physics2u
PhysicsMathsChemistryBiology
Tier
⌕ Search ⌘K
Derivation

Stokes Drag on a Sphere

Statement

For a rigid sphere of radius \(a\) held fixed in an unbounded incompressible Newtonian fluid of dynamic viscosity \(\mu\), streamed by a uniform velocity \(U\) at infinity in the vanishing-Reynolds-number (creeping-flow) limit, the steady hydrodynamic drag force exerted on the sphere is \(F = 6\pi\mu a U\), directed along the free stream and independent of the fluid density.

Why it matters

Stokes drag is the exact closed-form solution of a full three-dimensional viscous flow, and it is the benchmark against which every low-Reynolds-number transport result is measured: sedimentation of colloids and cells, the settling of aerosols and cloud droplets, the calibration of the falling-ball viscometer, and Millikan's determination of the electron charge all rest on it.

It is also the archetype of a linear, reversible, inertia-free flow. Because the drag scales linearly with \(U\) and with the linear size \(a\) — not with \(a^2\) as in high-Reynolds form drag — it defines the mobility \(1/(6\pi\mu a)\) that anchors the Einstein–Stokes relation for diffusion, the Péclet number, and the entire theory of microhydrodynamics.

Assumptions
Steady, incompressible Newtonian fluid.If the fluid is non-Newtonian or compressible, the stress is no longer \(\mu\) times the rate of strain and the constitutive relation feeding the momentum balance changes; the numerical coefficient \(6\pi\) is lost.
Vanishing Reynolds number, \(\mathrm{Re}=\rho U a/\mu \ll 1\).Retaining the convective inertia \(\rho(\mathbf{u}\cdot\nabla)\mathbf{u}\) makes the equations nonlinear; the drag acquires an \(\mathrm{Re}\)-dependent correction (Oseen: \(F=6\pi\mu aU(1+\tfrac{3}{8}\mathrm{Re})\)).
No-slip on the sphere surface.A partial-slip or shear-free (bubble) boundary reduces the tangential drag; a clean spherical bubble gives \(4\pi\mu aU\) instead of \(6\pi\mu aU\).
Unbounded fluid at rest at infinity.Nearby walls or neighbouring particles introduce long-ranged \(1/r\) hydrodynamic reflections; wall corrections (Faxén, Ladenburg) raise the drag and the far-field decay makes the correction non-negligible even for distant boundaries.
Rigid, non-rotating, impermeable sphere.Surface deformation, internal circulation (a liquid drop, Hadamard–Rybczynski), or through-flow modifies the surface traction and hence the coefficient.
Derivation
1
\[ \rho\left(\frac{\partial \mathbf{u}}{\partial t}+(\mathbf{u}\cdot\nabla)\mathbf{u}\right) = -\nabla p + \mu\nabla^2\mathbf{u} \;\xrightarrow[\ \mathrm{Re}\to 0\ ]{}\; \nabla p = \mu\nabla^2\mathbf{u},\qquad \nabla\cdot\mathbf{u}=0 \]
Nondimensionalising the incompressible Navier–Stokes equation on scales \(a\) and \(U\) shows the inertial terms carry a prefactor \(\mathrm{Re}\); in the creeping limit they are dropped, leaving the linear Stokes equations. A
2
\[ \nabla^2(\nabla p)=0,\qquad \nabla\times(\nabla p)=0 \;\Rightarrow\; \nabla^2\boldsymbol{\omega}=0,\quad \boldsymbol{\omega}=\nabla\times\mathbf{u} \]
Taking the divergence of the momentum equation and using \(\nabla\cdot\mathbf{u}=0\) gives \(\nabla^2 p=0\) (harmonic pressure); taking the curl removes the pressure gradient and gives a harmonic vorticity. Both are consequences of linearity. B
3
\[ u_r=\frac{1}{r^2\sin\theta}\frac{\partial\psi}{\partial\theta},\qquad u_\theta=-\frac{1}{r\sin\theta}\frac{\partial\psi}{\partial r} \]
The flow is axisymmetric with no azimuthal component, so continuity is satisfied identically by a Stokes stream function \(\psi(r,\theta)\) in spherical coordinates with the polar axis along \(U\). This reduces two velocity unknowns to one scalar. B
