Scroll to jaw. A scroll plate carries an Archimedean spiral r = r0 + Lθ/2π. Turning the plate by φ slides every jaw tooth along its radial slot by Lφ/2π, so with L = 5 mm one key turn at the scroll moves all six jaws 5 mm at once: that is why the part ends up centred.
Key torque to clamp force. The spiral flank is a very shallow wedge (pitch angle λ = atan(L/2πrs) ≈ 0.83°). With flank friction φs = atan μs the total radial force is Ft / tan(λ+φs), Ft being the scroll torque over rs; without friction it reduces to 2πT/L (virtual work). Since tanλ < μs the wedge is self-locking: the chuck stays tight after the key is removed.
Centrifugal grip loss. A jaw of mass m with its centre of mass at rcg flings outward with mω2rcg when gripping from outside, so Feff = F − mω2rcg; grip is gone at ω* = √(F/(m rcg)). The torque the grip can carry is nμFeffrw; if the cutting torque is larger the part slips.
Why more jaws give a rounder part. A thin ring (inextensible Euler ring, EI = Ebt³/12) squeezed by n equal radial jaws bends into n lobes. The bending moment under a jaw is M = (PR/2)(1/α − cotα), α = π/n. If the bore is turned while clamped it is round only under load: after release the springback leaves a lobed error equal to the loaded deflection difference between jaw and mid-gap. At the same total force it falls roughly as n−3 (about 19 times smaller for six jaws than for three here).
Simplifications: rigid chuck body and jaws, equal jaw forces, no jaw-to-part friction in the ring bending, thin wall (no shear or ring stretching), purely elastic (a red warning appears when the bending stress passes the yield strength), no vibration or machining forces. The deformation is drawn exaggerated. Illustrative parameters (not measured): scroll radius 55 mm, gear ratio 4, efficiencies 0.9 and 0.85, flank friction 0.10, jaw mass 0.9 kg, steel E = 200 GPa, yield 250 MPa. Educational model, not for choosing real clamping settings.
Running numerical checks...
3D twin: Six-Jaw Scroll Chuck 3D.