Relativity

The cyclotron that falls out of step

A charge in a magnetic field goes round at a rate that does not depend on how fast it is going: a faster particle runs a bigger circle in the same time. That coincidence is the whole of the cyclotron — a voltage reversing at one fixed rate can kick the particle every half turn for ever. Relativity breaks it. A proton's rate of going round falls one per cent for every nine and a half MeV it gains, each turn it arrives a little later at the gap, and after a hundred or so turns it is arriving when the voltage pushes the wrong way. The ceiling this sets rises only as the square root of the voltage, and getting past it took two ideas that are still how accelerators work.

Assumes: The speed that stops while the energy keeps coming · The force that does no work

The first cyclotron, built at Berkeley in the early 1930s, was small enough to hold in two hands, and it rested on a coincidence. A charged particle moving across a uniform magnetic field goes round a circle, and the time it takes to go round does not depend on its speed. A faster particle runs a bigger circle, and the extra distance exactly cancels the extra speed. So a voltage across a gap, reversing at the particle’s rate of going round, meets the particle every half turn with a push in the right direction. The particle spirals outward, gaining energy at each crossing, and never falls out of step, because its timing never changes.

The coincidence is a non-relativistic one. It holds well for slow particles and fails, slowly and then decisively, as they approach the speed of light. Within a few years of the first machines it was clear that the cyclotron had a ceiling, and that the ceiling was not set by the size of the magnet or the strength of the field but by relativity and the voltage across the gap. How it arises, and how it was got round, is the clearest case there is of a relativistic correction of a few per cent deciding what a machine can do.

The clock in a magnetic field

A charge qq moving at speed vv across a field BB feels a force qvBqvB at right angles to its velocity. That force does no work, as the force that does no work established, so it changes only the direction of motion, and the particle goes round a circle whose radius makes the force supply the centripetal acceleration: qvB=γmv2/rqvB = \gamma mv^2/r, with relativistic momentum γmv\gamma mv. The time for a turn is 2πr/v2\pi r/v, and the speed cancels:

ω=qBγm.\omega = \frac{qB}{\gamma m}.

For slow particles γ\gamma is one and the rate depends only on the charge, the mass and the field — about fifteen million turns a second for a proton in a field of one tesla. As the particle gains kinetic energy TT, γ=1+T/mc2\gamma = 1 + T/mc^2 grows, and the rate falls in proportion. The particle’s clock, in this sense, runs down as it gains energy, and the rate of running down is set by the rest energy: a proton is one per cent slower after gaining 9.5 MeV, a deuteron, twice as heavy, after 19.

A circling particle's clock runs down as it gains energy. How often a particle goes round in a fixed magnetic field, as a percentage of its slow-particle rate, against the kinetic energy it has gained: the rate falls as 1/γ = 1/(1 + T/mc²). A proton is 1% slower after 9.5 MeV, a deuteron, twice as heavy, after twice that. An electron is 1% slower after 5 keV and has fallen off the chart before it has gained anything worth having. A cyclotron's voltage reverses at one fixed rate, and every one of these particles drifts out of step with it.
Fig. 1 How often a particle goes round in a fixed magnetic field, as a percentage of its slow-particle rate, against the kinetic energy it has gained: the rate falls as 1/γ=1/(1+T/mc2)1/\gamma = 1/(1 + T/mc^2). A proton is 1% slower after 9.5 MeV, a deuteron after twice that; an electron is 1% slower after 5 keV and has fallen off the chart before it has gained anything worth having.

An electron’s rest energy is 1,836 times smaller than a proton’s, so it is 1 per cent slow after five kiloelectronvolts. For electrons the coincidence is never good enough to build a cyclotron on, which is why electron accelerators took a different path from the beginning.