4
\[ E^2\psi=0\ \text{for irrotational parts},\quad\text{but in general}\quad E^4\psi \equiv E^2\!\left(E^2\psi\right)=0,\qquad E^2=\frac{\partial^2}{\partial r^2}+\frac{\sin\theta}{r^2}\frac{\partial}{\partial\theta}\!\left(\frac{1}{\sin\theta}\frac{\partial}{\partial\theta}\right) \]
The single azimuthal vorticity component is \(\omega_\phi=-\dfrac{E^2\psi}{r\sin\theta}\); substituting into \(\nabla^2\boldsymbol{\omega}=0\) yields \(E^2(E^2\psi)=0\). This is the governing equation for creeping axisymmetric flow. C
5
\[ \psi(r,\theta)=f(r)\,\sin^2\theta,\qquad \psi\xrightarrow[r\to\infty]{}\tfrac{1}{2}U r^2\sin^2\theta \]
The uniform free stream \(\mathbf{u}\to U\hat{\mathbf z}\) corresponds to \(\psi_\infty=\tfrac12Ur^2\sin^2\theta\); the \(\sin^2\theta\) angular structure is preserved by \(E^2\), so separation with this single angular mode is exact, not an ansatz. B
6
\[ \left(\frac{d^2}{dr^2}-\frac{2}{r^2}\right)^{\!2} f = 0 \;\Rightarrow\; f(r)=\frac{A}{r}+B\,r+C\,r^2+D\,r^4 \]
With \(\psi=f\sin^2\theta\), \(E^2\psi=\big(f''-2f/r^2\big)\sin^2\theta\); applying \(E^2\) again gives the equidimensional (Euler) ODE. Trial \(f\propto r^n\) gives \((n-2)(n+1)\) per operator, so the four roots are \(n=-1,1,2,4\). C
7
\[ D=0,\quad C=\tfrac12 U\quad(\text{bounded far field});\qquad f(a)=0,\ f'(a)=0\ \ (\text{no-slip}) \]
The \(r^4\) term makes \(u\) diverge at infinity, so \(D=0\); matching the free stream fixes \(C=U/2\). At the surface both velocity components vanish, and since \(u_r\propto f\) and \(u_\theta\propto f'\), no-slip becomes \(f(a)=f'(a)=0\). B
8
\[ \frac{A}{a}+Ba+\frac{U}{2}a^2=0,\quad -\frac{A}{a^2}+B+Ua=0 \;\Rightarrow\; A=\frac{U a^3}{4},\quad B=-\frac{3Ua}{4} \]
The two no-slip conditions are a \(2\times2\) linear system for \(A,B\); solving gives the stream function completely. Purely algebraic. A
9
\[ u_r=U\cos\theta\left(1-\frac{3a}{2r}+\frac{a^3}{2r^3}\right),\quad u_\theta=-U\sin\theta\left(1-\frac{3a}{4r}-\frac{a^3}{4r^3}\right),\quad p=p_\infty-\frac{3\mu U a}{2}\frac{\cos\theta}{r^2} \]
Substitute \(A,B,C,D\) into \(u_r=2f\cos\theta/r^2\), \(u_\theta=-f'\sin\theta/r\). The pressure follows by integrating \(\nabla p=\mu\nabla^2\mathbf{u}\); note the slowly-decaying \(1/r\) velocity disturbance, the signature of Stokes flow. B
10
\[ \sigma_{rr}\big|_{r=a}=-p=-p_\infty+\frac{3\mu U}{2a}\cos\theta,\qquad \sigma_{r\theta}\big|_{r=a}=\mu\frac{\partial u_\theta}{\partial r}\Big|_{a}=-\frac{3\mu U}{2a}\sin\theta \]
On the surface \(u_r=u_\theta=0\), so \(\partial u_r/\partial\theta=0\) and the normal strain rate \(e_{rr}=\partial u_r/\partial r=0\); hence \(\sigma_{rr}=-p\) and \(\sigma_{r\theta}\) reduces to the single surviving shear term. C
11
\[ t_z=\sigma_{rr}\cos\theta-\sigma_{r\theta}\sin\theta=-p_\infty\cos\theta+\frac{3\mu U}{2a}\left(\cos^2\theta+\sin^2\theta\right)=-p_\infty\cos\theta+\frac{3\mu U}{2a} \]
The traction \(\mathbf{t}=\boldsymbol{\sigma}\cdot\hat{\mathbf r}\) projected on the flow axis \(\hat{\mathbf z}\). The angle-dependent surface stress collapses to a constant plus an odd term — the reason the integral is so clean. C
12
\[ F=\oint_{r=a} t_z\,dS=\int_0^{2\pi}\!\!\int_0^{\pi}\!\left(-p_\infty\cos\theta+\frac{3\mu U}{2a}\right)a^2\sin\theta\,d\theta\,d\phi=\frac{3\mu U}{2a}\,(4\pi a^2) \]
The \(-p_\infty\cos\theta\) term integrates to zero by fore–aft symmetry; the constant term multiplies the sphere area \(4\pi a^2\). This gives the final result. B
Result
\[ F = 6\pi\mu a U \]

