On the Method of Symmetric Products , and on Certain Circular Functions connected with that Method .
By the Rev. ROBERT HARLEY , F.R.A.S , Corresponding Member of the Literary and Philosophical Society of Manciester .
Communicated by A. CAYLEY , Esq. , F.R.S. Received October 18 , -Read December 13 , 1860 .
IN a paper printed in the second part of the fifteenth volume of the 'Manchester Memoirs , ' I have given a systematic exposition of Mr. COCKLE'S Method of Symmetric Products , and its application to the finite algebraic solution of the lower equations .
In that paper , to which I shall in future refer as my original memoir , I have also defined a new cyclical symbol , and I have by its aid succeeded in effecting the direct calculation of a certain sextic equation , on whose solution that of the general quintic may be made to depend .
In an Addendum I have pointed out the connexion between the circular functions which occur in my own researches and those to which we are led by the theory of LAGRANGE and VANDERMONDE , and , by means of the cyclical process , I have given a neat expression for the first coefficient of LAGRANGE'S reducing equation .
These researches I have followed up in an article " On the Theory of Quintics , " in the third volume of the ' Quarterly Journal of Pure and Applied Mathematics .
' My present purpose is not to repeat , but to endeavour to generalize and extend former results .
I shall therefore content myself with a very brief resume of my investigations , referring the reader for details to the above works .
Mr. COCKLE'S earlier researches on the subject were published in a series of five papers " On the Transformation of Algebraic Equations , " printed in the first and third volumes of ' The Mathematician* .
' SECTION I. The llilethod of ASymmetric Products , and a New Application of it to the Solution of the Lower Equations .
1 .
Any n symbols x , , x. , x3,. .
x* may be regarded as the roots of an equation of the form ( a , b , , ... ( X , l)n=a( x-xl)(x-ox)(x-X3 ) .
( )=Ot .
Let X1 , X3 , ... X. , be linear unsymmetric functions of x and of the form l+k1 2+k2x3+.* then , combining the conditions of symmetry , and rejecting those values of k ( k or indifferently ) which render X symmetric , we find k2+k+1-=O , that is , k is an unreal cube root of unity .
Represent this root by a , and put f(a ) =X1++ X2+2X3 ; then -f(a2)=-x , + a2+a X3 , and andr2(Xkf(OC ) , f(a2 ) ( :X)23x 2 nr(xf)=f(= ) , f( &lt; )= = ( Z)23x3Sx =2_(b23ac ) .
If , in evolving the roots by this method , we replace f(a ) by f , and r2(x ) by xr , there will result f=$l4:OG$2+:oc:$3 , f=x,1 + aX2 + X , ---=xi+ x2+ x3 ; 1a= f+ X2+ X3 , so that 17r f+ =a ( b+3ax ) , +7r ofy+a=a ( b+3ax2 ) , c , f+ c=f(b+3ax , ) ; and therefore f3+ = ( 1 ) --3b , 32ac , -33aa2db , 1)3 =(2b +9ae--27a2d ) whence , solving as for a quadratic in f3 , restoring the value of 'r , and extracting the cube root , we have fa = { -27ad+ 9abc--2b + 3as ( 27ad-18abed+ 4ac3+ 4b"d--b " ) and the roots of the complete cubic are included in the formula -3 ( 0f 4 mr a ) where m=l , 2 , or 3 .
In my original memoir I followed Mr. COCKLE , and in the application of the theory to the solution of the cubic and the biquadratic , I employed a subsidiary equation of the same degree .
This equation was obtained by eliminating x between the given one and y-2x)=o , 2z2 329 being rational and so constructed as to make r(y ) vanish .
It is true that the evanescence of t leads to an immediate solution , and that when y is known , x is also known .
But this evanescence is not essential to the theory ; and we are conducted to more significant results by dispensing with it .
4 .
For the quartic ( a , b , c , d , ex , 1)4-=0 , assume ( cyclically ) X1 = klx2+kx,3+kx4 , X2=xl +k2X2+k3X3+kX4 , X3= + k2+ kx 3+ kx , , and 3(Qx)=X X2 X3 .
Then , as before , combining the conditions of symmetry and rejecting those values of k which would render X symmetric , we are led to the cubic k+k-k1=-0 , of which the roots are -1 , 1 and -1 .
Let therefore Xl = x2+ -3 -X4 X2 = x , +x2-x3 ' ~-4X then r , ( X ) =X , X2Xa= ( )34Zx1x , + 8+1xx .
a=X ( + 4a c-8a2d ) .
5 .
The following solution is due to Mr. COCKLE , who communicated it to me in September of last year .
It may be considered as an extension of a solution of the complete cubic which I sent to him in January of the same year , and which , not essentially differing from the above , has a certain resemblance to that given by MURPiY in the 'Philosophical Transactions ' for 1837 .
