Fire a beam of electrons through empty space, somewhere with no electric or magnetic field at all. Most textbooks say nothing can change how they behave without a field to push them. In 1959 Yakir Aharonov and David Bohm showed that flipping a switch on a nearby coil changes them anyway. Proving it beyond argument took until 1986.
The road there runs from an 18th-century trick for the three-body problem, through the strange shape of a magnet's field, to the question Veritasium asks in the video this lesson follows: do we know how magnets work yet?
A map instead of arrows
Forces are arrows, and adding arrows that change direction with every move is hard. It's why three bodies pulling on each other have no general solution. In the 1770s Joseph-Louis Lagrange tried another way. Give every point in space a single number, a height, so that a star sits at the bottom of a pit. The force is then just the direction straight downhill, as strong as the ground is steep. Numbers are easy to add, so the landscape for any number of bodies is simply the sum of each one's.
- With the star, drag the point in close, then far out. Where is the push strongest?
- Switch to the sun and planet. Seen riding round with them, the spin flings things outward as well. Find the five places where the arrows vanish: Lagrange's balance points, where a small third body can keep its place.
- Switch to the charges. Masses only attract, so gravity only makes pits. Charges can repel too, so a + charge makes a hill.
- Now slide "Add the same number to every height". What changes, and what doesn't?
This height is the potential: \(V\) for gravity and \(\phi\) for electricity. The field is minus its gradient, the steepest way down:
A star of mass \(M\) gives \(V = -\frac{GM}{r}\). Potential energy is slightly different: it's the potential times the mass of whatever sits there, \(U = mV\).
Only the slope matters
Add 10 to every height and nothing changes. Every slope is the same as before, so every arrow and every force is too. Add a million, or take away a million: still nothing. Infinitely many potentials give exactly the same field.
You've met this in calculus. \(5x^2 + 5\) has the derivative \(10x\), but go back the other way and you get \(\int 10x\,dx = 5x^2 + C\). The derivative threw the 5 away, and nothing can bring it back. The field is the derivative of the potential, so it carries less information than the potential does.
So for nearly two centuries, physicists treated potentials as a handy shortcut and the fields as the real thing. Hold on to that.
A magnet's field lines are loops
- Compare the bar magnet with the + and − charges. From outside, can you tell them apart?
- Turn on "See inside the magnet". Which way do the amber arrows run inside it? Which way do they run between the charges?
From outside, a magnet's field looks just like the field of a + and a − charge. Inside, it's the opposite. Electric field lines start on + charges and end on − ones. Magnetic field lines carry straight on through the magnet, from S back round to N, and close up. They never start or end, because there's no magnetic charge for them to start on. Snap a magnet in half and you don't get a lone N and a lone S: you get two smaller magnets, each with both ends.
The loops come from electrons. Each electron is itself a tiny magnet, and in a piece of iron vast numbers of them line up together.
Loops break the height map. Follow a field line downhill all the way round its loop and you'd arrive back where you started, but lower, like an Escher staircase. No landscape can do that, so magnetism needs a different kind of potential.
Curl: does it turn a paddle wheel?
In the 1840s a young William Thomson, later Lord Kelvin, found one. To follow him you need one idea from divergence and curl. Picture the arrows of a field as flowing water and drop in a paddle wheel. How fast it spins is the field's curl at that spot. (The name came later, from Maxwell.)
- This whirlpool turns like a solid wheel, all of it together. Move the paddle wheel well away from the centre. Does it still spin?
- In the river, nothing goes round in a circle. Does the wheel turn?
- Choose "Round a solenoid". Put the wheel outside the coil, where the arrows plainly go round, then move it inside.
In a flat picture, the curl of \(\mathbf{F} = (F_x, F_y)\) is one number: how much \(F_y\) grows going across, minus how much \(F_x\) grows going up. In three dimensions it's an arrow along the wheel's axle, written \(\nabla \times \mathbf{F}\).
Thomson's insight was that the magnetic field is the curl of another field, the magnetic vector potential \(\mathbf{A}\). It's a potential with a direction at every point. Now each of the three has one:
The last thing you tried is the one that matters. Round a long coil carrying a current (a solenoid), \(\mathbf{A}\) circles the coil, but outside it has no curl at all. So outside there's no magnetic field, yet \(\mathbf{A}\) isn't zero. A field and its potential can disagree about whether anything is there.