Reading. The drag on a slowly-moving sphere is proportional to viscosity, radius, and speed — each to the first power. Of the total, two-thirds (\(4\pi\mu aU\)) is skin-friction from the tangential shear and one-third (\(2\pi\mu aU\)) is form drag from the fore–aft pressure difference. Crucially the fluid density does not appear: at \(\mathrm{Re}\to0\) inertia plays no role, so the drag is purely viscous.

Units check. \([\mu]=\mathrm{Pa\,s}=\mathrm{kg\,m^{-1}s^{-1}}\), \([a]=\mathrm{m}\), \([U]=\mathrm{m\,s^{-1}}\); the product is \(\mathrm{kg\,m^{-1}s^{-1}}\cdot\mathrm{m}\cdot\mathrm{m\,s^{-1}}=\mathrm{kg\,m\,s^{-2}}=\mathrm{N}\). The factor \(6\pi\) is dimensionless.

Limiting cases
  • Shear-free surface (clean bubble): replacing no-slip by zero tangential stress removes the friction contribution, giving \(F=4\pi\mu aU\); a general viscous drop interpolates via Hadamard–Rybczynski, \(F=2\pi\mu aU\,\dfrac{2+3\kappa}{1+\kappa}\) with \(\kappa=\mu_{\text{drop}}/\mu\).
  • Small but finite inertia: the Oseen correction gives \(F=6\pi\mu aU\left(1+\tfrac{3}{8}\mathrm{Re}\right)\), the first term of the systematic matched-asymptotic expansion in \(\mathrm{Re}\).
  • Terminal settling: balancing drag against net weight yields \(U_t=\dfrac{2a^2(\rho_s-\rho_f)g}{9\mu}\), the Stokes settling velocity.
  • Diffusive limit: the mobility \(b=1/(6\pi\mu a)\) combined with \(k_BT\) gives the Stokes–Einstein diffusivity \(D=k_BT/(6\pi\mu a)\).
Breaks when
  • Reynolds number is not small. Once \(\mathrm{Re}=\rho U a/\mu\gtrsim 1\) the neglected convective inertia matters; a laminar wake forms, then separates (\(\mathrm{Re}\sim 20\)), and by \(\mathrm{Re}\sim 10^3\) the drag coefficient plateaus near \(C_D\approx0.4\) — utterly unlike the \(C_D=24/\mathrm{Re}\) that \(6\pi\mu aU\) implies.
  • The continuum assumption fails. When the sphere size approaches the mean free path (Knudsen number \(\mathrm{Kn}=\lambda/a\) not small), as for sub-micron aerosols, the fluid slips at the surface and the drag drops by the Cunningham slip correction factor \(C_c>1\): \(F=6\pi\mu aU/C_c\).
  • Boundaries or neighbours are present. A wall at distance \(h\) raises the drag (e.g. \(F\approx 6\pi\mu aU\big[1+\tfrac{9}{16}(a/h)+\dots\big]\) for motion parallel to it); in a suspension the \(1/r\) hydrodynamic interactions couple all particles and no single-sphere formula applies.
  • Unsteady or accelerating motion. Rapid changes add the added-mass term \(\tfrac12(\tfrac43\pi a^3)\rho_f\dot U\) and the Basset history integral; the steady \(6\pi\mu aU\) is only the quasi-static part of the Basset–Boussinesq–Oseen force.
Failure modes
  • Using diameter for \(a\). \(a\) is the radius; substituting the diameter doubles the drag. A frequent factor-of-two error in viscometry.
  • Confusing \(\mu\) and \(\nu\). The formula uses dynamic viscosity \(\mu\), not kinematic \(\nu=\mu/\rho\). Plugging \(\nu\) leaves a stray factor of density.
  • Adding a \(\tfrac12\rho U^2\) dynamic-pressure term. Students import the high-\(\mathrm{Re}\) drag \(\tfrac12\rho U^2 C_D A\) reasoning; but \(6\pi\mu aU\) already is the total drag and density must not reappear.
  • Forgetting to verify \(\mathrm{Re}\ll1\). Applying Stokes to a raindrop or a marble in water: \(\mathrm{Re}\gg1\), so the answer is wrong by orders of magnitude.
  • Assuming terminal velocity is instantaneous. The approach to \(U_t\) has a relaxation time \(\tau=m/(6\pi\mu a)\) that is only negligible for tiny particles.
  • Mis-signing the buoyancy in settling. Using \(\rho_s\) instead of the density difference \((\rho_s-\rho_f)\) omits buoyancy and overestimates \(U_t\).
Discussion

