LIBRARY University of California. RECEIVED BY EXCHANGE Class ON THE PENTACARDIOID ABSTRACT OF A DISSERTATION Submitted to the Board of University Studies of the Johns Hopkins University in conformity with the requirements for the degree of Doctor of Philosophy BY EDWARD C. PHILLIPS, S. J. March, 1908 BALTIMORE, MD., U. S. A. 1909 ON THE PENTACARDIOID ABSTRACT OF A DISSERTATION Submitted to the Board of University Studies of the Johns Hopkins University in conformity with the requirements for the degree of Doctor of Philosophy BY EDWARD C. PHILLIPS, S. J. March, 1908 -■ OF 1 HE UNIVERSITI. OF ^ALIFOf BALTIMORE, MD., U. S. A. 1909 / The writer desires to express his gratitude for the constant interest and kindly encouragement and -assist- ance given to him by Professor Morley, not only "during the preparation of this paper but during his entire course at the University. ON THE PENTACAliDIOID. Bt Edward C. Phillips. S. J. In the discussion of the metrical properties of finite systems of lines in a plane there arises a series of curves called by Professor Morley Ennacardioids;* these curves, over and above their usefulness in connection with the system of lines from which they arise, have many interesting properties which seem worthy of some special investigation ; and it is proposed in this article to make a study of the Pentacardioid, the first in the series of Ennacardioids which has not as yet received any detailed treatment. The general symbol for an Bnna- cardioid is C", and we shall use throughout this paper the corresponding symbol C^ for the Pehtacardioid. The system of coordinates and the general method of analysis which will be employed are those described in § 1 of Professor Morley's Memoir on Reflexive Geometry. However, as the notation in the various articles on this and allied subjects varies considerably, I shall, in order to avoid confusion, here state briefly the symbolism I intend to follow. The conjugate of a complex number x will be represented by the letter y, and the conjugate of a, by h. A complex number of absolute value equal to unity is called a turn and will in general be denoted by the letter < or by a Greek letter; the conjugate of a turn being its reciprocal needs no special notation. There are certain special turns to which definite symbols have already been assigned, and these I shall retain ; they are the following : The square root of negative unity, designated by i ; the cube roots of unity, designated by w and its powers ; and the fifth roots of unity, designated by £ and its powers Thus these three symbols are defined by the equations i^ + 1 = 0, o' — 1=0, and e^ — 1 = 0. The modulus, or absolute value, of a complex number will be denoted by the letter p, or by placing the number between parallel strokes, thus : | a; | . A turn may also be expressed exponentially, *F. Morley: On Reflexive Geometry, Transactiont of the American Mathematical Society, Vol. VIII (1907), pp. 15 ff. 186967 2 ON THE PENTACARDIOID. and thus any complex number is of the form ft, or pe**, where 6 is the amplitude of the complex number. An equation in complex variables is said to be self-conjugate when it is identical with the equation obtained by replacing each quantity by its conjugate, or differs from this equation merely by some factor; such an equation corre- sponds to an equation in real variables with real coefficients. The roots of a self-conjugate equation in t are either turns or pairs of inverse points as to the base circle, and these latter bear to the self-conjugate equation in i the same relation that pairs of conjugate imaginary roots bear to the equation with real coefficients in one real variable. § 1. The Equation of the Curve. The Pentacardioid, or C"*, as defined by its line equation in its most general form, is X— ao -f ba^t -f IQa^^ + \Qh.f + bh^t^ + {V — hY = 0. (1) '» The turn t is here used as a parameter ; each value of t gives, for a varying x, a line or tangent of the curve, and the envelope of this system of lines is the point curve. The equation of the point curve can be obtained in map form by eliminating «/ between equation (1) and its derivative as to t\ the elimination gives x = ao— 4ai<— 6a2<2— 462<» — 6i<*. (2) The point Oq is the singular focus of the curve and will be called its center. There are in the above equation three independent complex arbitFary con- stants, and hence a C'° depends on six conditions. It takes two complex constants, equivalent to four conditions, to fix the reference system so that there are only two absolute conditions required to fix the shape of a C^ I shall in general consider two curves to be the same when they are similar ; or, stating the same thing analytically, two curves z=if(t) and x=.f^{t) will be considered the same when they are so related that there exists between the points of the two curves a one-to-one correspondence of the form z = mx -\- n, where m and n are arbitrary complex constants. A correspondence or transformation of this kind may be called a proportion ; it leaves the shapes of figures entirely unaltered, merely changing their size and position in the plane. In the method of discussion which I shall follow, it will be found convenient to take the center of the curve as the origin and to make a^ equal to unity ; the ON THE PENTACAKDIOID. 