ( 74 ) , of which only three are now independent , may be represented by , . . . . .
( 98 ) , ( 99 ) .
( 100 ) On substituting for , and from equations ( 91 ) , these become six linear equations , 1 , , and FIGURE OF EQUILIBRIUIsI OF A ROTATING OF LIQUII ) .
2:3 The solution of these equations is found to be 5365 .
These values been ecked insertion , not only in the six equations from which they were directly derived , but also in the remaining equations ( 73 ) and ( 74 ) .
We now have all the material necessary for the evaluation of whid ] , it wilI be remembered , are the right-hand members of equations ( 77 ) respectively .
These have been computed independently , nd I found TlIese values ought to satisfy ( cf. S16 ) in place of I find but the is greater thaIl might reasonably be expected in view of the very large number of operations in each computation .
* Thacoefficients in equation ( 90 ) are found to be , so that the equation itself becomes and the solution is found to be of 18 .
On substituting the value just obtained into tho values for foumd in the prcvious paper ( S34 ) , I find , thus completing the figure to of titics il it of on p. 21 .
The nlcly 1 it ) term .
Each of the quantities consisted of 326 snch such terms , each of which , it must bc remembered , is evaluated by fairly letIgthyscric.s of is by increases as we pass along the pear-shaped sel.ies , whereas in Sir G. solution was found to decrease .
Thus the present solution es in esscntials from that of DARWIN .
On the other hand the solution is similar to that for rotating cylinders*i the rotation was also found to increase as we passed along the pear-shaped series , we now see , th crease is at a very similar rate .
present figul.e , as far as order of small quantities , is the cylindrical , on replacing the p ) ueter 1 used in the second half of the two-dimensional by found to be A comparison of these two figures shows that the two may be taken to be very tely the same .
As velocity , alue in present threedimensional problem has been found to be , that in the nensional problems .
, in so far as it is possible to compare three-dimensional problem with its two-dimensional aualogue , we may say that the two rotations are in very close agreement .
of of Inertin .
19 .
We know will be or unstable according as momentum inCl.eases or decreases we the critical Jacobian ellipsoid along the se ] figures .
" " On the lindcrs ' Phil. , vol. I the of the third of titics ition oumbers giveIl origiunlly for the FIGURE OF EQUILIBRIUM OF A ROTATING MASS OF LIQUID .
The moment of inertia of the pear-shaped out the axis of rotation , say , will be given by where the integral is taken throughout the volume of the pear-shaped figure .
We have , by our choice of the coefficient , ensured that the volume of the pear-shaped figure shall remain always equal to that of the original ellipsoid , so that we have and therefore ( 101 ) Let us transform to co-ordinates iven by .
( 102 ) so that critical Jacobian ellipsoid is reduced to a sphere of unit radius , and the pear-shaped figure is reduced to a distorted sphere .
With this transformation , equation ( 101 ) becomes died . . . . .
( 103 ) where the integral is taken throughout the figure bounded by the surface .
( 104 ) Let be written for , and let us further put so that , are co-ordinates on a sphere of unit radius .
Lquation ( 104 ) becomes .
( 105 ) Let us suppose that , the radius vector to the boundary of this distorted unit sphere is given by VOL. CCXVII .
On substituting this value for .
into equation ( 105 ) and equating the coefficients of and , we obtain . . .
( 10C ) .
( 107 ) If is an element of solid angle , equation ( 103 ) may be written in the form Hence we find that may be written in the form , where , as ' as The integral is here taken over the sphere of unit radius , and so can be easily evaluated .
Carrying out the necessary computations , I find Thus the moment of inertia is given by 20 .
This again differs from Sir G. DARWIN'S result , in which it will be it was found that the moment of inertia of the pear increased as increased .
But we shall now see that the difference between the two results agrees exactly with what was to be expected from the different values of used in our two solutions .
Sir G. DARWIN chose his parameter ( which I shall denote by ) in such a way that the longest radius vector of the pear-shaped figure was .
In my solution , the longest radius vector is found to be , where denotes my parameter .
Hence I find as the relation between our parameters DARWIN'S rotation was given by .
FIGURE OF EQUILIBRIUM OF A ROTATING MASS O LIQUID .
The general solution , apart from special values for the rotation , is such that , in my notation , .