Let Jf=xl-2+x-x,4 9 and f2=x 1+XZ-3-X44 ; then f , f--x--x2-x3--x and as in cubics , f , +/ .--/ ==^^4^ ) , -fil-3f2= !
( b + 4ax3 ) , a --/ , -+f2-f a-(b-4ax , ) ; rIYIi2 830 therefore ( f 2+f -f = ( 1 , 4b , 42ac , -4a2d , 44a3eXb , 1)4 -= ( 256a3e-64a2bd+16ab2c3b4 ) ; and f +f+ 2j =fa2(b_+2a6+2 x+4a2x2 ) =(8ac+ 3b2 ) ; consequently ( i+ ?
2+f2f)2(f2 +f 2 ?
f+4f l { ( 8ac3b2)2-(256a3e-64a2bd+16ab2c-3b4 ) } ; or 7T2 ~r2 1 f2f+f* +F f2+f2*2 f=a4(-64a3e+162abd ? ?
16a2c2-16ab2+ 3b4 ) .
P2 It hence appears that f2 , f2 and,2 are the roots of the cubic ( a4 , 8aec3a2b , -64a3e+16a2bd+16a2216ab2 c+3 b4 , aI2f , 1)3 0 .
When f is known , x is given by X1=4 ( +f2 + a ) ' and the corresponding formule for x2 , x , x4 may be readily obtained .
6 .
Next , for the quintic ( are , b , c , d , e , flx , 1)5=0 , assume Xi =xI$k+x2+ k3+--4+-k45a , X2 = x , + k2x2 + k4x +++ k35 , X3=x +k3X2+k1X3+k4X4+k2X .
, X4=X.+k4X2+--f 3+X3^jk2X4+fk1X5 , and r4(x)=X , X2X3X4 .
In regard to the above distribution of the constants k , , ki , k3 , k , it will be observed that they are arranged according to the following scheme : 123424133142 4321 331 32fEV R. I.:ABREY ON THE METHOD OF SYMMETRIC PRO MCTS That is , the four horizontal rows read downwards are identical in value and order with the four vertical columns read from left to right , while k , and k4 lie in inverse symmetry upon , and k2 and k3 around , diagonals .
Probably no other distribution would render ra(x ) more nearly symmetrical* .
Combining , as in former cases , the conditions of symmetry and rejecting incongruous results , we arrive at the quartic k4+k3+k2+k+1=0 , of which the roots are the unreal fifth roots of unity .
Let then , x2 , a ?
, and W4 denote these roots , and let f ( ' )=x,1+ X2 , + +2X , ++W3X4+45 ; then f ( a2 ) =X +IX2 + 4X+3 , + + w3X , f ( X3)=x , +'2+0 x+3 +X4+'x5 , f( 4)=x , + w4 X2+X3 +2+X4+w x5 , and Tr4(X )=f(C ) , f(W(2).f(w3).f(O4 ) = ( x)4 5 ( x ) 2x , x+ 5($x , x^)2+ 5'= -4 ( b5ab2c+ 5a2c2)+5r ' , where -=X$12+X2X 3+XI3X4+XX6 T5 T$1 and 7J=XIX +XX .
IX3Xy X4X I 1X5X2 These two functions , 7 and ' , are circular , and complementary to each other .
Since their product is unsymmetric relatively to x , it follows that 4q(x ) cannot in general be rendered symmetric .
Before proceeding to discuss this product , it will be convenient to introduce some other matters connected with the general theory .
SECTION II.-Circular Functions and the New Cyclical Symbol .
7 .
In the transformation and general treatment of the higher equations circular functions occupy a conspicuous place , and play an important part .
An attentive consideration of the structure of such functions will enable us to devise a calculus whereby operations upon them will be materially abridged .
The theory is far from being complete , and its practical application admits of great improvement .
In my original memoir I have proposed and applied a symbol which not only helps , I think , to throw some light on the general theory , but also enables us to effect with ease and rapidity calculations which would otherwise be very laborious , if not wholly impracticable .
The method there employed exhibits to the eye and to the mind the various combinations of dimen* When all the divisors of n are even , a cyclical arrangement must be adopted , as in the case for quartics ( art .
4 ) .
When n is prime or odd , an arrangement similar to the above , or a modification of it , will probably be found available .
; :e32 sions without rendering it necessary to examine all the results of multiplication , or to make a hazardous selectimo of those which may bedeemed material .
Combining as many terms ( n ) in one as there ae roots involved , the cyclical symbol reduces the labour of multiplication alone by nth ; and in passing from the circular to the corresponding symmetric function the process also affords considerable facilities .
8 .
Let X(O ) be a function of x , and let X(g ) be derived from %(0 ) by advancing each of the roots contained in its g steps in a given cycle r. Ex. yr .
