= Solution
Put $\theta=x-t$. Taking the <gradient> of the <velocity potential> gives
$$
\boxed{\boldsymbol u=(-\epsilon y\sin\theta,\ \epsilon\cos\theta)},\qquad
\boxed{\nabla\cdot\boldsymbol u=-\epsilon y\cos\theta}.
$$
The motion is not generally incompressible; a <velocity potential> ensures irrotationality, not zero <divergence>. Both components have zero time average at each fixed point because their sine and cosine averages vanish.
The fluid acceleration is the <material derivative> $D\boldsymbol u/Dt=\partial_t\boldsymbol u+(\boldsymbol u\cdot\nabla)\boldsymbol u$. Direct differentiation gives
$$
a_x=\epsilon y\cos\theta+\epsilon^2(y^2-1)\sin\theta\cos\theta,
\qquad
a_y=\epsilon\sin\theta+\epsilon^2y\sin^2\theta.
$$
Averaging over one period at fixed $(x,y)$ therefore yields
$$
\boxed{\overline{\boldsymbol a}=(0,\epsilon^2y/2)}.
$$
This need not be the <derivative> of the mean velocity: the nonlinear advective term remains after averaging.
The dyed particle obeys the <Lagrangian trajectory> equations
$$
\dot x=-\epsilon y\sin(x-t),\qquad \dot y=\epsilon\cos(x-t),\qquad x(0)=y(0)=0.
$$
Expand $x=\epsilon x_1+\epsilon^2x_2+O(\epsilon^3)$ and $y=\epsilon y_1+\epsilon^2y_2+O(\epsilon^3)$, for bounded times. The first-order equations give $x_1=0$, $\dot y_1=\cos t$, so $y_1=\sin t$. At second order,
$$
\dot x_2=\sin^2t,\qquad \dot y_2=0.
$$
Integrating with zero initial values verifies
$$
\boxed{x(t)=\epsilon^2\left(\frac t2-\frac{\sin2t}{4}\right)+O(\epsilon^3),\qquad y(t)=\epsilon\sin t+O(\epsilon^3)}.
$$
The velocity averaged along this particle over $0\leq t\leq2\pi$ is its displacement divided by $2\pi$. Hence
$$
\boxed{\overline{\boldsymbol u}_{\mathrm{particle}}=(\epsilon^2/2,0)+O(\epsilon^3)}.
$$
This <Stokes drift> is a Lagrangian mean; it does not contradict the zero <Eulerian mean velocity> at a fixed point. The secular term is a small-parameter expansion, not an assertion of uniform accuracy as $t\to\infty$.
Back to article page
