Magnetic Klein Quartic

October 2, 2011

The Klein Quartic is a absolutely fascinating object and worthy of a post in its own right, or even a book. It is clear evidence of the explosion of imagination and creativity in geometry that was taking place in the nineteenth century, as it cut its ties to the “real world”. Since then it has turned up all over mathematics.

One way to consider the Klein quartic is as a generalisation of a regular polyhedron. The tetrahedron has three equilateral triangles meeting at each corner, the cube has three squares and the dodecahedron three pentagons. Three hexagons gives a tiling of the plane. Why stop there? What about three regular heptagons? There are important reasons why this does not work in a simple manner. By playing fast and loose with what we mean by “regular heptagon” however we can do something. One object we can make is the Klein quartic. It does not produce something like a sphere, as the tetrahedron, cube and dodecahedron do, instead it is more like a pretzel with three holes.

Combining these ideas with little spherical magnets, we can make a model of the Klein Quartic. To do this we obviously have to start by making a heptagon

You start with a ring of seven balls, then put another ring of 14 balls around it. Note as this happens the heptagon buckles into a saddle shape. This is because the balls naturally create angles of 120˚ at the corners. As we move round the shape therefore we turn through a total of 7*120 = 480˚, this is greater than 360˚. We say the resulting surface has negative Gaussian curvature. We may also consider the length of the second loop. It is roughly distance 2 from the centre of our shape, yet it has length 14. If it were a circle of radius 2 the circumference would be 2*2π, which is less than 14.

Two of these heptagons can fit together on an edge:

For fans of Indra’s pearls and sphere reflections the balls make a pretty pattern.


As the angle at the corner is 120˚ three will fit round a corner:

We could now continue this, bringing three heptagons together at each corner, but we want to create the finite object. Next attach an additional heptagon to each of the outer three:

Now connect the three outside heptagons together. to make a surface with three holes:

You need to repeat this four times, using a total of 24 heptagons. As you make them, be careful of one thing, the magnets line up so that you get  all N poles on one side of the surface and all S on the other. As you connect each surface, therefore, make sure that it agrees with the others:

When you have all four, put one at the center and then connect the others to each of its four holes

To finish, technically we should connect up the remaining six holes so each branch is connected to both the others. The resulting shape has three heptagons meeting at every corner, and a wonderful collection of symmetries many of which cannot be easily seen in this model, or any model in 3d!

Just for kicks, lets finish with the work of one of Klein’s contemporaries a Möbius strip:


CAMel

September 29, 2011

CAMel is a project to develop Rhino Grasshopper components for CAM (Computer Aided Manufacturing). Hence the silly name. It is very much work in progress, but if you are brave enough, here is a first release. All images and the video on this page are of a machine running GCode generated by CAMel.

Download CAMel 0.12
Download Rhino file (only needed if you want to see the example setup).

At present the components are just clusters with scripted components written within Grasshopper. The next major step will be to convert this into a proper grasshopper plug-in. This release has a grasshopper component with some documentation (there is a little more inside the clusters). All the code is CC-BY-SA licensed, and of course it should be noted that this is very much “use at your own risk”! My belief is that Grasshopper provides a natural environment to experiment with creating your own toolpaths. The purpose of CAMel is to make this process as easy as possible by giving the tools to convert simple toolpath ideas into usable paths and then exporting the GCode that will drive a machine.

The main components are as follows:

  • GCode Writer: Converts lists of points, vectors and feed rates into GCode for the machine.
  • GCode Checker: Reads GCode and checks and optimises it. For example a 5-axis machine can usually obtain any tool angle in two different ways. This selects the better angle. It will also give warnings of undesirable behavior in the GCode.
  • Surfacing: Creates a toolpath to cut an arbitrary surface (very rough version, designed to test others)
  • Swarf cutting: Creates toolpath from information about the movement of the tip of the tool and the point in which the tool enters the surface. For a 5-axis machine these paths can be quite different.

The code is currently set up for a single machine, I am happy to try to help adapt it to other machines (other commitments allowing) so get in touch if you are interested.

These components and code were developed with Santiago R Perez 21st Century Chair of Integrated Practice at the Fay Jones School of Architecture, University of Arkansas. I work in the Mathematics Department at the same university.

Earlier experiments with swarf cutting.


Hexayurt dome details and models

August 7, 2011

People are now starting to build my tri-dome and quad-dome versions of the hexayurt, so it is time to give some of the technical details. To start, however, here is an application of the intermediate value theorem!

