Why an Ion Trap Cannot Be a Simple Well: Earnshaw's Theorem and the Paul Trap — Epoche B2
The problem: no static electric cage exists It seems natural to believe that a charged particle can be held still if we surround it with the right static electric field — a kind of electric bowl. This belief is wrong, and the reason is a classical result known as Earnshaw's theorem, proved in 1842. The argument takes three steps, each of which a first-year student already has the tools for. First, what stability means. A ball rests stably at the bottom of a valley because its potential energy rises whichever way it is nudged. For a particle of charge $e$ in an electrostatic potential $\Phi$ — the potential being the energy per unit charge, so that the energy is $U = e\Phi$ — stable equilibrium at a point requires $U$ to curve upwards along every direction at once. In the language of calculus, the second derivatives $\partial^2 U/\partial x^2$, $\partial^2 U/\partial y^2$ and $\partial^2 U/\partial z^2$ must all be positive there. Second, what electrostatics permits. Gauss's law states that the electric field diverges only where charge sits: $\nabla\cdot\mathbf{E} = \rho/\varepsilon_0$. In empty space between the electrodes the charge density $\rho$ is zero, and since the field is the downhill slope of the potential, $\mathbf{E} = -\nabla\Phi$, the two facts combine into Laplace's equation, $$\nabla^2 \Phi = \frac{\partial^2\Phi}{\partial x^2} + \frac{\partial^2\Phi}{\partial y^2} + \frac{\partial^2\Phi}{\partial z^2} = 0 .$$ The Laplacian $\nabla^2$ is just the sum of the three curvatures, so Laplace's equation says something blunt: the curvatures must cancel among themselves. Third, the collision. Three numbers that sum to zero cannot all be positive. Any static field that pushes an ion inwards along two axes must push it outwards along the third; and the degenerate case where all three vanish is no help, since a flat direction is not a restoring one. A saddle — a mountain pass, rising in one direction and falling in the perpendicular one — is the best nature allows. The same fact has a more visual statement: solutions of Laplace's equation obey the mean-value property, in that the potential at any point equals its average over any sphere drawn around that point, and a quantity equal to its own surrounding average can have no strict maximum or minimum in the interior. Earnshaw's theorem is not a statement about the ingenuity of engineers but about the structure of the field equations. Two escape routes exist, and naming them shows exactly what the theorem does and does not forbid. Magnetic fields evade it for neutral particles, because what matters there is the field magnitude $|\mathbf{B}|$, which may possess a local minimum in free space even though $\Phi$ may not — this is how magnetic traps for neutral atoms and diamagnetic levitation work (Wing, 1984). And the magnetic force on a moving charge, $e\mathbf{v}\times\mathbf{B}$, depends on velocity and so is not the gradient of any potential at all; a strong axial magnetic field added to a static electric quadrupole bends the outward escape into a circular drift, which is the Penning trap (Brown & Gabrielse, 1986). What no arrangement can do is confine a charge with electrostatic fields alone. The solution: rotate the saddle in time Wolfgang Paul's solution, honoured with the 1989 Nobel Prize, was to make the saddle oscillate. The mechanical analogue is exact and worth holding on to: a ball placed on a saddle-shaped surface rolls off, but if the saddle is rotated about its vertical axis at the right speed, the ball is repeatedly caught by the rising side before it can escape down the falling one, and it stays near the centre. Paul's trap does the same electrically, by alternating which axes are confining. The field is a quadrupole : a potential quadratic in the coordinates, which gives forces growing linearly with displacement, like a spring. Write it as $\Phi \propto \alpha x^2 + \beta y^2 + \gamma z^2$ and impose Laplace's equation: $\nabla^2\Phi \propto 2(\alpha+\beta+\gamma) = 0$. So the coefficients must sum to zero, and the simplest choice with rotational symmetry about $z$ is $\alpha = \beta = 1$, $\gamma = -2$. This is why the standard trap potential has the shape it does — the $-2$ is not a design choice but Laplace's equation cashed out: $$\Phi(x,y,z,t) = \frac{U + V\cos\Omega t}{2d^{2}}\left(x^{2} + y^{2} - 2z^{2}\right),$$ with $U$ a static voltage, $V$ the amplitude of an oscillating voltage at radio frequency $\Omega$, and $d$ a length characterising the electrode spacing. At any instant the ion is confined radially and expelled axially, or the reverse; half an RF cycle later the roles swap. The motion along each axis follows from Newton's second law with the force $-e\,\partial\Phi/\partial x$. For the $x$ direction, $$m\ddot{x} = -\frac{e(U + V\cos\Omega t)}{d^{2}}\,x .$$ Now substitute the dimensionless time $\tau = \Omega t/2$, so that $d^2/dt^2 = (\Omega^2/4)\,d^2/d\tau^2$ and $\cos\Omega t = \cos 2\tau$. Dividing through by $m\Omega^2/4$ puts the equation into the standard form named after Émile Mathieu, who studied it in 1868 while analysing the vibrations of an elliptical drumhead: $$\frac{d^2 u}{d\tau^2} + \big(a_u - 2q_u \cos 2\tau\big)\,u = 0, \qquad \tau = \frac{\Omega t}{2},$$ with, for this potential, $$a_x = a_y = \frac{4eU}{m\Omega^{2}d^{2}}, \qquad q_x = q_y = -\frac{2eV}{m\Omega^{2}d^{2}}, \qquad a_z = -2a_x, \quad q_z = -2q_x .$$ (Sign conventions differ between texts; only the magnitudes matter for what follows.) Every symbol earns its place: $a_u$ measures the static voltage against the drive, $q_u$ the oscillating voltage against the drive, and both are suppressed by $\Omega^2$, because a faster shake gives the ion less time to move during each half-cycle. The factor of $-2$ relating the axial to the radial parameters is inherited directly from the $-2$ in the potential, which came from Laplace's equation — so the three axes cannot be tuned independently, and that constraint shapes every real trap de