The particle’s own clock still keeps the old time

There is a neat way to see where the factor γ\gamma comes from, and it says something about what has actually changed. In the particle’s own time, a turn takes 2πm/qB2\pi m/qB, the non-relativistic period, at every energy: the proper time between successive passes of the gap is exactly what Lawrence’s coincidence promised. What has changed is the relation between the particle’s time and the laboratory’s. A clock moving at speed vv runs slow by γ\gamma as seen from the laboratory, the effect the clock that has to slow derived from a light clock, and a particle going round a cyclotron is such a clock, carried round a circle and returned to the same place each turn like the travelling twin of the twin who comes back younger. The voltage on the electrodes keeps laboratory time. The particle keeps its own. The cyclotron fails because two clocks that agreed at low speed stop agreeing, and the disagreement is one of the oldest results in relativity, turned into an engineering limit.

The same factor can be read as inertia. In a magnetic field the force is always across the motion, and across the motion a fast particle resists being turned as if its mass were γm\gamma m, the transverse case that the push that does not point where the body goes separated from the longitudinal one. The two descriptions — a slowed clock, an increased resistance to turning — are the same arithmetic. Neither says that anything about the proton itself has changed.

Every turn a little later

A per cent sounds harmless. The difficulty is that it accumulates. The voltage across the cyclotron’s gap reverses at one fixed rate, and the particle, now going round one per cent more slowly, arrives at the gap one per cent of a turn later than the turn before. A hundred turns at that deficit add up to a whole turn, and long before that the particle is arriving when the voltage is passing through zero, or pushing backwards.

To put numbers on it, call φ\varphi the phase at which the particle crosses the gap, measured from the moment the voltage is at its peak. Each crossing adds qVcos⁡φqV\cos\varphi, two crossings a turn. If the voltage reverses at the rate of a slow particle, each turn the phase slips by 2π(γ−1)=2πT/mc22\pi(\gamma - 1) = 2\pi T/mc^2. Gaining energy makes the phase slip faster; the phase slipping makes the gain smaller. Treating the turns as a continuous variable and dividing one relation by the other gives

sin⁡φ−sin⁡φ0=πT22 mc2 qV,\sin\varphi - \sin\varphi_0 = \frac{\pi T^2}{2\,mc^2\,qV},

so the phase marches steadily towards +90°+90°, the edge of the swing, where the push vanishes, and beyond which the voltage pulls the particle back.

The phase slides to the edge of the swing, and the gain stops. The phase at which a proton crosses the accelerating gap, measured from the crest of a 100 kV swing, against the energy it has gained, turn by turn. Every turn the proton arrives a little later than the last, by 2π(γ − 1). Starting on the crest, it slides to the edge of the swing and stops gaining at 7.7 MeV; starting near the trough, it sweeps the whole swing and stops at 10.9 MeV; with the voltage reversing 1.14% slower than the slow proton circles, the phase first runs back and then forward, and the limit doubles to 21.8 MeV.
Fig. 2 The phase at which a proton crosses the gap, measured from the crest of a 100 kV swing, against the energy it has gained, turn by turn. Starting on the crest, it slides to the edge of the swing and stops gaining at 7.7 MeV; starting near the trough, it sweeps the whole swing and stops at 10.9 MeV; with the voltage reversing 1.14% slower than the slow proton circles, the phase first runs back and then forward, and the limit doubles to 21.8 MeV.

The figure follows the particle turn by turn rather than by the continuous formula, and the two agree. A proton launched on the crest of a 100-kilovolt swing reaches the edge after about fifty turns and 7.7 MeV. One launched near the trough of the swing — arriving early, so that it starts by gaining little but has the whole swing ahead of it — reaches 10.9 MeV before it falls out. Setting the voltage’s frequency a little below the slow proton’s rate does better still: the phase first slides backwards, because the particle is arriving early, then turns and slides forwards once the particle has slowed, sweeping the swing twice; the ceiling doubles, to about 22 MeV. No choice of starting phase or frequency does better than a small multiple of these numbers.

