ODE, ff^r^y^, SCIENCE ATSTD ART DEPARtFM^lJc)p| OF THE COMMITTEE OF COUNCIL ON EDUCATION, SOUTH KENSINGTON MUSEUM. A CATALOGUE OP A COLLECTION OF MODELS OF RULED SURFACES, CONSTRUCTED By M. FABRE DE LAGRANGE ; WITH AN APPENDIX, CONTAINING AN ACCOUNT OF THE APPLICATION OF ANALYSIS TO THEIR INVESTIGATION AND CLASSIFICATION, By C. W. MERRIFIELD, F.R.S., PRINCIPAL OP THE ROYAL SCHOOL OP NAVAL ARCHITECTURE AND MARINE ENGINEERING, AND SUPERINTENDENT OP THE NAVAL MUSEUM AT SOUTH KENSINGTON. r^ ^^V LONDON: PRINTED BY GEORGE E. EYRE AND WILLIAM SPOTTISWOODE, PRINTERS TO THE QUEEN'S MOST EXCELLENT MAJESTY. FOR HER MAJESTY'S STATIONERY OFFICE. 1872. 29992. 3^ The set of Models described in this Catalogue is deposited in the Educational Collection at the South Kensington Museum, Catalogue of a Collection of Models of Ruled Surfaces, constructed by M. Pabre de La~ grange, with an Appendix, containing an account of the application of Analysis to their investigation and classification, By €. W. MERRIFIELD, F.R.S., Principal of the Royal School of Naval Architecture and Marine Engineering ^ and Superintendent of the Naval Museum at South Kensington, INTRODUCTION. This collection illustrates the principal types of the class of surfaces which can be traced out in space by the motion of a straight line. These surfaces, on account of the facility with which they can be constructed and represented, and of the ease with which their intersections can be determined, are of more consequence than any others in the geometry of the Industrial Arts. It is only in small work, which can be put into the lathe, that the class of sm-faces of revolution approaches them in respect of general utility. The most important surfaces of all, the plane, the right cylinder, the right cone, and the common screw, belong to both classes. The representation of the surfaces by means of silk threads is of course only approximate ; an approximation of the same character as the representation of a curve by a dotted or chain line, or by a series of right lines touching the actual curve. Fio. 1. Fig. 1 is an example of the first, and Fig. 2 of the second. In both cases, the curve, although not actually drawn, is indicated with sufficient approximation for most practical purposes. Models Nos. 10 and 30 also afford illustrations of the principle exhibited in Fig. 2. The models are constructed with especial reference to the possibility of changing their shape, by moving some of the supports of the strings, by altering the lengths or positions of certain parts, or by converting upright forms into oblique. This possibility of deformation, as the process is technically called, greatly enhances the value of the models, by allowing them to represent a much greater variety of surfaces than if they were fixed. They are, however, too delicate to be much pulled about, and, unless they are very cautiously handled, the strings are apt to become entangled or break. They should never be used except by a person who understands them, and they should not be shifted without some good reason. In order to make this collection as useful as possible to the student of geometry, it has been thought advisable to give, in an appendix, a short account of the application of analysis to the investigation of these surfaces, and of their properties. The statement of these properties is scattered over a great number of treatises and tracts, and there exists no single work which gives a full account of ruled surfaces. The appendix, of course, requires some knowledge of analytical geometry of three dimensions. Any of the smaller modern treatises, such as Aldis or Leroy, contain more than is necessary as an introduction to the subject. The statements in the appendix have been chiefly taken from Monge's Applications de I ''Analyse a la Geometrie. Geometrical drawings of most of the surfaces represented by these models are contained in Bradley's Practical Geometry (2 vols., oblong folio, published by Chapman and Hall). Many of them will also be found in the French treatises on practical and descriptive geometry, such as Leroy, Adhemar, Lefebure de Fourcy, De la GouRNERiE, and in their treatises on Stereotomy and Stone- cutting {coupe des pier res). Many of them are also given in Sonnet's Dictionnaire des Mathematiques Appliquees. CATALOGUE. 1. Hyperbolic Paraboloid generated by a single system of right lines. Two bars each pierced with holes equally spaced. One bar is fixed, the other swings round an axis, which, moreover, can be inclined at different angles to the fixed bar. When the bars are parallel the strings indicate a plane. When they are inchned to one another, but still in the same plane, the strings still indicate a plane ; but when the bars are not in the same plane, the surface is the hyperbolic para- boloid. This surface is sometimes called the twisted plane. But it must not be supposed that it can be made by bending a plane. On the contrary, when the surface is twisted, no two of the strings lie in the same plane, and, therefore, no part of the surface is plane. It can neither be flattened nor made from a plane, without stretching or contraction. 'w.' The hyj^erbolic paraboloid is the natural surface proper for a ploughshare. 