Angles as Stretches of Rays, Areas, and the Metrical Hierarchy¶
Metrical Geometry (§§ 400–404, pp. 414–418)
§ 400. Definition of angle¶
f. §400 · p. p. 414400. I come now to a more difficult question than distance, namely the question as to the definition of angle. Here, to begin with, we must deal with rays, not with whole straight lines. The ray may be taken either as an asymmetrical relation, or as the half-line on one side of a given point on a line. The latter usage is very convenient, and I shall frequently employ it. Elementary Geometry assumes that two rays starting from the same point determine a certain magnitude, called the angle between them. This magnitude may, however, be defined in various ways. In the first place, we must observe that, since the rays in a plane through a point form a closed series, every pair of rays through a point defines two stretches of rays. Of these, however, one stretch contains the opposites of both rays, while the other stretch contains the opposites of neither—except, indeed, in the one case where the two rays are each other’s opposites. This case is met by Euclid’s postulate that all right angles are equal—a postulate, however, which is now known to be demonstrable [1]. Omitting this case, the angle between two rays may be defined as that stretch of rays through their intersection which is bounded by the two rays and does not contain the opposite of either, i.e. if A, B be the rays, and $\breve{A}$, $\breve{B}$ their opposites, the angle is the class of rays C which are separated from $\breve{A}$ or $\breve{B}$ by A and B. We might also, but for an objection to be mentioned shortly, define the angle as all the points on such rays. A definition equivalent to this last, but simpler in form, and avoiding the mention of the opposite rays, is the following [2]. Let a, b be any two points of the rays A, B, and let c be any point of the stretch ab. Then the class of points c, for all possible positions of a and b on their respective rays, is the angle between A and B. That is, every pair of intersecting rays divides the plane of the rays into two parts: the part defined as above is the angle. Or rather, the part so defined is the angle as a quantity: the angle as a magnitude is the divisibility of this part. But to these latter definitions we shall find fatal objections, and we shall find it necessary to adhere to the definition as a stretch of rays.
§ 401. Axioms concerning angles¶
f. §401 · p. p. 415401. Thus angle, like distance, is not a new indefinable, but like distance, it requires some new axioms. The angle between a ray A and its opposite $A'$ cannot be defined as above, but may be defined as the logical sum of the angles between A and B, B and $A'$ respectively. This limiting angle is greater than any other at the point, being in fact the whole half of the plane on one side of the straight line $AA'$. If the angles between A and B, B and $A'$ are equal, each is called a right angle. (That there are such angles, can be proved if we assume continuity.) Two intersecting straight lines make four angles, which are equal in pairs. The order of a collection of rays through a point in a plane may be obtained by correlation with the points where these rays intersect a given straight line, provided there is any straight line which all of them intersect. But since rays through a point in a plane form a closed series, while the points on a line do not, we require a four-term relation for the former order. The following definition seems adequate. Given four rays OA, OB, OC, OD through a point O and in one plane, if these all meet a certain straight line in A, B, C, D respectively, and A and C are separated by B and D, then OA and OC are said to be separated by OB and OD. In projective space this suffices. But in descriptive space we must provide for other cases. Thus if OA, OB, OC meet the given line, and B is between OA and OC, while OD does not meet the given line, then OA and OC are again said to be separated by OB and OD. If, finally, $OA'$ and $OB'$ be the opposites of OA and OB, then OA and $OA'$ are separated by OB and $OB'$. In virtue of the descriptive axioms of the preceding chapter, the order among the rays so obtained will be unambiguous, i.e. independent of our choice of the line ABC, and will cover all cases.
But now we need axioms analogous to those which, in the case of distance, were numbered (3), (7) and (8). At any given point in a given ray, there must be, in a given plane, two and only two rays, on opposite sides of the given ray (i.e. separated from each other by the given ray and its opposite), which make a given angle with the given ray; and angles must obey the axioms of Archimedes and of linearity. But in addition to these axioms, which insure that angles shall be numerically measurable, we must have some method of connecting the measure of angles with that of distances, such as is required for the solution of triangles. Does this require a new axiom? Euclid appears to obtain this, by means of I. 47, II. 12, and II. 13, without any fresh axiom. For this result we depend upon the propositions on the congruence of triangles (I. 4, 8, 26), which demand only, as we saw, the axiom that, with one angle at a given point, and one side along a given ray through that point, there exist two and only two triangles in a given plane through the ray (one on each side of the given ray), which are equal in all respects to a given triangle. Thus it would seem that no fresh axioms are required for angles in a plane.
§ 402. An angle is a stretch of rays, not a class of points¶
f. §402 · p. p. 416402. With regard to the definition of an angle as a portion of a plane, it is necessary (as in many other cases), if we retain this definition, somewhat to restrict the axiom that the whole is greater than the part. If a whole A has two parts B, C, which together constitute A, and if C be infinitesimal with respect to A, then B will be equal to A. This case occurs in a plane under the following circumstances. Let O, $O'$ be any two points, OP, $O'P'$ lines in one plane and making equal angles with the ray $OO'$ [3]. Then in Euclidean or hyperbolic space these lines OP, $O'P'$ will not intersect; thus the angle between $OO'$ and $O'P'$ will be part of the angle $O'$OP. Hence the above restriction is necessary as regards the axiom that the whole is greater than the part.
