]\gt ; [ 663 ] .
On the of Loci .
W. K. CLIFFORD , F.R.S. , ofessor of Applied Mathematics in Univcrsity College , London .
Received Apri18 , \mdash ; Read Ma .
, 1878 .
PART I. CURVES .
By a curve we mean a continuous one-dimensional aggregate of sort of elements , and therefore not merely a curve in the ordinary geometrical sense , but also a singly infinite system of curves , surfaces , complexes , \amp ; c. , such that condition is sufficient to determine a finite number of them .
The elements may be regarded as determined by coordinates ; and then , if these be connected by equations of any order , the curve is either the whole aggregate of common solutions of these equations , or , when this breaks up into ] distinct parts , the curve is one of these parts .
It is thus convenient to employ still further the language of geometry , and to speak of such a curve as the complete or partial intersection of loci in flat space of dimensions , or , as we shall sometimes say , in a -flat .
If a certain lmber , say , of the equations are linear , it is evidently possible by a linear transformation to make these equations equate of the coordinates to zero ; it is then ient to leave these coordinates out of consideration altogether , and only to regard the remnining equations between coordinates .
In this case the curve will , therefore , be regarded as a curve in flat space of dimensions .
And , in general , when we speak of a curve as in flat space of dimensions , we mean that it cannot exist in flat space of dimensions .
The whole aggregate of linear complexes may be regarded as constituting a space of dimensions , in which the special complexes , or aight lines , constitute quadric locus .
A ruled surface , or scroll , will be thus regarded as a curve in a quadric locus in a flat space of five dimensions .
If , however , the generators of the all belong to the same linear complex , the scroll must be regal.ded as a curve 1 } ing in a quadric locus in a flat space of dimensions .
And if , further , the scroll has two linear directrices , so that the generators belong to a linear congruence , then the scroll may be regarded as a curve lying on an ordinary quadric surface in three dimensions .
Thus , for example , quartic scrolls having two linear directrices COl.respond either to quadri-quadric curves of deficiency 1 ( that is , they are elliptic curves whose coordinates may be expressed as elliptic functions of one variable ) , to the curves of deficiellcy 0 which are the partial intersections of a quadric a cubic surface ( that is , they unicursal curves ) .
This view of ruled surfaces is made excellent use of by Voss , " " Sir Theory der windschiefen Flachen Math. Annalen , vol. viii .
p. 54 .
Similar rations apply to surfaces .
By a surface we shall mean , in general , continuous two-dimensional aggregate ( which may also be called a two-spread or twolocus ) of any elements whatever , curves , surfaces , complexes , \amp ; c. , defined by the whole or a portion of the system of solutions of equations among coordinates .
We shall assume that none of these equations are linear , and then speak of the surface as in a flat space of dimensions .
We shall in certain cases go further , and speak of an -spread or -way locus , viz. , a locus determined by the whole or an algebraically separate portion of the system of solutions of equations among coordinates ; if none of these equations are linear , the -way locus will be said to be in dimensions .
The general point of view is that of Professor CAYLEY , " " On the Curves which satisfy given Conditions Phil. Trans. , Vol. 158 ( 1868 ) , pp. 75-144 ; the methods of tion are those of Dr. ALMON , 'Solid Geometry , ' p. 261 .
Theorem A. Every proper of the order is in flat space of dimensions or less .
Fcr through points of it we can draw a flat space of dimensions , which must therefore contain the culve , since it meets it in a number of points greater than its order .
Thus , for example , there is no curve of the second order , in space of any number of dimensions , except a plane conic .
If , therefore , a system of curves , in a plane or on any surface , is such that two curves of the system drawn through an arbitrary point , then the coordinates of a varying curve of the system may be represented by , and the envelope of the system is , in the case of plane curves , a curve having the equation , where are three curves of the system ; in the case of curves on a surface , it is the intersection of the surface with another having an equation of that form .
Professor HENRICI has kindly written for me the notes in elucidation of this argument:\mdash ; ' ' In the first sentence of the paper it is stated that by a curve is meant any one-dimensional aggregate of any sort of elements .
The definitions given are algebraical , but the reasoning later on becomes more and more geometrical .
" " In this note the connexion between the algebraical definition and the geometrical reasoning will be shown in the case where the elements are plane curves of order ' ' If we suppose a cur ve given by its equation in point coordinates we may take the coefficients as homogeneous coordinates of the curve .