2. Hyperbolic Paraboloid. Two bars pierced with holes at equal distances, the holes being connected by two different systems of strings. The surface, as well as the arrangement, is very nearly the same as in No. 1, only that there are two paraboloids instead of one. As the movable bar swings round, the paraboloid opens out while the other closes up. If the bars are swung so as to be in the same plane, one system of strings describes a plane by parallel lines, and the other by lines radiating from a point. If one bar is now turned so as to be end for end, we still get a plane, the set of parallel lines now passing through a point, while the set which previously passed through a point has now become parallel. The pair of paraboloids intersect in three right lines. There is also a fourth intersection on the " line at infinity." 3. Hyperbolic Paraboloid. Two bars, equally spaced ; each turns on an arm perpen*- dicular to itself, and one arm swings on a pillar. These arms can be ranged in one plane, and also turned end for end. 4. llYFEa^BOLie. Pi^RABOLOiD generated by two systems of right liiiGS. A skew quadrilateral with four equal sides, each pierced with the same number of holes, equally spaced. The model exhibits the double generation of the surface. The plane containing two of the sides turns about hinges connecting it with the plane of the other two sides. By closing or opening this hinge the paraboloid opens out or closes. When com- pletely open, it forms a plane divided into diamonds. When completely closed it again forms a plane, but the division is no longer uniform. The strings then become tangents to a plane parabola. 5. Hyperbolic Paraboloid. A skew quadrilateral turning upon four hinges with parallel axes or pins. The difference between this and tlie last Is not in the kind of surface or mode of generation, but in the manner of deforming the surface. In No. 4 the lengths of the strings alter ; while in this model they remain unaltered. More- over, although the surface flattens in two ways, yet in both ways the strings become tangents to a plane parabola instead of parallel. This model is well adapted for showing the leading sections of the solid. All sections parallel to the pins of the hinges are plane parabolas, which degenerate into right lines when taken also parallel to the brass bars. Any other sections, whether perpendicular to the hinges or inclined to them, give hyperbolas, which degenerate into a pair of right lines when the plane of section is a tangent to the surface. It may be worth while to remark that there is nothing absurd in the tangent plane to a surface cutting that surface, as a student unaccustomed to those subjects might at first think. On the contrary, when a sm-face is bent one way in one direction and the other way in the opposite direction, the tangent plane must cut it. In this case, the plane passing through any two intersecting strings is a tangent plane, and evidently cuts the surface along each string. If we imagine two planes parallel to the hinge pins, and each bisecting a pair of opposite bars, we obtain the asymptotic planes of the paraboloid, each of w^hich is the assemblage of the asymptotic lines of the hyperbolas parallel to the principal hyperbolic section. Their being asymptotic . has reference to these hyperbolas, and not to the parabolic character of the surface. 6. Hyperbolic Paraboloid. A skew quadrilateral, with its opposite sides equal in length, and pierced with holes at equal distances. Nearly similar to No. 5, but differently mounted, and with the sides of different lengths, the alternate sides only being equal. It is virtually a slightly different aspect of the same sufface as No. 5. 7. Hyperbolic Paraboloid. A skew quadrilateral, with all its sides equal, and pierced holes at equal distances. As far as the curved surface is concerned, the same as No. 5. But the hinges are altered in direction, and the model shows plans and elevations of the right line generators of the surface. The rings also show parabolic sections of the surface. In consequence of the alteration in the direction of the hinges, the spacing of the inclined bars, although equidistant, is at a different pitch from that of the horizontal bars. 8. Hyperbolic Paraboloid. A skew quadrilateral with all its sides equal, and pierced with holes at equal distances. It shows the plans and eleva- tions of the right line generators. The rings show the parabolas of the principal sections. No. 7 represents one quarter of what is shown in No. 8. The upper corners of Nos. 7 and 8 correspond ; but the lower comer of No. 7 corresponds with the middle ring of No. 8. 9. Hyperbolic Paraboloid. A skew quadrilateral with all its sides Unequal. The surface is the same as Nos. 7 and 8, but the proportions and the portion of the surface chosen for representation are different. The quadrilateral base being Irregular, the strings alter in length as the surface is deformed by closing the hinges. O" 10. Hyperbolic Paraboloid. Skew quadrilateral, pivoting on a single hinge. Intended to show the construction of the parabola connecting two roads which meet obliquely. This constiiiction is used by engineers in lading out roads. 