3 equation in this form will be called the standard equation. The standard line and map equations of the general (7^ are therefore : x+ 5t+ lOat^ + 10bt^+ ^t^ + i/t^ = 0, (3) x = — it—6at^—'ibt^ — t\ (4) The equations in this form evidently depend upon and are entirely determined by the single complex number a, which T shall therefore call the determining coefficient of the curve. The equation of any C^ can be reduced to this form* by means of a proportion ; for let the equation of the curve be given in its most general form z = Oo — "^pi^it — Gp.^xr,t^ — 4(p2/x2)<^ — {pi/xi)t^ ; if in this we replace t by xfH (a change which leaves both the curve and the reference system unaltered), and then divide by p^xf^ and transpose the constant term, we have : (z — ao)/pi4'^ = — 4< — 6pxt^ — 4{p/xy — t* fi where p^pg/pi and x=:x^lx\^. And now the equation is in the standard form. This reduction is unique excepting as to the ambiguity introduced by the process of taking the cube root of xj, and hence it can be made in three, and only three, ways. We thus arrive at the following important theorem : Two G^'s are similar when, and only when, the determining coefficient in the standard equation of one curve differs from the determining coefficient of the other curve at most by a cube root of unity. Furthermore, since replacing a number (or a point which the number repre- sents) by its conjugate is equivalent to a reflexion in the axis of reals, it follows that if the determining coefficient of one G^ differs from the conjugate of the deter- mining coefficient of another C" at most by a cube root of unity, the two curves are inversely similar ; i. e., they are reflexions of each other in the axis of reals and so bear to each other the relation of an object and its image in a plane mirror. § 2. Construction of the Curve. A construction for any C" depending on certain properties of its osculants was described by Professor Morley in § 6 of the Memoir on Reflexive Geometry •There 1b the one exceptional case of the cnrves for which the coeflBcient Oj In equation (2) la zero, but we can consider this as a limiting case and include it In our standard equation by allowing the determiniDg coefficient a to take the special value co ; and It should be borne in mind that \a\ is really the ratio of |a,| to \a^\. 4 ON THE PENTACARDIOID. already referred to ; but there is another method, also mentioned by Professor Morley,* which is more easily applied to the case of the C^ and I shall here set forth a development of this second method. Refer to the standard equation of the C^ and consider the two special curves corresponding respectively to the values of the determining coefficient a = and a-=- (*:> . They are Xi + 5<+5<* + j/i<«=0, (5) a;3 + 10<'=+ 10<^ + y,<« = 0; (6) with map equations x, = — U-t\ (7) x^^ — Qe—U^. (8) These curves are two epicycloids easily constructed by simple mechanical means ; the first is the three-cusped epicycloid traced out by a point on the cir- cumference of a unit circle rolling about a circle of radius 3, whilst the second is the one-cusped epicycloid traced out by a point on the circumference of a circle of radius 4 rolling about a circle of radius 2. From these two curves all other C^'s can be built up very simply. We must first notice that any curve of the form x.i= — Qat^—Aht^ (9) is derived from (8) by multiplying x^ by px^, where a = pz = pe". For if in (9) we replace t by xH, we have aja — QpxH^ — 4px7' = fx% . (10) Now the effect of the factor px^ on the curve is merely to rotate it through the angle 50 and to enlarge it in the ratio p : 1. Looking again at the standard equation (4) of the general C^, we see that it is the sum of two simpler equations, namely (7) and (9) ; therefore any C^ can be constructed by adding the vectors that join the origin with corresponding points of two epicycloids properly placed. By corresponding points are meant points on the two curves given by the same value of t. Since the clinant of the tangent to all these curves at the point given by the parameter t is — t^, it is evident that at corresponding points of the two curves (7) and (9) the tangents are parallel, and owing to this fact we can readily determine as many pairs of corresponding points as we wish. We must, however, choose the proper starting points on the two curves, for there ♦Orthocentrlc Properties of the Plane »4-Llne : Transactions of the American Mathematical Society, Vol. IV (1903), pp. 7 and 8. a a ^ (6) 6 ON THE PENTACAEDIOID. are five tangents to a G^ in any given direction, and according as we associate one or another of these five points on the second curve with a selected point on the first we obtain five different resulting C^'s. This ambiguity can be obviated by choosing a definite value of the parameter t and determining the points given by this value on the two curves by the ordinary process of plotting. Thus, putting t = 1, say, we have aij = — 3 and a-g = — 6a — 46 ; these points are very easily found as soon as we know the value of a, and taking this pair as our starting points all further ambiguity is impossible. In practice, it is much more convenient to take the mean of two points than to take their sum, since to get the mean we need merely bisect the segment joining the points; and for our purpose the points (xj -\- x^l2 are just as good as the points (xj + a-g). Moreover, if we wish to alter the relative sizes of the two fundamental curves, which is equivalent to altering the modulus of the deter- mining coeflicient of the resulting curve, all we need do is to divide the stroke joining corresponding points in a different ratio than 1:1; thus if we wish the curve resulting from compounding (7) with the curve a; = — 6p<^ — 4p<^, we divide the stroke joining corresponding points of (7) and (8) in the ratio p: 1. Finally, when we wish to alter the amplitude of a, we merely rotate one of the curves (7) or (9) through an angle equal to five times the desired change in the amplitude. The combination of these two changes gives us every possible variation in the value of the determining coefficient, and this method has been found in practice to be a very simple and convenient one for obtaining all the distinctive types of C^ The shifting of one or both the fundamental curves without rotation does not affect the shape or size of the resulting curve and it is more convenient to place the component curves some distance apart than to have them concentric. Figure II shows a number of C^'s constructed by this method,* and Figure I illustrates clearly the carrying out of the method in two particular cases. The two broken curves (a) and {d) are the fundamental curves in this particular case ; their equations are (disregarding the constant term which has no effect on the shape or effective position) : xi = — 3^ — 2^ curve (a), arg = + 4< + <* curve {d). * The curves in Figure II are free-liaud reproductions, on a reduced scale, of larger figures constructed In the same manner as Figure I. M S.^^/g-'e. •fV (il S, = U" (c) S.'i'/v^c" U) S, = 31>, it" (e) S, = fiit'^ (/) S.-.^/^c'e b) s.^iU^ (c-J S, = %U c'l^ U) s, = :^'ixL^c> -t" c g XX.Y e, E. ') -S. - 6ci£.- (7) 8 ON THE PBNTACAEDIOID. On each of these curves are marked thirty points such that the angle between the tangents at any two successive points is 30 degrees ; the tangents at the points 1, 7, 13, 19, 25 on either curve are all parallel to the 'tangent at the point 1 on the other curve, and hence any one of those five points could be taken as a corresponding point to the point 1 of the other curve ; the points marked 1 on the two curves were chosen as a corresponding pair. The strokes joining these two points and the other corresponding points were divided in the ratio 1 : 2, and thus were obtained thirty points of the curve (6), of which the equation is 3a; = 2a;i -\- x^ z= 4t — 6f — 'ifi + <*. Next the strokes were divided in the ratio 2:1, giving the resulting curve (c), of which the equation is Sx = Xi + 2x2 = St—Sf— 2f+ 2tK The values of the determining coefficient for the four curves, taken in the order (a), (J), (c), {d), are respectively oo , — 1, — 1/4, 0. § 3. The Singularities of the C^. In order to determine completely the singularities, we need the ordinary equation of the curve, namely, an algebraic relation between x and y independent of <; we can best secure this by eliminating t between equation (2) and its con- jugate. Performing the elimination by Sylvester's dialytic method, we arrive at the desired equation in the following determinant form: 1 y 4b 4 6a 6b 4 4a Since x and y appear only in alternate rows, and one in each column, the developed equation will have as its highest term a;*?/*; hence a C" is of order eight. Since the four highest powers of x, and also of i/, are absent, the curve will have a four-fold point at infinity on each of the axes a; = and y^=0; i. e., at the X 4 6a 46 1 4a 6b 4 X 4 6a 1 •4a 66 X 4 1 4a X 1 1 y 46 1 4 y 6a 46 1 66 4 y = 0. (11) ON THE PENTACAEDIOID. 