III .
" On Scalar and Clinant Algebraical Coordinate Geometry , introducing a new and more general Theory of Analytical Geometry , including the received as a particular case , and explaining 'imaginary points , ' 'intersections , ' and 'lines .
" ' By ALEXANDER J. ELLIS , Esq. , B.A. , F.C.P.S. Communicated by ARCHIBALD SMITH , Esq. Received February 16 , 1860 .
( Abstract .
) Scalar Plane Geometry.-With 0 as a centre describe a circle with a radius equal to the unit of length .
Let OA , OB be any two of its unit radii , termed ( coordinate axes .
' From any point P in the plane AOB draw PM parallel to BO , so as to cut OA , produced either way if necessary , in M. Then there will exist some ' scalars ' ( 'real ' or 'possible quantities ' ) u , v such that GM=u-.GOA , and MP= v. OB , all lines being considered in respect both to magnitude and direction .
Hence GP , which is the 'appense ' or 'geometrical sum ' of OM and MP , or =GOM+MP , will =u.OA+v.OB .
By varying the values of the ' coordinate scalars ' u , v , P may be made to assume any position whatever on the plane of AOB .
The angle AOB may be takeni at pleasure , but greater symmetry is secured by choosilug GI auid OJ as coordinate axes , where IOJ is a right angle described in the right-handed direction .
If any number of lines OP , OQ , G)R , &amp;c. , be thus represented , the lengths of the lines PQ , QR , &amp;c. , and the sines and cosines of the aingles IOP , POQ , QOR , &amp;c. , can be immediately fturnished in terms of the unit of length and the coordinate scalars .
If P=X .
GI+y .
OJ , and any relation be assigned between the values of x and y , such as y=fx or p ( x , y)=0 , then the possible positions of P are limited to those in which for any scalar value of x there exists a corresponding scalar value of y. The ensemble of all such positions of P constitutes the 'locus ' of the two equations , viz. the 'concrete equatiol ' P=x .
GI ?
+y .
OJ , and the 'abstract equation ' y=f. .
The peculiarity of the present theory consists in the recognition of these two equations to a curve , of which the ordinary theory only fiurnishes the latter , and inefficiently replaces the former by some convention respecting the use of the letters , whereby the coordinates themselves are not made a part of the calculation .
A variation in either of these two equations will occasion a differeiice either in the form or position of their locus .
If the abstract equation be y= ax +6 , where a and b are any scalars , the concrete equation OP= x. GI+y.OJ becomes OP=x .
( GI+a .
OJ)+b .
OJ , which shows that OP is the appense of a constant line , and a line in a constant direction , and hence its extremity P must lie on a line in that direction drawn through the extremity of the constant line .
Also , since the length of OP is V ( x2+ y2 ) times the length of GI , the locus of the two equations OP=x .
0I1-y .
OJ and , v ?
2+y2=a must be a circle .
From these equations all the usual theory of the straight line and circle may be readily deduced , and all ambiguity respecting the representation of direction by the signs ( + ) and ( - ) may be removed .
Thus if the loci of OP=x .
OJ ?
y. OJ , y=ax+b , and OP'=OI .
OI+y'.OJ , ay'4 ?
x'=O intersect in the point Q , and OQ1 be the unit radius on the J side of 01 ( produced both ways ) , determined by the equation C/ ( 1+a2).0Q1=-a .
OI+OJ , then V ( 1+ a2 ) .
OQ will =b OQ .
And a line drawn from the locus of P parallel to OQ to pass through the point X , where OX= X. OI+ y. OJ , will be represented in magniitude and ditection by y , -ax,1-6 0 , V/ ( 1 +a2 ) *Q From this result the usual theory of anharmonic ratios is immediately deducible without any fresh ' convention ' respecting the signs ( + ) and ( - ) .
As every locus has two equations , each equation requires separate consideration .
The investigations colncerning the abstract equation remain nearly the same as in the usual theory .
When the abstract equation is given indirectly by the elimination of two constants between three equations , the result corresponds to the locus of the intersection of two curves varying according to a known law , ' coordinates proper , ' leading in its simplest forms , first , to Descartes ' original conception of curves generated by the intersection of straight lines moving according to a given law parallel to two given straight lines , and secondly , to Plucker 's ' point coordinates .
' The true relation of Plucker 's 'line coordinates ' to the ordinary system is immediately apparent on comparing the two sets of equations : ( 1 ) Concrete OP=x .
OI+y .
OJ Abstract F ( x , y ) = 0 , ( 2 ) First abstract y=ax ?