When I started working on the details for the tri-dome I realised I had made a bad assumption (thinking that the form was geometrically pure). This means that some of the details in my original write up were wrong. All a little embarrassing. Ironically, I might have missed a form that does actually work, had I not made the bad assumption. The shape, like the hexayurt, starts with a hexagonal based pyramid. In a traditional hexayurt this lies on top of a hexagon of vertical walls. Instead of this we attach a square to three of the edges and the classic hexayurt triangle (isocoles triangle with base and height the same length) to the other three. We can look at what happens as the pyramid is moved away from the ground, while the edges of the shapes rest on it:

This does not give a great building; there are holes. The holes are triangles and two of the sides have a fixed length. The final edge changes length, starting long, and ending short. We know we can fill the holes with classic hexayurt triangles. Two of the edges are the right length we just need the third. The length changes smoothly as we raise the roof, and starts shorter and ends longer than we want. Here we can apply the intermediate value theorem, so the correct length must be passed. As a mathematician I would stop there, the system works; however people are building the things…

So to calculate the correct angle for the square sides of the model we can look vertically down, calling the angle of the square face θ, (and assuming that the boards we are using are 8′ by 4′) needing as the classic maths problem asks to “find x”.In this case

x = 4 \sqrt{4 \cos(\theta)^2+1+2\sqrt{3}\cos(\theta)},

we want x = 4\sqrt{5} so:

4\sqrt{5} = 4 \sqrt{4 \cos(\theta)^2+1+2\sqrt{3}\cos(\theta)}

5 = 4 \cos(\theta)^2+1+2\sqrt{3}\cos(\theta)

0 = 2 \cos(\theta)^2 + \sqrt{3}\cos(\theta) - 2

Solving the quadratic:

\cos(\theta) = \frac{-\sqrt{3} \pm \sqrt{19}}{4}

Which gives an angle of about 49°, and the height of the roof (assuming 4′x8′ panels) is 8 \sin(\theta), just over 6′ at the edge and 10′ in the centre. We can use these, and useful facts about general tetrahedra to calculate all the angles between faces by using the lengths of their edges. If you want to follow the details yourself, you need to add vectors to get some of the edge lengths, then use the Cayley-Menger determinant to find the volume of the tetrahedron, and then the generalised Sine rule to (halfway down this page) to give the angle.

Technical details for TriDome: angles to nearest half degree, lengths to nearest inch (assuming 4'x8' panels). On the left the angles between faces and point heights, on the right lengths and angles of the base.

Technical details for QuadDome: angles to nearest half degree, lengths to nearest inch (assuming 4'x8' panels). On the left the angles between faces and point heights, on the right lengths and angles of the base.

Finally here are the hexayurt models (rhino 3dm and vrml formats) of the hexayurt, H13, TriDome, QuadDome, plus a couple of others, including a very large one.


Arrange whatever pieces come your way

May 14, 2011

(with apologies to Virginia Wolff)

A simple, classic puzzle is to give two shapes and ask if there is a way to cut one up so the pieces can be rearranged into the other. This game might seem to become silly if both shapes are the same;  if we insist that the new arrangement must be different the game becomes interesting again. Think about it, can you come up with ways to cut up a square so that the pieces can be formed into two different squares? Here is an example, not with a square, but with a rhombus:Having the same shape has an advantage. Think about the letter p below, it is part of the blue trapezium, when we rearrange the tiles the p moves with the shape. As the two shapes are the same we can think of this new p within the original rhomb. We can now repeat the process as many times as we want. In this case, it might be a little unsatisfying, however, as the next step for our p would cut it into two different pieces, as it lies on the edge. So where is it safe to put a p so that it will never get cut up? To answer this we have to follow the cutting lines, and a beautiful pattern emerges:The p would be safe within any of the pentagons, but if it crosses any of the edes it will, eventually be cut apart.

Puzzle: Can you work out the difference between the green and the blue pentagons? (Hint: it relates to the dotted and solid lines in the earlier pictures).

Studying what happens when we can move points or objects around in a space (in this case moving p around a rhomb) is studied in a part of mathematics called Dynamical systems the particular example here is called a Piecewise Isometry  (see this paper for a more formal account of their history and study). I have studied these systems myself, and recently submitted a paper looking at the behaviour and number theory that occurs within the pentagon generating system shown above (take a look! It has lots of pictures as well as more formal mathematics).

As you might have guessed from my preoccupations part of my interest in these systems is the pretty images that they produce; this system is particularly rich. This leads to the image at the top. You can take any rhombus and cut it up in a similar way. Take any rhomb (as shown below) and rotate until the side of the rhomb lines up with the top. This will leave a triangle and a trapezium that can be moved back on top of the original rhomb:Additionally this gives a system where the rotation on the two parts is the same, just around different points. You have to be a little careful, but you can use this to give a system for any angles. For any of these systems we can ask the question: Where is it safe to write p? Every angle gives a different pattern, and tiny changes in the angle leads to large changes in the pattern, however the patterns do relate to one another in some ways, as you can see in this video:


Follow

Get every new post delivered to your Inbox.

Join 203 other followers