Suppose that X(O)=-X X.21 and that we are following the first cycle , which , indeed , may be taken / as a type of all ; others , and is found in actual practice to be the most e( easy with which to operate .
Then , according to the definition , / x(1)==x.m , %(2 ) =X3X4 X(n1)=x , X(n ) =xx2=X(0 ) , Or again , if we suppose n-=5 , and %(0)= a'-(x +xX4 ) , $ then , following the same cycle , we have z(l)=x((x43 , +xax4 ) , y(3)=x42(x , 'x , ,+x )X ; '(4)S52lXX+X$2X3 ) : X(4 ) = ( xx4+x -X(a ) , ).(3 ) -=.X(X2a5-AX iX ) , z(4t)=x(xax +xa.xO , and so on .
Next , let / X(0 ) or , more simply , 'z(O ) * represent the ircular fuction x(0)+x(1)+ ... +X(-2)+z(-1 .
Then , since each root recurs at every nth step in the cycle , we have x(n ) =x(0 ) , X(f+l)=X(1 ) , and therefore 'X(l)= %(1)+X(2)+ ? .
+X(~-l ) +(- ) = { z(l)+X(2)+. .
+(n-l ) } +(o ) -x(l+X(n--l ) '(X(O ) Similarly , ~'x(2)=:g'x(l)=2'(o , and by generalization and induction , Hence THEOREM I.--A circular function is not affected in value by the simultaneous advancing or receding of the roots which it contains any number of steps in the cycle to which it belongs .
Whence it follows that 'Xi(0).'x2(O ) , or its equivalent , { xl(0)+xi(l)+.* +X1(n--2)+XI(nl ) }.'X2(O ) may be written XI( ?
) *:$XZ2( ) +X(l)x':X2(1l)+ ' ' +X1(a-2 ) .
1'X2(n-2)+xi(l( 1 ) .
x l ) which is equal to '{X , ( 0 ) .
X2(0 ) } ; and by simply interchanging Xi and X2 , we have also 'xl(0 ) .
'X2(0)-=$ ' { X2(o0 ) .
'Xi(0 ) } .
Which gives us THEOREM II.-The product of two circular functions belonging to the same cycle is itself a circular function to that cycle , and is given by the application of the cyclical symbol to the product of either function into the initial or leading terms of the other .
It should be remarked that : ' is a symbol of cyclical operation , and subject to the same laws , with certain obvious limitations as if it were a symbol of quantity .
Thus ; '{Xi(0)+X2( ) &amp; c 'X1()(0)+ &amp;c. }()+ ( )+ &amp; .
; and , in general , if we develope { Xl(0)+X2(0)+ &amp;c. } by the multinomial theorem , and then apply the symbol 2 ' to each term , the result will be equal to '({Xl(0)+X2(0)+ &amp;c. }m .
It should also be remarked that if C be a function or quantity such that it is not affected by the cyclical interchange of the roots , as when it is a constant quantity and therefore independent of the roots altogether , or as when it is a circular function and belongs to the same cycle as 2'x(0 ) , then will T'Cx(O)= ci'x(0 ) .
The foregoing theorems are true for any circular functions whatever , whether rational 334 or irrational , integral or fractional .
But the following holds only for those circular functions which are rational and integral .
THEOREM III .
If X be a circular function of the form : { ( XZ+Xb+Zc+ &amp;c. } , X being defined by Xm=aX'X l ... 4 a+a"X3 *.x. .
*x4+a " ' &amp;c. , where a , , y , ... I are positive integers or ( some of them ) zero , and m is the number of values of x ?
x ... xt or ( what is the same thing ) the number of terms contained in 'x021 , x..3 .
xt , then will x=nn ' aa +Z b+ c + &amp;c. ) , n ' being the number of values of X , and : EX being of the form ( a'+at"+a'"+ &amp;c. )Ex'x'x 2 ... x. For if , as we are permitted , we fix one of the roots , and permute the remaining n-1 roots in all possible ways , there will arise 1 .
2 .
3 ... ( n-1 ) , or ( say ) p , corresponding cycles ; and if , for the moment , we represent by 2 the sum of the p expressions formed by applying these cycles to any one of the values of X , then , since ' consists of n , and consequently 2 ' of np , expressions of the form Xa+Xb+%c+ &amp;c. , and since also nz=np , or a submultiple of np , it follows that X=+bP a b-pXb+CZP X+ &amp;c. But 2X=--2X .
Whence the theorem .
9 .
The symbol 2 ' admits of an easy extension to functions of the form fm( )=-Sm+ &lt; g^g2 + .
+ +-2L+1X g being an nth root of unity , real or imaginary .
For , if we represent this function by 2'5 , we shall have fv ( e).fp(={)=:( { mf ( rp( ) ) }= I{ apfFn(gq ) } or 2^r XM .