Hence I find DAIIWIN'S solution ought to coincide with the solution obtained in my previous paper to a given by tion g in present paper , which is believed to give a true of equilibrium , is derived from the by ssigning to a lue by In previous , it was that there was , so far as second-order terms only were concerned , a doubly infinite sel.ies of figures of , and it was that these could be defined in terms of two independent parameters , and where was the same thing as .
It accordin gly appears that any figure of series differs from figure of equilibrium the same value of ough his being different from mine .
The excess of my over his will be It is readily seen that the increase in the moment of inertia oyer that of the critical , which has been called , will be a linear function of and say Hence the exceHs of that of DAliwlN , will be of is easily obtained by ) to vanish in ysis given on p. 72 of my previous paper .
quantity hen appears of incr as as we pass along the cobia series of ellipsoids .
The value of , in the notation there used , , whence excess of my of over of RWIN o accordingly to be * I have changed the sigIl of , which , by an oversight , been with the oppositc sign in my formulae as printod in the My value of having been seen to be , it follows that DARWIN'S ought to bs so that of as deduced from my calculations ought to be .
In point of actually given by DAJiWIN was .
It appears , then , that DARWIN'S moment of momentum agrees exactly with mine , as it ought , except for the difference introduced by the different values we have taken for But , besides showing this , the calculations just given provide a check on the accuracy of the computations of both of us .
Although our figures , as far as the second order , have been calculated by very videly different methods , their moments of momentum have been found to agree very closely .
* 21 .
The moment of momentum in the cylindrical problem was announced in my -dimensional paper to increase with increasing On repeating the computations of this paper , I find that the coefficients in the equation of the surface were correctly given , but the final computation of was erroneous .
The corrected formula becomes .
The tability C 22 .
We have now found for the pear-shaped figure of equilibrium , * Il dded that before I had discovered this check on my )utations , I had calcuIated the of my figuro by diroct intcgration , using the method of S19 .
The total mass ought of course to come to exactly .
I find that the tcrmb resuIt in ] an incrcnse of mass , while thosc in , Nauce this with a decrease The error is in the very stage ; the on p. 95 ought to read and this lcads to the .
I have also calculnted .
by the method of S19 of the present paper , and found FIGURE OF EQUILIBRIUM OF ROTATING MASS OF LIQUID .
It follows that the moment of momentum is give by .
Thus so that the moment of momentum decreases as we pass along the series of -shaped figures , this series is therefore unstable .
AND DrscussIoN 23 .
hout the present paper , and my previous paper on the same subject , the cobian ellipsoid which bifurcates into the pear-shaped series of figures of equilibrium , been taken to be Any adjacent figure , whether of equilibrium or not , be supposed to be where is a function of in which the coefficients are numerically small .
For special values of , this figure will be one of equilibrium .
So long as we consider only figures which differ infinitesimally from the possi figures of equilibrium form a linear series , and is of the form , ( 108 ) where is a function of and is a parameter which must be so small that can be neglected .
In the previous paper it was shown that as soon as ) is taken into account , there must be supposed to be a doubly infinite series of figures of equilibrium .
The general form of is . . . . .
( 109 ) where is a second parameter of the same order of magnitude as but capable of varying quite independently of The of for this figure of equilibriunl is reater by than the ) critical .
The possible figul.es of equilibrium may be of as .
inside a rectangle as rectangular co-ordinates .
In the paper I have carried the investigation as far as , and find that the value of as far as third-order terms must be of the form . . .
( 110 ) where is a new function of and is a constant .
At first sight appears to be at our disposal , for if we replace the parameter by a new parameter , we can -11R .
J. H. JEANS ON INSTABILITy OF THE PEAB-SHAPED vary as much as we please .
But on examining the problem in detail it is found that is always infinite except for one special value of .
For this special value of we can eliminate altogether by a new choice of parameter .
But for other values of , our solution is only valid if is small , and , replacing by a new parameter , the solution reduces to and so returns to the original first-order solution ( 108 ) .
Tbus , except for one special value of , it is impossible to carry the linear series outside the second order rectangle ; if we attempt to do so , the solution ps lapsing back into the first-order solution .
In the previous paper I gave reasons for believing that Sir G. DARWIN , in his well- known investigation of this problem , had introduced a spurious equation of equilibrium .
