An Eighth Memoir on Quantics .
By A. CAYLEY , F.B.S. Received January 8 , -Read January 17 , 1867 .
TTHE present Memoir relates mainly to the binary quintic , continuing the investigations in relation to this form contained in my Second , Third , and Fifth Memoirs on Quantics ; the investigations which it contains in relation to a quantic of any order are given with a view to their application to the quintic .
All the invariants of a binary quintic ( viz. those of the degrees 4 , 8 , 12 , and 18 ) are given in the Memoirs above referred to , and also the covariants up to the degree 5 ; it was interesting to proceed one step further , viz. to the covarialits of the degree 6 ; in fact , while for the degree 5 we obtain 3 covariants and a single syzygy , for the degree 6 we obtain only 2 covariants , but as many as 7 syzygies ; one of these is , however , the syzygy of the degree 5 multiplied into the quintic itself , so that , excluding this derived syzygy , there remain ( 7 1= ) 6 syzygies of the degree 6 .
The determination of the two covariants ( Tables 83 and 84 post ) and of the syzygies of the degree 6 , occupies the commencement of the present Memoir .
The remainder of the Memoir is in a great measure a reproduction ( with various additions and developments ) of researches contained in Professor SYLVESTER'S Trilogy , and in a recent memoir by M. HERMIT * .
In particular , I establish in a more general form ( defining for that purpose the functions which I call " Auxiliars " ) the theory which is the basis of Professor SYLVESTER'S criteria for the reality of the roots of a quintic equation , or , say , the theory of the determination of the character of an equation of any order .
By way of illustration , I first apply this to the quartic equation ; and I then apply it to the quintic equation , following Professor SYLVESTER'S track , but so as to dispense altogether with his amphigenous surface , making the investigation to depend solely on the discussion of the bicorn curve , which is a principal section of this surface .
I explain the new form which M. HERMIT has given to the Tschirnhausen transformation , leading to a transformed equation the coefficients whereof are all invariants ; and , in the case of the quintic , I identify with my Tables his cubicovariants p , ( x y ) and p2(x , y ) .
And in the two new Tables , 85 and 86 , I give the leading coefficients of the other two cubicovariants p , ( x , y ) and p4(x , y ) .
In the transformed equation the second term ( or that in z4 ) vanishes , and the coefficient 2 of z3 is obtained as a quadric function of four indeterminates .
The discussion of this form led to criteria for the character of a quintic equation , expressed like those of Professor SYLVESTER in terms of invariants , but of a different and less simple form ; two such sets of criteria are obtained , and the identification of these , and of a third set resulting from a separate investigation , with the criteria of Professor SYLVESTER , is a point made out in the present memoir .
The theory is also given of the canonical form which is the mechanism by which M. HERMITE'S investigations were carried on .
The Memoir contains other investigations and formulae in relation to the binary quintic ; and as part of the foregoing theory of the determination of the character of an equation , I was led to consider the question of the imaginary linear transformations which give rise to a real equation : this is discussed in the concluding articles of the memoir , and in an Annex I have given a somewhat singular analytical theorem arising thereout .
The paragraphs and Tables are numbered consecutively with those of my former Memoirs on Quantics .
I notice that in the Second Memoir , p. 126 , we should have No. 26 =(No .
19)2_-128 ( No. 25 ) , viz. the coefficient of the last term is 128 instead of 1152 .
Article Nos. 251 to 254.-The Binary Qu'intic , Covariants and Syzygies of the degree 6 .
251 .
The number of asyzygetic covariants of any degree is obtained as in my Second Memoir on Quantics , Philosophical Transactions , t. 146 ( 1856 ) , pp. 101-126 , viz. by developing the function 1 _1Z ) ( I _XZ ) ( I _X2Z ) ( 1-LIZ ) X1-4Z ) ( -2z as shown p. 114 , and then subtracting from each coefficient that which immediately precedes it ; or , what is the same thing , by developing the function 1-X ( IZ ) X1Zz ) ( I X2Z ) ( I 3Z ) ( I 4Z ) X1-5Z ) which would lead directly to the second of the two Tables which are there given ; the Table is there calculated only up to Z5 , but I have since continued it up to z'8 , so as to show the number of the asyzygetic covariants of every order in the variables up to the degree 18 in the coefficients , being the degree of the skew invariant , the highest of the irreducible invariants of the quintic .
The Table is , for greater convenience , arranged in a different form , as follows : For the explanation of this I remark that the Table No. 81 shows that we have for the degree 0 and order 0 one covariant ; this is the absolute constant unity ; for the degree 1 and order 5 , 1 covariant , this is the quintic itself , being the Table No. 13 of my Second Memoir ; for degree 2 and order 10 , 1 covariant ; this is the square of the quintic , ( 13)2 ; for same degree and order 6 , 1 covariant , which had accordingly to be calculated , viz. this is the Table No. 15 ; and similarly whenever the Table No. 81 indicates the existence of a covariant of any degree and order , and there does not exist a product of the covariants previously calculated , having the proper degree and order , then in each such case ( shown in the last preceding Table by the letter N ) a new covariant had to be calculated .
On coming to degree 5 , order 11 , it appears that the number of asyzygetic invariants is only =2 , whereas there exist of the right degree and order the 3 combinations ( 13)(21 ) , ( 14)(18 ) , and ( 15)(17 ) ; there is here a syzygy , or linear relation between the combinations in question ; which syzygy had to be calculated , and was found to be as shown , ( 13)(21)+(14)(18)-(15)(17)=0 , a result given in the Second Memoir , p. 126 .
Any such case is indicated by the letter S. At the place degree 6 , order 16 , we find a syzygy between the combinations ( 13)2(21 ) , ( 13)(14)(18 ) , ( 13)(15)(1 7 ) ; as each term contains the factor ( 13 ) , this is only the last-mentioned syzygy multiplied by ( 13 ) , not a new syzygy , and I have written S ' instead of S. The places degree 6 , orders 18 , 14 , 12 , 10 , 8 , 6 each of them indicate a syzygy , which syzygies , as being of the degree 6 , were not given in the Second Memoir , and theywere first calculated for the present Memoir .
It is to be noticed that in some cases the combinations which might have entered into the syzygy do not all of them do so ; thus degree 6 , order 14 , the syzygy is between the four combinations ( 13)(15)(16 ) , ( 17)(18 ) , ( 14)(15)2 , and ( 13)2(20 ) , and does not contain the remaining combination ( 13)2(14)2 .
The places degree 6 , orders 4 , 2 , indicate each of them a new covariant , and these , as being of the degree 6 , were not given in the Second Memoir , but had to be calculated for the present Memoir .
254 .
I notice the following results : Quadrinvt .
( 6 No. 20 ) 3(19)2 , Cubinvt .
( 6 No. 20 ) = -(19)-3+54(19)(25 ) , Disct .
( as No. 14+g ] No.83)= ( -(19 ) , ( 25 ) , --3(29)3toa , 0 ) ' Jac. ( No. 14 , No. 20 ) 6(84 ) , Hess .
( 3 No. 16 ) ( 83 ) , the last two of which indicate the formation of the covariants given in the new Tables Nos. 83 and 84 : viz. if to avoid fractions we take 3 times the Table No. 16 , being a cubic ( a , ... )3(I , y)3 , then the Hessian thereof is a covariant ( a , ... )6(x , y)2 , which is given in Table No. 83 ; and in like manner if we form the Jacobian of the Tables Nos. 14 and 20 , which are respectively of the forms ( a,..)2(x , y)2 and ( a,..)6(x , y)45 this is a covariant ( a,..)6(x , y ) ' , and dividing it by 6 to obtain the coefficients in their lowest terms , we have the new Table , No. 84 .
I have in these , for greater distinctness , written the numerical coefficients after instead of before , the literal terms to which they belo~ng .
or , as this may be written , 1 , 2a6 , ?
2 0 .