Y , XP='I'(XM2:n- , ^^)=/ ( X^^Y ) , et en &gt; tZ-nq k enqL e ))V , where 2 ' is the simple or ordinary cyclical function to the first cycle ... 123 ... ( , n-1)123. .
( n-1 ) ... This may be proved thus:fm(gq)=+ex+ q x'+ ... + -'2)x_ + ( " -l ) ' m= 2 ' , and fP(eg-q)=-Xlp+e+-qx2P+ ?
2n-2qp+ +(n-q)(na-,2)Xp +(r-q)(n-1)i ='{n_ P ; whence , multiplying , bearing in mind that g"n_l , and arranging the multiplicand as below , + aXP ( X71+ ga-X+ ^+ * ** q+v(n I+ q( ?
- ' ' 'en +$(4+g2+gr+. .
and jfi(cV2 ) .
f3m()= -2m+( 2M + 3 ) !
( +( +4 )W ' ; and therefore to ?
( ).f'( ) f(.f ?
) .
f()=( x2s)2~2 &amp; mE ?
;t -(x x~ ) +52 'n ~ ' which , when rn=l , coincides ( as it ought to do ) with the above result .
11 .
The expansion of I ' ^me2'x^fx may , by Theorems I. and II .
, be effected thus , M:4( X MX37 +1 43 5+ 12 344 +4 el X24+ 4 5 ) ^XWX -x(x +++ x1 xm +X2 &lt; ) 12 nmI Sm nz I2z : m 2t41Z mmm -=%X2(x-x2m +xx4 +x+x +x ; x ) +Ix7x~x ; ; or since 12m(mtmsmm7 m Lm mm m\.5_ .
2mme mn lXJ mm XmX &lt; m4+x2 5m mm i ^v2a 1g " t-t4 a/ " t3 g 11 2 _3 4 225 3Consequently fj"( ?
) * f( &lt; V ) .
f " ( ?
3).f(W4)=(SX2n)2_ _zx2mx 2m in ^t rn)2+5^ r ^x^-5^ ' 2$( rn ) ; which , making ml=1 , gives the resolvent product for quintics s(x ) = ( rx2)2X2lX2+ 5 2x4xx(Xx2 x i)2 5 x , xx4-5 x +xx4 ) ; or , expressing the symmetric portion of this value in terms of the coefficients of the quintic , and writing X for the unsymmetric portion Z'x2(xx-5+xax4 ) , we have ra(x)4 ( --15a3e 5ac2bd+ 5a2a-5ab2c+b4)5x .
I remark also that r'r =x x 3+1xxx 1 +xaxx-3(-3+e bd ) .
12 .
The number of unequal values of the several functions r , , rr ' , X and z4(x ) may be determined by the following considerations .
Since the coefficients of the several terms in the expression for r are equal , we may , in forming its values , regard one of the roots as fixed , while the others are permuted inter se .
Thus we shall have 1.2.3.4 cycles , giving rise to 24 corresponding expressions for r ; but since , for the first cycle , and similar relations obtain for the other cycles , therefore these 24 expressions may be grouped in pairs , the members of each pair being equal .
Hence r has only 12 values .
In the same way it may be shown that r ' has only 12 values .
And since the several values of r ' may be referred to the same cycles which arise in the formation of the values of r , and since also r and r ' are complementary to each other ( for their sum =:X1ix , a one-valued function ) , it follows that rr , and therefore also X and T4(x ) , are six-valued functions .
An independent proof that the resolvent product Tr(x ) is six-valued may be found in my original memoir , Section II. .
art .
15 .
Fixing two of the roots , sx arise , viz.:"\ ; I ?
i ' ?
* ; 1I| II / \ { tt4 H , * THE METHOD OF SYMMETRIC PRODUCTS .
and 2 , , and permuting the others , the following cycles 1 / \ 4 , f i\\ / \ a , .
__ ' su , , oa~~~~~L ' s , ,.\a / aiI/ a/ SI ' '-- ?
--' 's ' ?
... . .
, / \ I @__ __*_ Six values of r will then be included in the formula to 6 x12 , and the other values , which are in fact the corresponding values of ' , will be included in the formulae , !
, !
s , 5 1x and $4 , 6 14The values of k are as follows : X1-=1I1(2x+34 ) , x4-:4x1(.23+x45 ) , X2= ; 12(x24+x3)X,5= ; x , ( x4+x3x ) , ~3= i(X2X5+X3=4)::64 ( X2X3+X4X5 ) .
And by Theorem III .
, art .
9 , 5X='.6 ( 1+1)Wx , x3=2:x2x , a result which may be verified by actual development and summation .
By Theorem II .