9 circular points, /and J. The symmetry of (11), with respect to x and y, shows that the curve is symmetric in its relations to the two circular points. A four- fold point being equivalent to six simple double points, these two singularities are equivalent to 12 double points; and as a rational curve of order eight has 21 double points, there must be nine other double points, and in general these all lie in the finite part of the plane. Now the condition for a cusp is the vanishing of the derivative of x as to <, or : 1 + 3a^ + 3i<2 + <3 = o. (12) As this is a self-conjugate equation in t, it has in general three turns as its roots; and hence every C* has three cusps, and only three, of which two, however, may be imaginary. Thus the nine double points comprise 3 cusps and 6 nodes. We are now in a position to resolve the singularities at /and /; for if we put d = number of nodes at each of the circular points, and Ic = number of cusps, and note that the C^ is of class five, as is shown immediately by equation (1), then Plucker's equation connecting the order and class of a curve gives us 5 = 56 — 2{2d + 6) — 3{2k + 3), or 2d + 3k = 15. (13) Moreover, as we have seen, d + k = 6, and hence d = k = 3; so that each of the four-fold points is equivalent to three nodes and three cusps. With regard to the line singularities, it is to be noticed that the clinant, being equal to — t^, has as its derivative — 5t*, and as this can not vanish for a turn, the C^ has no inflections; and by using the appropriate Pliicker equation, we find that there are six double tangents. This completes the enumeration of the simple singularities of the curve. By precisely analogous argument we can show that the general C" is of class n and order 2(n — 1), that it has an (w — l)-fold point equivalent to n — 2 cusps and (n — 2)(n — 3)/2 nodes at each of the circular points, that it has n — 2 cusps and (n — 2)(n — 3) nodes in the finite part of the plane, and finally that it has (n — 1)(« — 2)/2 double tangents and no inflections. It may of course happen that some of these singularities besides those at / and J are imaginary ; and I would note that there are two views which may be taken of such cases : The first is the projective view, which has been followed in the above discussion and treats all the singularities, both real and imaginary, as on the same footing; the second view is the one more proper to the present method of analysis, which may be called metrical geometry, and which acknowl- edges as properly belonging to the curve only the real nou-isolated singularities. 10 ON THE PENTACAEDIOID. In what follows, unless the contrary is stated, I shall confine myself to the second view and deal only with the real non-isolated singularities, and one of my main objects will be to determine how many such belong to any given curve and what effect their presence may have on the form or shape of the curve. Let us first consider the cusps. As we have seen, the cusp parameters are given by the self- conjugate cubic (12). Since this is of odd degree, one of its roots is necessarily a turn and hence every G^ has at least one cusp. The other two roots of (12) may be either turns or a pair of inverse points; in the latter case, the G^ has only one (real) cusp. If the three roots are turns, we have three cusps, in general distinct; two roots of (12) may become equal for special values of the determining coeflBcient a, and in this case two cusps come together; the resulting singularity, as will be shown immediately, involves also a node, and is a triple point with three coincident tangents and through which there passes only a single branch of the curve. Such a point does not differ materially in appearance from an ordinary point of the curve. For the proof it will be found convenient to have the standard equation expressed in terms of the cusp parameters. From equation (12) it is plain that if t^, t^, ta are the cusp parameters, then ti + tz + t3 = si = — 36, hh + ¥3 + ¥1 = Sg = 3a, (14) Hence the standard equation may be expressed in the form Sx = — 12t — 6s/ -f 4Sjf — 3<*. (15) This equation shows incidentally that the shape of the curve is determined uniquely as soon as the cusp parameters are given. If two roots of (12) become equal, then t^ =^2 ^^^ ^3 = — 1/^! so that the equation of the curve is, in this case, Bx = — -[2t — 6{tl— 2lfiy + 4{2ti—l/^y—3t\ (16) and the double cusp, given hy t = ti, is 3c = — 4ti—t\. (17) A pencil of lines through this point is given by the equation 3c—3x — r{Sd—3y) = 0, (18) T being the variable clinant and d the conjugate of c. If we substitute in this equation the values of a; and c given by (16) and (17), we get an octavic in t the ON THE PENTACAEDIOID. 