Pieri’s Projective Postulates, Lines, and Involutions

Projective Geometry (§§ 361–366, pp. 381–386)

§ 361. Recent threefold scrutiny of geometrical principles

f. §361 · p. p. 381

361. The foundations of Geometry have been subjected, in recent times, to a threefold scrutiny. First came the work of the non-Euclideans, which showed that various axioms, long known to be sufficient for certain results, were also necessary, i.e. that results inconsistent with the usual results but consistent with each other followed from the denial of those axioms. Next came the work of Dedekind and Cantor on the nature of continuity, which showed the necessity of investigating carefully the prerequisites of analytical Geometry. Lastly, a great change has been introduced by the Italian work on closed series, mentioned in Part IV, in virtue of which we are able, given a certain type of relation between four points of a line, to introduce an order of all the points of a line. The work of the non-Euclideans has, by this time, produced probably almost all the modifications that it is likely to produce in the foundations, while the work of Dedekind and Cantor only becomes relevant at a fairly advanced stage of Geometry. The work on closed series, on the contrary, being very recent, has not yet been universally recognized, although, as we shall see in the present chapter, it has enormously increased the range of pure projective Geometry.

§ 362. Projective, descriptive and metrical geometry

f. §362 · p. p. 381

362. In the discussions of the present Part, I shall not divide Geometries, as a rule, into Euclidean, hyperbolic, elliptic, and so on, though I shall of course recognize this division and mention it whenever it is relevant. But this is not so fundamental a division as another, which applies, generally speaking, within each of the above kinds of Geometry, and corresponds to a greater logical difference. The above kinds differ, not in respect of the indefinables with which we start, nor yet in respect of the majority of the axioms, but only in respect of comparatively few and late axioms. The three kinds which I wish to discuss differ both in respect of the indefinables and in respect of the axioms, but unlike the three previous kinds, they are, roughly speaking, mutually compatible. That is to say, given a certain body of geometrical propositions concerning a certain number of entities, it is more or less arbitrary which of the entities we take as indefinable and which of the propositions as indemonstrable. But the logical differences which result from different selections are very great, and the systems of deductions to which different selections lead must be separately discussed.

All Geometries, as commonly developed, agree in starting with points as indefinables. That is, there is a certain class-concept point (which need not be the same in different Geometries), of which we assume that there are at least two, or three, or four instances, according to circumstances. Further instances, i.e. further points, result from special assumptions in the various cases. Where the three great types of Geometry begin to diverge is as regards the straight line. Projective Geometry begins with the whole straight line, i.e. it asserts that any two points determine a certain class of points which is also determined by any two other members of the class. If this class be regarded as determined in virtue of a relation between the two points, then this relation is symmetrical. What I shall call Descriptive Geometry, on the contrary, begins with an asymmetrical relation, or a line with sense, which may be called a ray; or again it may begin by regarding two points as determining the stretch of points between them. Metrical Geometry, finally, takes the straight line in either of the above senses, and adds either a second relation between any two points, namely distance, which is a magnitude, or else the consideration of stretches as magnitudes. Thus in regard to the relations of two points, the three kinds of Geometry take different indefinables, and have corresponding differences of axioms. Any one of the three, by a suitable choice of axioms, will lead to any required Euclidean or non-Euclidean space; but the first, as we shall see, is not capable of yielding as many propositions as result from the second or the third. In the present chapter, I am going to assume that set of axioms which gives the simplest form of projective Geometry; and I shall call any collection of entities satisfying these axioms a projective space. We shall see in the next chapter how to obtain a set of entities forming a projective space from a set forming a Euclidean or hyperbolic space; projective space itself is, so far as it goes, indistinguishable from the polar form of elliptic space. It is defined, like all mathematical entities, solely by the formal nature of the relations between its constituents, not by what those constituents are in themselves. Thus we shall see in the following chapter that the “points” of a projective space may each be an infinite class of straight lines in a non-projective space. So long as the “points” have the requisite type of mutual relations, the definition is satisfied.

§ 363. Projective points and straight lines

f. §363 · p. p. 382

363. Projective Geometry assumes a class of entities, called points, to which it assigns certain properties [1]; In the first place, there are to be at least two different points, a and b say. These two points are to determine a certain class of points, their straight line, which we will call ab. This class is determined by b and a, as well as by a and b, i.e. there is no order of a and b involved; moreover a (and therefore b) is itself a member of the class. Further, the class contains at least one point other than a and b; if c be any such point, then b belongs to the class ac, and every point of ac belongs to ab. With these assumptions it follows [2] that, if c, d be any points of ab, then cd and ab coincide—i.e. any two points of a line determine that line, or two lines coincide if they have two points in common.