Y =Z { X 2X25+ X,4)2(X2X5+X3X4 ) }= '[ ( x25 +x+X4 ) { + x2 5+3 ( 4)+( + x+4x5 ) -+ ( x4X2+ 5X , X ) +X ( xx53+ x12X ) + ( X1x4+ x } ; or , multiplying out and reducing by the aid of Theorem I. * , we have s= I{(X4 +2 ) +2(Xx3xx5 +xI3x4Xs5 ) +2(3X4+x24x+ xi + 34 ) } + 2xxx3x4x.x 2e+ 2 4$ , xx4x ; and , therefore , by Theorem III .
, O Xp = 2= '4h+ 12xx a lxx2xx5 +4 x lxox , c -+4px4 + 12vxluoxf5 x Operating in the same way upon X2 , we are led to the corresponding value of 3 , and thence to dx3 .
Repeating the operation upon X3 , we find 2^4 ; and so on .
But it may be remarked here that in dealing with the higher dimensioned functions , it will be found convenient to drop the subject x and to work only with exponents .
Thus , following the first cycle and omitting , for convenience , the unit-suffix , we have x=-'{(21001)+(20110 ) } =(21001)+(20110 ) +(12100)+(02011 ) +(01210)+(10201 ) +(00121)+(11020 ) +(10012)+(01102 ) , and , = : '[{(42002)+(40220 ) } +2{(33101)+(30311 ) } +2{(32210)+(32021)+(31202)+(30122 ) } ] +8:424+4X43321 +23223 +22 { ~414+:-2312 } .
And the passage to &gt; 3 is easily effected .
Carrying forward the calculation and collecting results , we have* =X 2212 , x-2= 2X422 + 12Z414 + 403212 4223221+1222312 , xS3= 2X632 + 2X62212 + 3X5421 + 6X5322 + 8X53212+ 6X4231 +12X42212 +12~4322 + 8S43212 + 6~43221 +48X424 +24X3321 +1223223 -+2{4414 + -2312}X , x4= 2842 + 4X83212 +12X824 + 4X7531 + 4X7432 +16X74221 +16X73221 +12X6222 + 4$6541 +20X65312 +24X65221 +12X6422 +20X64212+24X6432 +28X64321 +52X63222 -245242 +24X52412 +48X5232 +16X52321 +72X5223 +44X54221 Thus if , in the expression 2 's x.x a4xx , we suppose ( ex. gr. ) that , of the five exponents , y is the greatest , then this function must be replaced by its equivalent 'VA xxsx:4 ' .
Or , suppose that the greatest exponent ( y ) is repeated , and that the function takes the form S1'aA3-x1x4x ; then this must be replaced by wi.xo4exSx ; and so on .
Following this method , the comparison of similar functions is greatly facilitated .
+ 12082422 +120X8754 +486X87432 +200o 86222 +804486432 +488X85423 +552X72622 +480X72422 + 707X76232 +660X76423 +744X6351 +476662543 +168X544 + 6X7431 + 6X62322 + 46:6533 + 26:52422 + 4X63222 44X4331 + 192X82322 + 288X87531 + 348X8733 + 364 ; 86541 +1188X8531 +1920X844 + 264X7252 +1024:72432 + 332X76521 + 504X75'2 + 444X6342 +1092Z6243 + { 60X834 + 14X74232 -+ ++t +IC 6X65231 20X6432 14X52432 12:5223 1244322 + 20s87212 + 380X87522 + 372X8624 + 702X86532 + 684X85242 + 360X733 + 776X72541 + 264X7625 + 942X76542 + 936X75243 +1380X6332 + 444X6533 + 30X7223 + 8X7433 + 12~65222+12165421 + 32X64232 + 24X5433}3X + 6X54321 + 8X42322 +236X87621 +240X87421 + 304X86231 +646X86422 + 844X85232 +-400X72631 +565X72532 +564X76241 +740O7653_ +4207543 + 768X62522 +228X65242 + 6X75322 + 20X6312 + 10X65432 + 18X5341 + { 2X824 + 6X54322 + 4862422 +58X5332 +6X5332 +4X434})$ 2"+ { 8424 + 4X3321 + 2S3223}X)3 +2 { 414 + l2312 } ; 4 .
The bar placed over a number consisting of two digits indicates that it is a single exponent .
Ex. gr. '-1262represes represents 2X , l751 represents , and so on .
14 .
We may regard X as the root of a sextic equation ; and since the coefficients of any equation may be considered as known , when the sums of the powers of its roots are known , therefore the formulae in the last article may be considered as giving implicitly the coefficients of the sextic in k. The equation in X being once obtained , we may , by a transformation linear in ?
, deduce the sextic in 0 or in rr ' .
But the high dimensions of the symmetric functions in x present practical ( almost insuperable ) difficulties in passing to the corresponding functions of the coefficients of the quintic .
Mr. CAYLEY has suggested to me a more convenient method , which , however , I must leave its author to expound himself .
