" On an Application of the Theory of Scalar and Clinant Radical Loci .
" By ALEXANDER J. ELLIS , Esq. , B.A. , F.C.P.S. Communicated by A. CAYLEY , Esq. Received February 20 , 1861 .
Read March 14 , 1861* .
1 .
The following investigation , which contains a correction and extension of that in Pluicker 's 'System der Geometrie ' ( § 3 .
art .
64 ) , is a direct application of the theories in the writer 's paper " On Scalar and Clinant Algebraical Coordinate Geometry " ( Proceedings , vol. x. pp. 415-426 ) , to which reference must be made for an explanation of the notation and terminology .
2 .
Letf(x , y ) be an algebraical formation ( function ) of x , y and of n+ 2m dimensions , such that when any scalar value is attributed to x , f0=O has n scalar ( possible ) and 2m clinant ( imaginary ) roots , and let A be the coefficient of in +2m .
Then f(X , y).(y-f- ) -f ) x ( y-f'x)(y-f2 ' ) ... ( y -f'(m))(Y -f()x ) .
3 .
Now let PQ be a line which , when produced , will cut the curve , whose equations are OM=x .
OI+y .
OB , f/ ( , y)=0 , ( where x and y are scalar , and OI is in the same straight line with OP , and OB is a line equal in length to 01 , and parallel to and in the same direction with PQ ) in the n points M , M , , ... M , ~ .
Then , PM1 PM PM PM I if OP=x1 .
01 , where x , is scalar , P-M * will repreIO OB ' GB sent the nscalarroots fx , f , x ... .f,1 .
n Hence ifPQ=y .
OB , the point Q not being necessarily a point in the curve , any one of the PQ--PMr MQ .
Q factors y1-f.x , will be represented by PQOBM MO Now for the clinant rootst , put y=r-ti.s , where the Roman i= V--1 , and r and s are scalar , then must f(x , y ) = Fl( , r , s ) + i. s. F( , r , s ) , because there are n scalar roots for each of which s=0 .
For the clinant roots , then , we have F=-0 , and F2=0 ; whence we find , by alternalely eliminating s and r , F3(x , r ) =0 , and F4(x , s)= ) .
Now describe the loci of the points R and S , where OR=x .
OI +r .
OB , F3(x , r ) = 0 , and OS=x .
OI+s .
OB , F4(x , s)=0 .
Then from the nature of the clinant roots , which will be in pairs of the form r1-i .
s , if we produce PQ to cut these curves , for every point RI in which it cuts the first there will be two points S ' , S'2 , in which it cuts the second .
If , then , from the point R ' we draw R'WM'=i .
PS't and R'M't=i .
PS ' , , two corresponding clinant roots of the equation will be IPM1 , PM2 , so that OB OB Y , ( r)"_ M PQ--PMri ) Ml( ) Q y~l-x= O -O ----B and f(r-~ PQ-PM( M ( r ) Q yi 2( B= OB See the example and figure in ( 11 ) .
Hence we shall have for the geometrical representation of any algebraical formation , by ( 2 ) , fi _ , iMI=X Q MnQ , 'Q M'x2Q M I(M ) Q M2( ) QBY -OB " *x OB G ' B OB OB OB 4 .
Next suppose that the origin of the coordinates be altered to Ot , so that Ott==OI , and O'I1 , when produced , passes through Q. This will be equivalent to putting y+b instead of y , and the result will not disturb the coefficient of in +2m .
Also since one of the radical loci for the scalar values remains unaltered , and the others ( which are the axes ) are parallel to the former , the points M , , M , , ... Main which PQ , when produced , will cut the curve , remain the same , and consequently there are the same number of scalar roots as before .
5 .
But as the terms themselves of the equation f(x , y)=0 are altered , the radical loci of R and S in ( 3 ) will be altered , and as the distances are now measured from Q and not from P , the lines QR ' +i .
QS'3 , QR ' , +i .
QS'4 will now be different and =Qm ' , , Qm ' , .
If then O'Q=xx .
OTI1 , and QP=y , .
BO= -y , .
OB , then M1P MnP m1 , P mt'P m(m)P m(m ) f( , - ) =X OB-OBOB OB " B ' O Or , replacing OB in each case by BO , we have f(x2 , Y2 ) =X ' M1P MPx 1Pt n2)p BO ' BO BO BO where , since there are n+-2m factors , X= ( 1I)f+2m .
X. 6 .
But if , instead of simply changing the origin , we change the inclination of PQ and make it parallel to OC , then if OB and OC make the angles 3 and y with OI , and OJ be at right angles to 01 , the lengths of 01 , OJ , OB , OC being the same , and M be any point , we have OM=x .
OI+y .
OB=(xy cos 3 ) .
01+y sin3 .