This extra equation could only be satisfied by assigning to a special value , IJanlely Sir G. RWIN accordingly gave this value to , so that the value of ( o ) decreased on passing his series of pear-shaped figures , and , assuming this value for , he showed his series to be stable .
But the investigation of the previous paper showed that there was need to assign this special value to , and the present investigation has further shown that with this value of it is impossible to extend the series beyond second-order terms at all .
There is only one value of which leads to a real linear series of configurations , and this is shown in the present paper to be as we pass along the true linear series continually increases .
The angular is vever found to decrease , so that the pear-shaped figure is shown to be unstable .
24 .
amount of colllputation involved in the problem has to be very great , as the whole question of stability instability depends on the sig of a single at the end of all this computation , the question of numerical accuracy becomes one of great importance .
The difference veen my second-order figure and that of DARWIN ises solely the diflerence in the value of .
The moment of inertia of such fures is linear fu1lction of , and a very simple calculation ives the rate at which it ought to vary .
Allowing for this difference in , I find that my computations give for the moment of inertia of DARWIN'S figure ( in terms of DARWIN'S parameter ) , , while DARWIN culated as the value of the same quantity .
FIGURE OF EQUILIBRIUM OF A ROTATING MASS OF LIQUID .
Since two independent sets of computations , conducted by entirely different methods , have been found to lead to precisely the same result , it seems highly probable that this result is accurate .
The agreement just mentioned may reasonably be regarded as guaranteeing the accuracy of all the second-order computations , both of DARWIN and myself .
actual criterion of stability , however , depends on the value assigned to , and since this depends in turn on the third-order terms , no check by compal.ison with DARWIN'S work is possible .
Some support is give to my value of by comparison with a parallel investigation of the ' ' Equilibrium of Rotatin Cylinders which I published some years ago .
Adjusting the parameters so that shall have , as closely as possible , the same physical interpretation in the two problems , I find for the factor expressing the increase of as we pass along the series of pear-shaped ures : for the three-dimensional problem , for the two-dimensional problem .
Apart from this , the checks I have used in the present paper are such that I believe it would have been very difficult for any error to escape detection .
25 .
The main object of the paper is achieved as soon as the pear-shaped figure is shown to be unstable .
It is at the same time of interest examine the bearings of this result on the wider question of which it is a part .
In fig. 1 let represent the series of Jacobian ellipsoids , the part PO ( drawn thick ) representing the stable part of the series , and the part OP ' ( drawn dotted ) representing the unstable part , so that is the point of bifurcation .
Let a diagram be drawn about this line having the angular momentum always represented by the vertical co-ordinates , so that for instance all systems , whether in equilibrium or not , which have the same angular momentum as the critical Jacobian ellipsoid , must be represented in the horizontal plane through ( this plane having as many " " dimensions\ldquo ; as are necessary ) .
In this diagram the pear-shaped series of figures will lie below this plane in the neighbourhood of O. There are two priori alternatives , represented on the right-hand and left-hand sides of.fig .
1 .
In the first place , it is possible that after passing a certain distance OR ' along the pear-shaped series with decreasing angular momentum , we come to a region ll'S ' in which the angular momentum again increases .
Any horizontal line in the diagram ought , on the principle of stable and unstable configurations of equilibrium occurring alternately , to lneet stable and unstable branches of the linear series alternately .
Thus the branch ought to be stable , so that would be a point of bifurcation on this series , and the other series through , say , would be unstable .
The alternative possibility is that the pear-shaped series of figures proceeds continually in the diagram , so that its angular momentum continually decreases .
26 .
Either of these two possibilities removes a difficulty to which Sir G. has drawn attention .
* In what is commonly referred to as ROCHE'S problem an attempt is made to discover the other end of the pear-shaped series of figures , on the supposition that this other end represents two detached masses revolving about one another .
Two such series of figures appear to exist ; in one the satellite is spheroidal except for tidal distortion , while in the other it is ellipsoidal .
The former series has been shown to be stable , the latter unstable .
As the angular momentum decreases on passing along these series , the distance between the two masses also decreases until a point is reached at which the two series coalesce , the configuration of bifurcation being one in which the radius vector from the centre of the primary to that of the satellite is equal to radii of the primary .
If the distance betwee1l the masses is decreased still further , the remaining configurations form an unstable series .
Sir G. DARWIN found a difficulty in the instability of this series , since he believed it to be the far end of the pear-shaped series which he thought stable .