15 .
But if , in place of the complete quintic , we take any one of the trinomial forms S+ex +f=O , X5+d2+f=O , x5 +cX3+f= 0 x. + bx4+f=0 , 341 to which , by the method of Mr. JERRARD , Mr. COCKLE , or Professor SYLVESTER , the complete quintic may be reduced , then the foregoing formulae may be made available .
For , any symmetric function of the roots may be resolved into a function of the sums of the powers of the roots , and these sums , in any of the above cases , can of course be easily calculated .
Thus , taking the second form , x5+dx2+f=0 , which in many respects is the most convenient , we readily find 21=0 , 22=0 , 23=-3d , 24=0 , and therefore x-= 2 212= ( 1)22 -2U1 3--(:2)2+214=0 .
In like manner we obtain:x2=4.52df , 2x3=6d4 , X4= 4 .
54df2 , x5 = 158.5df+ 56f 4 =x6=6d8+4 .
56df 3 .
Therefore the equation in k is 6_2.52dft4-2d43+54d2f 2X2_(58d5+55f3)X +d8=0 ; and since , in this case ( art .
11 ) , the corresponding equation for the resolvent product is 62 .
54df0'+2 .
53d43+58d2f2J2+ 55(58d5+55f8)fo+56d8=0 , where I have written a for r4(x ) .
This equation was first given by Mr. COCKLE .
See his paper entitled " Researches in the Higher Algebra , " printed in the second part of the fifteenth volume of the 'Manchester Memoirs .
' Some interesting and curious transformations of the equation may be found in the same paper .
SECTION III .
The Symmetric Product for Quintics .
16 .
We know that , for the perfect form ( art .
11 ) , a4= 15a3e4-5a2bd+5a2c2-5abc+-b4-5a4X And by definition ( art .
1 ) , nr(x ) = 01MA040 .
It hence appears that the symmetric product II is of twenty-four dimensions with respect to x. The partitions of twenty-four , for the quintic , are as follows:342 &amp; S 17 .
In order to facilitate operations it will be convenient to replace the coefficie of the first term by unity , and to suppose the quintic to be deprived of its second tem that is , in effect , to deal with the equation under the form ( 1 , ,0 , c , d , e , / f , x , 1)0 .
Then , if we assume f( m)=x+ + xm 2X+* Xm 4mx4+ x , = 5/ ~5,3 where , as in former articles , w is an unreal fifth root of unity , the following relation will obtain , viz. Xr.+I+1~ )'f3 W3rP2= rP3 +++r being 0 , 1 , 2 , 3 , or 4 indifferently .
Developing and reducing by the known properties of c , we find x2^=100(P,134+P1323 ) , ~3_1 5 1251 3 P+ P. +P2 2 20 +03P4 ) IX=15(32133+23,1+4+1 ) , :x4= 30(P204+p 23)+20(/ 3p2+P3:i4+P3`3+Pi3P ) +120,13I,2/ SX5= 5p+00( +I 33oo( p3 +P3P3+P P2p + 33 2 ) + 150(P21P +24P243 +3P 1+13 1+2 ) .
And by the method of the limiting equation or otherwise , x=--2c , x3=--3d , 2x4=2c--4e , x-5=5cd--f .
Whence by comparison with the above , -c =5(1P34+P1 3 ) , -==5(pf3 fi2+P3+W+ 4 ) 35+ _ 0(o ( 2r+ + 33+3 +,4/ 33 ++ 13 ; /32P)-5 cd* .
-=5cp6-t-lo(j3^+^ , +p~p:3P4 +32I)3+ 402 P43 18 .
If now we suppose one of the constituents , say,34 to vanish , and eliminate the remaining ones , the effect will be the same as if we supposed the symmetric product , expressed as a function of the coefficients , to vanish .
For 54/ 31/ 23334-8 ; so ; that the evanescence of / 3 involves the evanescence of 0 , and consequently of II .
Making 34 vanish , the equations at the foot of the last article become -c = 5f343 , d= 5G3Y3 + 21 ) -e 5 ( P 2+ J3 ) -ec2 -f=i3 +/ 32+ / ? ?
22 3I3 c+ od .
And the elimination of P3 , 32 , P3 gives H=kcx +I c12 65 c8d f10 c7 f2+ 625c3d f3 16 cl'e + 560 c6de +125 c5e -3125cde +2 c9d2 85 C5d3 +150 c4d2 + 96 c8e -1340 c4de2 --500 c3ee 9 c7d2e + 510 c3d3e -750 ced2e 1 C6d4 70 c2d5 -125 cd4 -266 c6e3 + 875 c2de +625 ce3 19 cSd2e2 175 cd3e2 +625 d2e2 + 23 c4de + 50 5e +336 c4e4 250 de4 2 C3d6 58 c3d2e3 + 14 c2d4e2 -160 c2e5 7 cde + 35 cde4 +1 d8 10 d4e3 + 25 e6 In order to determine k ( a constant numerical factor , dropped in the course of calculation ) , let us take the particular equation x +c 3=0 , for which we have ( art .