The defining feature of Stokes flow is that it is reversible: reverse the boundary motion and every streamline retraces itself, because the linear time-independent Stokes equations are symmetric under \(\mathbf{u}\to-\mathbf{u}\), \(p\to-p\). This kinematic reversibility is why a sphere experiences no lift, why the fore–aft pressure field is antisymmetric (contributing form drag but no net torque), and why micro-swimmers cannot propel themselves with reciprocal strokes — Purcell's "scallop theorem" is a direct corollary of the same linearity that produces \(6\pi\mu aU\).

The disturbance velocity decays only as \(1/r\), extraordinarily slowly. Physically the sphere drags a large volume of fluid with it; mathematically this long range is the Stokeslet — the flow of a point force \(\mathbf{F}\delta(\mathbf{r})\) — whose strength is exactly the drag \(6\pi\mu aU\). Every Stokes-flow field can be built from Stokeslets and their derivatives, making this result the fundamental solution of low-Reynolds hydrodynamics rather than merely one worked example.

The linear scaling in \(a\) (contrast the \(a^2\) frontal area of inertial drag) is what makes viscosity dominate at small scales: halving a particle only halves its drag but reduces its weight eightfold, so tiny particles settle glacially and diffusion wins. This single scaling underlies why bacteria live in a world of syrup, why fine dust hangs in air, and why the Stokes–Einstein relation ties a bulk transport coefficient to a molecular size.

Rigorously, the reason \(\mathrm{Re}\to0\) is a singular and not a regular perturbation is visible in the far field: the neglected inertial term \(\rho(\mathbf{u}\cdot\nabla)\mathbf{u}\sim\rho U^2 a/r^2\) and the retained viscous term \(\mu\nabla^2\mathbf{u}\sim\mu U a/r^3\) have ratio \(\sim(\rho U r/\mu)\), which exceeds unity once \(r\gtrsim a/\mathrm{Re}\). No matter how small \(\mathrm{Re}\) is, inertia eventually dominates far enough out. This is why the Stokes expansion fails to give the next term uniformly (Whitehead's paradox) and why Oseen's linearisation of the inertia in the far field — matched to the Stokes solution near the sphere — is required to compute the \(\tfrac38\mathrm{Re}\) correction. The same breakdown makes the two-dimensional cylinder problem have no bounded Stokes solution at all (Stokes' paradox).

Common misconceptions. Stokes drag is not "drag at low speed" but drag at low Reynolds number — a marble in glycerin at everyday speeds obeys it, while a dust grain in a fast wind may not. And the absence of \(\rho\) does not mean the fluid is massless; it means inertia is negligible relative to viscous forces, which is a statement about the ratio, not about \(\rho\) itself.

Worked examples
1
Drag on a bead in glycerin. A spherical bead of radius \(a=1.0\ \mathrm{mm}\) is pulled through glycerin (\(\mu=1.0\ \mathrm{Pa\,s}\), \(\rho=1.26\times10^3\ \mathrm{kg\,m^{-3}}\)) at \(U=1.0\times10^{-2}\ \mathrm{m\,s^{-1}}\). Find the drag and confirm Stokes flow applies. A
\[ \mathrm{Re}=\frac{\rho U a}{\mu}=\frac{(1.26\times10^3)(1.0\times10^{-2})(1.0\times10^{-3})}{1.0}=1.26\times10^{-2}\ll1 \]
\[ F=6\pi\mu a U=6\pi(1.0)(1.0\times10^{-3})(1.0\times10^{-2}) \]
\[ F=1.9\times10^{-4}\ \mathrm{N} \]

Reading. Because \(\mathrm{Re}\approx0.013\ll1\) the creeping-flow assumption is safe, and the drag is \(0.19\ \mathrm{mN}\), roughly the weight of a \(19\)-mg mass.