\(\mathbf{A}\) has the same freedom as a height, too. Add anything with no curl, such as the same arrow everywhere, and \(\mathbf{B}\) doesn't change.
Electrons carry a clock
In quantum mechanics an electron is a wave, and the Schrödinger equation says how the wave moves. Picture the wave's phase as a clock hand that turns as the electron travels, one full turn every wavelength. When two waves meet, what you see depends on how their hands line up: in step they add, and opposite they cancel.
The Schrödinger equation doesn't contain the fields \(\mathbf{E}\) and \(\mathbf{B}\). It contains the potentials, \(\phi\) and \(\mathbf{A}\). And \(\mathbf{A}\) changes how fast the hand turns along the way.
- Slide A so it points the way the electron is going, then against it. Watch the spacing of the crests.
For a charge \(q\) travelling along a path, \(\mathbf{A}\) turns its clock by
In words: add up the part of \(\mathbf{A}\) that points along the path, all the way along it. \(\hbar\) is Planck's constant divided by \(2\pi\). An electron's charge is negative, which is why \(\mathbf{A}\) pointing its way slows its clock.
Most physicists weren't troubled. You put potentials into equations because they make the maths easier, but surely the fields are what really act. Aharonov disagreed. You can't swap \(\phi\) for \(\mathbf{E}\) in the Schrödinger equation without losing the constant, just as \(5x^2 + 5\) became \(5x^2 + C\). So he went looking for a place with a potential and no field, where only the potential could be doing anything.
The Aharonov–Bohm experiment
- Turn up the flux. The electrons only ever fly where the magnetic field is zero. Do the fringes move?
- Find the flux that moves them half a fringe, then a whole one. Could you tell a whole fringe from no current at all?
- Slide "Add the same arrow to A everywhere". The crests change on both paths. Do the fringes?
Split a beam of electrons in two, send the halves either side of a long solenoid, and bring them together on a screen. Where the two waves arrive in step there's a bright stripe, and where they're out of step a dark one: interference fringes. An ideal solenoid keeps all its field inside, so the electrons never pass through any. But \(\mathbf{A}\) circles round the coil: above it, \(\mathbf{A}\) points against the electrons, and below it, along them. One clock runs fast and the other slow, so when the current comes on, the fringes slide.
How far? Follow the two clocks.
Each path turns its own clock:
The screen only shows the difference. Out along one path and back along the other is a loop round the solenoid, written \(\oint\):
Stokes's theorem turns the loop of \(\mathbf{A}\) into its curl added up over the area inside. That curl is \(\mathbf{B}\), and \(\mathbf{B}\) added up over the area is the magnetic flux \(\Phi\) through the coil:
So the fringes move by an amount set only by a field the electrons never touched:
And the arbitrary part of \(\mathbf{A}\) cancels. Add the same arrow \(\mathbf{C}\) everywhere: round a closed loop you go out and come back, so it adds up to nothing.
For an electron, the pattern repeats every \(h/e \approx 4.1 \times 10^{-15}\) webers of flux. (A weber is a tesla times a square metre, so that's a very small amount of field.)
Aharonov was Bohm's student. Bohm had been kept off the Manhattan Project as a security risk, lost his job at Princeton in the McCarthy era, and moved on to Brazil, Israel and then Bristol, where the two worked it out. They weren't quite the first: Werner Ehrenberg and Raymond Siday had predicted the effect in 1949, and almost nobody noticed. Many physicists thought it couldn't be true. Others, Feynman among them, later said it had been sitting in the Schrödinger equation since the day it was written.
Proving it
Real coils aren't infinitely long, so some field always leaks out, and a sceptic can blame the leak. In 1960 Robert Chambers, also at Bristol, used an iron whisker about a micrometre thick as his solenoid and saw the fringes shift. Critics pointed at the whisker's ends.
The answer came in 1986 from Akira Tonomura's team at Hitachi. They used a tiny ring-shaped magnet, which keeps all its field inside itself, and coated it in niobium. Cooled to 5 degrees above absolute zero, niobium becomes a superconductor and won't let magnetic field through, so none could leak. A copper coat on top kept electrons out of the ring. Then they sent electrons both through the hole in the middle and round the outside.
A superconductor can only hold magnetic flux in whole steps of \(h/2e\). One step gives the electrons through the hole an extra phase of
exactly half a fringe. That's what they saw: bright fringes outside the ring lined up with dark ones inside. Rings that trapped an even number of steps showed no shift, also as predicted. The effect is real.