" " As there are ratios of these coefficientst , it follows that all curves of order in , plane constitute a spread , and this will be a flat spread as no relation has been supposed between the coordinates .
' ' To determine in this spread a k-flat , , we have to assume a sufficient number of equations between the coordinates , or denoting by , . .
curves of order we may write down the equation of PROFESSOR W. K. CLIFFORD ON THE CLASSIFICATION OF LOCI .
To particularise still further , a system of conics having the characteristic must always have quadruple contact with a quartic curve ; and the different species may be enumerated by studying the successive degeneration of the curve , ending with the fundamental system , when it breaks up into four straight lines .
So again , there is no quadric scroll , in any number of dimensions , except the ordinary quadric surface which is in flat space of three dimensions .
A curve of the third order must be either the known skew cubic in three dimensions , or a plane cubic .
Hence , if a system of curves be such that three of them can be drawn through an arbitrary point , the equation of any curve of the system is of one of the two forms\mdash ; Wsnu cn dn where are parameters .
Hence it is easy to write down the equations to the envelopes in the two cases , and to enumerate the distinct species .
one element in the -flat in the form .
, and take the ratios of the as the coordinates of a variable curve .
" " For we get a net as the flat space of two dimensions or as the plane in this space , and for a pencil corresponding to the ' ' If , on the other hand , we assume in the k-flat equations between the coordinates , there remains a singly infinite number of curves , that is according to Professor CLIFFORD a curve ( with curves as elements ) , according to the usual nomenclature a series of curves .
' ' To determine the order of this curve we have to find the number of elements on it which satisfy a linear relation between the coordinates .
our case the condition that a curve shall pass through a given point gives such a relation , and the number of curves through a point is the order in question .
" " Hence , if we wish to extend a theorem relating to a curve ( in the ordinary sense with points as elements , but in any number of dimensions ) to a proposition relating to a series of curves , or if we wish to illustrate in a plane a theorem relating to a curve in more dimensions than three , we ve instead of a point on the curve to take a curve in the series , and to replace the order of the curve by the index of the series .
' ' The theorem that every curve of order two is a plane curve becomes thus\mdash ; the curves in a series of index 2 belong to a net .
' ' Further , the coordinates of a point on a conic may be represented as expressions of the second degree in a variable parameter , say by ; where , 2 , 3 , if the coordinates are taken in the plane of the conic , but if they are taken in space we have to take , 2 , , 4 , and so on for more dimensions .
The locus of these points , that is , the comic , is then given by an equation of the form where are three of the points .
" " If we apply this to our series we obtain the results stated in the text , viz. , the coordinates of any curve of a series such that two curves pass through a given point are of the form quoted , and the equation of the envelope is of the form being three of the curves .
" " Similarly , if the series is such that three pass through any point , then the series may be considered as 'curve ' of order three , and the statements made in the text follow at once from the known properties about cubic curves , which are either unicursal ( twisted , or plane nodal , cubics ) or they are plane curves of deficiency one \mdash ; January , 1879 .
] A cubic scroll must be of the nature of the skew cubic , because it is a curve with plexes for elements ) which is obliged to lie on a quadric locus ( that of the special complexes , or straight lines ) .
Theorem B. A curve of order in flat of dirnensions ( and no less ) be repres for point , curve of order in plane .
The proposition is obvious when .
The cone standing on a curve of order ( in ordinary space of three dimensions ) , and having its vertex at a point of the curve , is of order ; if then we cut this cone by a plane , we have the tortuous curve represented , point for poin , on a plane curve of order Now this process is applicable in general .
Starting with an arbitrary poin , of a curve in any number of dimensions , let usjoin this point to the other points of the curVe ; we shall thus get a cone of order .
For any flat locus of dimensions drawn through the point must meet the curve in points , of which is one ; and therefore it must meet the cone in lines .
Hence , if we cut this cone by such a flat locus not passing through , we shall get a curve of order in flat space of dimensions , which is a representation of the original curve .
By continuing this process we may go on diminishing the order of the curve and the number of dimensions by equal quantities , until we have subtracted fro1n each ; when we are left with a curve of order in a plane .
The reduction may , howeyer , be effected in one step .
A flat locus may be drawn throngh arbitrary polnts .
