]\gt ; .
I. Todbuntcr on Jacobi 's Theorem .
On Jacobi 's Theorem respecting the relative equilibrium of a Revolving Ellipsoid of Fluid , and on Ivory 's discussion of the Theorem By I. TODHUNTER , M.A. , F.R.S. late Fellow of St. John 's College , Cambridge .
Received November 23 , 1869* .
1 .
The late James Ivory contributed to the Philosophical Transactions various memoirs on the subject of the equilibrium of fluids and the figure of the earth : the memoirs will be found in the volumes for 1824 , 1831 , 1834 , and 1839 .
Ivory objected to the received theory of the equilibrium of fluids , and advocated some peculiar opinions at great length , and with much repetition .
I do not propose now to criticise these memoirs ; I will merely state that I consider them to be altogether unsatisfactory .
2 .
There is , however , one theorem in the general subject to which I now propose to draw attention , namely , Jacobi 's theorem respecting the possibility of the relative equilibrium of an ellipsoid of fluid having three unequal axes and revolving about the least .
Ivory discussed this theorem , and his errors are so numerous and so singular , that I have thought it would be desirable to place the rections before the Society which originally received and published Ivory 's communications .
In correcting Ivory 's errors and supplying his defects , I shall add something to the discussions which have hitherto been given of the theorem itself .
It will be seen as we proceed that one of Ivory 's errors has been already noticed and corrected .
3 .
Ivory first alluded to the matter in the memoir of 1834 , which was read to the Royal Society a few months before Jacobi announced his discovery of the theorem .
Of course at that date Ivory held the common opinion , that the relative equilibrium of a revolving ellipsoid with three unequal axes was impossible .
But he does not merely acquiesce in the erroneous opinion , he attempts to demonstrate it in the following manner:\mdash ; " " Further , the figure of the fluid in equilibrium can be no other than a spheroid of revolution .
Draw a plane through the axis of rotation and any point in the surface of the fluid .
This plane will contain that part of the attraction of the spheroid which is parallel to the axis of rotation , or to the coordinate : it will also the centrifugal force directed at right angles from the axis of rotation .
The same plane with also contain the resultant of the attractions parallet ?
and ; for if it did not , the resultant might be resolved into two forces , one contained in the plane , and the other perpendicular to it ; and the force perpendicular to the plane would pardy act in a direction touching the surface of the spheroid , which is inconsistent with the equilibrium of the fluid .
Wherefore , the whole attractive force at any point in the surface of the spheroid is contained in a plane passing through the point and the axis of rotation ; *Read Jan. 20 , 1870 .
See .
xviii .
Mr. I. hunter on Jacobi 's Theorem .
which obviously excludes ellipsoids with three unequal axes , and limits the figures of equilibrium to spheroids formed by the revolution of an ellipsis bout the axis of rotation ; . . .
The error here begins with the which I have put in italics ; the resultant of the attractions parallel to and ueed not act in the plane which Ivory specifies : the component which he obtains in a plane touching the surface may be balanced by a like component arising from the attraction parallel to and the the so-called ceutrifugal force .
4 .
To the Philosophical Transactions for 1838 Ivory contributed a memoir of ten pages on Jacobi 's theorem .
devotes a few sentences to the history of the matter .
He records the fact that Lagrange ] inferred that the figure of relative equilibrium must be a figure of revolution .
He makes no allusion , , to his own erroueous demonstratiou in the volume for 1834 .
5 .
The object of the memoir seems to be twofold\mdash ; to establish Jacobi 's the rem , and to deduce numerical results relating to the extreme possible cases analogous to those which had long been known relating to the extreme possible cases for an ellipsoid of revolution .
The first objecf is attained ; Jacobi 's theorem is demonstrated in a manner resembling that which had prcviously been used by Liouville .
The second object Ivory to attain , ving 1 an error in his process .
6 .
In the sccoud page of the memoir there is an error in mechanics resembliog that hich we have already noiiced in Art .
3 .
At any point in the surface of an ellipsoid , let the normal to the surface be ; and let it be terminated by the plane which is perpendicular to the axis of rotation : let be the length of this straight line .