11 roots of which are the parameters of the eight points in which the line intersects the C*; the result is (<-0'(4/«? + ^ + 0- J with the condition ^i^^g. Combining these two relations, we get the four following cubics whose roots give the twelve parameters of the six double tangents : 3 + 2S2(1 + £")< + 2sj(e3» + e*")<2 — Se'^'f = ; n = 1, 2, 3, 4. (28)„ It should be noted that the three roots of (28)^ are equal to the three roots of (28)j, each multiplied by s; and the three roots of (28)3 are equal to those of (28)2, each multiplied by e^ Thus the six double tangents are separated into two sets of three which are distinguished from each other by the following geometric property: Any tangent given by the equation (28)i is such that if we pass along the curve in the positive direction from the first point of contact to the second, we do not pass through any point at which the tangent is parallel to the double tangent; whilst for any double tangent of the second set, given by a root of (28)a, we always pass through one, and only one, point at which the 14 ON THE PENTACARDIOID. tangent is parallel to the double tangent. Since each of these equations is self- conjugate and of odd degree, it follows that every C* has at least two double tangents and, when there are only two, they belong to two different sets. It should be noted that if we consider Sj as the variable and i as a parameter, then (28)„ is the equation of a family of lines, or rather of two families of lines, and the envelopes of these lines are precisely the two deltoids (25)„ already found in connection with the tacnodes. That this is the case follows from the geometric connection between the double tangents of a curve and the tacnode; analytically it is shown very simply by the following form of the discriminant of (28)„, which as usual gives the values which the variable (the variable here being Sj) takes for all points on the envelope of the system. Calling t^, t^, t^ the roots of (28),,, we have ^= 1> 2. Note (added June, 1908). For completeness we should add the following compound singularities : Three coincident double tangents giving a triple tangent; the necessary and sufficient condition is sj = — (3/2)[(l + e")< + l/<2](e2" + f*") ; n = 1, 2, 3 or 4. (32) Four coincident double tangents; this singularity includes a tacnode and may be called a tacnodal triple tangent. The condition is s, = — (3/2)(e" + £*"), or Sj = — (3/2)(2 + e")(e^" + e*") ; w = 1, 2, 3 or 4. (33) ::?•- C4) ^; U) (e-l If} Ul U) (^) {^J Fio. III. — Compound singularities of tlie C. (15) 16 ON THE PENTACARDIOID. Six coincident double tangents giving a quadruple tangent. The condition is «i = — 3/2. (34) For convenience I append a table of all the compound singularities, including those mentioned in the above note, and give references to the figures in which occur the several singularities and the penultimate forms of the same. I should remark that the curves in Figure III are schematic, but resemble quite closely the actual curves. Singularities. Condition on Si- Figures. 1. Triple point, 1 branch *i=2<— l/i^ HI, (<^) = 2 cusps, 1 node 2. Triple point, 2 branches Sj = . ? . II, (i); III, (e) = 2 nodes, 1 cusp 3. Quadruple point, 1 branch Sj = — 3 III, [i) and (/) = 3 nodes, 3 cusps 4. Tacnode Si = — (3/2)(2< = 2 nodes, 2 double tgs. + f7<2)(e''" + 6*") I, (d) 5. Oscnode «! = — (9/2)(e" + e*") HI, (n) and (o) = 3 nodes, 3 double tgs. 6. Triple tangent Si = — (3/2)[(l + £")< = 3 double tangents + 1 /t^lie^" + f*") III, (a) and (/) 7. Tacnodal triple tangent s, = — (3/2)(e'" + f'") HI, (b), (j) and (0 = 2 nodes, 4 double tgs. or — (3/2)(2 + e'')(e^"+e*») 8. Quadruple tangent Si = — 3/2 HI, (»«) = 6 double tangents . § 4. The Complete System. Thus far I have considered the properties of the G^ in a somewhat isolated manner; it is my purpose now to take up the sets of C^'s which are connected together by the possession of certain special properties or certain common forms of singularities, and I shall devote this section to what may be called the com- plete system of C^'s. By the complete system I mean any collection of curves including within itself all the double infinity of forms or shapes which the C^ can have. In connection with this system of C^'a I shall consider certain singly ON THE PENTACARDIOID. 17 infinite sub-systems, and also certain interesting regions, loci and envelopes which naturally present themselves in the course of the investigation and some of which we have already come across in deriving the equations of condition treated in the previous section. The simplest analytical expression for this complete system is the standard equation which we have been using above, namely Sx = — 12t—6s2t^ + 4Sit^—3t\ (15) or, in line form, Sx + 15t + I0s2<2— 10s/ + 15<* + 3t/t^ — 0. (20) When in either of these equations we allow the determining coefficient s^, or, more conveniently for our purpose, Sj, to take the double infinity of values of the complex number of the binary domain, we get all possible types of (7^ We mp,y therefore consider Sj as representing a point of the plane, and two things are to be noted : first, that though s, is subject to the condition that it must always be the sum of three turns or of one turn and a pair of inverse points, this condition does not impose any restriction on the value of s-i ; and secondly, that