Suppose to contaln points of the curve at , and altother variable point , , of the curve .
Such a locus will meet an arbitrary plane in one point , R. As then moves about on the curve , will trace out on the plane a curve which corresponds to l-t , for point .
But this curve is of order , for a flat locus , passing through consecu- tive points of the original curve at , will meet that curve in other points , and therefore will meet also the locus of in points .
This locus is , therefore , of order , as was to be proved .
The fixed points through which the variable locus passes need not all be united at , but they may be any arbitrary points on the curye .
We will now consider some examples of this remark .
1 .
Unicursal curve of order in -dimensional space .
A curve of order flat of ( no less ) is We may prove this independently by considering a variable flat which passes fixed points on the curve .
Its equation will of the form where is a variable eter , and it will meet the curve in one other point , which is thus associated with a value of The ltions to such a may always be written in the form\mdash ; PROFESSOR W. K. CLIFFORD ON THE CLASSIFICATION OF LOCI .
where the , C. .
are linear functions of the coordinates , and the number of columns is .
For the homogeneous coordinates are proportional to rational integral functions of of the order .
Solving these equations for 1 , we find 1 , . .
, C. .
which is equivalent to the system written down above .
The more general system of equations- . .
( 2 ) where the A. . . .
are linear functions as before , may always and easily be reduced to the former , for they are got by eliminating from the equations .
, ( 3 ) We may , however , solve these equations for the ratios of the ates , which will thus be expressed as rational functions of of the order .
Solving these for 1 , . .
we come back to the previous system .
The equations ( 3 ) exhibit the curve.as the locus of the intersection of corresponding elements in projective pencils .
The equation to the flat which passes through the points whose parameters are , is easily seen to be \mdash ; . .
But this equation is manifestly divisible by the coefficient of , which is the product of the differences of all the .
If we write\mdash ; etc. then the equation is .
( 4 ) If we omit the suffixes of the in this formula we obtain the equation to the osculant flat at the point .
Namely ( beginning at the other end ) , it is\mdash ; ( 5 ) and we see at once that the class of such a curve is equal to its order .
We thus obtain a very useful representation ( Abbildung ) of the points of the -dimensional space by means of groups of points on such a unicursal curve , namely , each poin in the space is represented by the points of contact of the osculant flats which pass through it .
The use of such a representation of threedimensional space by means of a skew cubic was pointed out by Dr. HIRST , and the corresponding representation of a plane by means of a conic has been used by M. DARBOUX ( ' Sir un class remarquable de courbes et de surfaces Paris , 1873 , Note II .
, p. and by me ( ' ' On the Transformation of Elliptic Functions Proc. Lond. Math Soc. , vol. .
( 1875 ) , pp. 25-38 and 225-233 ) .
It may be wcrth while to mention that an extension to all space of the theory of the in-and- circumscribed polygon may be obtained by this means .
A curve of this kind determines also a dualistic correspondence in the space of dimensions .
Through every point may be drawn osculant flats , and through their points of contact another flat , hich shall be called the polar of the point .
If the point moves along a straight line its polar will pass through a fixed flat , the polar of the line .
And generally if the point lies in any flat the polar will pass through a fixed flat .
When we have the ordinary system of polar reciprocals in regard to a plane conic .
When we have that system in regard to a skew cubic which is described by , ' CreIIe , ' vol. lxv .
p. 39 .
These two systems are typical respectively of the cases in which is even and odd .
When is even , the relation between the coordinates of two points , which that each lies in the polar of the other , is a symmetrical one ; consequently those points which lie in their own polars are points on a certain quadric locus , and the system is merely that of the poles and polars in regard to this quadric locus upon which the curve lies .
The equation to this locus is at once obtained by equating to zero the quadrinvariant of the form which occurs in the equation ( 5 ) of the osculant flat , namely , it is -etc .
( 6 ) To prove this , observe that if in the equation ( 5 ) we substitute the coordinates of any point , the values of which satisfy the are the parameters of the points of contact of the osculant flats which pass through the point .
If be these values , the equation ( 4 ) represents the flat which passes through the points of contact , that is to say , the polar of the point .
Now if we denote by , .
the results of substituting the ates of the point iu , then we shall have\mdash ; PROFESSOR W. K. CLIFFORD ON THE CLASSIFICATION OF LOCI .
so that , when is even the equation of the polar is\mdash ; \fnof ; m \mdash ; etc. ( 8 ) that is , it is simply the polar of the point in regard to the quadric ( 6 ) .
