Spheres in a Stokes flow

Let us consider a single sphereS₁ with radius aembedded in a Stokes flow with a total non-hydrodynamic force ~F₁ and no torques acting on it. The sphere resides at position~r₁ with translational velocity~v₁ and angular velocity~ω1. The force~F₁ is equally distributed over the surface_∂S₁ ofS₁. Thus, the force density determining the fluid’s velocity via equation 2.10 is

(2.13) ~f₁= 1

4_πa^2δ(|~r−~r₁| −a)~F₁ .

Furthermore, the presence of the sphere imposes no-slip boundary conditions on the fluid flow, which read

which simply means that there is no relative velocity between fluid and surface elements on

∂S₁. While it is clear why the fluid cannot possess a velocity componentthroughthe sphere’s surface, the physical reason why there is no componentparallelto it is the following [92]. At the microscopic level, there are interaction (adhesion) forces between surface and water molecules which statistically reduce their relative velocity upon collision. As a result of the short mean free path of water molecules, i.e. the distance they travel before colliding with another water molecule, each water molecule near the surface collides with it many times. Therefore, theaveragefluid velocity in any small volumeV at the surface loses its velocity component parallel to the wall, even though individual water molecules within V may not be at rest. The no-slip boundary condition can thus be understood as a consequence of the continuum description of a dense fluid which microscopically interacts with the surface. It fails when the fluid is too diluted, or when its molecules only elastically bounce off a (hydrophobic) surface.

Stokes law Integrating the boundary conditions over_∂S₁eliminates the rotational term on the left hand side and, using equation 2.12, leads to

(2.15)

which is the famous Stokes law for a sphere with radiusain an overdamped fluid: The linear relation between its velocity and the force acting on it is quantified by the friction coefficient

(2.16) _γ=6_πηa .

Remarkably, this coefficient itself is a linear function of the sphere’s radiusa(rather than its cross sectional area).

Flow field around the sphere The velocity field~v(~r) caused by the sphere is given by insert-ing the force density 2.13 into equation 2.10. The resultinsert-ing integral can be solved analytically and, defining∆~r=~r−~r₁and∆r= |∆~r|, reads

Note that this flow is continuous and satisfies the no-slip boundary conditions. It is visualized in figure 2.1. One may think of the inside of the sphere as consisting of fluid which is moving with the same velocity as the sphere’s center and shell.

Dynamics of the sphere’s center position Considering the translational dynamics of the sphere’s center~r₁instead of the fluid, the Stokes law 2.15 replaces Newton’s second law (a second order, ordinary differential equation)

(2.18) md²_t~r=~F

with the first order, ordinary differential equation

(2.19) 0= −γd_t~r+~F .

Hence the widely used linear friction force−γ~vis not a heuristic term, but is as exact as the Stokes equation.

Faxén’s law So far we considered a sphere in a fluid which would be at rest without the force acting on that sphere. Now we are going to consider the same sphere embedded in a given velocity field~v₀(~r) (which satisfies the creeping flow equations) without specifying what this velocity field is caused by. Due to the linearity of the Stokes equation, we now have

(2.20) ~v(~r)=~v₀(~r)+ 1

This leads to modified no-slip boundary conditions at the sphere’s surface, reading (2.21) ~v₁+~ω1×(~r−~r₁)=~v(~r)=~v₀(~r)+ 1 Again, we integrate the boundary conditions over_∂S₁, but now we study how the new term~v₀ behaves inside the integral.

Expanding~v₀ as a Taylor series, terms of odd order do not contribute to the integral due to the spherical symmetry of_∂S. It can furthermore be shown in a lengthy but straightforward calculation (see pages 254–255 in [88]) which we omit here, that terms of even order higher than two vanish for all flow fields obeying the creeping flow equations.

Keeping only the zeroth and second order term, we can hence update Stokes law (equation 2.15) with two new terms introduced by the given velocity field~v₀ as follows:

(2.22) ~v₁=1

γ~F₁+ µ

1+a² 6~∇²

¶

~r=~r1

~v₀ .

This equation is called Faxén’s law for translational motion. For~v₀=0 it immediately reduces to the Stokes law. We treat~v₀and the force~F₁ acting on the sphere as given and use Faxén’s law to determine the sphere’s resulting velocity~v₁. Naturally, Faxén’s law can also be used the other way around – it determines~F₁when~v₁ is known.

Figure 2.1: The velocity field (red arrows) around a unit sphere (origin) moving with unit speed in x-direction as given by equation 2.17. The background color shows its absolute value in terms of the sphere’s velocity. The closed lines show contour lines of the velocity field’s absolute value. At the inner green line, the velocity has dropped to 50 %, and at the outer black line it has dropped to 25 %. A second sphere (bottom right) is embedded in the velocity field caused by the first sphere. It moves along with the fluid flow around it – this effect is called a “hydrodynamic interaction”. Its velocity can be quantified using Faxén’s law (equation 2.22), or equivalently via the Rotne-Prager matrices (equation 2.29). The velocities of both spheres are indicated with black arrows at their respective centers. The lengths of these arrows have the correct ratio, but are not in the proper scale with the fluid flow for visual purposes. The fluid velocity equals that of the first sphere at its boundary.

Two spheres in a Stokes flow Let us now consider two spheresS₁ andS₂ with radiia₁ and a₂, center points~r₁and~r₂, translational velocities~v₁and~v₂, rotational velocities~ω1and~ω2, and forcesF~₁ and~F₂ acting on them in a Stokes flow which would be at rest in the absence of the spheres (see figure 2.1). The no-slip boundary conditions for the fluid flow~v(~r) now read

(2.23) ~v₁+~ω1×(~r−~r₁)=~v(~r) where ~r∈∂S₁ and ~v₂+~ω2×(~r−~r₂)=~v(~r) where ~r∈∂S₂ and the force density is given by

(2.24) ~f(~r)= 1

4_πa₁^δ(|~r−~r₁| −a₁)~F₁+ 1

4_πa₂^δ(|~r−~r₂| −a₂)~F₂ .

Even for only two spheres of equal size, this boundary value problem for the creeping flow equations does not have a closed analytical solution. The solution can be constructed, however, using an iterative method which is called themethod of reflections. It can be found e.g. in chapter 5.12 of the textbook by Dhont [88], but exceeds the scope of this thesis. The method of reflections is in principle straightforward for two spheres. For three spheres, only the leading order terms in a series expansion of the correct flow field are known, and for more accuracy or more spheres one needs to retreat to expensive numerics. This makes the method of reflections impractical for most applications, including ours, which is why we do not discuss it in detail¹. Instead, we find an approximate solution to the given boundary value problem.