*Appendix A.1. Numerical Setup*

The mesh used in both simulations was a 0.41 m (8.2 *D*)-wide and 1.2 m (24 *D*)-long rectangle with a cylinder of diameter *D* = 0.05 cm placed 0.2 m (4 *D*) from the inlet (Figure A1). The top and bottom boundary conditions were symmetry-type boundary conditions. The outlet boundary condition for the reduced pressure was a homogeneous Dirichlet condition (*p* − *ρf gy* = 0 Pa). For the outlet velocity, a homogeneous Neumann boundary condition was used for outgoing flows, and a homogeneous Dirichlet boundary condition was used otherwise. A fixed value of *U* = 0.87 <sup>m</sup>·s<sup>−</sup><sup>1</sup> was given for the inlet velocity, giving a Reynolds number *Re* = *UD*/*ν* = 4.35 × 104.

m

**Figure A1.** Sketch of the geometry and the boundary conditions used for the computational domain.

For the simulation using a *k* − *ω*2006 turbulence model, the cells near the cylinder were 2×10−<sup>5</sup> m thick, giving a near-wall *y*<sup>+</sup> ≈ 1. Wall functions for smooth walls were applied on the cylinder surface.

For the simulations using the standard *k* − *ε*, the cells near the cylinder were 6 × 10−<sup>4</sup> m thick, giving a near-wall *y*<sup>+</sup> ≈ 30. A homogeneous Dirichlet boundary condition of 1 × 10−<sup>10</sup> <sup>m</sup>2·s<sup>−</sup><sup>2</sup> was applied on the cylinder surface for *k*, and a homogeneous Neumann boundary condition was applied for *ε*.

For both simulations, inlet boundary conditions for turbulent quantities were set following Table A1 with a turbulence intensity *I* = 2% and a turbulence length scale *l* = 0.07 *D*.

**Table A1.** Inlet boundary conditions for turbulent quantities.


Second order differentiation schemes and a time step equal to 2 × 10−<sup>4</sup> s were used. A probe placed on the top right of the cylinder (Point A in Figure A1) recorded the velocity signal along the simulations.