A ceiling that rises only as the square root

Starting from the trough, the phase can travel from −90°-90° to +90°+90°, a change of two in sin⁡φ\sin\varphi, and the energy at which it gets there is

Tmax=2mc2 qVπ.T_\text{max} = 2\sqrt{\frac{mc^2\,qV}{\pi}}.

The magnet does not appear. A larger magnet lets the particle go round more times before reaching the edge of the machine, but the particle has already stopped gaining. What sets the ceiling is the voltage across the gap, which decides how much energy is gained per turn and therefore how few turns are spent slipping, and the rest energy, which decides how fast the slip grows.

A ceiling that rises only as the square root of the voltage. The highest energy a fixed-frequency cyclotron can give, 2√(mc²qV/π), against the voltage across its gap, for electrons, protons and deuterons, with the proton's ceiling doubled by the best detuning of the frequency. At 100 kV a proton tops out at 10.9 MeV, or about 22 MeV detuned; quadrupling the voltage only doubles it. An electron's rest energy is 1,836 times smaller and so is its ceiling squared: 255 keV at 100 kV, which is why electrons are never put in an ordinary cyclotron.
Fig. 3 The highest energy a fixed-frequency cyclotron can give, 2mc2qV/π2\sqrt{mc^2qV/\pi}, against the voltage across its gap, for electrons, protons and deuterons, with the proton’s ceiling doubled by the best detuning. At 100 kV a proton tops out at 10.9 MeV, or about 22 MeV detuned; quadrupling the voltage only doubles it. An electron’s ceiling is 255 keV at 100 kV.

The square root is the hard part. Gap voltages of a hundred or two hundred kilovolts were what the radio-frequency technology of the 1930s could sustain across a cyclotron’s electrodes, and they put the ceiling for protons and deuterons in the range of ten to twenty-odd MeV — which is where the largest classical cyclotrons ended up. To double the ceiling the voltage must be quadrupled. Hans Bethe and Morris Rose pointed out the limit in 1937, and for a few years it looked like the end of the line for circular accelerators of this kind.

The electron’s ceiling, 255 keV at 100 kilovolts, is below the energy at which an electron’s speed is already three-quarters of light’s. The coincidence on which the cyclotron rests barely exists for electrons at all. One machine that does accelerate electrons in a magnetic field, the microtron, turns the problem round: it gives the electron so much energy per turn — about one rest energy — that each orbit takes exactly one more cycle of the voltage than the last, so the particle keeps step by skipping a whole cycle every turn instead of slipping a fraction.

The field that keeps step will not keep the beam

The direct fix is to make the field stronger where the particles are faster. If BB grows with radius exactly as γ\gamma does, qB/γmqB/\gamma m stays constant, the revolution rate is the same at every radius, and the voltage can stay at one frequency for ever. Such a field is called isochronous, and it is easy to describe. The trouble is that it does not hold the beam.

A cyclotron’s orbits are kept in their plane by the shape of the field lines. If the field weakens slightly with radius, the lines bow outward, and a particle drifting above or below the median plane meets a field component that pushes it back; if the field also does not weaken too fast, a particle drifting inward or outward is pushed back too. Both conditions together are expressed through the field index n=−(r/B) dB/drn = -(r/B)\,dB/dr, which must lie between zero and one. A field that strengthens outward has negative nn and bows its lines the other way.

The field that keeps step will not keep the beam. To keep a fixed frequency, the field must grow with radius exactly as γ does (solid). A field that grows outward bends its lines the wrong way to hold the orbit's plane: the field index n = −(r/B)(dB/dr), which must lie between 0 and 1 for an orbit to be stable both across and up and down (shaded), becomes −(γ² − 1), already −0.22 at 100 MeV. The beam spreads vertically and strikes the poles. Shaping the field into alternating ridges and valleys restores the vertical hold, which is how isochronous cyclotrons work.
Fig. 4 To keep a fixed frequency the field must grow with radius exactly as γ\gamma does (solid). The field index n=−(r/B) dB/drn = -(r/B)\,dB/dr, which must lie between 0 and 1 for an orbit to be held both across and up and down (shaded), then becomes −(γ2−1)-(\gamma^2 - 1), already −0.22-0.22 at 100 MeV. The beam spreads vertically and strikes the poles.