OJ , and OM=x'.OI +y'.OC=(x'+ycosy).OI+y'sin y.OJ , which give two simple equations to determine x , y in terms of x ' , y ' , so that on substitution f(x , y ) becomes 0(xt , yt ) , a formation of the same number of dimensions .
But the coefficient of y'n+2m will be different from that of in +2m Let it =- .
The curve for the scalar values of x ' , y ' in OM =x ' .
OI .
y ' .
0C , ( x ' , y')=O will be precisely the same as that in ( 3 ) ; but as PQ , having a different inclination , will cut it in very different places or not at all , the numbers of scalar and clinant roots of =0 may be individually different from , although their combined number will be the same as , those off= 0 .
If then there be h scalar and 2k clinant roots of -= 0 , where h+ 2k=n + 2m , and we determine N1 , N , ... N , N2 ' ... in the same way as we previously determined M1 , M , , ... M ' , M ' , ... we find N , Q NhQ Nx iQ N2cQ f(xv Y)= ( X ' Yv)= ' ff '"X U -o x o--U ... ... / ( ^)^^^^o^**oc-o^**oc- ' 7 .
This investigation shows that if there be any number of lines drawn from a point , cutting those radical loci of an equation for scalar and clinant roots which correspond to the inclination of these lines to the axis and to the position of the origin , coefficients X , p , &amp;c. can always be assigned such as to make the product of these ratios each equal to the other ; and that these coefficients have only to be multiplied by ( -l)n+2m , in order to give the coefficients corresponding to 2F r , 33 such a change of origin as would make the axis pass through the other extremities of the lines .
8 .
Now let 01,2 ... OnOX be a completely enclosed polygon , the sides of which , when produced , cut the radical loci of a known algebraical equation of n dimensions in known points .
Let the product of the series of ratios formed as above ( 3 ) , corresponding to lines drawn from the point 0r in the direction 0 , O,.+I ( on which lastnamed line the unit line O , r r.+i , corresponding to OI in ( 3 ) , is measured ) , be represented by [ Or , r+M -O .
0O'.+i1 ] .
And let X , ,r+I be the coefficient necessary to make this product a correct representation of the formation on the left-hand side of the given equation for the values of x and y , corresponding to the point 0 , .
and the direction 0 , .
0,.+ .
Then for the point 0 , +l and the direction 0,.+ Or ( for which the corresponding unit line is.+1 O ' , .
) , we shall have the product of the corresponding series of ratios represented by XA.+ , .
[ 0,.+ 1 , .
M -r0 , + l O ' , .
] where r,.+1,.= ( --l ) .
X , ,,.+1 .
Proceeding then round the polygon , and forming the products for each of the two sides terminating at each point , we have by ( 7 ) , X1 , n[O , , , nM o0ol ' ] =x , , , 2 [ ,2 M010'2 ] ( -I )nIl , 2[02 , 1M 020 1 ] = 2 , 3 [ 02 , 3M 02 3 ] , ( 1)ZXn-- , n[0 , n- , ,2M zM 01 0n--2]=Xn-i , n[0n- , nM 0 , _1 0 ' , ] ( -l)'nn-I , n [ 0n_n-I M.-n 0 ?
, n ] =-X,1 [ 0n , ,I M ?
o0,0 ' ] .
Hence multiplying all these n equations together , and remembering that Al , = ( -l1 ) .
X , 1 , and that ( -l)nn= ( l ) " , since nn and n are both odd or both even , we have [ 01 , nM -01,0 , .
[ 02,1 M.-0,0 1].- .
[ On , , , M-00 .
Onl ] =(-l).[Ol,2 M-OlO ',2 .
[ 0,2,3.M-020 ' ] ... [ 0 , , , i M -OO ' , ] .
Now since O , .
, ,+ M and O,.+1 , ,.M are on the same straight line , their ratio is scalar ; and since 0,0 ' , ,=-1.0,0 ' ] , we have , by an obvious extension of the notation , and by altering the order of the factors , [ 0 , ,2M- .
02,1 M ] [ 02 , 3 M03,2 MI ] ? . .
[ , I MO n , ,M ] =(l)n .
[ 01,0'2+ 20'1 ] [ o0203 030 ] O. .
[ 0o , 01 o 0o'1 ] = ( --1)n .
[ ( --1)n .
( --1)n. .
( n factors ) ... ( --1 ) " ] =(-1 ) .
( -l ) ( -1)n 1 ) .
( -.l1)-=(1)2n1 * .
9 .
Hence if all the roots are scalar , that is , if each side of the polygon cuts the scalar radical locus in as many points as its abstract equation has dimensions , we have the following fundamental proposition of the theory of transversals : " If we take the points at the extremities of a side of a polygon , and find the ratios of the distances of these points from the intersections of that side with a curve , and proceed in this manner regularly round the polygon in one direction , the product of all these ratios will be unity .