It is to obseryed that the quadric is completely determined when the curve is given .
I reserve the question of the conditions to which the curve is subject when the quadric locus is giyen , or , say , the discussion of the problem to represent the relation of poles sJld polars in regard to a quadric locus ( in space of an eyen number of dimensions ) by means of a unicursal curve .
But when is odd , the last term of equation ( 4 ) is negative , and the equation of the polar is\mdash ; \mdash ; etc. ( 9 ) that is , it is skew symmetrical , and every point lies upon its .
It is convenient to use the term co-flat for points , which are in the same flat ; with this nomenclature we may say that when is odd every point is the points of contact of the , which can be through it .
This will be recognised as an extension of the property of a skew cubic , that point in space is co-planar with the points of contact of the three osculating planes which can be drawn through it .
A case of this skew symmetrical relation is given by any arbitrary state of motion of the whole space as a rigid body , the relation between two points being that the line joining them moves perpendicularly to itself .
The polar of any point is an flat drawn through it perpendicular to the direction of its motion .
When is even there is always one point which remains at rest , and all the polars pass through this point .
Thus the general motion of a solid in an even number of dimenslons always depends in this simple way the motion in one dimension less .
In an odd number of dimensions , however , every point moves in the general case ; but if any point is at rest , then all the points in a certain straight line are at rest .
Besides its order and class , a curve has , in general , characteristic numbers intermediate to these , which may be called its first rank , second rank , etc. The first rank is the order of the locus traced out by straight lines through two consecutive points of the curve ; the second rank , of that traced out by planes through three consecutiye points ; and generally the rank is the order of the wide locus traced out by k-flats tbrough consecutive points .
For the curve just consldered these numbers are 2 , 3 , .
; it is convenient to derlve them from the corresponding numbers for its projection , the unicursal curve of order in dimensions , to which we now proceed .
2 .
Unicursal curve of order in \mdash ; 1 dimensions .
Every curve of order in flat space of is either unicursal or .
For it may be represented point-for-point on a plane cubic .
We shall treat these two cases in succession .
They are exemplified by the two species of quartics in ordinary tri-dimensional space .
The coordinates of a point on the unicursal curve are proportionaL to rational integral functions of a parameter .
This representation may be simplified in a manner due to ROSANES , 'Crelle , ' vol. lxxv .
p. 166 .
We have binary quantics of order ; now these may be linearly combined in different ways so as to produce a perfect power .
Hence the original quantics may be expressed each as a linear function of the powers of the same linear quantics .
Thus , for example , three binary cubics may be simultaneously reduced to the forms .
where identically .
If the of a point in a plane are respectively proportional to these cubics , we may , by solving the equations for , obtain three linear functionoe of , he coordinates , which are respectively proportional to .
Transformin them to the new triangle whose sides are we must have the equation of a unicursal cubic expressed in the form It is clear that the lines are tangents at the three points of inflexion .
In general , let the quantics be . . . . . . . .
then the linear quantics .
are the factors of PROFESSOR W. K. CLIFFORD ON THE CLASSIFICATION OF LOCI .
Since there are identical relations between quantics , the equations of the unicursal curve may be written in the form ' it is evident that the equations represent stationary osculant flats , that is to say , flats which pass through consecutive points of the curve .
The properties of this curve may be very conveniently studied by regarding it as a projection of the curve considered in the last section .
If all the points of that curve be joined to a point , not situated upon it , the joining lines will form a cone of order ; and on cutting this cone by an flat we shall obtain the curve now under discussion .
The points of superosculation , whose existence has just been proved , are then clearly the projections of the points of contact of osculant fiats to the full-skew curve drawn through the point O. It follows that is odd , these points of are on the same .
but when is even this is not the case , unless the point lies on the quadric locus a , ssociated with the full-skew curve , in which case we have a special variety of the projection .
Thus the three points of inflexion of a nodal cubic are in one straight line ; but a unicursal skew quartic in ordinary space has not in general the property that the points of contact of four stationary osculating planes are in one plane .
The property established above for the full-skew curve shows that this will be the case if the four points form an equianharmonic system , or if the quadrinvariant of the quartic which determines them is equal to zero .