Hence writing x+ &amp; c ' for x , the last-mentioned equation is the condition in order that the equation X X~~~~~~ ( -M)(X-A)(-7)(x -)(XS )=may be transformable into X(X3')(x7')(X8)(x')-= 0 , where '+r'=0 , b ' -A'_ 0 , that is , into the form x(x'2 P2)(X2I 82)= 0 .
Or replacing y , if we have ( a , b , c , d , el f Asx , y ) = a(X ay ) ( xPy ) ( x -yy)(x y)(x y ) , then the equation in question is , the condition in order that this may be transformable into the form ( a ' , 0 , c ' , 0 , e ' , 0Xx , y)5 , that is , in order that the 18-thic invariant I may vanish .
Hence observing that there are 15 determinants of the form in question , and that any root , for instance a , enters as a ' in 3 of them and in th6 simple power o in the remaining 12 , we see that the product a181 1 , 2e a2 1 , 13+y , mr 1 , P+s , ap contains each root in the power 18 , and is consequently a rational and integral function of the coefficients of the degree 18 , viz. save as to a numerical factor it is equal to the invariant I. And considering the equation ( a,..Xx , y)'-= 0 as representing a range of points , the signification of the equation I=0 is that , the pairs ( j3 , y ) and ( 8 , 6 ) being properly selected , the fifth point ax is a focus or sibiconjugate point of the involution formed by the pairs ( f3 , y ) and ( I , s ) .
Article Nos. 257 to 267.-Theory1 of the determination of the Character of an Equation ; Auxiliars ; Facultative and Non-facultative space .
257 .
The equation ( a , b , c ... .x , y)fl= O is a real equation if the ratios a : b : c,. .
of the coefficients are all real .
In considering a given real equation , there is no loss of generality in considering the coefficients ( a , b , c. .
) as being themselves real , or in taking the coefficient a to be =1 ; and it is also for the most part convenient to write y= 1 , and thus to consider the equation under the form ( 1 , b , C..)~X , 1)n=0 .
It will therefore ( unless the contrary is expressed ) be throughout assumed that the coefficients ( including the coefficient a when it is not put =1 ) are all of them real ; and , in speaking of any functions of the coefficients , it is assumed that these are rational and integral real functions , and that any values attributed to these functions are also real .
MDcCcLxYII .
4 B The equation ( 1 , b , c. .
j(x , 1)n:=-O , with a real roots and 23 imaginary roots , is .
said to have the character oar+2i3i ; thus a quintic equation will have the character 5r , 3r+2i , or r+4i , according as its roots are all real , or as it has a single pair , or two pairs , of imaginary roots .
259 .
Consider any mn functions ( A , B , ... K ) of the coefficients , ( mn= or &lt; n ) .
For given values of ( A , B , ... K ) , non constat that there is any corresponding equation ( that is , the corresponding values of the coefficients ( b , c , ... ) may be of necessity imaginary ) , but attending only to those values of ( A , B , ... K ) which have a corresponding equation or corresponding equations , let it be assumed that the equations which correspond to a given set of values of ( A , B , ... K ) have a determinate character ( one and the same for all such equations ) : this assumption is of course a condition imposed , on the form of the functions ( A , B , .
K ) ; and any functions satisfying the condition are said to be " auxiliars .
" It may be remarked that the n coefficients ( b , c , ... ) are themselves auxiliars ; in fact for given values of the coefficients there is only a single equation , which equation has of course a determinate character .
To fix the ideas we may consider the auxiliars ( A , B , ... K ) as the coordinates of a point in m-dimensional space , or say in r-space .
260 .
Any given point in the r-space is either " facultative , " that is , we have corresponding thereto an equation or equations ( and if more than one equation then by what precedes these equations have all of them the same character ) , or else it is " non-facultative , " that is , the point has no corresponding equation .
261 .
The entire system of facultative points forms a region or regions , and the entire system of non-facultative points a region or regions ; and the r-space is thus divided into facultative and non-facultative regions .
The surface which divides the facultative and non-facultative regions may be spoken of simply as the bounding surface , whether the same be analytically a single surface , or consist of portions of more than one surface .
262 .
Consider the discriminant D , and to fix the Ideas let the sign be determined in such wise that D is + or according as the number of imaginary roots is 0 ( mod .
4 ) , or is 2 ( mod .
4 ) ; then expressing the equation D= 0 in terms of the auxiliars ( A , B , ... K ) , we have a surface , say the discriminatrix , dividing the rn-space into regions for which D is + , and for which D is - , or , say , into positive and negative regions .
263 .
A given facultative or non-facultative region may be wholly positive or wholly negative , or it may be intersected by the discriminatrix and thus divided into positive and negative regions .
Hence taking account of the division by the discriminatrix , but attending only to the facultative regions , we have positive facultative regions and negative facultative regions .
Now using the simple term region to denote indifferently a positive facultative region or a negative facultative region , it appears from the very notion of a region as above explained that we may pass from any point in a given region to any other point in the same region without traversing either the bounding surface or the discriminatrix ; and it follows that the equations which correspond to the several *528 points of the same region have each of them one and the same character ; that is , to a given region there correspond equations of a given character .
264 .
It is proper to remark that there may very well be two or more regions which have corresponding to them equations with the same character ; any such regions may be associated together and considered as forming a kingdom ; the number of kingdoms is of course equal to the number of characters , viz. it is =-(n+2 ) or 1(n+1 ) according as n is even or odd ; and this being so , the general conclusion from the preceding considerations is that the whole of facultative space will be divided into kingdoms , such that to a given kingdom there correspond equations having a given character ; and conversely , that the equations with a given character correspond to -a given kingdom .
Hence ( the characters for the several kingdoms being ascertained ) knowing in what kingdom is situate a point ( A , B ... K ) , we know also the character of the corresponding equations .
265 .
Any conditions which determine in what kingdom is situate the point ( A , B , ... K ) which belongs~to a given equation ( 1 , b , c ... )Cx , 1 )f= 0 , determine therefore the character of the equation .
It is very important to notice that the form of these conditions is to a certain extent indeterminate ; for if to a given kingdom we attach any portion or portions of non-facultative space , then any condition or conditions which confine the point ( A , B , ... K ) to the resulting aggregate portion of space , in effect confine it to the kingdom in question ; for of the points within the aggregate portion of space it is only those within the kingdom which have corresponding to them an equation , and therefore if the coefficients ( b , c , ... ) of the given equation are such as to give to the auxiliars ( A , B , ... K ) values which correspond to a point situate within the above-mentioned aggregate portion of space , such point will of necessity be within the kingdom .
266 .
In the case where the auxiliars are the coefficients ( b , c , ... ) , to any given values of the auxiliars there corresponds an equation , that is , all space is facultative space .
And the division into regions or kingdoms is effected by means of the discriminatrix , or surface D=O , alone .
Thus in the case of the quadric equation ( 1 , x , y31d , 1)2=O the m-space is the plane .
We have D= 2_ y , and the discriminatrix is thus the parabola x2 y=O .
There are two kingdoms , each consisting of a single region , viz. the positive kingdom or region ( _2-y= + ) outside the parabola , and the negative kingdom or region ( x2-y= - ) inside the parabola , which have the characters 2r and 2t , or correspond to the cases of two real roots and two imaginary roots , respectively .
And the like as regards the cubic ( 1 , x , ) , so , 1)3=O ; the m-space is here ordinary space , D -4X3z + 3Xiy2 + 6xyz 4y3 -z2 , and the division into kingdoms is effected by means of the surface D 0 ; but as in this case there are only the two characters 3r and r+2i , there can be-only the two kingdoms D+ and D =having these characters 3r and r+2i respectively , and the determination of the character of the cubic equation is thus , effected without its being necessary to proceed further , or inquire as to the form or number of the regions determined by the surface D=O : I believe that there are only two regions , so that in this case also each kingdom consists of a single region .