2
Terminal settling velocity of a steel sphere. A steel sphere (\(\rho_s=7.8\times10^3\ \mathrm{kg\,m^{-3}}\), radius \(a=0.50\ \mathrm{mm}\)) falls under gravity through glycerin (\(\rho_f=1.26\times10^3\ \mathrm{kg\,m^{-3}}\), \(\mu=1.0\ \mathrm{Pa\,s}\)). Find its terminal speed and verify Stokes validity. Balance net weight (gravity minus buoyancy) against drag. B
\[ \tfrac{4}{3}\pi a^3(\rho_s-\rho_f)g = 6\pi\mu a U_t \;\Rightarrow\; U_t=\frac{2a^2(\rho_s-\rho_f)g}{9\mu} \]
\[ U_t=\frac{2(0.50\times10^{-3})^2(7.8\times10^3-1.26\times10^3)(9.81)}{9(1.0)} \]
\[ \mathrm{Re}=\frac{\rho_f U_t a}{\mu}=\frac{(1.26\times10^3)(3.6\times10^{-3})(0.50\times10^{-3})}{1.0}=2.3\times10^{-3} \]
\[ U_t\approx 3.6\times10^{-3}\ \mathrm{m\,s^{-1}}=3.6\ \mathrm{mm\,s^{-1}} \]

Reading. The sphere settles at a few millimetres per second, and \(\mathrm{Re}\approx2\times10^{-3}\ll1\) confirms Stokes drag is the correct law — this is exactly the falling-ball viscometer principle, invertible to solve for \(\mu\).

Units check. \(\dfrac{\mathrm{m^2}\cdot\mathrm{kg\,m^{-3}}\cdot\mathrm{m\,s^{-2}}}{\mathrm{Pa\,s}}=\dfrac{\mathrm{kg\,m^{-2}s^{-2}}\cdot\mathrm{m^2}}{\mathrm{kg\,m^{-1}s^{-1}}}=\mathrm{m\,s^{-1}}.\)