And generally when is even , the points of superosculation will be co-flat if , and only if , the quantic in which determines them has its quadrinvarlant zero .
By using the values of the coordinates of a variable point of the curve expressed terms of a parameter , we may obtain an expression of this quadrinvariant and also of its product by the in terms of the roots of the quantic .
Let be the values of which belong to the points of superosculation , and the coordinates of a variable point on the curve .
Then we may write , 2 , . .
and the coordinates of the point are , . . .
If for shortness we write instead of , then the condition that the points shall be co-flat is This is obviously always satisfied if is odd , for then the determinant is skew symmetrical , and being of odd order it necessarily vanishes .
If , however , is even , the determinant is a symmetrical function of the roots which vanishes when any two of them are equal ; and consequently it must contain as a factor the product of the squares of their differences .
Now the determinant is of the order in each root , and the discriminant is of order ; therefore the remaining factor is of order 2 in each root , and being a symmetrical invariant must be a function of the squares of their differences .
It can therefore be no other than 2 ; this is , to a factor equal quadrinvariant form .
The equation to the flat passing through two consecutive points of the curve at , and through other points , is clearly where the are current coordinates , and the determinant is expressed in umbral notation .
Writing in this for , and for , we may observe that the determinant so that the equation is of order in .
It thence follows that different flats may be drawn through arbitrary points to touch the curve ; or that the developable traced out by the tangent lines is of the order .
Similarly , from the value of the determinant which is equal to the product of the differences of , . .
multiplied by we may conclude that the number of flats which can be drawn through conPROFESSOR W. K. CLIFFORD ON THE CLASSIFICATION OF LOCI .
secutive points of the curve and through other arbitrary points is ; or that the -wide locus which is traced out by flats passing through consecutive points is of the order .
For the equation of an flat passing through consecutive points is clearly where we must substitute for the , etc. , the descending powers of beginning at the .
Making equal to we obtain the equation of the osculant flat at any point of the curve ; it is where product of the differences of all the except .
Thus the class of the curve is .
3 .
curve of order in n\mdash ; k dimensions .
The characteristic numbers belonging to this curve may at once be obtained by regarding it as a projection of the full-skew curve .
The number of ranks is and the numerical values of them are respectively 2 , 3 , .
; the class is ) and the number of points of superosculation is .
For example , the unicursal quintic in three dimensions is of rank , and of class , and it has superosculant planes .
Convenient forms of the equations may be got by eliminating some of the variables the equations of the full-skew curve ; but care must be taken to select these variables so that the resulting system is sufficiently general .
4 .
Elliptic curve of order in dimensions .
We have proved already that a curve of order in dimensions can be represented , point for point , on a plane cubic .
If , therefore , it is not unlcursal , its coordinates can be expressed in terms of elliptic functions of a single parameter .
Now , it follows from the investigations of CLEBSCH , 'Crelle , ' vol. lxiv .
( 1864 ) , pp. 210-270 , that if points of the curve are co-flat , the sum of their parameters will differ from a certain constant by a sum of integer multiples of the two periods of the elliptic function .
Let the periods be and , then if are the parameters of the po nts , where is a constant , and are integers .
To find the points of superosculation , we must suppose the points to become identical , or the , still satisfying this equation , to become equal .
We thus obtain and values of , representing distinct points , will be got by giving to the numbers the values , 1 , .
independently .
Hence there are points of superosculation .
Thus a plane cubic has nine inflexional tangents , and a quadri-quadric curve has sixteen superosculant planes .
Propositions hold good in general in regard to the grouping of these points , which are analogous to those which relate to the inflexions of a cubic .
Thus , an through \mdash ; 1 of them will lrvays peither through another besides , or through the line one of the .
This is obvious from the values already given for the parameters of points of superosculation .
Through any point of the curve can be drawn osculant flats .
This is proved in the same way as the preceding proposition , which is , in fact , the ection of it ; for if through the given point we draw a cone containing the curve , and cut it by an flat , the section will be an elliptic curve of order in dimensions , and the projections of the points whose osculant flats pass through the given point will be points of superosculation on the projected curve .
Hence , also , the lines joining the given point to the points of contact are grouped in respect of coflatness in the same way as the points of superosculation .
in the curve of next lower order .
More generally , through given points of the curve there can be drawn flats which have pointic contact with the curve .
