Artifacts of Research: On Singularities

Jonathan D. Chertok (Universal Joint) 1

Figure 1: Computer images of recreated series of Rodenberg models with lines of ruling inscribed in their surface. Chertok 2008.

1. Introduction
The models presented at SIGgraph are 1:1 recreations of a classical mathematical model collection originally made by hand in plaster in the 1860s (fig. 2 and 3). Unlike the original collection, these models were created using both mesh-based and NURBS-based mathematical modeling, Computer Aided Design (CAD) software.

Fig 2. Original collection housed at the University of Goettingen. The physical models were then created utilizing plasterbased and resin-based Rapid Prototyping (RP) technologies, with each creating the models from the 3D files. These technologies use the CAD file to either drop a liquid binder onto a powder or to direct a laser as methods to solidify the model material (fig 4). The collection of 23 white plaster-based RP models are recreations of a series of models designed to represent the types of singularities possible on a cubic surface. With a singularity effectively being an abrupt change in shape one can look for the "double points" (in German "doppelpunkt") where a surfaces basically passes through to the other side. A close look at the series' nomenclature (described elsewhere) will show that each model represents either an individual or combined examples of the various types of singularities.

Figure 3: Double contact sheet of the author's photographs of the original plaster models c. 1860. This series was created by Carl Rodenberg for his thesis under the direction of Felix Klein (1849 1925) the founder of modern topology. Effectively this was an attempt to catalog a part of the mathematical universe. They exhibit an elusive beauty that I would characterize as "sublime". In a sense they appear so rational as to almost be facts of nature, while at the same time one has the distinct sense that this rationality just slightly eludes comprehension.

website: http://www.universaljointdesign.com blog: http://www.formpig.com CONTACT: press@universaljointdesign.com

Fig 4. Rapid Prototyping (RP) recreations being produced Additionally presented in the Addendum are unique new models of the Clebsch Diagonal Cubic - a so-called "smooth" surface which is the "god-head" of the series from which the series of singularities originate. Effectively the Clebsch Diagonal Surface shows the 27 total possible lines on a smooth cubic surface while the 23 models shown in the Rodenberg series show "aesthetically pleasing" examples of the singularities that result when you remove some of these lines.

2. Exposition
Beyond their sublime beauty, the new models proved to be a wonderful entry point for contemporary architecture and contemporary construction issues related to digital technology. In this vein, three observations regarding technique are germane. First, the fact that the surfaces had to be modeled as mesh in mathematical modeling software (the only way to generate these zero sum implicit equations of the surfaces) while the straight lines were generated as NURBS lines in CAD software (from their parametric equations) resulted in the lines which were inherently more malleable than the surfaces. That is, their NURBS nature means that we could parametrically control them which is a possible avenue for future work.2 Second, the relative precision of the parametrically derived lines - when literally placed against the complicated curving mesh surfaces (generated by the Marching Cube Algorithm of the software) - allowed for a visual check of the algorithm, which would have been difficult to discern otherwise.3 Third, the ability to "flythrough" using a 3D mouse provided unequalled power for querying both the model and the relationships inherent in it. It also proved to be particularly useful for creating geometry particularly when adding the sphere to show the various types of intersections on the Clebsch Diagonal Surfaces Configuration. Thus, the use of CAD to work with these models provided obvious visualization and interactive benefits as compared to conventional and more static mathematical modeling software. Similarly, the ability to work with actual models provides obvious benefits (fig. 5 and 6). Utilizing newer two color transparent RP model technologies can provide even more unique feedback in these and other respects (fig. 9 and 10). Figure 6: Ruled Cubic Surface. Fig 5. Cubic with 4 - A1 Type Singularities

website: http://www.universaljointdesign.com blog: http://www.formpig.com CONTACT: press@universaljointdesign.com

3. Future Research
I'd like to explore the tendencies of the marching cube algorithm in most mathematical modeling software to either follow or not follow lines of principal curvature. In the case of the cubics presented here there is a curious tendency to appear to follow lines of principal curvature but to then throw in facets which "dirty" the mesh. Similarly an understanding of when one can expect planar quads using this algorithm has eluded me.; sometimes mathematical modeling software provides obviously planar quads and sometimes it does not. Also, there is a worthwhile kinematic exercise to explore with respect to the 27 lines on the Clesbch Diagonal Cubic. 15 of these lines may collapse to the plane (like those in a hyperbolic paraboloid) and I'd like to make some constructive experiments for large scale supporting formwork for concrete supports for tranportation infrastructure. On a computational note, it appears to me that the lines on the original surface (see fig. 7) may have been numbered by hand when they arrived at the various Universities in the 1800's or that the original numbering was destroyed over the years and redone. As in some cases this numbering is reported to be incorrect (more on this numbering in the Addendum). With help it should be possible to run a "script" that would generate all possible 36 combinations for the numbering of the lines on this surface based upon the actual CAD configuration and in conjunction with the historical documentation of the geometric relationships of these lines. This could then be used to check the numbering of the original models.