11. Hypeeboloid of one Sheet. Two rings or circles in parallel planes are pierced with equally spaced holes. In a certain position the threads give, 1st, a cylinder, and 2ndly, a cone. The upper ring turns round a pin at its centre. In turn- ing it, the cylinder closes in and the cone opens out, each altering into a hyperboloid of one sheet. We can go on turn- ing the ring until these coincide in one hyperboloid, of which we thus get both systems of generating lines. If the rings are set on a slope the hyperboloid is elliptic. If the. rings are horizontal the hyperboloid is one of revolu- tion. Sloping one ring so as not to be parallel with the other, gives rise to some curious iniled surfaces, but these are not in general hyperboloids. 12. Hyperboloid of one Sheet. Two rings of different radius in parallel planes are divided into the same number of equal parts. The smaller and upper ring turns round a pin at its centre. In a particular position of the rings, the threads give two cones. Turning the ring transforms each of the cones into a hyperboloid, and when the two hyperboloids coincide, we get the two systems of right line generators. The same stand also has a model of a hyperboloid with only one set of strings. By turning the upper ring either way it deforms into a cone, in the one case with its vertex between the rings, and in the other with its vertex at a con- siderable height above the rings. Both these can have their upper rings moved along the top bar so as to incline the surfaces. We still get cones and hyperboloids, but it is only when the rings are horizontal and centre to centre, that we get surfaces of revolution. 13. Hyperboloid of one Sheet ; with its asymptotic cone. 14. Hyperboloid of one Sheet ; with its asymptotic cone. The tangent plane to the cone is also drawn. It meets the hyperboloid in two parallel right lincfit. One of these right lines is the line of contact of a hyper- bolic paraboloid witli the hyperboloid, and the tangent plane is one of the director planes of the paraboloid, both systems of generating lines of which are exhibited. 9 15. Hypekboloid of one Sheet. A slight variation from No. 14. The paraboloid only shows one system of right line generators, and the tangent ])lane is made by parallel instead of radiating lines. 16. Hyperboloid of one Sheet, and its tangent paraboloid. This shows the transformation of a cylinder and its tangent plane into a hyperboloid and its tangent paraboloid. 17. Conoid with its director plane. The director curve is a plane curve. By shifting the position of the brasses, the conoids deform into different conoids or other allied surfaces. 18. Conoid with a director cone. The director curve is of double curvature. By shifting the position of the brasses the conoids deform into different conoids or other aUied surfaces. 19. Conoid showing both sheets of the surface. By shifting the position of the brasses the conoids deform into different conoids or other aUied surfaces. 20. Conoids. Model showing the transformation of a cylinder into a conoid and back again. Also model showing the transformation of a cone into a conoid and back again. It is to be noticed that the head-lines of the two conoids, that is to say, the right hne in which the two sheets of each conoid meet, are perpendicular to one another. The transformation is effected by making the upper semi- circle turn through two right angles. 21. Conoids. Intersection of two equal conoids having a common director plane. The horizontal intersection is a plane eUipse. 22. Conoid, in contact with a hyperboUc paraboloid. 23. Conoids. Two equal circles in parallel planes, divided equi-distantly, are connected by threads, so as to form four surfaces. A cylinder. A conoid. A cone. A second conoid. The director planes, as well as the head Hnes, of these conoids are at right angles to one another. 10 24. Conoids. Two equal circles in parallel planes are connected by threads so as to form four surfaces. A cylinder. A cone. A conoid. A second conoid, with its director plane and line at right angles to those of the former. Same arrangement as No. 23, except that the lower ring is replaced by a plane of section a little higher up. The section gives, — For the cone, a circle smaller than the upper ring. For the cylinder, a circle of the same size as the upper I'ing. For the conoids, two ellipses turned crosswise. 25. Model exhibiting the simultaneous transformation of a conoid into a cylinder, a cylinder into a conoid, the paraboloid touching the conoid into the tangent plane of a cylinder, and the tangent plane of a cylinder into the tangent paraboloid of a conoid, and reciprocally. The changes may be arranged as follows : — From. Conoid. Tangent paraboloid. Cylinder. Tangent plane. Into. Cylinder. Tangent plane. Conoid. Tangent paraboloid. These changes are all effected simultaneously by one move- ment, which can be reversed. 26. Model exhibiting the transformation, first, of a conoid into a cylinder. Second, of the tangent paraboloid of the conoid into the tangent plane of the cylinder. 