If , . .
are the parameters of the given points , those of the required points are given by where the integers may take independently the values , 1 , . .
From these results we may now determine the various ranks and the class of the curve .
Suppose we know the number of flats which can be drawn through bitrary points in space\mdash ; or , which is the same thing , through an arbitrary flat P to touch a certain curve .
Then , if the arbitrary flat meets the curve in any point , two of these will cid at that point .
For taking an flat in the flat , and joining it to all the points of the curve by flats , we may cut this figure by a plane or 2 flat .
Every flat will cut this plane in a single point .
The problem is then reduced to drawing tangents from a point , the intersection of by the plane ) to a plane curve ; and we know that when this point lies on the curve , two of the tangents coincide at it .
PROFESSOR W. K. CLIFFORD ON THE CLASSIFICATION OF LOCI .
In general , a certain number of flats can be drawn through an itrary flat to have -pointic contact with a given curve ; this number fact , the rank of the curve .
If the arbitrary flat meets the curve at any point , then of these flats coincide at that point .
For we may project the whole figure from an flat lying in the flat on to a flat .
The problem is then reduced to drawing flats through a given point to have k-pointic contact with a curve in dimensions .
Now we know , from the example of the full-skew curve , that , when the point lies on the curve , of these coincide at the point .
If the arbitrary flat meet the curve in more points than one , of the osculants will coincide at each of them ; and this result is not affected by the union of the points into one .
In particular , if it meet the curve in coincident points , the number of osculants which there coincide is .
Applying now these general considerations to the elliptic curve , we find at once that the rank of it is .
For we must add to the number , just obtained , the number , , given by the theory of elliptic functions for the flats drawn through consecutive points of the curve to have -pointic contact elsewhere .
In particular , the class of the curve is ; we have observed already that the number of superosculants is Thus , a plane cubic is of order 3 , class 6 , and has 9 inflexions ; a quadri-quadric is of order 4 , rank 8 , class 12 , and has 16 superosculant planes .
We learn , moreover , that a quintic curve in four dimensions , when not unicursal , is of first rank 10 , second rank 15 , class 20 , with 25 points of superosculation .
Hence a quintic in three dimensions , with five apparent , is of rank 10 , class and has 20 superosculant planes ; this follows by projection from the former case .
A curve of this kind , viz. , an elliptic curve of order in an flat has its coordinates determined by the equations , . .
last term right being else according even ) , where and .
[ If is even , we may write instead of ] The condition for points to be co-flat is then See LINDEMANN ; CLEBSCH'S ' Lectures on Geometry , ' vol. [ Theory of Derived Points on an Elliptic Curve .
SYLVESTER'S theory of derived points on a plane cubic is as follows : ffom any given point on the curve , we may construct its tangential , or point where the tangent at the original point meets the curve again ; similarly we may construct the tangential of the tangential , or second tangential , and so on .
By joining any two non-consecutive points on this series , we can find their , the point where the joining line meets the curve again .
In this way we obtain an infinite group of points derived from ( and including ) the original point , such that the lin joining any two of them is either tangent at one of these or passes through a third point of the group .
It is to be observed that all points on the curve uniquely derived from the given point by any geometrical process , the point where the conic of five-pointic act meets the curve again , the point where cubics of eight-pointic contact meet the curve again , \amp ; c. ) are included in the group .
The coordinates of any derived point may be expressed rationaly in terms of the coordinates of the original point , and the order of the functions to which they are proportional is always a square number .
Thus the three coordinates of the tangential are proportional to quartic metions of the coordinates of the original .
If the square root of the order of these functions be called the order of the derived point , then we have the theorem that when three derived points are in a straight line , the order ol one of them is equal to the sum of the orders of the other two .
It is observed that there is no derived point whose order is divisible by 3 .
By help of this observation it is easy to make out a scheme of the orders ; for when we join two points , the order of their residual must be the sum or the difference of the orders of the points , and one or the other of these is always divisible by 3 .
This theory is really a geometric representation of the multiplication of elliptic functions .
The coordinates of any point on the cubic curve may by proper choice of axes be made proportional to elliptic functions of a parameter , so that : : sn cn dn .
This being so , if are parameters of three points in a straight lin , we shall have .
If be the tangential of , the three points are in a straight line , and , or .
Hence the series of tangentials has for parameters \amp ; c. : and in general the parameter of any derived point is of the form nu , where is a positive or negative integer .