But pro ceeding in the same manner , that is , with the coefficients themselves as auxiliars , to the case of a quartic equation , the r-space is here a 4-dimensional space , so that we cannot by an actual geometrical discussion show how the 4-space is by the discriminatrix or hypersurface D=O divided into kingdoms having the characters 4r , 2r+2i , 4i respectively .
The employment therefore of the coefficients themselves as auxiliars , although theoretically applicable to an equation of any order whatever , can in practice be applied only to the cases for which a geometrical illustration is in fact unnecessary .
267 .
I will consider in a different manner the case of the quartic , chiefly as an instance of the actual employment of a surface in the discussion of the character of an equation ; for in the case of a quintic the auxiliars are in the sequel selected in such manner that the surface breaks up into a plane and cylinder , and the discussion is in fact almost independent of the surface , being conducted by means of the curve ( Professor SYLVESTER'S Bicorn ) which is the intersection of the plane and cylinder .
Article Nos. 268-273.-Application to the Quartic equation .
268 .
Considering then the quartic equation ( a , b , c , d , eXo , 1 ) 4O ( I retain for symmetry the coefficient a , but suppose it to be =1 , or at all events positive ) , then if I , J signify as usual , and if for a moment 2=a2d -3abc+2063 , X= 3aJ+2(b2ac)I , we have identically 9 ( 3a2J2+X2)a2 = 9(b2_ ac)'X2-a2(b2ac)3(P -27J2 ) -a2X3 ( see my paper , " A discussion of the Sturmian Constants for Cubic and Quartic Equations , " Quart .
Math. Journ. t. iv .
( 1861 ) pp. 7-12 ) .
And I write X=b2_aC , y= 3aJ +2(2 ac)I , z= 27J2(= D ) .
269 .
I borrow from STURM'S theorem the conclusion ( but nothing else than this conclusion ) that ( x , y , z ) possess the fundamental property of auxiliars ( that is , that the quartic equations ( if any ) corresponding to a given system of values of ( x , y , z ) have one and the same character ) .
The foregoing equation gives 9X3y2-XZ-_y3 =a square function , and therefore positive ; that is , the facultative portion of space is that for which 9x2_y2x3z-y3 is + .
And the equation x3(9y2_Z ) _y3=O is that of the bounding surface , dividing the facultative and non-facultative portions of space .
270 .
To explain the form of the surface we may imagine the plane of Xy to be that of the paper , and the positive direction of the axis of z to be in front of the paper .
Taking z constant , or considering the sections by planes parallel to that of xy , -Z=0 , gives y2(9x3-y)=O , viz. the section is the line y=O , or axis of s twice , and the cubical parabola y=x3 .
Z= + , the curve x3= 9Y-z has two asymptotes y= / z , parallel to and equidistant from the axis of a , and consists of a branch included between the two parallel asymptotes , and two other portions branches outside the asymptotes , as shown in the figure ( zz + ) .
z= - , the curve x3has no real asymptote , and consists of a single branch , resembling in its appearance the cubical parabola as shown in the figure ( z=- ) .
( Z= + ) ( z=- ) / ~~~~~~~y Il A~no't.faxc~. .
, C.)0C .
) Taking x as constant , or considering the sections by planes parallel to that of zy , the equation of the section is z=9y2-E , which is a cubical parabola , meeting the plane of vi in a point on the cubical parabola y= 9X3 , and also in a twofold point on the axis of x , that is , touching the plane of xy at the last-mentioned point .
271 .
The surface consists of a single sheet extending to infinity , the form of which is most easily understood by considering the sections by a system of spheres having the origin of coordinates for their common centre .
These sections have all of them the same general form ; and one of them is shown ( Plate XX .
fig. 1 ) , the projection being orthogonal on the plane of xy or plane of the paper , and the spherical curve being shown , the portion of it above the plane of the paper by a continuous line , that below it by a dotted line ( the double point in the figure is thus of course only an apparent one ) : the same figure shows also the sections by planes parallel to that of xy previously shown in the figures ( z + ) and ( z-- ) .
272 .
Now considering the discriminatrix D=0 , in this case the plane z= 0 , it appears that the bounding surface and this plane divide space into six regions , viz. above the plane of the paper we have the four regions , A non-facultative , B facultative , A ' facultative , B ' non-facultative , and below it the two regions , C facultative , C ' non-facultative .
There are thus in all three facultative regions A ' , B , C , and since A ' and B correspond to D= + , these must have the characters 4r and 4i , and it is easy by considering a particular case to show that B has the character 4r , and A ' the character 4i ; C corresponds to D= - , and can therefore only have the character 2r+2i .
Hence , for any given equation , ( x , y , z ) will lie in one of the regions ( B , A ' , C ) , and if ( x , y , z ) is in the region B , the character is 4r , , , All , , VI a , C , , , 2r+2i .
273 .
It is right to notice that the determination of the character is really made in what precedes ; the determination of the analytical criteria of the different characters is a mere corollary- ; to obtain these it is only necessary to remark that z=+ , x= + , y= + includes the whole of facultative region B , that is , ( x , y , z ) being each positive , the character is 4r ; Z -+3 , X= + , Y=-A l include each a part and together the whole X_ of facultative region A ' , x=- , y that is , z being + , but ( x , y ) not each positive , the character is 4i ; X-+5 =+ include each a part and together the whole X=Y=of facultative region C , , , Xa , y z=- , x=- , y= + does not include any facultative space , that is , z being- , the character is 2r+2i ; and the combination of signs z= - , r=- , y+ is one which does not exist .
The results thus agree with those furnished by STURM'S theorem ; and in particular the impossibility of z= , x= - , y= + appears from STURM'S theorem , inasmuch as his combination would give a gain instead of a loss of changes of sign .
Article Nos. 274 to 285.-Petermtination of the characters of the quintic equation .
274 .
Passing now to the case of the quintic , I write J No. 19 , K= No. 25 , D= No. 26 , L --No .
29 , I= No. 29A ; viz. J is the quartinvariant , K and 1D are octinvariants ( D the discriminant ) , L is 12-thic invariant , and I is the 18-thic or skew invariant .
Hence also J , D , 2"L-J3 are invariants of the degrees 4 , 8 , 12 respectively ; and forming the combinations a 2'1LJJ3 DD = curve p(x , y ) =O , which is Professor SYLVESTER'S Bicorn : this curve divides the plane into certain regions , and if we attend to the solid figure and instead of the curve consider the cylinder , then to each region of the plane there correspond in solido two regions , one in front of , the other behind the plane region , and of these regions tn solido , one is facultative , the other is non-facultative ( viz. for given values 'of ( x , Y ) , whatever be the sign of p(x , y ) , then for a certain sign of z , zp(x , y ) will be positive or the solid region will be facultative , and for the opposite sign of z , zp(x , y ) will be negative or the region will be facultative ) .
It hence appears that we may attend only to the plane regions , and that ( the proper sign being attributed to z , that is to J ) each of these may be regarded as facultative .
It is to be added that the discriminatrix is in the present case the plane y=O , or , if we attend only to the plane figure , it is the line Y=0 ; so that in the plane figure the separation into regions is effected by means of the Bicorn and the line 9=O .
277 .
Reverting to the equation of the Bicorn , and considering first the form at infinity , the intersections of the , " curve by the line infinity are given by the equation y3 ( 2y-x)O viz. there is a threefold'intersection Y= 0 , and a simple intersection 2yx= 0 ; the equation Y3=0 indicates that the intersection in question is a point of inflexion , the tangent at the inflexion ( or stationary tangent ) being of course the line infinity ; the visible effect is , however , only that the direction of the branch is ultimately parallel to the axis of x. The equation 2y-x=O indicates an asymptote parallel to this line , and the equation of the asymptote is easily found to be 2y-x+5=0 .
278 .