In Euclidean space this answer is sufficient, since, if OP makes with $OO'$ a less angle than $O'P'$ does, OP and $O'P'$ will intersect. But in hyperbolic space, OP and $O'P'$ may not intersect even then. Hence if we adhere to the above definition of angle, we shall have to hold that the whole may be less than the part. This, however, is intolerable, and shows that the definition in question must be rejected. We may, however, still regard angle as the stretch of rays; for the rays in the angle at $O'$ are not part of the rays in the angle at O. Hence it is only as a stretch of rays, or as the magnitude of such a stretch, that an angle can be properly defined.
As showing, in a curious manner, the increased power of deduction which results from the above axioms concerning distances and angles, we may remark that the uniqueness of the quadrilateral construction, which before could not be proved without three dimensions, can now be proved, as regards all constructions in one plane, without any assumption of points outside that plane. Nothing is easier than to prove this proposition by the methods of elementary coordinate Geometry. Thus although projective Geometry, as an independent science, requires three dimensions, any projective proposition concerning plane figures can be metrically proved, if the above axioms hold, for a two-dimensional space.
§ 403. Areas and volumes¶
f. §403 · p. p. 417403. As regards figures of three dimensions, angles between planes and solid angles can be defined exactly as rectilinear angles were defined. Moreover fresh axioms will not be required, for the measurement of such angles can be deduced from the data we already possess.
With regard to areas and volumes some remarks seem necessary. Areas and volumes, like angles, are classes of points when taken as quantities, and divisibilities when taken as magnitudes. For areas and volumes we do not require afresh the axioms of Archimedes and of linearity, but we require one axiom apiece to give a criterion of equal areas and volumes, i.e. to connect their equality with that of distances and angles. Such an axiom is supplied, as regards areas, by the axiom that two congruent triangles have the same area, and as regards volumes, by the corresponding axiom concerning tetrahedra. But the existence of congruent tetrahedra, like that of congruent triangles, demands an axiom. For this purpose, Pasch [4] gives the following general axiom: If two figures are congruent, and a new point be added to one of them, a new point can be added to the other so that the two new figures are congruent. This axiom allows us to infer congruent tetrahedra from congruent triangles; and hence the measurement of volumes proceeds smoothly.
§ 404. Right and left¶
f. §404 · p. p. 417404. In three dimensions, a curious fact has to be taken account of, namely, the disjunction of right and left-handedness, or of clockwise and counter-clockwise. This fact is itself of a descriptive nature, and may be defined as follows. Between two non-coplanar rays, or between four non-coplanar points taken in an assigned order, there is always one of two opposite relations, which may be called right and left. The formal properties of these relations have been explained in Part IV (§222); for the present I am concerned with their geometrical consequences. In the first place, they cause volumes to become magnitudes with sign, in exactly the way in which distances on a straight line have sign when compounded with their sense. But in the case of distances, since not all are on one straight line, we could not thus compound distance and sense generally: we should require, for a compound, some more general notion than sense, such as vectors supply. Here, on the contrary, since, in a three-dimensional space, all volumes have one or other of two senses, the compound can be made for all volumes. Thus if the volume of the tetrahedron abcd has one sign, that of bacd will have the opposite sign. This is the familiar geometrical fact that the determinant giving the volume of a tetrahedron abcd has one or other sign according as the sense of abcd is the same as or different from that of OXYZ, where O is the origin and X, Y, Z any positive points on the axes. It is this fact, also, which gives signs to angular momentum in Dynamics. The importance of the fact (which itself seems to be an independent axiom) is this, that it makes a distinction between two figures whose metrical properties are all identical. It is this distinction which puzzled Kant, who, like most of his contemporaries, supposed all geometrical facts to be metrical. In itself, the fact would be no more puzzling than, the distinction between the stretches AB and BA, which are metrically indistinguishable. But it becomes puzzling when metrical equality is supposed to result from motion and superposition. In our former definition of motion (§390) we omitted (as was then observed) a condition essential to its definition. Not only must two congruent figures be metrically equal, but there must be a continuous series of equal figures leading from the one to the other. Or, what amounts to the same thing, if a, b, c, d and $a'$, $b'$, $c'$, $d'$ be homologous non-coplanar points in the two figures, the tetrahedra abcd, $a'b'c'd'$ must have the same sense. In the case of equal and opposite tetrahedra, these conditions fail. For there is no gradual transition from clockwise to counter-clockwise; thus at some point in the series a sudden jump would be necessary. No motion will transform abcd into a tetrahedron metrically equal in all respects, but with the opposite sense. In this fact, however, there seems, to my mind, to be nothing mysterious, but merely a result of confining ourselves to three dimensions. In one dimension, the same would hold of distances with opposite senses; in two dimensions, of areas. It is only to those who regard motion as essential to the notion of metrical equality that right and left-handedness form a difficulty; in our theory, they are rather a confirmation than a stumbling-block.
With this we may end our brief review of metrical Geometry, leaving it to the next chapter to discuss its relation to projective Geometry and the projective theory of distance and angle.
Footnotes