27. French Skew Arch (biais passe). The inner drum, of yellow thread, represents this surface. It is a skew sm^face, with a right line du-ector ; and its faces, the planes of the two semicircles, are usually parallel, although the model permits them to be placed obliquely to one another. The horizontal line joining the centres of the two large semicircles is the right line director. The construction for any one of the generating lines is as follows : — Draw a plane tlu'ough the right line director at 11 ' any selected obliquity. It will, of course, give the radii of the outside circles, and the line joining the points at which it cuts the inside semicircles will be a generator of the surface. This line will evidently pass through the du'ector line, because it is in the same plane with it. In stone or brickwork, the sides of the voussoirs will be given by the auxiliary plane in question. When the openings are parallel the voussoir joints are therefore plane, and the simplicity thus gained is the chief reason for adopting this form of skew arch. It is usual to take the right line director perpendicular to the openings, and sym- metrical to them ; that is to say, passing through the middle point of the parallelogram of the springing plane. When the openings are not parallel, the voussoir joints shown by the model are deformed into hyperbolic parabo- loids. This deformation, is, however, very slight, and in practical work w^ould be avoided altogether by adhering to the principle of drawing a plane through the director line. The spacing of the voussou's is usually determined by dividing the outer semicircle into equal parts. This form of arch is inconvenient w hen the obliquity, and the length of the barrel are excessive, for the generators are not generating lines of the cylinder containing the opening semicircles, but chords of it, and, therefore, at the middle, falling considerably inside it. The arch therefore droops in the middle, and this would be ugly and inconvenient if the proportions were excessive. It is interesting to compare this surface with the skew vault of Marseille {arriere voussure de Marseille), an example of wliich is shown in the set of plaster models contributed by the '* Brothers of the Christian Schools," and another, in imi- tation brickwork, among M. Schroder's models of fm-naces. In this case the curvihnear directors do not tally with one another, although they remain parallel, and the right line director is a Vertical hue behind the smaller arch. The con- struction for the right line generators is the same for both, namely, to consider an auxihary plane pivoting about the right line director. 28. Staircase Vault for a square well {vis St Gilles carree), 29. Staieoase Vault. Model for exhibiting some pro-^ perties of this ruled surface, by showdng how it is obtained from the defonnation of a cylinder {douelle de .la vis St, Gilles carree). 12 30. Cylindek with Helix and DEVELorABLE He- LIXOID. The helix is simply a screw thread. The developable helixoid, shown by the purple threads, is the surface swept out by the right line tangents of the helix. If we consider that each gore can be turned a very little bit about the thread which separates it from the next gore, we see that the surface can be flattened out or developed into a plane, without any crumpHng. This happens because every two consecutive generating lines meet one another on the heHx. That is why its surface is called developable. Its section by a horizontal plane is the involute of the circle. The model allows the pitch of the heUx to be shortened by lowering the upper plate, and the cyUnder can also be incHned. When oblique, however, the curve which replaces the helix is not such a screw thread as can be turned in the lathe. 31. Skew Helixoid. This surface is described by a right line which always passes through the axis of a cylinder and makes a constant angle with that axis. It also passes through a hehxor screw thread traced on the cylinder. The model only shows the surface, not the cyHnder. It is the surface of what is known as the screw with a triangular thread. The section by a horizontal plane is the spiral of Archimedes. This is not the commonest form of the skew helixoid ; that is best seen on the underside of a screw staircase, or on the driving face of a common screw propeller. In these, two generating lines are at right angles to the axis. The surface may also be considered as generated by a line which makes a constant angle with a given fixed line, and moves up that line, and at the same time turns round it, at uniform rates. 32. Skew Surface with its tangent paraboloid, capable of transformation into another skew surface while the para- boloid deforms into a plane. This is (for a certain position of the lower semicircle) a skew surface with a director plane, the plane being vertical* The director curves are : one of them a circle divided equi- distantly, the other a semicircle divided so as to keep the strings parallel to the director plane. 13 Intersections of ruled Surfaces. 33. Intersection of two cones having double contact with one another, that is to say, having a pair of tangent planes in common. The consequence of their having double contact is that their curve of intersection breaks up into two plane ellipses. The vertices of the cones slide along a rule which turns on a universal joint. See also model No. 38. 