+b Secolid abstract F ( a , 6)-=( .
The second set of equations determines a curve by supposing a and b to vary , then eliminating a and b , and referring the ultimate abstract 2G M. equation to the concrete equation ( 1 ) .
On comparing these two sets of equations , we see that x and y in the first are involved in precisely the same manner as a and 6 in the second , so that if any equation F ( u , v ) =0 were given , it might determine one or the other by precisely the same algebra , according as x , y or a , 6 were substituted for u , v. Whence flow Plucker 's theories of collineation and reciprocity .
The investigations respecting the concrete equation , on the other hand , are altogether new .
The most general form of the two equations , is Concrete , OR1=f1(x , y ) .
OA +f , ( x , y ) .
OB Abstract , p(x , y)=O , which will clearly determine a curve as definitely as before .
If in lieu of the abstract equation p ( x , y ) = 0 , we were given the locus of the two equations OM=x .
OA+y .
OB , q(x , y)=O , and from any point M in this curve we drew MN parallel to OB , cutting OA produced in N , so that ON=x .
OA , NM=y .
OB , we could find x and y from this curve , and consequently form OL=f1(x , y ) .
OA , and LR=f(x , y ) .
OB , and by this means determine the point R , where OR=OL+LR , in the locus of the general equations , corresponding to the point M in the particular locus .
The general form is therefore the algebraical expression of a curve formed from another curve by means of operations performed on the coordinates of the points in the latter , as when an ellipse is formed from a circle by altering the ordinates in a constant ratio .
The algebraical treatment of this case consists in putting p=f2 , Y ) , q=f2(x , Y ) , and between these equations and 0 ( x , y ) = 0 , eliminating x and y , to find 4 , ( p , q ) =0 .
The locus is then reduced to that of OR-=p .
OI+q .
OJ , at ( p q ) =0 , which is the ordinary simple case .
But the whole of this latter locus does not in all cases correspond to the locus of the general equations , because not only x and y , but also p or/ i ( x , y ) , and q or 2(x , y ) must be all scalar .
Thus , in the case of the parabola derived from a circle , by substituting for the ordinate of a point in the latter its distance from a known point in the circumference of the same , we know that it is impossible to derive more of the parabola than can be obtained by taking the diameter of the ' cirele as the ordinate .
The algebraical process gives first ( 1 ) OR=X .
OI+ ( X2+y2).OJ , y2+x2=2ax ; and then , putting p=x , q= +V y2 ' ) we find ( 2 ) OR=p .
OI+q .
OJ , q2= 2ap .
In this case q is always positive and =+/ ( 2ax ) .
But x , and therefore p , always lies between the limits 0 and 2a , and hence q must lie between the same limits .
Consequently the only part of the curve ( 2 ) represented by the equations ( 1 ) is the semi-parabola contained between the origin and the ordinate 2a .
OJ .
This general form of the concrete equation , therefore , furnishes an elementary method of representing curves or parts of curves .
Thus OR=(h+a cos a+x cos a).OI+(k(+a sin a+ sin a)).OJ X2 + y2= a2,1 are the equations of a line HK=2a ( cos a. OI+sin a. OJ ) , and having one of its extremities determined by the equation OH=h .
0I+k .
OJ , so that its length is 2a times that of 01 and the angle ( 01 , HK)=a .
To determine the intersections of two such finite curves , given by the equations OP=f1(X , y ) * OI+f2(X , y ) .
OJ , q(4X y)= and OP'-=f( Y/ y ) 0* I+f2'(X ' , y ' ) .
OJ , 0'(X ' , y')=0 , we have OP=OP- , and consequently f1=-fI and f2 =f2f , which , with p=0 and 0'=0 , give four equations to determine x , y , XI , y ' .
The curves will , however , not intersect , unless not only four such scalars exist , but they make f , f2 , f/ , f2f all scalar .
The transformation of coordinates may now be investigated more generally in the form of the two problems : 'given a change in the concrete ( or abstract ) equation , to find the corresponding change in the abstract ( or concrete ) equation respectively , in order that the locus may remain unaltered .
' The ordinary theory only comprehends an exceedingly simple instance of the first problem .
The second is indeterminate so far as the representation of portions of curves is concerned , so that any abstract equation may , by the help of a properly selected concrete equation , represent any curve whatever .
A curve by which the scalar value of y is exhibited corresponding to any scalar value of x in the equation O(x , y ) =0 , and which in this simple case is furnished by the locus of the two equations OP= .