owing to the relation between the determining coefEcient and the shape of the curve established at the end of § 1, we may without any loss of generality restrict Sj to any third portion of the plane bounded by a pair of straight lines or rays through the origin. We shall make use of this restriction on Sj later on. I shall begin this investigation with the cusps. The first fact that strikes our attention is that though we have a double infinity of curves yet they have but a single infinity of cusp tangents ; for on combining the condition for cusps, which is simply l-\-8zt — Sit^ + i^=0, (35) with the line equation (20) of the system, both Sj and s^ are eliminated together, giving as the envelope of the cusp tangents the curve Sx+ 5t+ 5t^ + Syt^=0.. (36) This is a three-cusped epicycloid concentric with the whole system (20) and has for its map equation Sx = — 4t — 1\ (37) Each line of this curve, therefore, must be a cusp tangent for a whole infinity of C^'s of the system. But it must be noted that not any three tangents of (36) can be chosen as cusp tangents of a C^ of the system ; for the cusp parameters 18 ON THE PENTACARDIOID. of any curve of the system are subject to the involution determined by equation (35), namely Sg -]- 1 = 0, and therefore when two of the tangents are chosen, the third is uniquely determined. The epicycloid (36) has some further interesting connections with the cusps, which I shall here set forth. For this purpose I shall consider the behavior of the three cusps when we hold one of the cusp parameters fixed. The cusps of the complete system are all included in the formula 3c„ = - 9<„ — Ss^tl + Sitl, n = l,2, 3, (38) where <„ is a root of (35). To study the behavior of the cusps, it is best to replace Sj and s^ by their values in terms of <„ and to simplify the equations by means of the involution Sg + 1 = 0. We thus obtain, as the equations of the three cusps, 3ci = — 6^1 + 2tl/t^ — 2t% + t{, ' Sc^ = — 6<2 + 2tl/t, — 2tlt, + 4, (39) 3c3 = — 6^3 + 2tyti — 2tlt, + tl . This of course is not the only form in which the equations can be put ; and by replacing one of the parameters by its value derived from the involution Sg + 1 = 0, we get another form which will be found useful to us, namely 3c3 = 6/t,t, + 2/tltl + 2lt\tl + \lt\tl, (40) and two similar equations obtained by cyclically interchanging the subscripts. Now let us hold one of the parameters, t-^ say, fixed and allow the other two to vary, replacing them by t. Under these conditions the three cusps will have to move along definite curves, since their positions depend on the single variable t. The cusp Ci is by this means separated, as it were, from the other two, and these latter lose their identity and may be treated as being a pair interchangeable at will. The cusp Cj moves along one curve, namely 3ci = — 6o, it acquires a pair of nodes. Now when Sj ^ 0, we have the three-cusped epicycloid, which we know independently has no nodes ; therefore when «i is within Dq the C^ has an even number of nodes, and when Sj is without Dq the G^ has an odd number of nodes. We can now make out a table of singularities for the various regions, the number of simple singularities possessed by a given C^ being placed opposite the region in which the coefiGcient Sj of the given C^ happens to lie. Region. Cusps. Db. Tgs. Nodes R, 3 6 Even R, 3 4 Even Rs 3 4 Even R, 3 2 Even A 1 4 Odd i?« 1 2 Odd This table of regions is incomplete in so far as the number of nodes is con- cerned and must remain so until the equation of condition for the singularity consisting of two nodes and a cusp is found. I shall here leave the subject, with the hope that it may be completed at a later date. Johns Hopkins Univkbsitt. March, 1908. • Cf. Table of singularities, p. 16. The completeness of this table, upon which rests the validity of the argument here used, has not as yet been rigidly proved; but careful investigation makes it almost certain that there are no other compound singularities occurring on any &'. BIOGRAPHICAL NOTE. Edward C. Phillips was born in Germantown. Pennsylvania, on November 4, 1877. His early education was secured in the Parochial Schools, and he made his collegiate studies at the College of St. Francis Xavier, New York City, graduating from that institution with the degree of Bachelor of Arts in 1898. He then entered the Novitiate of the Society of Jesus at Frederick, Maryland. PVom 1901 to 1904 he was at Woodstock College, Maryland, engaged chiefly in graduate studies in Philosophy. In October, 1904, he came to the Johns Hop- kins University and entered the department of Mathematics as a graduate student. Since then, with the exception of the year 1906-7, he has been following courses of Mathematics, Physics and Physical Chemistry. March, 1908. BERKELEY ^'^^^' ^HIS BOOK IS due"^ ^^_ ^^ ao «-^ no. rl^^=^ BeS^^^«^ bate 20m-ll,'20 I- ^.