Published online by Cambridge University Press: 30 March 2017
A comprehensive study of the classical linear spin-down of a constant-density viscous fluid (kinematic viscosity   $\unicode[STIX]{x1D708}$ ) rotating rapidly (angular velocity
 $\unicode[STIX]{x1D708}$ ) rotating rapidly (angular velocity   $\unicode[STIX]{x1D6FA}$ ) inside an axisymmetric cylindrical container (radius
 $\unicode[STIX]{x1D6FA}$ ) inside an axisymmetric cylindrical container (radius   $L$ , height
 $L$ , height   $H$ ) with rigid boundaries, which follows the instantaneous small change in the boundary angular velocity at small Ekman number
 $H$ ) with rigid boundaries, which follows the instantaneous small change in the boundary angular velocity at small Ekman number   $E=\unicode[STIX]{x1D708}/H^{2}\unicode[STIX]{x1D6FA}\ll 1$ , was provided by Greenspan & Howard (J. Fluid Mech., vol. 17, 1963, pp. 385–404). For that problem
 $E=\unicode[STIX]{x1D708}/H^{2}\unicode[STIX]{x1D6FA}\ll 1$ , was provided by Greenspan & Howard (J. Fluid Mech., vol. 17, 1963, pp. 385–404). For that problem   $E^{1/2}$  Ekman layers form quickly, triggering inertial waves together with the dominant spin-down of the quasi-geostrophic (QG) interior flow on the
 $E^{1/2}$  Ekman layers form quickly, triggering inertial waves together with the dominant spin-down of the quasi-geostrophic (QG) interior flow on the   $O(E^{-1/2}\unicode[STIX]{x1D6FA}^{-1})$  time scale. On the longer lateral viscous diffusion time scale
 $O(E^{-1/2}\unicode[STIX]{x1D6FA}^{-1})$  time scale. On the longer lateral viscous diffusion time scale   $O(L^{2}/\unicode[STIX]{x1D708})$ , the QG flow responds to the
 $O(L^{2}/\unicode[STIX]{x1D708})$ , the QG flow responds to the   $E^{1/3}$  sidewall shear layers. In our variant, the sidewall and top boundaries are stress-free, a set-up motivated by the study of isolated atmospheric structures such as tropical cyclones or tornadoes. Relative to the unbounded plane layer case, spin-down is reduced (enhanced) by the presence of a slippery (rigid) sidewall. This is evidenced by the QG angular velocity,
 $E^{1/3}$  sidewall shear layers. In our variant, the sidewall and top boundaries are stress-free, a set-up motivated by the study of isolated atmospheric structures such as tropical cyclones or tornadoes. Relative to the unbounded plane layer case, spin-down is reduced (enhanced) by the presence of a slippery (rigid) sidewall. This is evidenced by the QG angular velocity,   $\unicode[STIX]{x1D714}^{\star }$ , evolution on the
 $\unicode[STIX]{x1D714}^{\star }$ , evolution on the   $O(L^{2}/\unicode[STIX]{x1D708})$  time scale: spatially,
 $O(L^{2}/\unicode[STIX]{x1D708})$  time scale: spatially,   $\unicode[STIX]{x1D714}^{\star }$  increases (decreases) outwards from the axis for a slippery (rigid) sidewall; temporally, the long-time
 $\unicode[STIX]{x1D714}^{\star }$  increases (decreases) outwards from the axis for a slippery (rigid) sidewall; temporally, the long-time   $(\gg L^{2}/\unicode[STIX]{x1D708})$  behaviour is dominated by an eigensolution with a decay rate slightly slower (faster) than that for an unbounded layer. In our slippery sidewall case, the
 $(\gg L^{2}/\unicode[STIX]{x1D708})$  behaviour is dominated by an eigensolution with a decay rate slightly slower (faster) than that for an unbounded layer. In our slippery sidewall case, the   $E^{1/2}\times E^{1/2}$  corner region that forms at the sidewall intersection with the rigid base is responsible for a
 $E^{1/2}\times E^{1/2}$  corner region that forms at the sidewall intersection with the rigid base is responsible for a   $\ln E$  singularity within the
 $\ln E$  singularity within the   $E^{1/3}$  layer, causing our asymptotics to apply only at values of
 $E^{1/3}$  layer, causing our asymptotics to apply only at values of   $E$  far smaller than can be reached by our direct numerical simulation (DNS) of the linear equations governing the entire spin-down process. Instead, we solve the
 $E$  far smaller than can be reached by our direct numerical simulation (DNS) of the linear equations governing the entire spin-down process. Instead, we solve the   $E^{1/3}$  boundary layer equations for given
 $E^{1/3}$  boundary layer equations for given   $E$  numerically. Our hybrid asymptotic–numerical approach yields results in excellent agreement with our DNS.
 $E$  numerically. Our hybrid asymptotic–numerical approach yields results in excellent agreement with our DNS.