" 10 .
Thus if the triangle O0,0,0 cut a straight line in A , B , C so that these points lie on 0,03 , 0,0 , , 00,2 respectively , then OIC 02A 03B 02C 03A OltB .
If the same triangle 20,0,3 cut a conic in the points A , A2 by 0203 ; B1 , B2 by OO3 ; C , C2 by 0,0 , respectively , then 0,0 , o , c , O OAOA 0A B1 OoB , ,i O , C , ' , C2 , x ?
A. '2 , A , x 01 0 3B If the triangle touches the conic in A , B , C , then A , A2 coincide with A , &amp;c. , and the above equation becomes OiC 0A 03B _o 02C 03A 01B The first case shows that the three points lie on a straight line ; the second , that the three lines 01A , 02B , 03C meet in a point* .
If the point C were to lie at an infinite distance from 0 , and 02 , then since 0,1C==002+02C , and 1C =1+ ?
102 where the last GOC o2C term is infinitesimal , we have merely to consider 0C = I , or omit it from the productt .
Thus if we take as two sides of the triangle the asymptotes of an hyperbola meeting at O , and suppose the third side 0203 to cut the curve in A , and A2 , then since both 0102 and 0103 cut the curve at an infinite distance , the equation reduces to OA OA .
-2A1 02A2 = 1 ; 03A1 OA2 whence we readily find 03A2= O2A2 , or =-O2Alo The first value is impossible , because A2 lies between 02 and 03 ; the second gives the well-known property of the hyperbola{ .
11 .
As regards the clinant roots , or the intersections of the sides of the polygon with the clinant radical loci , the wording of the proposition ( 9 ) requires to be changed to adapt it to the more complicated form of the product ; and perhaps it may be considered sufficient to refer to the forms of the products in the final equation of ( 8 ) , and append a single simple example in which the product admits of being easily formed .
Ltta a0 Jol Let the triangle 00 03 touch a circle BC in the points B and C , and let the side 0203 be perpendicular to the line 010 drawn through the centre of the circle , 0 , but be wholly exterior to the circle .
Take O as the origin of coordinates , and suppose that the equations to the circle are OP=x .
OI+y .
OJ , x +y2=c2 .
Now change the origin to 0 ' , where 00'=-10302=-b .
OJ .
Then the equations to the circle become OtP=x .
OI+y'.OJ , x'2+ ( y'--6)2=2 .
And for the radical loci , when y ' is clinant , we have ( 3 ) O'R x. OI + r. OJ , r-b==0 , that is , the line 010 , and 0tS=Xr .
OI+S .
OJ , -2 _8s-2= C , that is , the rectangular hyperbola S'1 Sf2 .
The roots will then be 02R+i .
02St2 _ 02R+ RM'1 02 , M1 OJ J OJ and OR +i.O0S2 ' O0R + R+BM ' _ 0 , M ' OJ OJ -~ OJ ' and the factors corresponding to them in the final product in ( 3 ) will be 0M 03--O , M^ _ MM '3 MO '2 OJ OJ OJ O Similarly , by transferring the origin to 0 " , where 00"=b .
OJ , the equations to the circle become O"P=z .
01+ y " .
OJ , 2+ ( y " + 6)2=c2 .
The radical loci , when y " is clinant , will have for their equations O"R=x " .
OI+r " .
OJ , r"+b=O , that is , the line 010 , and O"S=x " .
01+s " .
OJ , " -2 s"2 =c , that is , the rectangular hyperbola S " , , S " , .
Hence the roots will be 03R +i .
03S " _ OR+ Rm n_ 03m ' OJ -OJ-OJ and O3R+i .
S " , OR+ Rm ' , 3 2m OJ 0oJ OJ- ' and the factors corresponding to them in the product ( 5 ) will be , after replacing OJ by JO , as there shown , 08020 3mo 0,30-0 _m m , l 0 , mI , 0 JO JO JO JO Hence the final equation in ( 8 ) gives O )2 II M 3Ml 20 B2=12.0 , C } 60 , m2 02 , m ) T 1 .
Now the second factor of the product on the left-hand side of this equation , is itself a product of which each factor =-1 , and it may consequently be omitted from the equation , so that OiC _+ O1B,2C 03B The under sign cannot be taken , because it would necessitate one of the points B or C lying without , and the other within the corresponding side of the triangle .
Hence we must have O , C _ O1B 0,0-O 0B ' that is , BC will be parallel to 0302 , a result which readily follows from other considerations .
12 .
The great complexity of the process , and the necessity of knowing the algebraical form of the equation to the curve , will probably prevent this employment of clinant roots from having any practical value beyond the completion of the theoretical harmony between the abstract and the concrete , the algebraical formula and the geometrical figure .

