Teaching a Catalan Dome

How we built Catalan vaults in Kolhapur with masons who had never turned an arch

Most conversations about computational design end at the screen. Ours started there and had to survive contact with a construction site in Kolhapur, a team that had never built a vault, and a budget that ruled out almost everything a Catalan vault is supposed to need.

The brief was two domes - a studio and office for an architect. The typology we chose, the Catalan or tile vault, is one of the oldest compression structures in the world and one of the least forgiving. It carries load through its geometry, not through steel. Get the curvature wrong and it doesn't sag; it fails. That is exactly why we wanted it, and exactly why it should have been impossible with the team and resources we had.

So we split the problem in two: resolve the difficulty in software, and remove it from the site.

In Rhino, we modelled the domes as what they physically are - thrust surfaces. We studied the line of compression, the bonding pattern, the curvature at every course. We were not chasing a shape. We were chasing a structure that would stand in tile and mortar with nothing holding it up while it was built. Once the geometry was honest, we stopped designing and started translating.

This is the part the industry skips. A model is not a method. A mason cannot build a mesh. So we decomposed each dome into discrete, repeatable steps - the way you would break a complex movement into drills. The computational logic left the computer as physical instruments: 1:1 curvature layouts, arch templates, and sequencing diagrams that made the thrust lines and surface gradients legible to a hand, not a processor. Complex geometry became simple scaffolding, grids, and analog markers a person could check by eye.

Then we removed the cost. A Catalan vault conventionally leans on shuttering and CNC formwork. We had neither. We replaced heavy shuttering with lightweight coconut-leaf mats and MS rod grids. We introduced earthen pots as filler in the soffit - structural efficiency and material pragmatism in the same move. The first dome followed traditional tile-vault layering in brick, establishing a baseline for curvature, bonding and stiffness. The second pushed further, with bamboo reinforcement and pots forming the soffit, paired with lime and mud mortars applied in successive layers.

None of the masons had built a vault before. By the end, they were turning one. That is the result I am proudest of - not the domes, the transfer. Material behaviour, bond strength and timing were tested continuously on site, and every correction fed back into the drawings. The knowledge did not flow one way from studio to labour. It circulated.

The finished domes read as quiet sculptural volumes. Their soffits reveal layered mortar, filler pots and lime finishes rather than hiding the method of making. They are low-carbon, built without industrial composites, and - this is the point - repeatable. We did not produce a one-off feat. We produced a technique a local team can now carry to the next site.

Advanced geometry does not require advanced machinery. It requires abstraction, communication and sequencing handled with enough care that the complexity lands on the drawing, and never on the worker.