Context
The following project was completed in early 2009, during the second stage of a three month residency at EKWC (European Ceramic Work Centre) in The Netherlands where the focus was to explore Ceramics in Architecture in a collaboration between an artist and a designer. I'm an architect and I collaborated with Del Harrow, a ceramic artist. Our particular aim was to explore digital methods for design and/or production along with traditional methods. Provided here are some insights into the process and results of one of the projects.

Research
The project shown here extends the work performed during the first stage of residency, in the summer of 2008. Del and I were inspired by the research of physicist and Harvard doctoral student Peter Lu into Girih tiles, commonly known as Penrose tiling. Lu described a conceptual breakthrough that occurred around 1200CE when tile patterns were “re-conceived as tessellations of a special set of equilateral polygons” in Islamic architecture. This allowed for precise patterns to be developed over large surfaces using aperiodic (non-repeating) Penrose patterning.

Our goal then was to create a tessellated screen wall for extremely arid surfaces. We decided to explore Penrose shapes to create interlocking clay extrusions. Our hope was to use these shapes to provide a visually engaging geometry while simultaneously meeting the performative goals of a load-bearing system that would provide cooling, ventilation, and light filtration using a method of horizontal stacked open-celled units.

Second Stage Exploration
During the intervening months before we returned for our second stage at the EKWC, I was increasingly intrigued by the interlocking nature of Penrose geometries and their potential to create tiles that would tessellate over potentially doubly-curved facades. What follows is the work I did independently when we returned in early 2009.

Rather than create shallow tiles that might be pressed, I wanted to explore the potentials afforded by CNC methods with deep patterning, yet being lightweight. This suggested a production process utilizing slip-casting.

Special thanks to EKWC and all the staff for a superb experience!

• 
• Analysis of tiling at Gunbad-i Kabud tomb tower, Maragha, Iran (Lu, P.: 2007, Steinhardt, P., “Decagonal and Quasi-Crystalline Tilings in Medieval Islamic Architecture”, Science 315, pp 1106-1110.)
• 
• Analysis of tiling at Gunbad-i Kabud tomb tower, Maragha, Iran (Ibid.)
• 
• Geometry Creation: Wire Deformations in Maya of Penrose Shapes. My first step was to create the desired vector geometry in 2D CAD. I then imported these into Maya, turning each tile into NURBS surfaces using the Wire Deformer function to deform the geometry to create the desired shapes and depths. Note that these are very simple surfaces at this stage.
• 
• Digital rendering in Maya of tile aggregation. This pre-visualization was enough to move ahead, even without showing the detail I envisioned for each tile.
• 
• CAM Tooling: Visual Mill Preview of Dodecahedron - radial pattern with bullnose bit and large stepover. Each tile was then converted into a polygon STL file which was imported into Visual Mill - a CAM (Computer Aided Manufacturing) software where the tooling for a CNC (Computer Numeric Control) mill was performed. My experience has been that CAM software packages permit significant design opportunities for creating effects which are often overlooked and exploring these was my intent. Among other manipulations, one can easily control the ways in which the CNC mill will move, the stepover of the milling paths, and the various contours which can be created when using different bits.
  
  I needed a "positive" shape of each tile (a piece that resembled the tile) to then cast a "negative" in plaster, creating the appropriate void for the slip-casting process. This required milling a "negative", which I then would cast a positive into. To achieve the desired outcome, I then explored a variety of tooling paths, stepovers, and bit shapes.
• 
• Digital rendering in Maya of inverted Visual Mill preview. My previous tooling work was done using MasterCAM for the CNC file preparation. However, for this project, Visual Mill was the software that was available. Initially doubtful, I discovered that I rather like Visual Mill, finding it almost as powerful as MasterCAM and significantly more intuitive. Like other good CAM software, it allows you to pre-visualize the final result which will be produced by the CNC tool. Another superb benefit is that one can export a polygon STL file of a tooled file (like the one above), then import this into your choice of 3D modeling software for manipulation and/or rendering.
  