11 ) 0= 5c2 and consequently HI= 56C2 .
Then since , in this case , the coefficients d , e , f severally vanish , the foregoing formula gives n=kc"2 ; and therefore =5 5 .
* The symmetric product for the quintic , deprived of its second term , x5-5PX3_ 52 -R +E=0 , was irst calculated by Mr. CocKILE .
See his paper " On Equations of the Fifth Degree , " published in the Appendix to the Lady 's and Gentleman 's Diary for 1858 .
In the second section of my original memoir I have verified his result by an independent calculation and supplied the constant numerical multiplier .
Mr. CocI : LE presented the product in the form of a function of P , Q , S and E , S being given by S=P+ R. Following the notation of my friend , I gave the product in the same form .
But the passage to the We have thus obtained the symmetric product for the quintic wanting in its second term ; but it seems desirable on many grounds to calculate the product for the perfect form .
A variety of methods of performing this calculation might be suggested .
The reversal of the problem of transformation appears , at first sight , the most easy and practicable , and if in the foregoing expression for II the following substitutions be made , viz.2b62 --5 +c for c , 4b3 3bc 52 +d ford , 3b6 3b2 2bd -52 " 5--+ e for e , 45 b3c ba d be 4 -b 6+cb2d- ?
e +f for f , 55-58+52-5Sffo , the result will be the symmetric product for the complete quintic ( 1 , b , c , , de , f , 1)5=0 .
But it will be found on trial that this process , though apparently simple , does in point of fact involve prodigious labour .
Mr. SAMUEL BILLS of Hawton , who kindly undertook to assist me in the calculation , communicated to me in the early part of the present year that portion of the expression into which f3 enters .
But the difficulties of the calculation and the want of means of verifying successive results have led him to abandon the work as impracticable .
An equally effective and a much more expeditious process is supplied by the following considerations .
20 .
It occurred to Mr. COCKLE that the symmetric product for the perfect quintic x -5 M5555 5Qx -5R+E= corresponding expression in P , Q , R and E is easily effected , and I find that the result ( as yet unpublished ) of the transformation is1= 514x + 625 P12 +325 P8Q E+2 P7 E 2+1 3Q 3 +2000 P0 ?
R +560 P6QR +5 P5R +1 PQRt 250 P9Q2 85 P5Q3 +6 P4Q2 +2400 P8R2 +268 P4QR2 +4 P3R2 225 7Q2 R -102 P3Q3R +6 P2QRt 25 P6W + 14 p2Q5 +1 PQ4 +-1330 p6R3 + 35 2QR3 +-1 PR3 + 95 P5Q2R2 7 PQR2 +1 Q2Rt2 115 P4Q4R +2 5QR + 336 P4R +2 QR4 + 10 P3Qs 58 P3Q2R3 + 14 p2Q4R2 + 32 P2R5 7 PQ6R 7 PQ2R4 +1 Qs +2 Q4R3 +1 RG 347 RE4:t EV .
HALEY ON THE METHOD OF SYMMETEIC PRODUCTS .
is given by the expression M2 M3 xr-D-.M +D2 .
r.D3r1.- &amp;c. , where X is the symmetric product for the imperfect form treated of in the foot-note under paragraph 18 , D is the differential symbol ddddd dzlx + dx2+ dX3 + dX4 + WX-5 and the relations DP= D(1xx,2)= -4 x= --4M , and others corresponding to them , hold .
Mr. COCKLE had further noticed that DTr is to be regarded as free from M , a condition substantially equivalent to the expunging of the portion 5b , + 4bbC of the operator V which will be presently considered .
Although he had not accurately completed this process of derivation , yet he had made a near approach to its completion , when Mr. CAYLEY , to whom as well as to myself Mr. COCKLE had communicated it , showed that the same results might be more ifmmediately and conveniently obtained by means of the quantical calculus , and in so doing he incidentally corrected an oversight which I had already pointed out to Mr. COCKLE .
In a letter under date September 28 , 1859 , Mr. CAYLEY called my attention to the circumstance that the several coefficients of the resolvent equation of the quintic are leading coefficients of a covariant .
Mr. COCKLE had previously suggested that the symmetric product H1 , or the last coefficient , was such a term .
The test that a function of the roots may be such a term is that it is reduced to zero by the operation a.+ -a2+ba3+b4+'a5 .
It is clear that this is the case with respect to each factor of the product = , f(0v).f(O2).6fl3).f(o3 4)= 54i1234 4 ' Therefore it is also the case with the product itself ; and since the like is true with respect to the other five values of 6 , it is also true with respect to any symmetrical function of the six values .