OX+y .
OJ , 0(w , y)=0 , is termed the ' scalar radical locus ' of the abstract equation p(x , y)=O , and corresponds to what has been hitherto insufficiently designated as the ' locus of the equation ' 9 ( x , y ) = 0 .
It presents a necessarily imperfect image of that equation .
Clinant Plane Geomnetry.-Reverting to the original pair of equations OP==x .
O+-y .
OJ , O(4 , y)=O , and remembering that even if clinaiit ( 'impossible ' or ' imaginary ' ) values were substituted for x and y in the expressions x. 01 , y. OJ , they would still represent definite lines ( see abstract of Paper on 'Laws of Operation , &amp;c. , ' Proceedings , vol. x. p. 85 ) , and conisequently the line OP would still be perfectly determined , we see that the limitation of x and y to scalar values in the previous investigations was merely a matter of convenience .
We may therefore give x any clinant value , and after determining the correspondent elinant value of y from q(x , ' y ) =0 ( which will always exist if the equation is algebraical ) , substitute these values in the concrete equation , and thus find OP , and consequently the locus of the equations .
We may observe , however , that as a clinant iilvolves two scalars , we must have some relation given betweeni them , directly or indirectly , in order that there may be only one real variable , without which limitation the locus would in every case embrace the whole plane .
The general algebraical process is as follows , the Roman letter i being used for +/ ( 1 ) .
Let OP=P(X1 , X2 .
0O , where X1 =p +i q1 , i X ?
, =p+i * q , , , and all the p , q are scalar .
We can reduce this expression to OP= ( P1+i .
P2 ) .
01 , where P , and P2 are scalar 'formations ' ( or 'functions ' ) of the 2n scalars p1. .
pp , ql .
q , .
As this is equivalent to OP=P , .
OI+ P2 OJ , it is precisely the same as the general scalar concrete equation lately investigated .
But one abstract equation will now no longer suffice ; for if we put P1=r , P2=y , we must have 2n-1 additional equations , in order ultimately to find f(x , y ) =0 by eliminating 2n variables between 2n +1 equations .
The result , OP= O ?
+y .
OJ , f(x , y)=0 , with the conditions of scalarity , will then enable us to determine the locus by the usual process .
If there be only two clinants , X=_p + i. q , and Y=r+ i. s , and we have given OP=P .
( X , Y ) .
0I ; C. ( X , Y)=0 , where C=0 may be termed the ' curve equation , ' these reduce to OP=P1.OI+P2.OJ , Ci = 0 , C2=0 , where P1 , P2 , C1 , C2 are formations of p , q , r , s , so that , on putting P1-= , P2=y , we have only four equations , between which we cannot eliminate p , q , r , s. This again shows the necessity of some additional relation , A(p , q , r , s ) =0 , which may be called the I assignant equation , ' in order finally to discover ( x , y ) =0 , and thus determine the locus .
The only case ordinarily considered is where X is scalar .
This corresponds to putting q=0 for the assignant equation .
Hence q disappears and we have OP=P1 .
0+P2 .
OJ , C , =0 , C2=0 , where PV , P2 C ) , C2 are formations of p , r , s , and hence , putting P1= x , P2=y , we immediately eliminate P , r , s , and determine the locus .
This general theory is illustrated by numerous examples , and in particular Pluckier 's 'involutions ' by means of 'imaginary lines ' are fully explained by help of the really existent lines of this theory .
The general theory of the intersection of two 'clinant loci ' is precisely analogous to that of two scalar loci with general concrete equations .
In the particular case where the concrete equations are the same for both , or the reduced equations are OP= [ P1(p , q , r , s)+ i. P2(p , q , r , s)]j 01 , C1(p , q , , S ) 0 , C2(p , q , r , 8 ) = 0 , A(p , q , ax , s)_)0 , and OP'= [ Pl(p ' , q ' , r ' , s')+i .
P2(P ' , q ' , r ' , St ) ] .
01 , C1 ( p_ 2t , _ 7'_ 8s)=0 , CA2(pF , q ' , rs ) ' 8=)0 , Al(p/ , q rf , sl=0 , the initersection will evidently be determiined by putting p=p ' , q-= q , r=r t , =8 which will give six simultaneous equations between four variables , p , q , r , s or pl , q ' , r ' , st. This gives two equations of condition .
If , however , no assignant equations were given , we might determinie the values of p , q , r , s from the four reduced curve equations , and then assume any assignant equations compatible with these solutions .