At the same point in the surface a straight line in the direction of the resultant of the attraction of the whole mass of the ellipsoid , and let it be termlnated by the same principal plane ; let be the length of this straight line , then Ivory says : \mdash ; " " Let denote the lbird side of the has and for its other sides : then will represent the only force which , together with the attractive force , will produce a resultant in the direction of at right angles to the SUl.face of the ellipsoid This statement is quite wrong .
Any straight line which is in the same plane as the normal at any point , and the direction of the resuItant attraction at that point , may be taken for the direction of such an additional force Ivory requires ; and the magnitude of the force can then be properly determined .
7 .
In order to render the discussion of Ivory 's memoir readily intelligible , it will be necessary to indicate briefly the demonstration of Jacobi 's theorem .
Let the equation to an ellipsoid be Mr. I. Todhunter on Jacobi 's Theorem .
The attractions which the ellipsoid exerts on a point parallel to the axes of coordinates are known to be respectively , Cz , where JIere denotes the mass of the ellipsoid , and stands for : see the ' Mecanique Celeste , ' .
No. 7 .
Take the axis of for that of revolution , and let be the elocity .
Then the necessary and sufficient for relative equilibrium is that the equation should coincide with the differential equation to the surface of ellipsoid , namely , Hence we have the conditions ( 1 ) lf we eliminate between these we obtaiu .
But from the values of and we hnve Thus ( 2 ) becomes This may be satisfied by putting which gives us an ellipsoid of revolution .
Or it may be satisfied by making ; may be reduced to . . . .
( 3 ) have then to show that for suitable values of and this relation will hold .
This will be in the .
We must also show that ralue of is positive .
From ( 1 ) we have ' Ir .
I. Todhunter on Jacobi 's Theorem .
so that Put in the values of , and this reduces to so that is positive .
8 .
Before considering Jacobi 's theorem , we will advert to the case in which We have from ( 1 ) ; let be the density of the ellipsoid , then Put for ; thus . . . . . .
( 4 ) Here we have nged u into , and into , in order to have the same notation as Ivory has .
Ivory says : " " From the equation ( 4 ) we learn that will be known when is given , or that every spheroid of a determinate form requires an appropriate velocity of rotation .
" " The inspection of the same equation is sufficient to show that is positive for all values of as it vanishes both when is zero and infinitely great , it must pass at least once from sing to decreasing , it will admit of at le one maximum value .
By differentiating with to we obtain , . . . . .
( 5 ) from formula learn that is positive between the limits ; that it will consist of a positire and a negative part wheu is reatcr t ; and the part decreasing while the part increases , that it 1 ' ill ultimately bc negave lvhen is infinitely great .
It follows therefore that can be only once equal to zcro , and conscquently that can ve only one maximum , while ases f to This is quite unsound , because the ords u I have put in italics / untrue .
are two } of separating the into a ) flrt a negatire part .
We may take for the positive part the integral Mr. I. Todhunter on Jacobi 's Theorem .
between the limits and , and for the negative part the integral between the limits and 1 .
Or we may put the int in the form I do not know which of these two ways Ivory adopted .
It is true that in each of them the positive part decreases as increases ; but in each of them the negative part is finite when is finite , and is infinitesimal when is infinite , and so does not always increase with as Ivory supposes .
9 .
However , although Ivory 's reasoning is unsound , the resuIt which he lvishes to establish is correct , as I shall now show .
Consider the integral .
We have to show that as increases from unity to infinity , the integral vanishes and changes sign once , and only once .
It is obvious that the integral must vanish once , because it is positive when l , and legative when : it is indeed infinitesimal in the lntter case , but the sign is certainly negative .
Put for lx ; thus the integral becomes , say .
We have to show that the equation can have only one root between and .
lf the equation could have more than one root it must have three roots at least ; and then the equation must have two roots at least .
Now , say , and Thus is positive when , and while increases is always negative ; thus continually diminishes algebraically , so cannot change sign more than once .
Hence cannot vanish more than once ; it will vanish once because is positive when , and has a finite value when namely , , that 10 .
There is also another way in which the resuIt may be established .
For from ( 4 ) , by the known method of integration , have Mr. I. Todhunter on Jacobi 's Theorem .
; therefore Thus shes n .
( 6 ) Ivory gives these formulae with some misprints .