4. Summary
The intellectual seed that started this work was a posthumously published work by Robin Evans called The Projective Cast: Architecture and its Three Geometries (1998). And since I first photographed these model in 1999, Ive been intrigued by their sublime beauty. In the end I have concluded that this is partly a consequence of Evans' characterization in another context that by using more geometry they appear to have less.

Acknowledgements
I want to thank Professor S.J. Patterson for allowing me to photograph the original collection and for his continuous encouragement and assistance. A big thanks to Richard Morris for all his continuing help and the use of his software program SingSurf. Thanks to Stuart Dickson for his expertise and guidance. Finally I want to thank Giorgio Ferrarese whose great efforts allowed me to see this work to the finish line. The models were produced with support and funding from ZCorporation, Objet Geometries and the Imaging Technology Group at the Beckman Institute for Advanced Science & Technology at the University of Illinois.

Figure 8: Jim Blinn studying the author's documentation and taking a close look at the work at SIGgraph 2008.

Figure 7: New models on Flickr c. 2010.

website: http://www.universaljointdesign.com blog: http://www.formpig.com CONTACT: press@universaljointdesign.com

the surface continues upward, downward and beyond the extent of the cylinder that has been used to trim the exterior portion, then you have a good jump on understanding this surface. Just think of the model as hollow. Unlike the other models in this exhibition it has no singularity type associated with it, as it is non-singular; it is smooth. In the following we will describe the geometric configurations of this surface. I trust you will find both the models and the animation instructive.

Figure 9: Clebsch Diagonal Surface c. 1860.

Addendum: The Clebsch Diagonal Surface Configuration

1. Introduction
In the case of the Clebsch Diagonal Cubic we see a "smooth" cubic surface, as opposed to the singular cubic surfaces, which is in fact the "god-head" from which the other models of the singularities spring. Effectively, by removing lines from the Clebsch Diagonal Cubic, one can arrive at the models for the other singularities. The Clebsch Diagonal was originally discovered and documented by Alfred Clebsch (1833 1872). Both the transparent models and white plaster-based models you see are in some sense recreations of the original plaster model from the model collection that was originally made by hand in plaster in the 1860s. This model in particular has historical significance and has been touched upon by a number of early machine designers in addition to mathematicians.4 While no documentation of the original fabrication exists, we do have the original recipe for the modeling clay. In addition to serving as an interesting analogue to contemporary RP materials, it is quite charming as it calls for white blotting paper, 1 Liters of river water, 10 Marks of essence of lavender, 10 Marks of essence of clove, a tin container and a porcelain bowl among other things. Like the other models in this exhibition, you are looking at the outside of the surface. If you can visualize the fact that

Figure 10: Clebsch Diagonal Surface c.2008

2. 27 Lines
The surface has 27 lines on it. This is the maximum possible number of straight lines on a cubic surface if, as a geometer would say, "the lines are finite". If there are infinite lines on the surface then it could be any number of ruled cubics - four of which are shown in the Rodenberg Series. This is one of the reasons this model is so famous. Anyway, these lines are straight lines but for mathematicians it is only necessary to say lines. Oddly in NURBSNURBS-CAD we say "curves". The lines are either skew - they pass each other without intersecting - or they are co-planar - and thus they intersect. There is a special case of co-planar, which are parallel lines, but there are no parallel lines on the Clebsch Diagonal Surface. 12 of the lines from the 27 were studied by Ludwig Schfli (1814 1895). These are called Schflis Double-Six in English. This arrangement consists of six pairs of skew lines (remember this means they do not intersect). These are numbered 1, 2, 3, 4, 5, 6, and 1, 2, 3, 4, 5, 6. A pair would be 1 and 1 or 2 and 2 and so on.

website: http://www.universaljointdesign.com blog: http://www.formpig.com CONTACT: press@universaljointdesign.com

Most of the lines however cross each other. Lines numbered 1 6 get crossed by lines numbered 1 6. The only exception here is that lines with the same number (e.g. 1 and 1) do not touch each other. Again, these are the pairs of skew lines. Thus each line from 1 6 is coplanar (they intersect) with each line from 1 6 (excepting its prime as stated earlier). Similarly, each line from 1 6 is coplanar with each line from 1 6 (again excepting its non-prime). In the case of the Clebsch Diagonal Cubic, only four of these five intersecting lines intersect within the model and one intersects lower down beyond the base of the model.