34. Common groin. Intersection of two cylinders having a pair of common tangents. The model may be set square or oblique. 35. Intersection of two cylinders, one piercing the other so as to give two separate loops of intersection. 36. Intersection of two cylinders, having a common tangent, 80 as to give a curve having a double point at the point of contact. 3Y. Intersection of two cylinders, neither completely piercing the other, so as to give only one loop of intersection. 38. Intersection of two cones, having double contact, along a pair of plane ellipses. 39. Groin. Oblique intersection of two splayed vaults of the same spring. 40. A pair of intersecting planes, which, by pulling the brass ball so as to give simultaneous rotation to the two upper rods, deform into paraboloids first, and then into planes described by radiating strings. 41. Intersecting cyhnder and plane. By pulling the brass ball the head brasses rotate together, and the cylinder deforms into, first, a hyperboloid, and then a cone, while the plane deforms into, first, a paraboloid, and then again into a plane with radiating hues. 42. A pair of intersecting cyhnders on circular bases. By puUing the brass ball the head brasses rotate together, and the cylinders deform, first, into hyperboloids, and then into cones. 43. A pair of intersecting cylinders on irregular bases. By pulling the brass ball the head brasses rotate together, and the cyhnders deform, becoming at last cones. 14 44. Groin. Model showing the deformation of a common groin, both obliquely, and by splaying the vaults. The model shows not only the intersection, but the plans of the intersection and of the generating lines. 45. Helix or Screw-thread. Model showing the transformation of the right line gene- rators of a right cylinder into screw threads of various pitch or obliquity. The pitch of p, screw is the distance between two succes- sive turns, measured in a direction parallel to the axis. When this distance is small, the screw is said to have a fine pitch, when great, a coarse pitch or high pitch. 15 APPENDIX. By C. W. Merrifield, F.R.S., Principal of the Royal School of Naval Architecture and Superintendent of the Naval Museum at South Kensington. This Appendix contains an account of the application of analysis to the investigation and classification of mled eur- faces. It is not proposed to follow all the defoimations which those surfaces can be made to undergo in the arrangements illustrated by the models. That would take a very large volume, and, even so, could hardly be given completely. The analysis has been kept as simple as possible, and has been written out in the form which appeared best adapted to the consideration of surfaces, not with a view to their general properties, but specially to the particular mode of generation by means of straight hues. For that reason, no mention has been made of the cones and cylinders of the second degree. These are treated with sufficient fulness in all the ordinary books. The student may extend much of what is stated here by introducing the principles of elUptic deformation and oblique deformation. The latter is frequently equivalent to a change in the direction of coordinates. Both these transformations are applicable to nearly all that follows, and the student should bear this application always in mind. He will do well to work it out to its consequences in some of the simpler cases, for which the formulae are not unmanageably long, as they are apt to become if used injudiciously. Motion of a right Line. A right line is completely defined by the condition that it shall meet four fixed curves in space, to the extent that there is not an infinite number of right fines which will satisfy this condition. The condition, in fact, gives rise to an equation, which may have more roots than one, but in which each root will have a definite and not a variable value. Moreover, the root of the equation may be imaginary, and it may happen 16 that the geometrical Hne may also be imaginary. Admitting imaginary quantities, however, the line is definite. If the hne be only conditioned to meet three curves in space, it is free to move so as to trace out a curved surface, v^^hich is called a ruled surface. It may, as before, happen that the director curves are so chosen that no real hne can meet all' three in real points. If, therefore, there are three directors, one of them must be taken within certain Hmits of position or direction in order that the problem may be really possible instead of imaginary. The point is to see distinctly that a right line must have three directors, and three only, to trace out a surface. The surface so traced is called a ruled surface. It is possible to replace one of the director curves by some equivalent condition. But the conditions taken altogether must be such as are equivalent to the restriction imposed by three director curves, neither more nor less. The classifica- tion depends upon the selection of these conditions. Passing by the common cone and cyhnder, let us proceed to consider the next in order of simplicity. The Hyperbolic Paraboloid. This surface is traced out by a variable right hne, which always meets three fixed right fines, which are parallel to one plane.