Ghost Field


“Ghost Field” was inspired by Paul Harrison’s pattern seeking program, “Ghost Diagrams”. Like Harrison’s software, our project builds from a discreet set of components, but it adds a third dimension in order to allow a second pattern to combine with the first. A Rhinoscript implementation of “Ghost Diagrams” was developed in order to test variations and build a virtual mock-up of the design before assembly. The project culminated in a piece composed of more than 100 individual glazed ceramic components, weaving and interlocking to achieve varied densities and porosities. Ultimately, we hope to have acheived a whole that is in fact greater than the sum of its parts.

In the spring of 2010 at the University of Pennsylvania, Ghost Field was realized as an expanse of variegated pattern built of three types of ceramic components which had been designed and tested digitally. While a product of small-scale mass production itself, it sought to inject difference into the commonly repetitive aggregation of a modular component system, drawing inspiration from a number of sources. Dr. Paul Harrison’s pattern seeking program “Ghost Diagrams” (available at supplied a conceptual framework for the non-linear aggregation of multiple component types, while the work of Erwin Hauer and Evan Douglis offered formal inspiration. A prototype was assembled and exhibited at the final review of Jenny Sabin’s spring course entitled “Experiments in Fabrication: Digital Ceramics,” of which the authors were a part.

Digital techniques were used at a few key stages of the design and fabrication process. Essentially, the process of casting ceramic requires a few parts to be created in sequence: a positive, which is used to create a mold, which is used to create greenware, which is fired to create bisqueware, which is then glazed and fired again to create a finished piece of ceramic. The positives used for this project were designed first in Rhinoceros and then printed using a Z-Corp 3d printer. Digital fabrication of the positive allowed us to preserve both the precision of the 3d digital model and its correspondence with its real-world counterpart, which in this case was key to the second stage of the digital design process. Over the course of the semester, the authors developed a Rhinoscript application which allowed them to test complex patterns of aggregation using the three component types which were concurrently being cast. These cast components were eventually glazed and fired, then assembled according to a digital prototype created with the Rhinoscript application.

Though the development of the project progressed fairly smoothly over the course of the semester, there were a few noteable obstacles along the way–both digital and physical. First was to translate Harrison’s two-dimensional components into three dimensions, and next to translate the patterns themselves. The work of Hauer and Douglis gave us a formal language which meshed nicely with Harrison’s two-dimensional geometry, and using a combination of lofting, sweeping, and patching, the team developed a set of nurbs models for the components. A simple Grasshopper plugin aided in the conversion to mesh models for printing by helping to control the size and shape of each mesh face. Still, the construction of the nurbs components tended to leave some characteristic ridges and valleys in the finished models that were not exactly consistent with the formal expression sought, but due to constraints involved in connecting the components they were unable to be smoothed out. Translating Harrison’s patterns into three dimensions would not have been possible (in the time allotted) without a 3d implementation of his “Ghost Diagrams”, and though our final application did not manage to incorporate the automated pattern-seeking intelligence of ghost diagrams, it was able to rely on a human operator to help predict patterns while it took care of collision detection and component selection. The collision detection feature was incredibly useful when trying to weave a second pattern through the first, but even with algorithmic aid that particular task was exceedingly difficult.

Back in the real world, fabrication of the components proceeded quickly once the team completed a second set of molds, allowing us to cast about six components a day for the last month of the semester. As mentioned above, 3D prints of the components were used to create these two-part plaster molds, taking note of the component geometry when planning the mold seams and location of the pour spout. Liquid clay (slip) was poured into the molds and left for 30 minutes, after which the excess slip was emptied out in order to form a hollow component. The components were taken out of the mold after 18 hours, allowing them to be cleaned before becoming too dry. Each component was cleaned by hand, using clay tools and a laser-cut stamp to ensure consistency at the connection points. After cleaning, the components were bisque fired at Cone 03/04, a low fire chosen based on our clay body. The bisque fired components were strung along metal wires, dipped into a glaze and hung to dry. This method was used to maintain uniform, even coverage, both for aesthetic value and to ensure that the final components would accurately connect in the final assembly. After glazing, the components were put back into the kiln and fired a second time. Ceramics can be fired and glazed multiple times, but when working with component assemblies, it is critical to take into account additional shrinkage and glaze thickness of the final parts. The components were assembled using laser-cut Plexiglas and cork hardware. There were two connection types – a simple part to part connection and an interlocking connection that also braced the two layers together for structural integrity. Cork was used to ensure a tight connection while also providing a tolerance for inaccuracies.

In the end, the project culminated in a piece composed of more than 100 individual glazed ceramic components, weaving and interlocking to achieve a field of varied densities and porosities. When viewed straight on it is a distinct pattern of two interwoven layers, but in perspective it forms a field, where the undulating geometries create unanticipated patterns of movement of their own. Ultimately, we hope to have achieved a whole that is in fact greater than the sum of its parts. Special thanks go to Jenny Sabin, Annette Fiero, Ferda Kolatan, and Franca Trubiano for invaluable advice and insights; to Michelle Miller for her expertise in running the kilns and her patience with a bunch of architecture grad students who couldn’t tell glaze from slip before taking this class; and to LAB Studio and the Institute for Medicine & Engineering at the University of Pennsylvania for providing much of the equipment and funding necessary to make this project a success.