So what's real?
The measurement is settled. What it means isn't, and physicists split two ways.
- The potentials are real. It's \(\mathbf{A}\) that appears in the Schrödinger equation, and \(\mathbf{A}\) is right there where the electrons fly. Aharonov and Bohm called their paper "Significance of electromagnetic potentials in the quantum theory", and Feynman argued in his lectures that \(\mathbf{A}\) is a real field.
- The field acts at a distance. The field is shut inside the coil, yet it affects electrons that never go there. Aharonov himself came round to this view, calling it a non-local effect of the field. Many physicists dislike it, because the point of fields was that causes act only where they are.
Either way, nothing you can measure depends on the arbitrary part of \(\mathbf{A}\). Only the loop \(\oint \mathbf{A}\cdot d\mathbf{l}\), which is the flux, ever shows up.
There's a gravitational version. In 2022 a team at Stanford launched rubidium atoms up a 10-metre vacuum tower with a 1.25 kg block of tungsten at the top. Each atom's wave was split into two pieces about 25 centimetres apart, and only the upper piece passed close to the tungsten. When they recombined, the phase had shifted by an amount the tungsten's pull on the atoms' paths can't account for, but the gravitational potential can.
So, do we know how magnets work? The equations predict what magnets do to many decimal places. What physicists still argue about is what's doing the work: the field \(\mathbf{B}\), or the potential \(\mathbf{A}\) beneath it.
Problems
A potential is \(V = 3x + 2y\). Find the field \(-\nabla V\). Then do it for \(V = 3x + 2y + 7\).
Answer
\(-\nabla V = (-3, -2)\) both times. The 7 has no slope, so it makes no difference to the field.
Find the curl of \(\mathbf{F} = (-y, x)\) and of \(\mathbf{F} = (y, 0)\). Which is like the whirlpool and which like the river?
Answer
For \((-y, x)\): \(\frac{\partial x}{\partial x} - \frac{\partial (-y)}{\partial y} = 1 + 1 = 2\), the whirlpool.
For \((y, 0)\): \(0 - 1 = -1\). Nothing goes round, but it still turns a wheel clockwise, like the river.
Outside a solenoid, \(\mathbf{A}\) has the shape \(\mathbf{A} = \frac{(-y,\; x)}{x^2 + y^2}\). Show that its curl is zero. Then find \(\oint \mathbf{A}\cdot d\mathbf{l}\) once round a circle of radius \(r\) about the middle.
Answer
\(\frac{\partial}{\partial x}\frac{x}{x^2+y^2} = \frac{y^2 - x^2}{(x^2+y^2)^2}\) and \(\frac{\partial}{\partial y}\frac{-y}{x^2+y^2} = \frac{y^2 - x^2}{(x^2+y^2)^2}\). The curl is the first minus the second: zero.
On the circle, \(\mathbf{A}\) points along the path with length \(\frac{1}{r}\), so the loop adds up to \(\frac{1}{r} \times 2\pi r = 2\pi\), not zero. Stokes's theorem isn't broken: the circle goes round the middle, where the curl isn't zero. That's where the flux is.
What flux moves an electron's fringes by exactly half a fringe? Use \(h = 6.63 \times 10^{-34}\) J s and \(e = 1.60 \times 10^{-19}\) C.
Answer
Half a fringe is \(\Delta\theta = \pi\), so \(\frac{e\Phi}{\hbar} = \pi\) and \(\Phi = \frac{\pi\hbar}{e} = \frac{h}{2e} = \frac{6.63 \times 10^{-34}}{3.20 \times 10^{-19}} \approx 2.07 \times 10^{-15}\) webers. It's the same step a superconductor holds flux in, which is why Tonomura saw exactly half a fringe.
A solenoid 1 micrometre in radius holds a field of 0.01 T. How many fringes does the pattern slide when it's switched on?
Answer
\(\Phi = B \times \pi r^2 = 0.01 \times \pi \times (10^{-6})^2 \approx 3.14 \times 10^{-14}\) webers. One fringe is \(h/e \approx 4.14 \times 10^{-15}\) webers, so it slides about \(7.6\) fringes.
The flux is \(3h/e\). What does the screen show, compared with the current off?
Answer
\(\Delta\theta = 3 \times 2\pi\): three whole turns of the clock, so the fringes land exactly where they were. You'd only know by watching them slide past while the current rose.