^ Suppose that the plane in question is made to pass through one of the lines, which is taken as the axis of x, and that this plane is taken for the plane of (x, z). Suppose, also, that one position of the variable line is taken for the axis of y. Then the other two fixed lines may be written as — y = *i \ y = ^h \ and the surface must evidently contain these. But thes^. equations give 1^X1/ -^ n^b^z — l^ xy + n.;^h^z — * This only introduces one condition among the director lines, for the plane is arbitrary, and therefore may be taken parallel to any two given lines. It is, therefore, only the third director which is restricted. 17 and these will satisfy the equation cz — xy, if "^ = "" IT ~ " IT conditions which we may always satisfy by a suitable choice of tiie axis of z. Moreover, it is clear that the equation cz = xy will also be satisfied by the other system of lines ar = « my ■\- nz ■=■ 0. . , , na proviuecl c = . ^ m Hence the equation of the surface may always be written as cz = xy, whatever point of the surface be chosen as origin. There is always one point on the surface for which, when the equation is so written, z will be perpendicular to {xy) ; but when that is so, it will not be generally true that x and y are at rio;ht angles to one another. -F A) 5 + (C;7 n Aj^ f = 0. (7.) which reduces to the same form as (()) if C r= 0. 30 • Ruled Surface with right line Director, The axis of z being taken as the director, the equations of the variable line may be written as whence and X =^ ky mx -\- n = z. dy \ y j^ = -,- = — = constant, ax k X _ dz dz dy dx dy dx Diiferentiating again, = ^+2^^ + ^^^^ dx'^ dxdy dx dy'^ \ dx) t , y dij Substituting - for -j- we obtain '^ X dx „ d^z ,^ d^z „ d^z ^ , ^ which is the equation of the surface. If we had taken any other right line than the axis of z for the director, we should have had the constants of its equa- tions appearing in the differential equation (as in equation No. 7 of the previous section). Equation (8) is therefore not a general one, but a restricted form, in which the con- stants have received particular values which simplify the result. Ruled Surfaces considered generally. We have seen that we cannot express the equation of a ruled surface in a functional form, without differentiation. This happens because the simplest expression of a ruled sur- face is, that its tangent plane at any point shall meet it in a right line, or, what is the same thing, that one of its tangent lines at any point shall be wholly in the surface. Now the question of tangency is emphatically a question of differen- tiation. In what follows it will be convenient to use the ordinary abridged notation of partial differential coefficients, namely, _^dz _dz ^ ~~ dx ^ ~~ dy 31 dx^ ' ~~ dxdy ~~ dy^ ~~ dx^^ ~ dx^dy d^z ^-^ ~ dxdy'^ ~ dy^ A ruled surface is expressed with complete generality by considering it to be traced out by the motion of the gene- rating line 'O y = c^x + c^ in which the four quantities c, which are constant so far as the equation of the line in any one position is concerned, but variable parameters, when considered with reference to the position of the line, are to be made to disappear. The obvious way of effecting this is to obtain, by means of implicit differentiation, a relation between the partial differential co- efficients of 2, which, when thus cleared of what is special to the particular generating line, will be tlie differential equa- tion, in partial differentials, of the surface. Operating implicitly upon the above equations we get dy dy whence p -\- qc^ = q (1.) Differentiating again upon the same suppositions we have r + 25^2 + tc,;^ = 0. (2.) Differentiating a thkd time, we get a -f 3/3^2 + 370.2 + ^c.^ = 0. (3). If now, by ordinary algebra, we eliminate c^ from equations (2) and (3), we get the ditferential equation of ruled surfaces. It is not worth while to write it down, as it is more conve- niently used in the impHcit form given above, than in its very cumbrous expHcit form. We liave already noticed that the general form of a ruled surface cannot be expressed as a single functional equation. It follows that the differential equation has no general primitive. 32 Developable Surfaces, If we consider a stiff card, of the form a X Yf E in the figure above, to be deeply scored along the right lines B A h, C B c, D C f/5 E D e, so that we can bend it along each of them, the broken line ABODE will form a polygon, at ■first plane, but a skew polygon when we come to bend the the surface, which will then form a polyhedron, every edge of which will run into every successive edge, along the broken Kne ABODE. It is evident that this condition is necessary to our being able to deform the surface. For if one of the scores (say c B) stopped short of the edge ABO, the card would not bend, and if A B were not an actual edge, in that case it would not bend either. Now if we consider a surface which can be formed by the gradual bending of a plane surface, the only departures from this type are, — (1.) That the polyhedral surface is replaced by a curved surface. (2.) That the polygon A B D E is replaced by a cui^e. (3.) That instead of a finite bending along a few lines A «, B J, c, &c. we have an infinitesimal bending along an infinite number of such lines infinitely close together. - (4.) That all these lines are tangents to the curve ABO... which replaces the polygon. (5.) That the consecutive lines A «, B Z>, meet one another ; that is to say, the shortest distance between them is an infinitesimal of a higher order than the dis- tance between any other two points of them. This is imphed in their being tangents to the same curve. 