Problems
  1. A water droplet of radius \(a=10\ \mu\mathrm{m}\) falls through air (\(\mu=1.8\times10^{-5}\ \mathrm{Pa\,s}\), \(\rho_f\approx1.2\ \mathrm{kg\,m^{-3}}\), droplet density \(\rho_s=1.0\times10^3\ \mathrm{kg\,m^{-3}}\)). Estimate its terminal velocity and check \(\mathrm{Re}\).
    Solution \(U_t=\dfrac{2a^2(\rho_s-\rho_f)g}{9\mu}=\dfrac{2(10^{-5})^2(1.0\times10^3-1.2)(9.81)}{9(1.8\times10^{-5})}\). Numerator: \(2\times10^{-10}\times998.8\times9.81=1.96\times10^{-6}\); denominator \(1.62\times10^{-4}\); \(U_t\approx1.2\times10^{-2}\ \mathrm{m\,s^{-1}}=1.2\ \mathrm{cm\,s^{-1}}\). Then \(\mathrm{Re}=\rho_f U_t a/\mu=(1.2)(1.2\times10^{-2})(10^{-5})/(1.8\times10^{-5})=8\times10^{-3}\ll1\). Stokes is valid; this is the classic cloud-droplet result.
  2. Show that the Stokes drag coefficient, defined by \(F=\tfrac12\rho U^2 C_D(\pi a^2)\), equals \(C_D=24/\mathrm{Re}\).
    Solution Set \(6\pi\mu aU=\tfrac12\rho U^2 C_D\pi a^2\). Solve: \(C_D=\dfrac{6\pi\mu aU}{\tfrac12\rho U^2\pi a^2}=\dfrac{12\mu}{\rho U a}=\dfrac{12}{\mathrm{Re}/2}\)... carefully: \(\dfrac{6\pi\mu aU}{\tfrac12\pi\rho U^2 a^2}=\dfrac{6\mu U}{\tfrac12\rho U^2 a}=\dfrac{12\mu}{\rho U a}=\dfrac{12}{\mathrm{Re}}\cdot\)? Re-do: \(\dfrac{6\mu}{\tfrac12\rho U a}=\dfrac{12\mu}{\rho U a}\). Since \(\mathrm{Re}=\rho U a/\mu\), and here the length is the radius; using diameter \(d=2a\) as the conventional length, \(\mathrm{Re}_d=\rho U d/\mu=2\rho U a/\mu\), giving \(C_D=\dfrac{24\mu}{\rho U d}=\dfrac{24}{\mathrm{Re}_d}\). Thus with the standard diameter-based Reynolds number, \(C_D=24/\mathrm{Re}\).
  3. A sphere of radius \(a=2.0\ \mathrm{mm}\) settles at terminal velocity \(U_t=5.0\times10^{-3}\ \mathrm{m\,s^{-1}}\) in an unknown oil of density \(\rho_f=0.90\times10^3\ \mathrm{kg\,m^{-3}}\); the sphere density is \(\rho_s=2.5\times10^3\ \mathrm{kg\,m^{-3}}\). Find the oil's viscosity.
    Solution From \(U_t=\dfrac{2a^2(\rho_s-\rho_f)g}{9\mu}\), \(\mu=\dfrac{2a^2(\rho_s-\rho_f)g}{9U_t}=\dfrac{2(2.0\times10^{-3})^2(2.5\times10^3-0.90\times10^3)(9.81)}{9(5.0\times10^{-3})}\). Numerator: \(2\times4.0\times10^{-6}\times1600\times9.81=0.1256\); denominator \(0.045\); \(\mu\approx2.8\ \mathrm{Pa\,s}\). (Check \(\mathrm{Re}=\rho_f U_t a/\mu=(900)(5\times10^{-3})(2\times10^{-3})/2.8\approx3.2\times10^{-3}\ll1\), valid.)
  4. Verify the split of the total drag into friction and form contributions by integrating separately. Using the surface stresses \(\sigma_{rr}=-p\) and \(\sigma_{r\theta}=-\tfrac{3\mu U}{2a}\sin\theta\), show the pressure (form) drag is \(2\pi\mu aU\) and the shear (friction) drag is \(4\pi\mu aU\).
    Solution Form drag: \(F_p=\oint(-p\cos\theta)\,dS\) with \(-p=-p_\infty+\tfrac{3\mu U}{2a}\cos\theta\). The \(p_\infty\) term integrates to zero; \(\oint\tfrac{3\mu U}{2a}\cos^2\theta\,dS=\tfrac{3\mu U}{2a}a^2\!\int_0^{2\pi}\!\!\int_0^\pi\cos^2\theta\sin\theta\,d\theta\,d\phi=\tfrac{3\mu U}{2a}a^2(2\pi)(\tfrac23)=2\pi\mu aU\). Friction drag: \(F_\tau=\oint(-\sigma_{r\theta}\sin\theta)\,dS=\oint\tfrac{3\mu U}{2a}\sin^2\theta\,dS=\tfrac{3\mu U}{2a}a^2(2\pi)\!\int_0^\pi\sin^3\theta\,d\theta=\tfrac{3\mu U}{2a}a^2(2\pi)(\tfrac43)=4\pi\mu aU\). Sum \(=6\pi\mu aU\), confirming the one-third / two-thirds split.
  5. Two identical spheres of radius \(a\) are glued in contact and towed broadside (line of centres perpendicular to \(U\)) at low \(\mathrm{Re}\). Without solving the two-body problem exactly, argue whether the total drag is more or less than \(2\times6\pi\mu aU\), and name the physical effect.
    Solution The drag is less than \(12\pi\mu aU\). Each sphere generates a \(1/r\) Stokeslet velocity field that entrains the surrounding fluid; at its neighbour this induced flow is partly co-directed with \(U\), so the neighbour moves through fluid already dragged along and feels reduced relative velocity, hence reduced drag. This hydrodynamic screening (the leading two-body Rotne–Prager / method-of-reflections correction, of order \(a/d\) with \(d\) the centre separation) is why sedimenting clusters fall faster than isolated spheres and why suspension sedimentation is hindered collectively. The exact broadside pair drag per sphere is \(6\pi\mu aU(1-\tfrac{3}{4}(a/d)+\dots)\), confirming the reduction.