Now it has been showu by Laplace that the equation ( 6 ) has one positive root , and only one , namely , when .
See the ' Mecanique Ce'leste , ' Livre III .
No. 20 .
11 .
We now return to Jacobi 's theorem .
We take the integral in ( 3 ) , and put.1 ' for ?
; we put , and we may suppose greater than , so that is posive ; and we put .
Thus ( 3 ) becomes Denote the left-hand } ) of by ; then we propose to examine the range of values of and which make vanish , supposing both and positive .
If we regard as an abscissa , and as an ordinate corresponding to , we in effect propose to trace in the first quadrant the curve determined by the cquation 12 .
If we put , the equation ( 7 ) becomes .
( 8 ) Ivory says : " " It is obvious that there is only one value of that will verify the equation just found ; for the integral can pass only once from being positive to be ative while increases from 1 to be inimitely great I cannot admit that this assertion is obvious ; the result , however , may be established by the following investigatiou:\mdash ; Denote the integral by $s long as is less than unity , is positive , and when is infinite is negative .
Thus must vanish once as changes from unity to infinity : we have to that can anish only once .
If the equation could have more than one root , it must have three roots at least ; and then the equation must have two roots at least : this we shall show to be impossible .
We have ; .
I. Todhunter on Jacobi 's Theorem .
it is obvious then that cannot vanish until is greater than unity , that is , until is greater than 3 .
Thus is negative when ; and we can see that is positive when .
Hence changes sign once as increases from 3 to infinity .
Put , and let ; thus , say .
Then we know that is negative when , and positive when If , then , the equation could have more than one root , it must have three roots at least , and then the equation must have two roots at least .
Now Thus is always positive lvhen is greater than 3 , and therefore continually increases , and so cannot vanish more once .
Hence cannot vanish more than once , and therefore cannot vanish more than once .
Thus is always negative when has any value greater than that which makes vanish .
13 .
Therc is also another way in which the resuIt may be established .
It will be found that Then it may be that the last expression will vanish once , and only once , as changes from 1 .
This method is adopted by Liouville in an article in his 'Journal de Mathe'matiques ' for April 1839 : the article consists of observations on the memoir by Ivory , which we are discussing .
Liouville says that the value of , which makes the Iast expression vanish , is a little less than 2 .
Ivory has a formula which is equivalenC to this , but he does not employ it to show that there is ouly one value of ; for he had already , as we have seen , stated this to be obvious .
From the circumstance that Liouville gives a slrict demonstration , it is plain that ] rees with me in thinking that Ivor.v 's slatement is not obvious .
to Ivory , the of which satisfies ( 8 ) is We will detlote this by 14 .
I now that if we ibe to value than a corresponding value of exists , which will make vanish . .
I. Todhunter on Jacobi 's Theorem .
Let such a value be asclibed to ; then from Art .
12 it follows that , with this value of and with , the value of will be negative : we shall now show that by taking large will be positive , thus the sigtl of is the same as that of where stands for When is made small , the term may be neglected in comparison with so the sign of the integral become the same as that of The first term is infinite when , and the second term is finite ; thus the sign of the iu tegral is positive when is small enough .
Since is negative when , and is positive when is large enough , it must vanish for some intermediate of 15 .
Moreover , when is very small , we have approximately Thns to make ish when is very , we have approximatcly ' therefore so that approximately .
shadl ncxt ding to given vnlue of is one ?
of ill make V vanish .
; then Thus we have ; Mr. I. .
on Jacobi 's Theorem .
and therefore Hence .
( 9 ) The right-hand member is necessarily positive .
If we put , we may wrile the equation thus , This shows that if be kept constant , continually increases with and therefore cannot vanish more than once , and therefore of course V cannot vanish more than once .
17 .
Ivory makes the followiog statement:\mdash ; Let V stand for the integral in the equation ( 7 ) ; and supposing that and vary so as always to satisfy that equation , we shall have Now , representing any positive quantity , we may conceive it to increase from zero to be infinitely great ; in which case it follows from the nature of the function , that during the whole increase will be negative : wherefore the other term will be positive ; which requires that decrease continually It is here in fact asserted that is negative for such values of and as make vanish ; this is , however , wrong , as equation ( 9 ) shows that is positive when V. The mistake was pointed out by Liouville in the memoir already cited , and the correction was accepted by Ivory in the Philosophical fransactions for 1839 , pages 265 , 266 .