33 If the curve ABODE degenerates into a single point, there will occur some singularities which will mask the general properties of these surfaces. This degeneration gives conical surfaces. When, further, the single point in question moves off to infinity, we get the cylindrical surfaces. This is not all. If we take two cards, counterparts of one another, and glue them together along the polygon A B C P E, we can deform the double plane into a double polyhedron, the two sheets of which will be placed back to back. There will then be a finite angle between the corresponding gores in the two sheets. When we replace this polygon by a curve, this angle will be infinitesimal, and a section at right angles to A B C D E will be a cusped curve. Each sheet will evidently be a ruled surface, the variable right lines of which will touch the curve which replaces the polygon ABODE. It is evident that the right line generators must otherwise stop abruptly, for they could not get from one sheet to the other on any different conditions. A good imitation of this can be made by taking two sheets of paper, fitted over one another, as A B D, cutting out a curved piece (E F G) from both, and laying a smear of strong glue along E F G, so as to fasten them together ?'J'J •J2, 34 along tliat curved side only. Then take one sheet up by the comer A or B, and let the other sheet hang freely. If there be no crumpling or stretching, we shall thus obtain a developable surface of two sheets. A tangent to the curve E F G, such as K F L, will be partly on one sheet and partly on the other. The curve E F G which connects the two sheets is called the edge of regression. It is not a very significant name, being a bad translation of the French arete de rehroussement. Still it is the usual name. In the case of cones, this curve degenerates into a point It then ceases to be a necessity that the angle between the two sheets should be infinitesimal, and, accordingly, this angle is then generally finite. The analytical criterion of a developable smface is derived from the consideration that its generating lines must be tan- gents to a curve in space. In analytical language, they must have an envelope. It must, therefore, be possible to get rid of the parameter which distinguishes one line from another by differentiating with regard to that parameter. Going back to equation (2), namely, r + 2sc2 + tc^^ = 0, in which there is only one parameter, Co, and differentiating with regard to that, we get s -\- tC.^=zO Now eliminating Ca by ordinary algebra, we obtain, as the differential equation of ruled surfaces, rjf-/-= 0. The geometrical interpretation of this is that a section parallel to the tangent plane at any point, and infinitely near to it, is always a parabola. In other words, the curvature at every point is parabolic. To resume then, the differential equation of ruled siu'faces* is the result of eliminating c between r + 2sc + ^c^ = and a + 3/3c + Syc^ + hc^ = * It seems at first sight rather singular that the elimination of a single constant should introduce four new partial differentials ; but it must be recol- lected that the partial differentials are here only accidental. The work we are doing is implicit differentiation, and the variation of the whole equation simply amounts to the introduction of one new element. This consideration seems to dispose (affirmatively) of the question whether the resulting equation represents ruled surfaces only. The converse question, whether ruled surfaces are included in it, is evident at first sight. 35 while that of developable surfaces is rt = s\ It will be observed that this is the condition that tha first equation should give two equal values for c. Cones and Cylinders, The differential equations previously obtained are not those which distinguish these surfaces from other ruled or developing surfaces. On the contrary, they are restricted by conditions which settle the position of the vertex of the cone, or the direction of the generators of the cylinder. They are consequently expressed by differential equations of the first degree. If we desire to eliminate the constants of the vertex, or of direction, we must proceed to third clifferentials. Writing the equation of the cone as and regarding x and y as independept variables, it is easy to verify that «y - f _ a§ — h __ iSi_^■_^^ r ""* 25 ■" T~ ' but these equations are also derivable from the general equation of developable surfaces, rt = s-, and do not distinguish conical surfaces from all others. Jf again we write the equation of the cylinder as ny — mz = Y {mx — ly) we find that the numerators of the above fimctions vanish, or, as the I'esult may be more simply stated — = -= - i3 7 5 and this appears to distinguish the cylindrical from other developable surfaces. Tlie theory of this part of the subject is recent, and far from complete. Meanwhile it is certain that what goes before is strictly true as stated ; but the student must be cautious of drawing inferences which go beyond the text. For instance, Mr. Cayley has shown that the equations do not represent cylinders only, or even ruled surfaces only. It is when taken in connexion with the equation of develop- able surfaces that they represent cylinders. 