The number , taken positively , coincides with what was caHed the order of the derived point .
For the elliptic functions of nu are of the order in the elliptic functions of In this way all points of the theory are explained , excepting the fact that no derived point has its order divisible by 3 .
derived points , only certain forms present themselves .
Let that number which , with othel numbers , makes up zero , be called the of those numbers ; it is , in fact , their sum taken negatively .
Then the process of forming the group is to start from unity , and add the residual of every numbers of the group , repetitions being allowed .
I say that by this process we shall only get numbers of the form .
For let , \amp ; c. be such numbers , then their residual is , which is a number of the same form .
Now as unity , with which we start , is of this form , it follows that all the numbers of the group must be of the form January , 1879 .
] OF DEFICIENCY Theorems relating to Abelian Functions .
It will be convenient to put together shortly those propositions relating to the application of Abelian functions to curves which will be wanted in the sequel .
The aggregate of the real and imaginary points on a curve constitutes a two-way spread , or surface , which may be transformed , by stretching without tearing , into the surface of a body with holes in it .
On this surface there are distinct closed curves which cannot without breaking be shrunk into a point , namely , one round each hole , and one through each hole .
Any other irreducible circuit must be made up of combinations of these .
If any rational function of the coordinates be integrated from one point to another along the curve-spread , the value of the integral will depend upon the path of the integration .
If the integral becomes infinite at any points , it may be altered in value by making the path go round one or more of these ; but in any case it may be altered by incorporating into the path any of the closed circuits just mentioned .
It is found that there are distinct rational functions of the coordinates whose integrals do not become infinite at any point of the curve-spread .
Any other integral which is everywhere finite must be a linear combination of these .
Of such linear combinations it convenient to take a certain set as the set ; they are so chosen that each of them , when integrated along a closed path which goes round a hole , gives zero for all the holes but one , and for that one ; thus , the integrals , which we may call , . .
, are associated one by one with the holes 1 , 2 , . .
If they are integrated along a closed curve passing through the hole , let the values be called ; then it is found that , or the integral of through the hole is equal to the integral of through the hole If we now take all the integrals from a point to a point along the same path , and if , . .
are the set of values for one such path , and , . .
for another path , then we must have PROFESSOR W. K. CLIFFORD ON THE CLASSIFICATION OF LOCT .
where the numbers are integers ; namely , is the additional number of times the new path has gone round the hole , and is the additional number of times it has gone through that hole .
We shall write these equations thus , . . ?
, and shall say that the quantities are to the quantities ?
respect of the periods Suppose now that the curve has no actual nodes , and that a locus of any order intersects it in the points . . .
Then , if another locus of the same order intersects in the points , . .
we take any one of the integrals , say from to , from to , . .
from to , the sum of these results will be congruent to zero .
That is to say Here the 2 refers to the suffixes of the and , not to .
There are such equations .
This is ABEL'S Theorem .
When the curve is in a -flat and of the order , we shall use this theorem chiefly for its intersections with flat .
If we regard the lower limits of the integrals ( the points x ) as fixed , the integrals for any poin may be regarded as parameters belonging to that point , and then ABEL'S Theorem gives us equations between the parameters of points which lie on flat .
The truth of these equations is necessary to the points .
on flat , but it may not be sufficient .
Thus in a bicircular quartic curve , , we have one equation to express that points are in a straight line , and if the points are collinear the equation is true .
But it does not follow from the equation that the collinear ; in fact , the equation holds equally good if the points are a circle .
If the sums of the parameters of points are given , that is , if we have the equations\mdash ; , 2 , . .
the being arbitrary constants , and the lower limits of the being sl.lpposd constant ; then the upper limits , . .
may be ) ressed in terms of the quantities namely , they the roots of an equation of degree whose coefficients are -fnnctions of the .
If , . . .
then the indicating that each of the integers , . .
is to take all integral values positive and negative .
When the lower limits are so chosen that the sum of the parameters is zero for the complete intersection by any locus , this -function has remarkable properties .
If we sum each of the parameters for any points on the the -fimction whose arguments are these sums will vanish .
That is if . . .
then these sums are taken for any points , not only will the vanish , but also its rential coefficient in regard to any one of the points .
And generally , if we take for the the sums of the parameters for points , the and its first differential coefficients in regard to any of the points will vanish .