Consequently each coefficient of the sextic in 6 is the leading coefficient of a covariant .
At present , however , we have only to deal with the last coefficient , that is , the symmetric product .
21 .
11 is a seminvariant* reduced to zero by the operation PODUCaS 349 5-V=5a +4bIC +Sc+ 2de +:e ; and if we write nl=11no+lb+ n+ , b + ... .
+1n bS , Io , is known , being what 11 becomes when the quntic is deprived of its second term , and H , II2 , ... 1T may be found from it by means of the formule n , =(v'n+ , , n=--I(v'n , + cn , ) , r-5 0l I ( 7l:l4 ?
O wherI Assume and then and if we write then will so , in general , if we shall have =s ( v'TII , _4a+4TS_,2 V'-=V-(5^+4 1b , )=3Cae+2da+ea r. no=Ao++ Bof+ Cof2+D30f V= Vef= 3c3d+2d , ; --5n , =1-'n=V'A , +Be +f ( Bo+2Coe ) +f2(V"C , +3Doe ) +fS V"D , ; nT , =A+Bf+ Cf 2+D , f3 , --2 .
5n2 , =v ' , +4 , cfn , = VI"A , + B , e+4CeAo +f ( V/ " B +2C , e+4d , Bo ) +f2(V"C + 3D , e+ 4 , C0 ) +3V"D +4^ , Do ) ; n= A+B+ Btf C Df2 ++D,3 , -5s(t + l)n , + , -= v'rn+ 4a n , _ , = v"IAt+ Bte+4BcAt , +f ( V"Bt+2Cte+4aBt , , ) +f2(v"C , + 3Dte+ 4Ct-..,1 ) +f3(v"Dt+ 4baD , ) These formulae will enable us with comparative ease and great rapidity to derive the symmetric product for the complete quintic from that for the quintic wanting in its second term .
But it is noteworthy that the complete value of a seminvariant or critical function may always be deduced from the result obtained when anty coefficient ( excepting only a the first ) vanishes .
Thus , for the equation ( a , b , e , f ix , 1)5=0 , the operator is V=5abb +44bb+ 3ca , + 2de +eaf ; and if ex. gr. we write I=-I+Id+I2d+ ... and suppose Io given , all the other terms can be found .
For , in this case , let V'=7-V(3dced+2d 5e)= 5aa+4boc+eCbf ; then 'Io +d7'I+d 2VI2 ... VI= +3c(I1 +2dI2 + 32I3 + ... ) |=0 , t +2 ( d3eI+d21 eI +. .
) J and therefore 3cIl=--V'Io 6c= --VtI1-2b , Io , 9cl3=-V I2-2Ia , j , &amp;c. , which give I , , I,2 I3 , &amp;c. And it is easy to see how to obtain the corresponding formuluI for the other cases , viz. when the given function is free from either G , e , orf .
The extension of the process to equations of other degrees than the fifth presents no difficulty .
23 .
In applying the formulae in art .
21 , it will be convenient to omit the constant numerical factor 56 .
The calculation may be thus conducted : Terms in no. Ao Bo Co Do cT2 +1 2cd 6s5 cI0 cd + 625 cIe -16 ede + 560 c5e + 125 ede -3125 cd 2 c5d3 85 c4d2 + 150 ce2 + 96 cede2 1340 c3e2 500 c7d2e 9 c3d3e + 510 c2d2e 750 c6d4 1C 2d 70 cdC4 125 cY3 266 cde3 + 875 ce + 625 c5d"e219 cd3e 175 dae2 + 625 c4de + 23 d5e + 50 4e4 + 336 de4 50 c3d6 2 cSd2e358 cd4e 2+ 14 c2e5 -160 cde 7 cde4 +35 ds +1 d4e3 10 e+ 25 But a more complete verification is obtained by taking the sum of all the numerical coefficients that enter into the expression for 1 .
That sum is --4761104 , as it ought to be .
For the roots of ( 1 , 1 , 1 , 1 , 1 , 1x , 1)5= are --I , o5 --%5 a2 , '- ?
a , a being an unreal cube root of unity ; and for this particular form we have 0==11-10a , 0 , = 21+10m , 0=1 , d0= -29 , 04=4+20a , 06=-24-20a ; or 103=-331 , 8A=--29 , 06A=496 .
Consequently nI= 0 102030Ad=-4761104 .
26 .
Another convenient verification is afforded by writing a , 5b , 10c , 10d , 5e , f for a , b , c , d , e , f respectively , when the sum of the numerical coefficients of the several powers off should be zero* .
And the transformed result will be worth having for its own sake , as belonging to Mr. CAYLEY'S standard ( a , b , c , a , e , fXc , I)5 .