The mistake vitiates the remainder of Ivory 's memoir .
The investigation of Art .
16 is taken in substance from Liouville 's memoir ; he also demonstrates the proposition of Art .
14 , but not in the way which I have adopted .
18 .
The extract given in the preceding article from Ivory 's memo involves another error , which Liouville does not notice : the words " " which requires that decrease continually contain an arbitrary unproved assertion .
We have ; hence if were negative , would be positive ; then might be Mr. I. Todhunter on Jacobi 's Theorem .
positive or it might be negative : we cannot assert at once , as Ivory does , that must be negative .
Suppose , for example , thatV slood ; then would be negative for positive values of , and would continually increase with 19 .
As I already stated , Ivory accepted correction made by Liouville ; but in the two in which the mist is other assertions advauced .
It in effect uecessary for Ivory 's purpose to trace the curve determined by in the lirst quadrant ; but instead of demonstration such as we have supplied in Art .
14 , Ivory gives unwarranted assertions of the kind already noticed in Art .
18 .
We have from which it follows that , whatever positive number stands for , is negative for all values of that make positive , that is , for values of less than 3 .
This lvory gives , and so far he is correct ; is certainly negative , and uot zel.o , so its is less than 3 .
Ivory vishes to show that can nevcr be zero .
IIe lakes the diffelential equation ; he says , " " Further , in the differential equation cannot be zero ; because , increasing without limit , is essentially positive This is quite unsound .
Regard as abscissa and as the corresponding ordinate of a curve determined by ; then that always positive , yet may anish ; that is , there may be point or points on the curve at which the tangcnt is parallel to the axis of ) scissae .
Suppose , for example , that stood for ; then is always ) ositive , is when is less than , $)nd when or I do uot assert that can anish in hc case ; I only intai that Ivory 's lment to } -ound .
I .
I. Todhunter on Jacobi 's Theorem .
presently demonstrate that cannot vanish .
20 .
Ivor .
concludes the two pages in the Philosophicnl Transactions for 1839 thus : " " If the sign of be changed , the result will be positive ; and hence .
; that is contained between and 1 , we obtain a condition between aoy two values of and that satisfy the equation namely , the expression must be a positive qnantity , or , which is the same thing .
This is obscure in its commencement ; but the statement to which it proceeds is intelligible , but is unwarranted .
For , granting that is always to be , this will be secut'ed if is always positive ; that is , if is positive ; that is , if is always positive .
And as lies between and this condition is certainly secured if is greater than According to Ivory 's statement it vould be necessary that should be greater than ; whereas we see that it would be sufficient that should be greater than Ivory 's second statement is inconsistent with his } ) ecomes consistent with if we change to ; if , however , his secoud statement is to be taken as what he intended , and his first statement corrected to agree with it , his error is aggravated .
In fact , however , I doubt whether any such necessarrJ criterion as Ivory proposes can be easily deduced from the value of .
For , granting th is always negave ttnd never zero , it will not follow that every meant in the which expresses must be negative , but only that the Mr. I. Todhunter on Jacobi 's Theorcm .
aggregate of the positive elements , if such there be , shouId fall s of tlJe ggregate of the negative elements .
However , be this as it may , can be no doubt that the specific criterion which I proposes is quitc unsupported by demonstration .
When and very large , the relation veen them is approximately that given in 21 .
In Arts .
14 and ] 6 it is that for every value of eater than , there is one , and only one , value of hich will make ranish .
We shall now corresponding to every alue of there is one , and only one , value which will make Whatever be the value of , it is obvious that V is } ) ositive , and when is large enough : thus there must be some intermediate vfdue ofp which makes anish .
We have to show that there is only one such value ; to show this we shall demonstrate the proposition which Ivory asserted on insufficicnt grounds , namely , that can never anish simultaneously with V. ; and we know that cannot vanish so long as is less than 3 .
See Art .
19 .
We shall first show that the limit may be changed to a larger limit .
If the integral is positive , the integral will also be positive wheu is a divisor under he integral sign ; for by introducing this denominator diminish all the elements of the integral , but diminish evcry element in a higher degree than any positive element .