Conversely , if the quantities are such that and its first differential vanish , then it is possible to find points such that . . .
Although here the number of equations is greater by .
than the number of unknown quantities , yet it is possible to satisfy them all in virtue of the relations existing between them .
Relation betmeen the Order and Deficiency of a Curve .
We shall now apply these theorems to the study of curves existing in dimensions , of the order ?
and deficiency flat cuts such a curye in points , such that the sum of each of the parameters , for the points , is zero .
But flat is determined by points ; so that , arbitrary points being chosen on the curve , it is always possible to find other points , so that the sum of each parameter for the whole points shall be zero .
Let then be the sums of the parameters for the given points ; then to find the remaining points we have the equations If is not greater than , we know that these equations can be solved , although the solution may be indeterminate .
But if , the equations cannot be solved unless certain conditions are satisfied by the .
Let ; then conditions must be satisfied ; namely the must be sums of the parameters of not more than points .
But they are sums of the parameters of points ; therefore is not greater than , or is not greater than .
We have proved then that PROFESSOR W. K. CLIFFORD ON THE CLASSIFICATION OF LOCI .
If , then is not than Conversely , if is at most to may also state the proposition in this way .
A curve of order and , not greater than , can at most imensio , [ It appears , therefore , that the theorems at the beginning of the paper may be extended , and that dimensions we have the curve of which is unicursal , the curve of order , and deficiency at most 1 , of order , and deficiency at most 2 , and so on till we come to the order , which is the first case of exception , and may have deficiency .
This uive is the natural geometric representation of the general Abelian functions , its multiple tangent flats playing the same part as the double tangents of the quartic in RIEMANN'S beautiful paper on the case H. EBER has noticed that in four dimensions this curve is the complete intersection of three quadric loci .
nuary , 1879 .
] PROFESSOR W. K. CLIFFORD ON CLASSIFICATION OF LOCI .
Moreover , we see once that the theory can be extended to other curves of deficiency 1 ; as , for example , the quadri-quadric curve .
Starting with any point on this curve , we may find the pomt where the osculating plane at that point meets the curve again ; then repeat the process with the point so found , and so on .
The plane joining any three of these points will meet the curve in another derived point , or else touch it at one of the three points .
The plane drawn through one derived point to touch the curve at another derived point will meet it again in a derived point , or touch at the first point , or osculate at the second .
The coordinates of any derived point are of the order in those of the original point , where may be called the order of the derived point .
In this case the order of no derived point is divisible by 2 .
I was desirous of finding a similar representation of the multiplication of hyperelliptic and Abelian functions ; and therefore sought for cases in which derived elements might be found on curves ( in the sense explained in the beginning of this paper ) of deficiency greater than 1 .
For this purpose I considered scrolls .
Taking an arbitrary generator on a quartic scroll having two linear directrices , we may draw a one-@heeted hyperboloid through three consecutive generators at that place ; this will meet the quartic scrol in one other generator , which is thus uniquely derived from the one .
Similarly on a quintic scroll contained in a linear complex , the two tractors of consecutive generators meet the scroll in two other points lying on a generator .
And on a sextic scroll not contained in a linear complex , the.linear complex having five-line contact at a given generator ( containing five consecutive generators ) will contain one other generator of the scroll .
In these three cases , then , from any three , four , or five generators we may uniquely derive a fourth , fifth , or sixth generator respectively ; and the whole theory of derived elements may be applied to the generators of these scrolls .
Unfortunately , however , each of the scrolls considered is at most of deficiency 1 , so that we merely get more illustrations of the multiplication of elliptic .
And it may be shown , in general , that a curve on which such a theory of derived points is possible , is at mcst of deficiency 1 .
Suppose that it has no singular points , and that points on it being given , there is uniquely determined one other point .
If this is effected ( as in the above examples ) by drawing a flat space through the points , which meets the curve in one other point , then it must be of the order .
Moreover , it must be in a flat space of so many dimensions that the flat of one dimension less is determined by points .
Now flat is determined by points ; therefore , the curve is in a flat .
Thus the impossibihty of extending the theory of derivation to curves of deficiency greater than unity is equivalent to the proposition that curve of order in dimensions is at most of deficiency 1 .
This failure was the starting point of the present paper .
It remains to explain why , in the group of numbers expressing orders of the