The discussion of the equation would show that the curve has an ordinary cusp ; and a cusp of the second kind , or node-cusp , equivalent to a cusp and node ; the curve is therefore a unicursal curve , or the coordinates are expressible rationally in terms of a parameter p ; we in fact have ( v + 2 ) ( p3 v2+ 2p-4 ) ( p+ 2)2(p -3 ) ZZ Y--3 p +'P(1 +1 ) whence also die 2 279 .
The curve may be traced from these equations ( see Plate , fig 2 , where the bicorn is delineated along with a cubic curve afterwards referred to ) : as 0 extends from an indefinitely small positive value s through infinity to -1-A , we have the upper branch of the curve , viz. 0 - , gives w= oo , y s , point at infinity , the tangent being horizontal .
= = , gives x= -1 , y + , the node-cusp , tangent parallel to axis of y. 0= -2 , gives x =O , y=O , the tangent at this point being the ' axis of x. = -1 1 , gives x= s , y + , point at infinity along the asymptote .
And as p extends from xZ= -1-jI+ to x= -- , we have the lower branch , viz. 0= -1-H , gives x= -so , yi= -so , point at infinity along the asymptote 4= x=-762 , y= -412 ; the cusp , shown in the figure out of its proper position ( observe that for x= 7 7623 , we have for the asymptote y 4027- , so that the distance below the asymptote is 7- ; Professor SYLVESTER'S value ye -25 for the ordinate of the cusp is an obviouserror of calculation ) .
A= - , gives =o , = co , point at infinity , the tangent being horizontal .
The class of the curve is =4 .
280 .
The node-cusp counts as a node , a cusp , an inflexion , and a double tangent ; the node-cusp absorbs therefore ( 6+8+ 1= ) 15 inflexions , and the other cusp 8 inflexions ; there remains therefore ( 24-15-8= ) 1 inflexion , viz. this is the inflexion at infinity , having the line infinity for tangent ; there is not , besides the tangent at the node-cusp , any other double tangent of the curve .
281 .
The form of the Bicorn , so far as it is material for the discussion , is also shown in the Plate , fig 3 , and it thereby appears that it divides the plane into three regions ; viz. these are the regions PQR and S , for each of which { ( x , y ) is = - , and the region TU , for which 4(x , y ) is + ; that is , for PQR and S we must have J= - , and for TU we must have J= + .
Hence in connexion with the bicorn , considering the line MyO.we have the six regions P , Q , R , S , T , U. It has just been seen that for P , Q , R , S we have J= , and for T , U we have J + ; and the sign of J being given , the equations 2"1L-J8 ID z=2 LsJJ .
y= D , then fix for the several regions the signs of 2"L-J3 and D , as shown in the subjoined Table ; by what precedes each of the six regions has a determinate character , which for R , S , and U ( since here D is- ) is at once seen to be 3r+2i , and which , as will presently appear , is ascertained to be 5r for P and r+ 4i for Q and T. 282 .
We have thus the Table P , D_+ , J=- , 2'1L-J3=+ }5r , Q , D=+ , J=+,2L-J3 = T , D=+ J=+ , 211L J3= + RI D=J- , 21LJ3= ?
US D-- , J+ , 2"1L-J-+= + so that we have the kingdom 5r consisting of the single region P , the kingdom r+4i consisting of the regions Q and T , and the kingdom 3r+2i consisting of the regions R , Sand U. 283 .
Fora given equation if D is = , the characteris 3r+2i ; ifD=+ , J=+ , the character is r+4i ; if D= + , J , then , according as 2"L -J3 is -+ or is - , the character is 5r or r+4i .
But in the last case the distinction between the characters 5r and r+4i may be presented in a more general form , involving a parameter p , arbitrary between certain limits .
In fact drawing upwards from the origin , as in Plate , fig. 3 , the lines x2~y=O and x+y= 0 , .
and between them any line whatever x+yi= , the point ( x , y ) , assumed to lie in the region P or Q , will lie in the one or the other region according as it lies on the one side or the other side of the line in question , viz. in the region P if a+ vi is =- , in the region Q if x+ pie is =+ .
But we have 2"L-J3 ?
uJD -J3 and J being by supposition negative , the sign of 211LJ3+JD is opposite to that of x+ pie .
The region is thus P or Q according to the sign of 2"L-J3+pJD ; and completing the enunciation , we have , finally , the following criteria for the number of real roots of a given quintic equation , viz. If D , the character is 3r+2i , If D=+ , J=+ , then it is r+4i .
But if D + , J=- , then p being any number at pleasure between the limits +1 and -2 , both inclusive , if 21L-J+1+pJD=+ , the character is 5r , 2"L J3+PJD - , , , , , , , r+4i .
284 .
The characters 5r of the region P and r+4i of the regions Q and T may be ascertained by means of the equation ( a , 0 , c , 0 , e , 0)(d , 1)5= 0 , that is 0(a04+10c02F+5e)=0 ; there is always the real root d= O , and the equation will thus have the character 5r or r+4i according as the reduced equation a04+10ce'+5e=0 has the character 4r or 4i .
It is clear that ( a , e ) must have the same sign , for otherwise 02 would have two real values , one positive , the other negative , and the character would be 2r+2i .
And ( a , Me ) having the same sign , then the character will be 4r , if 2 has two real positive values , that is , if ae-5c0 is , and the sign of c be opposite to that of a and e , or , what is the same thing , if ce be - ; but if these two conditions are not satisfied , then the values of 02 will be imaginary , or else real and negative , and in either case the character will be 4r .
285 .
Now , for the equation in question , putting in the Tables b=d=f=0 , we find D=256 ae3(ae-5c')2 , J= 16 ce ( ae+3c2 ) , 2"L-J3= 212 ce3{2(aec2)4-c2(ae+ 3c3)3 } 212 ce ( ae-5c2)(2a e +a2c2e2+8ac4e-+5e6 ) We have by supposition D + , that is , ae= + ; hence J has the same sign as ce ; whence if J= + , then also ce= ?
, and the character is 4i ; that is the character of the region T is r+4i .
But if J=- , then also ce=- .
But ae being + , the sign of 2"L -J3 is the same as that of ce(ae-5C2 ) , and therefore the opposite of that of ae-5c2 : hence D + , J= the quartic equation has the character 4r or 4i according as 2"L_ J3 is + or - .
Whence the region P has the character 5r and the region Q the character r+ _L-4i ; and the demonstration is thus completed .
Article Nos. 286 to 293.-HERMITE 's newform of TSCHIRNHAUSEN s transformation , and application thereof to the quintic .
286 .
M. HERMIT demonstrates the general theorem , that if f(x , y ) be a given quantic of the n-th order , and p(x , y ) any covariant thereof of the order n -2 , then considering the equation f(x , 1 ) = 0 , and writing ( x , 1 ) t(n1 ) ( where f.(x , 1 ) is the derived function of f(x , 1 ) in regard to r ) , then eliminating x , we have an equation in z , the coefficients whereof are all of them invariants off(x , y ) .
287 .
In particular for the quinticf(x , z ) =(a , b , c , d , e , f)[x , y)5 , if 9(X5 BY ) , P2(X5 Y)l P3(5 BY)l P4(5 BY ) are any four covariant cubics , writing tfl I(XN )+ f2(X 1)+ ( 3 1)+WP4(XI ( viz. the numerator is a covariant cubic involving the indeterminate coefficients t , U , v , w ) then , in the transformed equation in z , the coefficients are all of them invariants of the given quintic .