We now see , however , that this series may , without change of stability , join up with either the series on the right-hand of our diagram , or the series RO on the left .
'Coll .
' III .
, pp. 515-524 .
problem has only been solved strictly by imposing sphericity on the and assuming the satellite to be .
Sir G. DARWIN'S work ( Coll. Works , ' III .
, p. 436 ) , leaves little room for doubt that result may be extended in the way I have stated .
If the satellite is not infinitesimal , the radius vector depends on the ratio of the masses , but always lies between the uarrow limits 2 .
457 and 2 514 times the radius of the primary ( see DARWIN , loc. cit. , p. 607 ) .
FIGURE OF EQUILIBRIUM OF A ROTATING MASS OF LIQUID .
27 .
We must now consider what motion is to expected a Jacobian ellipsoid which has reached the point of bifurcation at which instability sets in .
POINCARE remarks*that if the pear-shaped figure proved to be unstable , " " la mass fluide devrait le dissoudre par un cataclysme subit After reaching the poin in our diagram , the mass cannot move along the pear-shaped series , since this would involve a decrease of angular momentum .
It may be thought of as moving along the unstable branch OP ' of the series of Jacobian ellipsoids for an infinitesimal time until some slight disturbance brings its instability into play .
Now of all the vibrations of this figure , it is known that one only is unstable , namely that corresponding to the third zonal harmonic of the ellipsoid .
The initial motion of the figure must then be one in which the displacement at every point of the surface is proportional to the third zonal harmonic .
Thus the fluid begins by describing exactly the pear-shaped series , but as soon as the changes in angular momentum become appreciable , it leaves this series , and passes through a series of configurations represented in the region above in fig. 1 .
These may at first be thought of as lying parallel to the pear-shaped series , but above If there is a stable branch such as which ultimately passes above , it is conceivable that the series of non-equilibrium configurations might ultimately coalesce with the series of equilibrium configurations , and the motion would be continued along this series .
In this case , M. POINCARE'S " " cataclysme subit\ldquo ; would consist in a jump from the stable series PO to the stable series Judging from the sults of my parallel investigation on the configurations of rotating cylinders , this possibility does not seem at all likely .
Tt is , I think , much more probable that the pear-shaped series lies like the series OR in my figure .
In this case also the liquid would move through a series of configurations which would initially be close to the series OR , but would get continually futher removed from configurations of equilibrium .
) rotuberance resulting from the initial third harmonic displacement would develop in a manner somewhat similar to that of the aped figure , but as the motion would necessarily be possessed of a considerable amount of kinetic energy , the phenomenon would be a dynamical and not statical one .
If the co1lfiguration represented at in fig. 1 is the highest stable configuration possible for a single mass of liquid , this kinetic process can end in only one way , namely in the separation of the mass of liquid into two parts .
As the third harmonic displacement develops , the region of the pear which moves with greatest velocity is known to be the extreme end of the protuberance .
It is , therefore , natural to suppose that this part of the figure would be shot away first .
Moreover as the departure from a figure of equilibrium is probably pretty prollounced before the separation takes place , it is likely that the mass in question will be shot away with a considerable velocity .
* Letter to Sir G. in the 's ' Coll. Works , ' III .
, p. 31 VOL. CCXVII .
After this projectile has left the main mass , the angular momentum of the latter ( measured of course per unit mass ) will be reduced , and as the tidal influence of the newly-bol satellite is gradually withdrawn , the primary settle down to a state of stable equilibrium in which its figure is again that of a Jacobian ellipsoid .
cycle of processes such as this might very conceivably constitute the life-history of a .
body after once it had passed the critical state represented by the point of bifurcation on the Jacobian series .
There seems to be no reason why the protuberance should always develop at , and be shot off from , the same end of the Jacobian ellipsoid .
Thus it appears that the instability of the pear-shaped figure leads us to contemplate a series of processes of much the same nature as would have been expected if the pear had proved to be stable , except that we are now led to assign a much shorter time to these processes .
The problem has ceased to be one of statics and has become one of dynamics ; the phenomenon is no longer one of slow secular change but of collapse and explosion .
The mechanism of planetary birth which is now suggested is so rapid that there need be no difficulty in supposing a planet to have been through the cycle several times ; had the pear proved to be stable the cycle would probably have been so slow as to create a real difficulty .

