Flow Around a Sphere
Flow Around a Sphere: Theoretical Foundations
Overview
Professor, how is the flow around a sphere different from that around a cylinder?
A sphere is a three-dimensional object, so its wake structure is inherently 3D. There is no region where a 2D calculation suffices, unlike with a cylinder. Also, the abrupt change in $C_D$ depending on Re, called the drag crisis, is practically very important. The dimples on a golf ball utilize this phenomenon.
Stokes Flow (Very Low Re)
First, please explain the most basic part.
When Re $\ll 1$, the inertial term can be neglected, and an exact solution to the Stokes equations exists.
This is Stokes' drag law. $a$ is the sphere's radius, $U_\infty$ is the freestream velocity. Rewriting it as a drag coefficient,
So $C_D = 24/Re$ means the drag coefficient decreases as Re increases, right?
However, the absolute drag force $F_D$ increases proportionally with $U_\infty$. The reason $C_D$ decreases is that the inertial force reference $\frac{1}{2}\rho U_\infty^2$ increases even faster.
Flow Transition by Reynolds Number
Let's organize the classification of flow around a sphere.
| Re Range | Flow State | Characteristics |
|---|---|---|
| Re < 20 | Steady Axisymmetric | No separation ~ tiny recirculation |
| 20 < Re < 210 | Steady Axisymmetric Wake | Growth of annular recirculation region |
| 210 < Re < 270 | Steady Non-axisymmetric | Loss of planar symmetry (regular bifurcation) |
| 270 < Re < 800 | Periodic Vortex Shedding | Regular shedding of hairpin vortices |
| 800 < Re < $3 \times 10^5$ | Subcritical Regime | Turbulent wake, $C_D \approx 0.44$ |
| $3 \times 10^5$ < Re < $4 \times 10^5$ | Drag Crisis | $C_D$ drops sharply from $0.44$ โ $0.1$ |
| Re > $4 \times 10^5$ | Supercritical Regime | Turbulent boundary layer separation |
Compared to a cylinder, the Re at which vortex shedding begins is higher.
Correct. Because a sphere is 3D, the flow can wrap around the object, delaying wake destabilization. Also, instead of a clear periodic pattern like a vortex street for a cylinder, the structure is more complex, involving the shedding of hairpin vortices.
Empirical Drag Correlation Formulas
How do you estimate $C_D$ at intermediate Re numbers?
The Schiller-Naumann correlation is commonly used.
For a wider range, there is Morrison's formula.
This formula provides a continuous fit from $Re = 10^{-1}$ to $10^6$, including the drag crisis.
Morrison's formula is quite complex. But it seems useful to be able to represent even the drag crisis.
The Paradox: Golf Ball Dimples Double the Flight Distance
If you launch a perfectly smooth sphere and a dimpled sphere at the same speed, which one flies farther? Intuition might suggest "the smooth one has less air resistance," but the answer is the opposite. The dimples induce boundary layer turbulence, pushing the separation point rearward, which drastically reduces the wake width. As a result, the drag coefficient becomes about half that of a smooth sphere. It is said that the flight distance of a typical golf ball is about twice that of a ball without dimples. This is a design that artificially induces the "drag crisis" at lower Re numbers, an aerodynamically elegant solution. When solving for a sphere's drag coefficient in CFD, the influence of surface roughness models is directly linked here.
Computational Methods for Flow Around a Sphere
Selection of Numerical Methods
What numerical method is suitable for calculating flow around a sphere?
A sphere is inherently a 3D problem, so computational cost is significantly higher compared to a cylinder.
| Re Range | Recommended Method | Mesh Scale Guideline |
|---|---|---|
| Re < 300 | DNS | 0.5 to 2 million cells |
| 300 < Re < $10^4$ | LES | 5 to 50 million cells |
| $10^4$ < Re | RANS (SST) / DES | 2 to 20 million cells |
Mesh Strategy
How do you create a mesh for a sphere?
Using an O-type (spherical shell) mesh around the sphere surface is ideal.
- First Wall Layer: $y^+ < 1$ (wall-resolved). For Re = 1000, $\Delta r / D \approx 5 \times 10^{-3}$
- Spherical Direction Division: Finer near the equator (to track separation point movement), coarser near the poles is acceptable
- Wake Region: Ensure at least $30D$ downstream from the sphere center. To track hairpin vortex development
- Computational Domain Outer Radius: At least $20D$ from sphere center (blockage < 0.25%)
How do you handle the singularity at the poles (the problem where mesh converges to a single point)?
Good question. A standard spherical coordinate mesh collapses cells at the poles. Countermeasures include:
- Cubed Sphere: Project each face of a cube onto a spherical surface. No polar singularity.
- Unstructured Mesh: Prism layers near the sphere surface, tetrahedral/polyhedral elsewhere.
- Overset Mesh: Cover the polar region with a separate patch.
Practically, STAR-CCM+'s polyhedral mesh or Fluent's poly-hexcore are convenient. They automatically generate prism layers on the sphere surface and fill the outside with polyhedral cells.
Utilizing Axisymmetric Calculations
For low Re, can we use an axisymmetric calculation?
For the range of steady axisymmetric wake at Re < 210, calculation with axisymmetric (2D cross-section + no swirl) is possible. In OpenFOAM, you can run simpleFoam with a wedge mesh (5-degree wedge shape). Computational cost is less than 1/100th of 3D.
However, for Re > 210, axisymmetry breaks, so full 3D calculation is absolutely necessary. It is dangerous to extrapolate to high Re just because "it worked with axisymmetric."
Coupling with Particle Tracking
Flow around a sphere is related to particle sedimentation and such, right?
Correct. The terminal velocity of a settling sphere from Stokes' law is,
In CFD-DEM coupling, fluid forces (drag, lift, added mass force) are applied to particles. The Schiller-Naumann or Gidaspow models are standardly used as drag models. This can be implemented using Fluent's DPM module or OpenFOAM's DPMFoam.
Related Topics
Experience the theory firsthand with the interactive simulator for this field
All Simulators