Conducting the investigation by means of a certain canonical form , which will be referred in the sequel , he fixes the signification of his four covariant cubics , these being respectively covariant cubics of the degrees 3 , 5 , 7 , and 9 , defined as follows ; viz. starting with the form 3 y2X yXq2 q3 3 , Yf Y -2 a , 6 , c , d 6 , c , d , e c , d , e , f =-3 No. 16 , -3(A , B , C , Dix , y)3 , or ( 3A , -B , -C , -3DCx , y)3 , suppose , and considering also the quadric covariant ( c , i , raw ) y)2 , = No. 14 , then i , P2 P35 p4 are derived from the form ( A , B , C , Djx-n(px+2yy ) , 'y+H(2acx+i3y))3 , viz. we have pi(x , y ) 3(A , B , C , Dix , y)3 , p , ( x , y ) =+3(A , B , C , Dix , y)2( -Px2yy , 2ax+ py ) , { P3(X , y)}-3(A , B C , Dix , y ) ( -Ax-2yy , 2ax+Py)2 , 4(Xy ) } =+3(A , B , C , DI -3x-2ryy , 2ax+Py)3 , where { p3( y ) } and { 94(X , y)}are the functions originally called by hiini -(x , ) and 4 !
c 2 p4(x , y ) : those ultimately so called by him are 93(x , y)=4 { 3(x , y ) } + Jpl(x , y ) ( J = No. 19 ) , P4(X , y)=4{ 4(x , y ) } +-3JP2(X , y)+964,1(x , y ) , where 4 , ( x , y ) is the cubicovariant ( -27A2D+9ABC-2B3,.JX , y)3 of p , ( 0 x y ) , = ( .3A , 'B -C , 3D]Cx , y)u , ut supra .
The covariant p2(x , &amp; ) has the property that if the given quintic ( a , ... Cx , y)5 contains a square factor ( lx+my)2 , then P2(X , y ) contains the factor lx +ny : { 3(X , y ) } and JP4(X , y)t are covariants not possessing the property in question , and they were for this reason replaced by p3(X , y ) and P4(X , Y ) which possess it , viz. 3(x , &amp; ) contains the factor lx+ my , and p4(x , y ) contains ( lx+My)3 , being thus a perfect cube when the given quintic contains a square factor .
288 .
The covariants l(x , y ) and p2(x , y ) are included in my Tables , viz. we have pl(x , &amp; y)-3No .
16,.2(x , y)= No. 23 ( observe that in No. 23 the first coefficient vanishes if a= , b =0 , which is the property just referred to of p2(x , y ) ) ; the other two covariants , as being of the degree 7 and 9 , are not included in my Tables , but I have calculated the leading coefficients of these covariants respectively , viz. Table No. 85 gives leading coefficient ( or that of x3 ) in 03(X , y ) , and Table No. 86 gives leading coefficient ( or that of X3 ) in 4(X , y ) .
The coefficients in question vanish for a= 0 , b=0 , that is , C(X , y ) and ( 4(X , y ) then each of them contain the factory ; if the remaining coefficients of 4(x , y ) were calculated , it should then appear that for a=0 , b= 0 , those of X2y , Xy2 would also vanish , and thus that 4(X , y ) would be a mere constant multiple of y3 .
Table No. 85 .
a3cef ? ?
1 a262ef2 a63df2 + 64 64f2 o , a3 df2 + ?
15 a2bcdf2 94 awe e2f 54 b4def 144 a3de2f 32 a2bce2f + 86 a62cif2 48 64e3 + 135 a3e4 +16 a2bd2ef +106 ab2cdef + 184 b3c2ef + 108 a2bde ' 96 ab2ce3 135 b3cd2f + 288 a2c'f2 + 63 ab2d3f 272 63cde2 450 a2c2def -188 ab2d2e2 + 243 b63d3e + 80 2ac2e3 + 32 abc3ef 66 b6c3df 360 a2ed'f + 60 a6c2d~f + 212 b2c'3e + 135 a2cd2e2 + 68 abc2de ' + 148 b2cd'e + 360 a2d4e 36 abcd3e 412 62cd4 160 abd ' + 144 bc~f + 108 ac4df 36 bc4de 180 ac4e2 48 bc3d 3+ 80 ac3d2e + 124 ac2d 4 48 +32 +415 +1119 +1294 e M. HERMIT , P. 17 , has erroneously written N(X , y ) +4A1(x , y ) , instead of 493(x , y ) +Af1( , y ) ; the latter expression is that which he really makes use of , and the formula in the text is correct .
If by establishing two linear relations between the coefficients ( t , u , v , w ) the equation A0 can be satisfied ( which in fact can be done by the solution of a quadric equation ) , then these quantities can be by means of the relations in question expressed as linear functions of any two of them , say of v and w ; and then the next coefficient 3B will be a cubic function ( v , w)3 , and the equation 13 =0 will be satisfied by means of a cubic equation ( v , W)3= 0 , that is , the transformed equation in z can be by means of the solution of a quadric and a cubic equation reduced to the trinomial form z5+Z+ = ?
0 and M. HERMIT shows that the equation A_= 0 can be satisfied as above very simply , and that in two different ways , viz. 291 .
10 .
2=O if Dt2 -6BDtv-( D 1-OAB )v2=O , Bu2 -2Dluw -(9BD -1OAD1)w2 =O , that is , N denoting as above , if t 3BD+ A/ ND v U= D1 + UN DI ~~B 292 .
20 .
Writing the expression for A in the form D1(t2iDv2+2Duw I OAiDw2 ) + BD(lOAv26tv -u2+ 9DW2 ) , then A O. if t2DV2+2Duw-10ADw2 0 , lOAv26tv -U2+ 9 DW22=O .
These equations , writing therein t= VDT , u-U+5AW , v=k-V , W. become T2-V2 +4UW-O ?
-5AV2 + 3VFDTV + U2 +IlOAUW +(25A 2 9D)W2 =0 , the first of which is satisfied by the values T=gW--U , V-=W+-U ; and then substituting for T and V , the second equation will be also satisfied if only,2 5A+ 3/ D Article Nos. 293 to 295.-HERMITE'S application of the foregoing results to the determination of the Character of the quintic equation .
293 .
By considerations relating to the form D{[Djt2-6BDtv-D(D , -1OAB)v2 +D[-Bu2+2Djuw + 9BD-1OAD w2 ] } , M. HERMIT obtains criteria for the character of the quintic equationf(x , 1)=0 .
continuous line , .
passing from a point at infinity , through the ; cusp of the bicorn .
on to the node-cusp , then forming a loop so as to return to the node-cusp , again meeting the bicorn at the origin , and finally passing off to a point at infinity , the initial and ultimate directions of the .
curve being parallel to the axis of x. 300 .
It may be remarked that , inasmuch as one of the branches of the cubic touches the bicorn at the node-cusp , the node-cusp counts as ( 4+2= ) 6 intersections ; the intersections of the cubic with the bicorn are therefore the cusp , the node-cusp , and the origin , counting together as ( 2 +-6 +1= ) 9 intersections , and besides these the point at infinity on the axis of x , counting as 3 intersections .
This may be verified by substituting in the equation of the cubic the bicorn p-values of x and y. We must , however , to include all the proper factors , first write the equation of the cubic in the homogeneous form ( 9x+ 8g , + 5z)2z(y-z)2(25z9y ) 0 , and herein substitute the values X : : Z= --(@ +2)(qp 3_ 2+2 4 ) : ( 0 +2 )2(o 3)4p .
( 0 +1 )93 ; the result is found to be 03{(0 +1)(4p2+66_9)2_ ( 20+3)2(4 3 +4 2+ 18P + 27 ) } = 0 , that is 90',3(o + 2'(40 + 3 ) 2-0 ; and considering this as an equation of the order 12 , the roots are 4p 0 , 3 times , A_=2 , 1 time ; 4p 4 , 2 times , and 4=co , 6 times .
301 .
The cubic curve divides the plane into 3 regions , which may be called respectively the loop , the antiloop , and the extra cubic ; for a point within the loop or antiloop , '4 , ( x , y ) is = -- , for a point in the extra cubic 4+(x , y ) is = + .