The three horizontal planes are simply pointing out three sets of three lines that are part of the 15 Diagonal Lines. The three angled planes are formed by three pairs of lines, which are pointing out the other remaining lines from the 15 Diagonal Lines.

Figure 12: Clebsch Diagonal Surface with imprinted and numbered lines c.2008 James Joseph Sylvester (1814 1897) discovered the Sylvester Pentahedron, which is a five sided object described by the bottom two horizontal planes and the three angled planes. One triangle at the top and one at the bottom and four isosceles trapeziums as the sides (just think of these as triangles with their tops chopped off). Apparently this is important because the planes of the sides of this pentahedron can be described by equations which can also describe the equation for the actual surface of the Clebsch Diagonal Surface. For more on this we would need to consult a geometer.

Figure 11: Clebsch Diagonal Surface c.2008 There are 15 other lines called the Diagonal Lines. These are 12, 13, 14, 15, 16, 23, 24, 25, 26, 34, 35, 36, 45, 46, and 56. These run horizontally and vertically. These lines are given two digit numbers. Line 12 intersects the plane described by lines 1 and 2 and the plane described by lines 1 and 2. Similarly line 23 intersects the plane described by lines 2 and 3 and the plane described by lines 2 and 3. One does not need a line 21 as it is coincident with line 12. Similarly for 32 as it would be coincident with line 23.

4. Passages
Lets return briefly to this idea that this is a surface with two sides and look at the idea of Passages which was a classical idea that was apparently used "intuitively".5 There are seven passages in the model a fact that is related to the Coxeter-Dynkin Diagram which we will discuss in another paper. Three are just obvious. They are the three holes in the surface. The other three are a little higher, rotated by 60 degrees and up at the ears. Actually these three passages help us find the seventh passage. Just jump into one of the ears, continue toward the middle and then drop down through the waist this is the seventh passage.

3. Additional Geometric Objects

There are 30 points through which 2 pairs of lines intersect. 24 of these exist within the extent of the model. 6 are beyond the extents of the model. This makes for a total of 12 lines that have 5 intersections along their full length. We discussed these earlier when we talked about the numbering. There are also 10 points at which 3 of the 27 lines meet. These are called Eckardt Points. 7 of these intersections exist within the extent of the model and 3 are outside the model.

website: http://www.universaljointdesign.com blog: http://www.formpig.com CONTACT: press@universaljointdesign.com

14. Coxeter-Dynkin Diagram

Figure 13: Clebsch Diagonal Surface c. 1860.

5. Numbering Script
In total there are 36 different ways to number the lines on the Clebsch Diagonal Surface according to the rules above. Research indicates that the numbering on the original models was done at the Universities that received the model, and not at the point of manufacture. As a final exercise I intend to run a computer based script in Rhinoceros 3D modeling software defining all the possible combinations of numbering by querying the CAD geometry and the documented relationships inherent in the geometry. This exercise would provide an opportunity to verify the various numberings labeled by hand on the various models in collections around the world. 14. Still from Animation of the Clebsch Diagonal.

6. Conclusion
In summary, this portion of the research is model based exercise, rooted in mathematics and CAD, and I trust that it will provide interesting points of departure for future work. Jonathan Chertok May 2008 Austin, Texas

chertok@universaljointdesign.com http://www.universaljointdesign.com There is a range of mathematical software that will generate singularities in a parametric way along with their lines and this is definitely an area which we would like to research.
2 3 As

Professor Samuel Patterson pointed out to me, the lines on the original plaster models are in fact highly curved and meandering due to the imprecision of the plaster model surfaces, which was related to practical issues in making the model.
4

Ferguson, Eugene S. 1962 Kinematics of Mechanisms from the Time of Watt.

5

Fischer, Gerd (1986) Mathematical Models: Photograph Volume and Commentary. Braunschweig / Wiesbaden: Friedr. Vieweg & Sohn.

website: http://www.universaljointdesign.com blog: http://www.formpig.com CONTACT: press@universaljointdesign.com