  This technique was extremely useful in helping me select my final milling strategy and I created a number of different toolpaths to evaluate. With milling being a subtractive method, one can obviously only view the negative results. Clearly, this is a pre-visualization challenge when one is going to cast into this milled mold to create a "positive" result. To see the digital "positive", I exported the tooled Visual Mill preview file (as a polygon STL) and brought it into Maya. I then flipped it over and deleted all of the surfaces that are orange in the previous image (which represent the solid milling stock), leaving behind the white contoured tile, but upside-down (see above). This process also helped me discover the little “peaks” around the edges, which I removed by making an important tooling adjustment before the final CNC milling.
• 
• Proof of concept milling test number one. This involved testing my milling strategy and attempting to cast into it - a process I planned to do with plaster. At this point, I was not entirely confident of the digital preview as a reliable method to forecast my desired stepover results, nor was I sure I would be able to remove my plaster cast from the foam mold. Similar to a photographer using one f-stop that is higher and lower than the chosen value, I picked the stepover I liked best and then created two more milling files, with both larger and smaller stepovers. I then created three different CNC milling files, thinking I would pick the one I liked best. Papa Bear, Mamma Bear, and Baby Bear, if you will.
• 
• Casting the Positives: Plaster positive cast into milled foam negative. To facilitate removal of the plaster casts from the foam mold, I embedded wire into the back of the plaster. However, even with the wires in the plaster, I was unable to remove the cast pieces from the foam mold. Upon the advice of experienced mold makers, I carefully drilled a hole from the back into the foam, then blew air compressed air into it and the plaster parts popped out.
• 
• Three Proof of Concept molds in plaster of various stepovers. Note the holes in each mold to permit removal with compressed air. While I was pleased to find that the middle “Mamma Bear” version (selected from the preview of the Visual Mill tooling file) was in fact still the one I wanted to use, the problem I faced was that due to the combined nature of the foam and the scalloped tooling pattern, the plaster molds had many edges, and all surfaces were “fuzzy.” Conferring with the excellent staff at EKWC, it was felt that both the scallops and all of the little “fuzzy edges” were tightly gripping the mold, preventing the plaster from coming out. They feared that if I used these plaster positives when casting the larger “negative” plaster molds for the slip-casting process, I would not be able to remove them - especially for the larger tiles.
• 
• Silicone Molds: Peter Oltheten and me performing vacuum casting with silicone rubber. The final negatives were milled in foam, but instead of using plaster to cast the positives, we cast them in silicone rubber. This way, there would be some flex in the parts, allowing them to be removed from the plaster slip-casting molds. Since silicone easily entraps air bubbles, the only way to remove bubbles from the two-part silicon rubber material is to perform the casting in a vacuum. The odor from the silicon rubber was highly noxious, requiring pressurized air filtration as you can see above.
• 
• Five silicone rubber positives of Penrose tiles for slip casting molds - one for each Penrose tile type - which I used to create the two-part slip-casting molds in plaster. The results were quite nice and very flexible; feeling a bit like doggie chew toys. I had a few air bubbles to contend with, but these proved not too debilitating.
• 
• Casting Plaster Slip-Casting Molds: Cottle with “bowtie” shape about to be cast for front of tile. Next up was the creation of the two-part plaster molds. The first step was to build a form out of “cottle” boards of the proper size on a smooth table. Then place the piece to be cast inside and pour plaster to the desired thickness surrounding the part.
• 
• Cottle filled with plaster for “bowtie” shape, with a smaller wooden key for the back of the tile. Then, I removed the cottle boards and flipped the plaster mold, creating keys in the still-soft plaster for proper registration of the two halves. You'll notice in the image above a wooden “bowtie” piece sticking out. The reason for this is that I wanted to create a lip on the bottom of each tile of about 1/2” to increase structural integrity and to provide greater surface area for installation. Thus, I had to create a smaller piece for each tile and place it in the mold prior to pouring the back half. This opening also provided a hole through which the clay slip would be poured in and out of the mold.
• 
• Slip-Casting the Tiles: My slip-casting plaster forms after very much (ab)use!
• 
• Drilling holes for air pressure removal of the clay slip from the mold. While the silicon rubber helped in their removal from the plaster slip-casting molds. However, the rubber molds also captured the “fuzzy” nature of the CNC foam pieces. When slip casting, the clay slip found all these nooks and crevices and were reluctant to let go when I tried to get them out of the mold. Once again, air pressure came to the rescue. I ended up drilling tiny holes in the plaster molds - actually using MIG welding rod because it was thin and long - so I could pressurize the mold with air to remove the very damp clay once it had been cast. Every one of the peaks required a hole, as did the center of the largest piece. Also, you can see in the image above the “keys” for registering the two separate halves of the mold.
• 
• Backside of leather hard slip-cast tile. Left: Proper casting time. Right: Excessive casting time.
• 
• Bisque firing of some of the tiles.
• 
• Glaze experimentation on left, shallow casting on right. Following the bisque firing, I began to experiment with various glazes. Because I liked the white color of the stoneware, I wanted a translucent glaze that would settle into the valleys and be thinner at the ridges - providing a range of hues. The nature of glazes is that when they are applied they are entirely opaque - yet when fired they become somewhat vitreous (like glass) allowing for various degrees of translucency. The number of applied coats of a glaze has a dramatic effect on the depth of color, as well. In the image above, one can see tests with a blue glaze and the effects of differing numbers of coats.
• 
• Luster Glazes: Gold (left), pearlescent (right). Luster glazes required three separate firings in total. Worth noting is that ventilation is absolutely key when working with lusters.
• 
• Final Pieces. In addition to different glazes, I also experimented with dying the slip. In the final results you will see a variety of these techniques. Also, it is worth noting an error that was made, but not discovered until after firing. Given that all clay slip has a uniform shrinkage ratio, I failed to account for the fact that the large dodecahedrons would shrink more than the smaller tiles; thus a gap was formed around these that was slightly larger than anticipated.
• 
• 
• 
•