If in conjunction with the cubic we consider the discriminatrix , or line y 0 , then we have in all six regions , viz. y being _ + , three which may be called the loop , the triangle , and the upper region ; and y being three which may be called the right , left , and under regions respectively ; the triangle and the under region form together the antiloop .
302 .
It is now easy to discuss HERMITE'S two sets of criteria ; the first set becomes y , =-+ , iY-=1- , J(x+y ) =4+ , 4( , y)= , character 5r J(.y-1)(-+y)=+,.4xv , Y ) + character r+4i , Y=+o~~~~~~~~~x +(Y , =+J y=+ , y-1 + , J(v+y ) _- , ( x , y)=- , cannot exist .
Referring to the Plate , fig. 4 , which shows a portion of the cubic and the bicorni , then 10 the conditions y= + , 4(x , y)imply that the point ( x , y ) is within the loop or within the triangle of the cubic ; the condition y1brings it to be within the triangle , and for any point within the triangle we have x +y- , whence also the condition J(x+y)= + becomes J= - ; hence the conditions amount to J= - , ( x , y ) within the triangle ; but by the general theory ( x , y ) , being within the triangle , that is , in the region P or T , if J =- &gt; will of necessity be within the region P ; n so that the condi-* MDcccLxvJJ .
4 D tions give J=- , ( x , y ) within the region P ; the corresponding character being Sr , which is right .
20 .
y= + , 4(x , y)=- , the point ( a , y ) must be within the loop , or within the triangle ; if ( x , y ) is within the loop , theny-1=+ , x+ y=1 , and the condition J(y1)(++y)=+ becomes Jthat is , we have J and ( x , y ) within the loop , that is , in the region T. And again , if ( x , y ) be within the triangle , then y1 , +y= + , and the condition J(y1)( x+y)+ still gives J= ; but J=- , and ( x , y ) within the triangle , that is , in the region T or P , will of necessity be in the region T ; so that in either case we have j - , ( a , y ) in the , region T , which agrees with the character r+4i .
3 ? .
y= + , ( x , y)= + , ( x , y ) is in the upper region , that is , in the region Q or T ; if ( x , y ) is in the region Q , then of necessity J= - , and if in the region T , then of necessityJ=+ , that is , we have J==-.- , ( x , y ) in the region Q , or J= + , ( x , y ) in the region P , which agrees with the character r+ 4 .
And it is to be observed that the portions of T under 20 and 30 respectively make up the whole of the region T , and that 30 relates to the whole of the region Q , so that the conditions allow the point ( x , y ) to be anywhere in Q or T , which is right .
40 .
y=+ , t(x , y)=- , ( x , y ) is in the loop or the triangle , and then y-1= + implies that it is in the loop , whence x+y= + , and the conditionJ ( x +y)=becomes J= ; we should therefore if the combination existed have J= - , ( x , y ) within the loop , that is , in the region T ; but this is impossible .
303 .
HERMITE'S second set of criteria are y= ?
, 25y=+ J=_ , 4(xy)= - , character 5r .
y=+ , 29-y=+ , J=- , '+(x , y)=+l + 29 Jo character r+4i .
25 10 .
If y= + , 4( , y ) - , then the point ( x , y ) must be situate within the loop or within the triangle ; and recollecting that at the highest point of the loop we have y=295 , the condition 9-y + is satisfied for every such point , and may therefore be omitted .
The conditions therefore , are J=- , ( x , y ) within the loop , that is , in the region T , or within the triangle , that is , in the region P or the region T ; but for any point of T the general theory gives J + , and the conditions are therefore J= - , ( x , y ) within the region P ; which agrees with the character 5r .
20 .
y= -+- , 4{$ , y ) _ + , that is , ( x , y ) is within the upper region , that is , in the region o2 or T ; and 2-y= +(x , ( y ) will be within the portions of Q and T which lie beneath the line y=_295 ; but J , and therefore ( x , y ) cannot lie in the region T ; hence the conditions amount to J=- , ( x , y ) within that portion which lies beneath the line y=9 of the region Q. Y= +7 -9f-y + , ( X , y ) lies beneath the line y= =- , viz. in one of the regions P , Q or T ; but J= + , ( x , y ) cannot lie in the region P or Q ; hence the conditions give J=-+ , ( x , y ) within the portion which lies beneath the line y=29 of the region T. 40 , 9 + , 2-y= , that is , ( x , y ) lies above the line y296- , and therefore in one of the regions T or Q ; and by the general theory , according as ( x , y ) lies in T or in Q , we shall have J+ or J= , hence the conditions give J ( x , y ) within the portion which lies above the line y of the region Q. X=:+7 ( x , y ) within the portion which lies above the line y=295 , of the region T. 20 , 30 , and 40 , each of them agree with the character r+4i , and together they imply J_- , ( x , A ) any where in the region Q , or else J= + , ( x , y ) anywhere in the region T ; which is right .
Article Nos. 304 to 307.-HERMITE'S third set of Criteria ; cominarson with No. 283 , and remarks .
304 .
In the concluding portion of his memoir , M. HERMIT obtains a third set of criteria for the character of a quintic equation ; this is found by means of the equation for the function a4(00-Ol)(01-02)(02-03)(03-04)(04-0O ) of the roots ( 00 , 01 , 02 , 031 04 ) of the given quintic equation ( a , b , c , d , e , fXO , 1)5=0 .
The function in question has 12 pairs of equal and opposite values , or it is determined by an equation of the form ( u2 , 1)12=0 , which equation is decomposable , not rationally but by the adjunction thereto of the square root of the discriminant , into two equations of the form ( u2 , 1)6=0 ; viz. one of these is q12 +u1O(a+ 3VA ) + u48 [ T'(a V/ j)2+ , A ] u6 d + ?
4 [ 1(a+VA)2+A]L +U2 ( a 23A)A2 + As-=0 and the other is of course derived from it by reversing the sign of V/ A. I have in the equation written ( a , d ) instead of HERMITE 's writing capitals A , D ; the sign of the terni in u6 instead of +7 as printed in his memoir , is a correction communicated to me by himself .
The signification of the symbols is in the author 's notation a=54A , d=4 .
59(AD -8 D1 ) , A=55D 4 whence , in the notation of the present memoir , the expressions of these .
symbols are a= 54J d= -1510(211LJ3 3 JD ) , A= 55D .
305 .
From the equation in u , taking therein the radical A/ A as positive , M. HERMIT obtains ( d &lt; 0 a mistake for d &gt; 0 ) the following as the necessary and sufficient conditions for the reality of all the roots , A= + , a+ 3VA= - , d= + , character 5r ( HERMITE'S third set of criteria ) .
306 .
It is clear that a+3VA-A is equivalent to ( a and a2_ 9A + ) , and we have a2-9Z=55(125J2-9D ) , so that these conditions for the character 5r are D= + , J=- , 125J2 9D= + , 2"LJ3 3JD=Now , writing as above , 211-J3 D X=J3 , J2 ' these are y=+ , J= , , 1%5 y=+ , x -y= ; the conditions y+ , J=imply that ( x , y ) is in the region P or the region Q ; and the condition x-4y= ( observe the line x--5y= 0 lies between the lines x+y= 0 , x-2y=0 , and so does not cut either the region P or the region Q ) restricts ( x , y ) to the region P ; and for every point of Pie is at most =1 , and the condition '9-2 =+ is of course satisfied .
The condition , 125J2_ 9D + , is thus wholly unnecessary , and omitting it , the conditions are D=+ , J=- , 2L-J 33JD=0 , character Sr , which , -3 being an admissible value of jh , agrees with the result ante , No. 283 .
307 .
