Some experiments use polarized light. Between two crossed polarizers a perfect window stays completely black — all light is blocked. But stress in the window (from a pressure difference across it, an over-tight mount, or a thermal gradient) makes it leak a little light: the black brightens. This model takes the resulting in-plane principal stress difference Δσ = σ₁−σ₂ as its input — it does not calculate the pressure-to-stress conversion, which needs a structural model of the specific window. Drag the sliders and watch the two windows.
What the picture shows & the physics. The "light bending" is the optical retardation, set by the stress-optic law δ = C · h · (σ₁−σ₂), where σ₁−σ₂ is the in-plane principal stress difference and C is the stress-optic coefficient (fused silica ≈ 3.55 × 10−12 Pa−1, about 4 nm of bending per mm of thickness per MPa of stress; calcium fluoride is lower). Between crossed polarizers the fraction of light that leaks through is sin²(πδ/λ) for a window at 45°, shown here for green light (λ = 550 nm) — that is the number on the right and the meter fill. The right-hand glow is brightened so the trend is easy to see; the exact leak fraction is the number. Gentle mounting (well under 1 MPa of principal stress difference) leaks almost nothing — the higher values on the slider stand for the larger stress a big pressure difference, thermal gradient or over-tightened mount would induce. Converting a pressure difference into an actual stress field needs a structural model (clear aperture, thickness, edge support, elastic constants, pretension), which this teaching model does not do. A real window's stress is also uneven (worst at the rim), so a real view would be brighter near the edges; this even swatch is a simplification, not a substitute for a proper finite-element stress analysis.