On the surface the hull only has to keep water out. Submerged, the whole envelope is squeezed inward by hydrostatic pressure, and the failure mode changes completely: not flooding, but buckling - the shell snapping inward all at once. Geometry, not raw strength, is what saves you, and a sphere is the best geometry there is.
Water adds ~10 kPa (about 0.1 atm) of pressure per metre of depth. A sphere converts that into uniform membrane compression - a stress concrete is delighted to carry. Slide the crown depth from 1 to 10 m and watch the compression and the buckling safety factor.
Reference geometry: 8.5 m internal Ø, 200 mm wall. Buckling uses the classical sphere formula P_cr = 1.21·E·(t/R)² with a 0.2 imperfection knockdown.
Under external pressure, the shape you choose matters far more than a few extra MPa of material strength. Here is the ranking, and why the icosahedron from the early sketches had to go.
| Geometry under external pressure | Behaviour | Verdict |
|---|---|---|
| Flat panel / box wall | Bends inward, ~90 t on a 3×3 m panel at 10 m | avoid |
| Faceted icosahedron | Bending + stress concentration at every edge and corner | poor |
| Cylinder + domed ends | Membrane compression, easy to ring-stiffen | good |
| Sphere | Uniform membrane compression, least material for a given depth | best |
A thin shell fails by going unstable long before the material is crushed. That is why stiffness (Young's modulus E) and roundness matter more than compressive strength here: the buckling pressure scales with E·(t/R)², and with the square of how out-of-round the shell is. Build it round.
At 1-10 m the pressures are tiny (~1.1-2 atm). The 200 mm wall is driven by rebar cover and buildability, not pressure, and ends up dozens of times over-safe against implosion - leaving headroom to certify deeper later.