It may be remarked in passing that if 12345 is a function of the roots ( x , , X2 , X3 , X4 , 5X5 ) of a quintic equation , which function is such that it remains unaltered by the cyclical permutation 12345 into 23451 , and also by the reversal ( 12345 into 15432 ) of the order of the roots , so that the function has in fact the 12 values ccz=12345 , 01=24135 , 2=13425 , P2 -32145 , Z3 =14235 , =3343125 , eZ4=21435 , f4 13245 , 5= 31245 , P5 -14325 , &amp; 6=41325 , P6-12435 , then { ( a , 1 ) being any unsymmetrical function of ( ex , 13 ) , the equation having for its roots the six values of p5(M , X0 ) ( viz. p(oc , t31 ) , @(os2 , / 32 ) ... { 136 , ) ) can be expressed rationally in terms of the coefficients of the given quintic equation and of the square root of the discriminant of this equation .
In fact , v being arbitrary , write L=l71{v-q(~ , j3 ) } , M-FIo{v-Nf3 , a ) } , then the interchange of any two roots of the quintic produces merely an interchange of the quantities L , M ; that is , L-+ M and ( L-.M ) ?
:(x , e2 X3 I 5 ) are each of them unaltered by the interchange of any two roots , and are consequently expressible as rational functions of the coefficients ; or observing that 42(X1 , X32 X4 X , ) is a multiple of VD , we have La function of the form P+QV/ D ; the equation L=0 , the roots whereof are v=zz(oca f ' ) ... v=5(Psa , / 6 ) , is consequently an equation of the form P+ QVD=0 , viz. it is , a sextic equation ( *)Cv , 1)6= 0 , the coefficients of which are functions of the form in question .
Hence in particular u= =12345 =(x1.2)Q2(x3 ) ( ' _ r)2('rxQr ) .
is determined as above by an equation ( *t(2 , 1)6=0 .
Another instance of such an equation is given by , my memoir " On a New Auxiliary Equation in the Theory of Equations of the Fifth Order , " Phil. Trans. t. 151 ( 1861 ) , pp. 263-276 .
Article Nos. 308 to 317.-HERMITE'S Canomicalform of the quintic .
308 .
It was remarked that M. HERMITE 's investigations are conducted by means of a canonical form , viz. if A ( -J , = No. 19 as above ) be the quartinvariaxit of the given quintic ( a , b , c , d , e , f(x , y)5 , then he in fact finds ( X , Y ) linear functions of ( x , y ) such that we have ( a , b : C , d , e , f ]AX , y ) 5-i a , P k ) N3Vi , P111 7'1x ) Y ) 5 ( viz. in the transformed form the two mean coefficients are equal ; this is a convenient assumption made in order to render the transformation completely definite , rather than an absolutely necessary one ) ; and where moreover the quadricovariant ( Table No. 14 ) of the transformed form is =s1~/ AXY , or , what is the same thing , the coefficients ( X , p , 4/ k , / yi , x ) of the transformed form are connected by the relations f_4pdVk+3k=0 , ; .f , _41p,64p , / V+3k= ?
kk'-3 &amp; qd +21c=VA , the advantage-is a great simplicity in the forms of the several covariants , which simplicity arises in a great measure from the existence of the very simple covariant operator dd -y-(viz , operTatingtherewith on any covariant , we obtain agaMin a covariant). .
or putting for P ' , Q ' their values , this is =4C into 4(oa , 3 , rXQ , P)2(ax2 + 2jxy + ryy2 ) , that is , we have AT2-_ U2=c x2+PXy+yy2 ; and we have also AT2 U2=4 jA[(X+Y)2 ( XXY)2 ] VAXY consequently oax2+oxy+ry2=AT2_U2=VAXY .
311 .
We have * SC Q'TQU ) this gives and thence -~~~~~~( ' 2cy 0ba -6.2)2(a , boc oef6 ) =AXay anY t pi / V , , Xi Y)5 120A(vX+V'Y ) ; the left hand side is a linear covariant of the degree 5 , it is consequently a mere numerical multiple of Px+Qy ; and it is easy to verify that it is =120(Px+Qy ) .
( In fact writing b-d=e=O , the expression ise(3c2b-afy)2(ax5+10cxy2+ft ) , and the only term which contains x is a2f2 .
b2.10cx3y2 120a2cf2 .
x ; but for 1id-e O , Table No. 22 gives Px=a2cf2x , and the coefficient 120 is thus verified .
) But Px+Qy is ( X+Y ) , and we have thus AP=A &gt; =A4 " ' whence not only v= Vk suppose , C but we have further k= a result given by M. HERMIT .
' 315 .
Substituting for '=Y ' the value 4/ k , we have ( a , b , c , d , e , f 3(n , I5 =(a , b , Con d , elf3 CA+Q/ A ) X+/ C( WA )2 s/ t WAP A)X+2 vC(W +P WA)X Y)5 pi -v/ k p ' I ](Xi Y)55 and we have then MX2+I3ty+yY2=VAXY , viz. the left-hand side being the quadricovariant of ( a , b , c , d , e , fj)x , y)5 , the equation shows that the quadricovariant of the form ( x , , v/ ik , Vk p'5 , A'X , Y)5 is =VAXY , and we thus arrive at the starting-point of HERMITE'S theory .
316 .
The coefficients ( X , { , s / k5/ i , ' , X ' ) of HERMITE'S form are by what precedes invariants ; they are consequently expressible in terms of the invariants A , B , C ( and I ) .
M. HERMITE writes and he finds BC N/ A= q-3h+2k1 , 3= h-k- ` , v5= k , or , what is the same thing , A3+ 3AB ?
.C AB+C C g= .
y A--,5 k= bA which give q , h , k in terms of A , B , C , and then putting 2+ 23 ~~~~12 A =(9k1 16h-k-gh)2 24hc3 -k ( the equation T2= A7A is in fact equivalent to the before-mentioned expression of T2 M1DCCCLXII .
4 EE It is in fact by means of these comparatively simple canonical expressions that M. HERMIT was enabled to effect the calculation of the coefficient A. Article Nos. 318 to 326.-Theory of the imaginary linear transformations which lead to a real equ ation .
318 .
An equation ( a , 6 , c , ... :X , y)nl=O is real if the ratios a : b : c , &amp;c. of the coefficients are all real .
In speaking of a given real equation there is no loss of generality in assuming that the coefficients ( a , b , c , ... ) are all real ; but if an equation presents itself in the form ( a , b , c , ... 3)w , y)flO with imaginary coefficients , it is to be borne in mind that the equation may still be real ; viz. the coefficients may contain an imaginary common factor in such wise that throwing this out we obtain an equation with real coefficients .
In what follows I use the term transformation to signify a linear transformation , and speak of equations connected by a linear transformation as derivable from each other .
An imaginary transformation will in general convert a real into an imaginary equation ; and if the proposition were true universally , -viz .
if it were true that the transformed equation was always imaginary-it would follow that a real equation derivable from a given real equation could then be ' derivable from it only by a real transformation , and that the two equations would have the same character .
But any two equations having the same absolute invariants are derivable from each other , the two real equations would therefore be derivable from each other by a real transformation , and would thus have the same character ; that is , all the equations ( if any ) belonging to a given system of values of the absolute invariants would have a determinate character , and the absolute invariants would form a system of auxiliars .
But it is not true that the imaginary transformation leads always to an imaginary equation ; to take the simplest case of exception , if the given real equation contains only even powers or only odd powers of x , then the imaginary transformation x : y into ix : y gives a real equation .
And we are thus led to inquire in what cases an imaginary transformation gives a real equation .
319 .
I consider the imaginary transformation x : y into ( a + bi)x + ( c + di))y : ( e +f i)x +(+ hi)y , or , what is the same thing , I write X ( a+bi )X +(c+ di)Y , y ( e +f i)X+ ( g + hi)Y , and I seek to find P , Q real quantities such that Px+Qy may be transformed into a linear function RX+SY , wherein the ratio R : S is real , or , what is the same thing , such that RX+SY may be the product of an imaginary constant into a real linear function of ( X , Y ) .