Now the value of the last integral is we are certain that this is positive if is not less than that is , if is not less than On trial it will be found that this condition is satisfied , even when is zero , providedp be not greater than .
Thus have only to cousidc } Mr. I. Todhunter on Jacobi 's Theorem .
the case in which both is greater than , and is less than ; and we have to show that is negative in this case .
Put ; thus becomes a function of and ; and we have to show that cannot vanish 1vhen lies between and , and is less than , so that is less than .
We have wbcre stands for ; therefore Separate the integral into two parts , one extending between the limits and , and the other between the limits and .
It is obvious that the second part is positive , so that we have only the first part to examine ; this is and we shall show that this is positive .
Let stand for ; then we shall first show that is positive ; and next that is positive .
Now will certainly be positive if is .
The last integral is ' Mr. I. Todhunter on Jacobi 's Theorem .
that is , and this is positive , for does not exceed Now the coefficient of in is positive , since is less than thus cannot change sign more than once ; it changes sign once , for it is negative when .
, and positive when .
The factor will be found to increase continually as increases from to .
Hence in passing from , a larger fraction is taken of every positive element than of any negative element ; so that since the former integral has been shown to be positive , the latter is positive .
To demonstrate that continually increases as increases from to take the differential coefficient with respect to ; the sign of this differenti } coefficient is the same as that of that is , of , that is , of ; this is when ; when it becomes ) , which we shall now show to be positive .
We have by supposition less and ; and as does not exceed , this is greater than and so is positive .
Hence we that when , the value of necessarily positive , and cannot be zero .
22 .
We have shown in the preceding Article , that cannot vanish simultaneously with V ; the demonstration is rather complex , and perhaps for this reason Ivory attempted to establish the proposition by unsound general reasoning .
Liouville does not give the proposition , although it is naturally required to make the discussion of the admissible lues of and complete .
23 .
Thus we may state the results of Arts .
14 , 16 , and 21 in the folMr .
J. M. Heppel on the Theory of Continuous Beams .
lowing manner : the curve determined by the equation be traced in the first quadranl , every straight line parallel to the axis of meets the once , and only once , and every straight line parallel to the axis of and at a greater distance from it than , meets the curve once , and only once .
vanishes with , the meets the axis ofp at right angles ; and from Art .
15 it follows that when and are indefinitely great , the angle which the tangent to the curve makes with the axis of is very nearly a right aIlgle .
Thus , for sm values of the curve is concave to the axis of , and for vely large values of the curve is to the axis ofp ; so that the curve must have a point or points of inflexion .
24 .
In the very careful account of Ivory 's mathematical researches , which is given in the fourth volume of the ' Abstracts of the Papers . . .
of the Royal Society , ' it is said , with respect to 's theorem , " " In a paper in the Transactions for 1838 , Mr. Ivory has with great elegance demonstrated this .
theorem , and has given , with greater detail than its authors had entered on , several statements regarding the limitations of the proportions of the axes The language is cautious , but seems to imply some suspicion with regard to the accuracy of the statements .
As we have now seen , many of Ivory 's statements are inaccurate , and though accurate , are based on unsound reasoning .
On the Theory of Continuous l ; cams.\ldquo ; By JOHN IIORTIMER HEPPEL , M. Inst. C.E. Communicated by Prof. W. J. MACQU0RN RANKINE .
Received December 9 , 1869* .
In venturing to present to the Royal Society a paper on a subjcct which has engaged the attention , more especially in France , of some of the most eminent engineers and writers on Iical P , the feels it to be incumbent on him to state the nature of the claim to ttention which he hopes it may be found to possess in poiot of originality or improvement on the method of To do this clenrly , however , it will be ecessary to advcrt to the principal steps by which progress in the vledge of this subject been made , both in France and in this country .
The theory of continuous beams appears to hare first attracted attention in rnnce about 1825 , when a method of determining all the conditions of eqnilibrium of a straight beam of uniform section throughout , on any number of level supports at any distances apart , each span being loaded uniformly , but the iform loads itrying in any nner from to another , was investigated and published by M. Navier .
This method , although perfectly exact for the assumed objectionuble from the great 1 bour and of the ] aticns it tlUltry 2 See .
p. 1