29992. D 36 The complexity of these results was to be expected. The general equation of cones involves an arbitrary function and three arbitrary constants, while that of cylinders involves a function and two constants. Now the ordinary practice in the formation of differential equations is to consider arbi- trary functions on the one hand, as yielding what is called a general solution, or primitive ; and arbitrary constants on the other hand, as yielding a complete solution or primitive. The simultaneous elimination of functions and constants would naturally be more complex, and the result at which we have arrived, namely, that it leads to the existence of simultaneous equations in differentials, is only reasonable. The equations of the cones ay — /3^ _ aS — ^7 __ /3S — y^ V~ ~ ~^s - t in virtue of the relation rt =5^, lead to the equation of the third degree 4 {ay - f) (^S - y') = (.S - ^yY- This equation is more general than that of conical surfaces. Its integration has not yet been effected, but it is known to be satisfied by surfaces traced by common parabolas inter- secting consecutively ; moreover, it is easily shown that it is also satisfied by the equation of the ruled surface with a di- rector plane, a surface wliich is not generally developable. This can easily be verified by taking the director plane parallel to the axis of z^ when it may be wTitten as z =. x¥ {ax -f hy) + yf {ax + by) + (p (ax + by). The result, being invariant, will not be affected by a change of co-ordinates. The question lias not yet, however, been fully studied. The Screw, and Surfaces connected with it. If a cylinder is put into the lathe and turned with a steady motion while a tool travels along in a direction parallel to the axis of the cylinder, also with a steady motion, the curve traced upon the cyhnder is a screuj thread, or helix, as it is sometimes called. It is the only curve in which any portion can be super- posed upon any other portion so as to fit exactly. The only curve, because the circle and straight line are simply extreme cases of it. If the tool is still as well as steady we get a 37 circle. If the cylinder is still while the tool travels we get a right line. The most convenient form of its equations is x = a cos ^, 2/ = aein dy z =cS. Its tangent, of com^se, touches the cylinder x^ -{- i/^ = c? and makes an angle with the axis whose trigonometrical tangent is a -* The radius of absolute curvature is evidently the same as that of the elliptic section of the cylinder passing through the tangent. The minoi' semi axis of this ellipse is «, and its major semi axis s/ d^ + c^. c? + c^. Hence the radius of absolute curvature is * a If we eliminate a and d we get a surface which is indepen- dent of the diameter of the cylinder, or of the inclination of the particular helix got by varying the diameter of the cylinder at the same time that the pitch remains constant for all values of a. This is called the shew helixoid. It is the underside of a common screw staircase, or the driving side of a common screw propeller. A mooiing screw and the twisted surfaces of a square-threaded screw are other examples. The equation is y . z ^ = tan -. X c Practical reasons made it inconvenient to include the common form qf this surface in this series of models. More- over, the examples of it are too frequently met with in practice to render its exhibition worth encountering an incon- venience. There is no difficulty in representing it by strings or threads; but there is a difficulty in deforming it with regularity. The skew helixoid selected for representation (model No. 31) is one in which the generating line passes through the axis, and is inclined to it at a constant angle instead of being at right angles to it. It is the surface of what is called a screw with a triangular thread. If we call ^ the angle which its generating line makes with the axis, this generating line will pass through the point of the helix (a cos ^, a sin ^, c 6), and also the point of the axis »= c6 — a cos x" + y- ay When the surface is actually developed or flattened, the helix becomes a circle whose radius is that of the absolute c? + c2 curvature of the helix, that is to say, — . The French Skew Arch. {Biais passe.) See model No. 27. Take the line joining the centres of the large settiicircles as Jc, the middle point of it for origin, and the axis of z, vertical. Then if we call the radii of the small circles c and their distance from the axis of x, b, their eq[uations may be wi'itten as a:=a, (j/ — by -{- z'^ z=z c- x=^a, (i/-\-by-}-z^ = c^ A plane through the axis of x, inchned to the vertical at an angle (p, will he i/ = z tan ^ : substituting y cot p for z in the above equations and solving for y, we find for the extremities of a generating line X = Of 1/ s= i) sin '^^ + sin