Before proceeding further, let us consider for a moment what is meant by saying that two points determine a class of points. This expression is often thought to require no explanation, but as a matter of fact it is not a perfectly precise statement. The precise statement of what is meant is this: There is a certain definite relation (K say) which holds between any couple of points and one and only one corresponding class of points. Without some such definite relation, there could be no question of two points determining a class. The relation K may be ultimate and indefinable, in which case we need the above properties of the class ab. We may obtain, however, a derivative relation between two points, b and c say, namely that of being both collinear with a given point a. This relation will be transitive and symmetrical, but will always involve reference to a term other than those (b and c) which are its terms. This suggests, as a simplification, that instead of a relation K between a couple of points and a class of points, we might have a relation R between the two points a and b. If R be a symmetrical aliorelative, transitive so far as its being an aliorelative will permit (i.e. if aRb and bRc imply aRc, unless a and c are identical), the above properties of the straight line will belong to the class of terms having to a the relation R together with a itself. This view seems simpler than the former, and leads to the same results. Since the view that the straight line is derived from a relation of two points is the simpler, I shall in general adopt it. Any two points a, b have, then, a relation $R_{ab}$; a, c have a relation $R_{ac}$. If $R_{ab}$ and $R_{ac}$ are identical, while b and c differ, $R_{bc}$ is identical with both $R_{ab}$ and $R_{ac}$; if not, not. It is to be observed that the formal properties of any such relation R are those belonging to the disjunction of an asymmetrical transitive relation and its converse—e.g. greater or less, before or after, etc. These are all symmetrical aliorelatives, and are transitive so far as their being aliorelatives will permit. But not all relations of the type in question are analyzable into a transitive asymmetrical relation or its converse; for diversity, which is of the above type, is not so analyzable. Hence to assume that the straight line can be generated by an asymmetrical relation and its converse is a new assumption, characteristic of what I shall call Descriptive Geometry. For the present, such an assumption would be out of place. We have, then, two indefinables, namely point, and the relation R or K [3]. No others are required in projective space.

§ 364. Definition of the plane

f. §364 · p. p. 384

364. The next point is the definition of the plane. It is one of the merits of projective space that, unlike other spaces, it allows a very simple and easy definition of the plane. For this purpose, we need a new axiom, namely: If a, b be two distinct points, there is at least one point not belonging to ab. Let this be c. Then the plane is the class of points lying on any line determined by c and any point x of ab. That is, if x be any point of ab, and y any point of cx, then y is a point of the plane cab; and if y be a point of the plane cab, then there is some point x in ab such that y is a point of cx. It is to be observed that this definition will not apply to the Euclidean or hyperbolic plane, since in these two lines may fail to intersect. The exclusion of Euclidean and hyperbolic space results from the following axiom [4]: “If a, b, c be three non-collinear points, and $a'$ be a point of bc other than b and c, $b'$ a point of ac other than a and c, then there is a point common to $aa'$ and $bb'$.” By means of this axiom we can prove that the plane cab is the same as the plane abc or bac, and generally that, if d, e, f be any three non-collinear points of abc, the plane def coincides with the plane abc; we can also show that any two lines in a plane intersect.

§ 365. Harmonic ranges

f. §365 · p. p. 384

365. We can now proceed to the harmonic range and von Staudt’s quadrilateral construction. Given three collinear points a, b, c take any two points u, v collinear with c but not on ab. Construct the points of intersection au . bv and av . bu; join these points, and let the line joining them meet ab in d. This construction is called the quadrilateral construction. If we now assume that outside the plane abu there is at least one point, we can prove that the point d is independent of u and v, and is uniquely determined by a, b, c. The point d is called the harmonic of c with respect to a and b, and the four points are said to form a harmonic range. The uniqueness [5] of the above construction—the proof of which, it should be observed, requires a point not in the plane of the construction [6]—is the fundamental proposition of projective Geometry. It gives a relation which may hold between four points of a line, and which, when two are given, is one-one as regards the other two. Denoting “c and d are harmonic with respect to a and b” by cH$_{ab}$d, the following properties of the relation are important: (1) cH$_{ab}$d implies dH$_{ab}$c, i.e. $H_{ab}$ is symmetrical; (2) cH$_{ab}$d implies aH$_{cd}$b, i.e. the relation of the pairs ab, cd is symmetrical; (3) cH$_{ab}$d implies that c and d are different points, i.e. $H_{ab}$ is an aliorelative. This last property is independent of the others, and has to be introduced by an axiom [7].

Having obtained the harmonic range, we may proceed in two different directions. We may regard the harmonic relation as a relation of two pairs of points: hence, by keeping one of the pairs fixed, we obtain what is called an involution. Or we may regard the harmonic relation, as in the symbol cH$_{ab}$d, as a relation between two points, which involves a reference to two others. In this way, regarding a, b, c, as fixed, we obtain three new points d, e, f on the line ab by the relations cH$_{ab}$d, aH$_{bc}$e, bH$_{ac}$f. Each of these may be used, with two of the previous points, to determine a fourth point, and so on. This leads to what Möbius [8] calls a net, and forms the method by which Klein [9] introduces projective coordinates. This construction gives also the method of defining an harmonic ratio. These two directions in which projective Geometry may be developed must be separately pursued to begin with. I shall take the former first.

§ 366. Involutions

f. §366 · p. p. 385

366. By means of the harmonic relation, we define an involution. This consists of all pairs of points which are harmonic conjugates with respect to two fixed points [10]. That is to say, if a, b be the two fixed points, an involution is composed of all pairs of points x, y such that xH$_{ab}$y. If four points x, y, $x'$, $y'$ be given, it may or may not happen that there exist two points a, b such that xH$_{ab}$y and $x'H_{ab}y'$. The possibility of finding such points a, b constitutes a certain relation of x, y to $x'$, $y'$. It is plain that this relation sometimes holds, for it holds when x, y are respectively identical with $x'$, $y'$. It is plain also that it sometimes does not hold; for if x and y be identical, but not $x'$ and $y'$ then the relation is impossible. Pieri [11] has shown how, by means of certain axioms, this relation of four terms may be used to divide the straight line into two segments with respect to any two of its points, and to generate an order of all the points on a line. (It must be borne in mind that, in projective Geometry, the points of a line do not have an order to begin with.) This projective order is obtained as follows.

Footnotes