This is more readily done by determining X and Y from the two unreduiced curve equations C(X , Y)=O , C , ( X , Y)=O .
The process then corresponds to that for the simplest scalar case of intersection .
If the values of X and Y prove to be scalar , then the assignant equations are q=0 and s=0 , and we have an ordinary scalar case of intersection .
But if this is not the case , and we find X=al+i .
b , ... .
, + i.6b " ; Y=e , +i .
d , ... .
qi .
d , we may take ( q bl ) ... ( q b ) _ 0 , ( s -dl ) ... ( 8 dn)=Q , amnong others , as assignant equations and determine the corresponding loci .
These loci will be found to intersect in all the ( perfectly real ) points determined by the values of X and Y , but not necessarily in these only .
Such points will of course not belong to the curves derived from putting q=O , s=0 , and hence cannot in any sense be called points of intersection of these curves , although they have hitherto been termed ' imaginary points of intersection .
' The discovery of equations to loci described according to some geometrical law , furnishes a convernient illustration of this clinant theory .
From any point 0 , draw radii vectores OU , OR to any curves , and make RP=h A OU , where A is scalar .
Put OP=(p+i q ) .
01 , OR=(r ?
i s ) .
01 , OU=(u+i v).01 .
Then the condition OP= OR+ RP gives p= r+ hue , q= ?
+hv , which correspond to the two reduced curve eq'uations .
We now require three more equations in order to eliminate four of the six scalars p , q , r , s , u , v , and find a relation between the two remnaining scalars , sufficient with one of the above concrete equations to determine the locus of one of the points P , R , U. Two of these three equations will amiount to assigning the locus of two of these points , and the third equLation will amount to assigning somxie i'elation between the angular motions of OU and OR , without which the loci can clearly not be deterninied .
This third condition is frequently givenl in the form of requiring M ( a point in RP , produced either way if necessary where RM=k .
OU and k is scalar ) to lie on a given curve .
It is then most convenient to introduce two new scalars m and n , so that OM= ( m+i .
n ) .
OI , when the condition OM=OR+ RM , gives the two additional equations m=r+ ku , n=s+ kv , which with the five others will serve to eliminate six of the eight scalars , m , n , p , q , r , s , u , v , and leave the required abstract equation between the remaining two .
The eliminations are very simple in a great variety of curves .
This theory is fully illustrated by examples .
The first problem in the transformation of coordinates , ' given an alteration in the concrete and curve equations , to find the corresponding alteration in the assignant equation , so that the curves may remain identical , extent excepted , ' is solved thus .
Given OP= ( P1+i * P2 ) blem , 'given an alteration in the curve and assignant equations , to find that in the concrete equation , ' admits of a similar solution .
Given OP=.0O+y.OJ , 0(Q , y)++(Zy)= ) , as the equations to the locus , and also C1(p , q , r , s)=O , C2(p , q , r , s)=O , A-q'(p , q , r , )+4'(p , q , r , as the new curve and assignant equations : put = 0 ' , 4= 4/ , and determine xc= ( p , q , r , 8 ) , Y-=7(p , q , r , s ) , and then use OP=(4+i .
II ) .
01 as the new concrete equation .
The result is independent of the form of the curve equations .
The geometrical significance of these transformations is that there are no 'families ' of plane curves .
'Clinant radical loci , ' or curves which furnish a sensible geometrical picture of the relations of the corresponding clinants which satisfy any abstract equation , may be obtained thus .
Let the given equation be C ( X , Y ) = 0 .
This may be regarded as two reduced curve equations , C , ( p , q , r)0 , 0C2 ( p , q , -O The values of X are assumed by drawing radii vetorees to points in any curve , and the corresponding values of Y have to be pictured .
We must therefore have some equation A-0 , which in combination with the other two will give the curve by which X is thus determined .
Eliminating the scalars , p , q , r , s , two and two , between these three equations , we obtain the following six , of which the first and last , and one of the intermediate ones , are in general only required : f ( p , q)=0 , ( p , r ) 0 , f3 ( p,8 ) =0 , 4 ( q , r)=0 , f5( q,8 ) =0 , f2 ( r,8 ) =s0 .
We now construct the curves containing the points X , S , Y , as the loci of the equations OX=OP+PX =(p+i .
q ) .
O , f1(p , q)=O , OS =OP+PS =(p+i .
8 ) .
0I , f ( p , S)-= , OY=OR+RY=OR+PS=(r+i .
) .
01 , f6 ( r , 8)=0 .