This will be the case if PX+Qy ( 1+Oi){P(aX cY)+Q(eX+gY ) } that is if , P(X die )+Q(fX+hY)=z 0 { P(aX+cY)+Q(eX+fY ) } , which implies the relations 6P +fQ=4(aP+eQ ) , dP+hQ= 4(cP+gQ ) , or , what is the same thing , ( b -a4)P+(fe)Q= 0 , ( dc)P+(h-gO)Q=0 , and if the resulting value of P : Q be real , the last-mentioned equations give ( ag-.ce)02-(ah+bg-cf -de)O+bh-df--0O and 0 being known , the ratio P : Q is determined rationally in terms of 0 .
320 .
The equation in 0 will have its roots real , equal , or imaginary , according as ( ah+ b-cf-de)2 -4(ag -ce)(bhdf ) , that is a2h2 -+ b2g2 + cf 2+ d2e2 2ahbg 2alwf2 ahde 2bgcf2bgde 2cfde + 4adfqh + 4bceh is A+ , -= 0o ; and I say that the transformation is subimaginary , neutral , and superimaginary in these three cases respectively .
In the subimaginary case there are two functions Px+Qy which satisfy the prescribed conditions ; in the neutral case a single function ; in the superimaginary case no such function .
But in the last-mentioned case there are two conjugate imaginary functions , Px+Qy , which contain as factors thereof respectively two conjugate imaginary functions UX+ VI .
321 .
Hence replacing the original x , y , X , Y by real linear functions thereof , the subimaginary transformation is reduced to the transformation x : y into kX : Y , where k is imaginary ; and the superimaginary transformation is reduced to x+iy : x-iy into k(X+iY ) : ( X-iY ) , where k is imaginary .
As regards the neutral transformation , it appears that this is equivalent to x= ( a+ bi)X+ ( c + di)Y , Y= ( g+hi)Y , with the condition 0=(ah+bgq)2 -4agbh , = ( ah -bg)2 , that is , we have ahby= 0 , or without any real loss of generality g =a , h= b , or the transformation is x= ( a+ bi)X+(e + di)Y , ( a+bi)Y , that is , x : y=X+kY : Y , k being imaginary .
322 .
The original equation after any real transformation thereof , is still an equation .
of the form ( a,.(X .
9u~ )1 0 ; and if we consider first the neutral transformation , the transformed equation is ( a , .
* XX + lie , Y ) = 0 ; this is not a real equation except in the case where k is real .
For the superimaginary transformation , starting in like manner from ( a,. .
x. B ) , = 0 , this may be expressed in the form ( ac + pit , 'Y + SU p. .
i:( , MAtA + y.-iy iX= n= , viz. when in a real equation ( x , y)n % we make the transformation x : y into x+i , : the coefficients of the transformed equation will form as above pairs of conjugate imaginaries .
Proceeding in the last-mentioned equation to make thetransformation x+iy : x-iy into k(X+iY ) : X-iY , I throw k into the form cos 24i+i sin 24 , = ( cos p+i sin q5 ) .
( cos pi sin ) ( of course it is not here assumed that 5 is real ) , or represent the transformation as that of x+iy : xZ-Y 'into ( cos P+i sin 4p)(X+iY ) : ( cos 0 -i sin 9)(X-.iY ) ; the transformed equation thus is ( cs+I3i,..a .
oPi 3(cos p+i sin p)(X+iY ) , ( cos p-i sin p)(X-iy))n=O .
The left-hand side consists of terms such as ( X2+Y2)n-2s into ( ry+i)(cos sp+t sin sp)(X+iY)s+(ryi)(cos so-i sin sp)(XiY ) , viz. the expression last written down is ( y cos spi sin sp){(X+iY)8+(X-iY)S } -(y sinSp+ cossp){(X~iY)(XiY ) i and observing that the expressions in { } are real , the transformed equation is only real if ( y cos sq5sin sp ) +(r sin Sp+Z cos sp ) be real , that is , in order that the transformed equation may be real , we must have tan sp=real ; and observing that if tan sp be equal to any given real quantity whatever , then the values of tan p are all of them real , and that tan p real gives cos p and sin p each of them real , and therefore also p real , it appears that the transformed equation is only real for the transformation x+iy : x-i y=(cos p+i sin p)(X+iY ) : ( cos p-i sin p)(X-iY ) , wherein p is real ; and this is nothing else than the real transformation x.y into X cos p-Y sin @ : X sin 'P+Y cos p. Hence neither in the case of the neutral transformation or in that of the superimaginary transformation can we have an imaginary transformation leading to a real equation .
324 .
There remains only the subimaginary transformation , viz. this has been reduced to a : y into kX : Y , the transformed equation is ( a,..ktXl Y ) f=O and this will be a real equation if some power / P of k ( p not greater than n ) be real , and if the equation ( a,..j(x , y)nf= 0contain only terms wherein the index of x ( or that of &amp; ) is a multiple of p. Assuming that it is the index of y which is a multiple , the form of the equation is in fact X0(xP , yP)m=0 , ( n=mp +a ) , and the transformed equation is X4(kPXP , YP)m=0 , which is a real equation It is to be observed that if p be odd , then writing 1 =K ( K real ) and taking hI the real p-th root of K , then the very same transformed equation would be obtained by the real transformation xy into AX : Y ; so that the equationobtained by the imaginary transformation , being also obtainable by a real transformation , has the same character as the original equation .
326 .
Similarly if p be even , if K be real and positive , the equation kP =K has a real root k ' which may be substituted for the imaginary k , and the transformed equation will have the same character as the original equation ; but if K be negative , say K= -1 ( as may be assumed without loss of generality ) , then there is no real transformation equivalent to the imaginary transformation , and the equation given by the imaginary transformation has not of necessity the same character as the original equation ; and there are in fact cases in which the character is altered .
Thus if p=2 , and the original equation be a(X2 , y2)1fl0 , or ( X2 , y2)m =o , then making the transformation x : y into iX : Y , the transformed equation will be X(X2 , Y2)m= 0 or ( X2 , _Y2)m 0 , giving imaginary roots X2+aY2=O corresponding to real , roots x2-ay2=0 .
Article No. 327.-Application to the auxiliars of a quintic .
327 .
Applying what precedes to a quintic equation ( a,. .
0 z , y)5= 0 this after any real transformation whatever will assume the form ( a',. .
( X ' , yl)50 ; and the only cases in which we can have an imaginary transformation producing a real equation of an altered character is when this equation is ( a ' , 0 , e , 0 , e ' , )02x ' , yl)5= 0( ' not 0 0 ) , or when it is ( a ' , 0 , 0 , 0 , e ' , 0)2a/ , y')5=O , viz. when it is x'(a'x'4+10c'x'2y'2+5e'y4)=O , or '/ ( a'x'4+5e'y'4)=0 .
In the latter case the transformationx , y'intoX v/ :Y givesthe real equation X(a'X4-5e'Y4)=0 .
I observe however that for the form ( a ' , 0 , 0 , 0 , el , Oj(X , y)4 , and consequently for the form ( a , ... .3x , y)5 from which it is derived we have J 0 ; this case is therefore excluded from consideration .
The remaining case is ( a ' , 0 , c ' , 0 , e ' , O ) ' , y')5=O , which is by the imaginary transformation a !
: y ' into iX : Y converted into ( ad , 0 , -c ' , 0 , e ' , O0X , Y)5=0 ; for the first of the two forms we have J=16a'de'2 , and for the second of the two forms J= -16a'c'e'2 , that is , the two values of J have opposite signs .
Hence considering an equation ( a , b , c , d , of )3x , y)5=O for which J is not = 0 , whenever this is by an imaginary transformation converted into a real equation , the sign of J is reversed ; and it follows that , given the values of the absolute invariants and the value of J ( or what is sufficient , the sign of J ) , the different real equations which correspond to these data must be derivable one from another by real transformations , and must consequently have a determinate character ; that is , the Absolute Invariants , and J , constitute a system of auxiliars .