Set off OP at pleasure on OI ( produced both ways ) , and draw the ordinate PXS , cutting the two first curves in X and S. Through S draw SY parallel to 01 , cutting the third locus in Y : then if OX=X .
01 , and OY=Y .
01 , X and Y are corresponding solutions of C ( X , Y)=0 .
When only scalar solutions are required , the abstract equations to the first and last curve become q=0 , s=0 , so that both of these curves coincide with 01 produced both ways , and the intermediate or connecting curve is the ordinary scalar radical locus .
Scalar Solid Geometry.-The same theories apply with proper modifications .
The concrete equation OP=x .
OI+y .
OJ+z .
OK , with one abstract equation , f ( x , y , z ) = 0 , gives a surface for its locus , and with two abstract equations , fi ( V , Y , y)=0 , f2(x , Y , z)=0 , a curve .
From these equations all the ordinary theories are most readily deduced .
We may also take the concrete equation more generally in the form OP=f1 ( x , y , z ) .
0I +f , ( Q , Y , Z ) .
OJ +f3 ( x , y , z ) .
OK , the abstract equation being F ( x , y , z ) =0 .
We obtain the surface by putting x1 =f , y1 =f,2 ' z =f , , and eliminating x , y , z between these equations and F= 0 , thus finding 0 ( x1 , y , zl ) = 0 .
The locus of ?
P=Z1 . ?
y1 . ?
J+Z , *OK , 0 ( x , y1 , ,z , )=0 must be limited by the condition that not only x , y , z , but also fi ( x , y , z ) f2 ( X , Y , z ) , f ( XI y , Z ) must all be scalar .
If three variable parameters a , 6 , c are introduced , we require in addition three equations of condition , F , ,=0 , F2=O , F , =0 , between x , y , z , a , 6 , e , in order to eliminate all six and find 0 ( xl , y , , z , ) = 0 .
Clinant Solid Geometry.-Some precaution is now necessary to indicate the plane on which the quadrantal rotation symbolized by i is to take place .
This is effected by introducing all three coordinate axes into the concrete equations .
Let the abstract equation be F ( X , Y , Z)=0 .
Then , supposing X=p+iq , Y=r+is , Z=u+iv , this equation reduces to F , =0 , F , =0 , where F , and F , are formations of the six scalars p , q , r , s , u , v. We now require two assignant equations to determine the curves on which OX and OY are to be taken .
These may be given in the form of the single clinant equation A ( X , Y , Z ) = 0 , which reduces to the two scalar equations A1=O , A2=O .
Now take OR=f1(X , Y , Z ) .
OI+f2 ( X , Y , Z ) .
OJ .
Then , since i. 01 OJ , because the assumption of the two axes determines the rotation to be in the plane IOJ , we can reduce this to OR= R1 .
OI+R2 .
OJ , where R1 , R2 are formations of the six scalars .
By virtue of the equations A1= 0 , A2=0 , F1= 0 , F2=0 , which will give q , s , u , v in terms of p , r , the line OR is perfectly determined by the assumption of p and r. Next take V/ ( R12 + R22 ) .
OR1 = OR , so that OR1 is a unit radius in the direction OR .
Put OP=f3 ( X , Y , Z ) .
OR1 ?
+4(X , Y , Z ) .
OK , which reduces to OP=R3 .
OR1 , +P3 .
OK , where R3 , P3 are formations of the six scalars , because i. OR1 =OK on the plane R , OK .
The locus of P will now manifestly be a surface , the concrete equation of which becomes of the usual form on putting for OR1 its value .
Determine P1 , P2 by the equations P1 .
V(R12 + R22 ) _= R,3 , P2 .
V(R12+Y)=R2 Rt , and we find OP=P1 .
01+P2 .
OJ+P3 .
OK .
Putting x= P1 , y= P , z= P3 , and eliminating the six scalars between these three equations and A , = 0 , A2 = 0 , F1 = 0 , F2=0 , the locus becomes that of OP=x .
OI+y .
OJ+z .
OK , 0 ( x , y , z)=0 , which is limited by the conditions of scalarity .
After illustrating this theory by an example , the theory of intersection , 'when the elimination gives clinant values to determine the points , is discussed .
The general theory is further illustrated by the determination of equations to loci , leading to very simple and extremely general modes of finding families of surfaces .
The problems of the transformation of coordinates and of radical loci are shown to be precisely analogous to those of plane curves .
It will be evident that these investigations merely open out a new field for algebraical geometry , of which it is impossible to foresee the extent .