For the isochronous field the index is exactly −(γ2−1)-(\gamma^2 - 1): a small negative number at low energy and a large one at high, and vertically defocusing throughout. The classical cyclotron is in fact built the other way — with a field that falls slightly with radius to provide focusing — which makes its timing problem worse, not better. Timing and focusing pull in opposite directions, and a uniform azimuthally symmetric field can have one or the other but not both.

Llewellyn Thomas showed in 1938 how to have both. If the field is made to vary round the circle as well as with radius — strong over ridges, weak over valleys — a particle crossing the boundaries between them at an angle gets a vertical push at each crossing, and the pushes can be arranged to focus. The average field can then rise with radius to keep step, while the alternation provides the focusing the rising field takes away. The idea waited until after the Second World War to be built, and it underlies the sector-focused, isochronous cyclotrons that now produce protons of hundreds of MeV for physics and for cancer therapy. Spiral ridges, which add a further alternating component, are why the poles of such machines look like a pinwheel.

Sweep the frequency instead

The other fix leaves the field alone and changes the frequency. If the voltage’s frequency is lowered as the particle gains energy, in step with 1/γ1/\gamma, the particle can be kept at the right phase all the way.

Sweep the frequency down and the ceiling disappears. The frequency at which a synchrocyclotron's voltage must reverse, as a fraction of its starting value, to stay in step with a proton of each energy: 1/γ. It is 0.904 at 100 MeV, 0.728 at 350 and 0.573 at 700. Sweeping the frequency down once per bunch, and relying on the phase stability that holds a bunch near the right phase while the sweep is slow, removes the ceiling of the fixed-frequency machine; the price is that only one bunch at a time is accelerated, so the beam arrives in pulses.
Fig. 5 The frequency at which a synchrocyclotron’s voltage must reverse, as a fraction of its starting value, to stay in step with a proton of each energy: 1/γ1/\gamma. It is 0.904 at 100 MeV, 0.728 at 350 and 0.573 at 700.

That sounds as though it needs impossible precision — matching a frequency to a particle’s energy turn by turn — and it would, except for a self-correcting effect found independently by Edwin McMillan and Vladimir Veksler in 1944 and 1945. If the frequency is swept slowly, and the particles cross the gap while the voltage is falling rather than at its peak, a particle that arrives a little early gets a little more energy, which slows its revolution and makes it arrive later next time; one that arrives late gets less, speeds up, and arrives earlier. The bunch oscillates about the right phase rather than drifting away from it. That is phase stability, and it is the same mutual adjustment the clock that is pulled into step followed in pendulum clocks on a common beam, now acting between a particle’s energy and its timing.

The synchrocyclotron, a cyclotron with its frequency swept down once for each bunch, removed the energy ceiling: Berkeley’s large magnet, originally built for a classical cyclotron, was rebuilt as one in 1946 and accelerated deuterons and protons to hundreds of MeV. The price is that only one bunch is accelerated at a time, during a sweep that takes thousands of turns, so the beam comes out in pulses and its average intensity is far below an isochronous machine’s. Synchrotrons, which sweep the field as well as the frequency and keep the radius fixed, took the same principle to the energies of the speed that stops while the energy keeps coming.

The same clock, read another way

The slowdown that ruins a cyclotron is not a nuisance everywhere. A particle’s revolution rate in a known field is a measurement of γm\gamma m, and Penning traps, which hold a single ion in a magnetic field and listen to it go round, weigh ions to parts in 101110^{11} by exactly this formula at low energy. At high energy, the turns a muon makes at any energy followed a muon’s spin precessing in a storage ring a little faster than its orbit, both rates slowed by the same γ\gamma, so that their difference measures the muon’s magnetic anomaly. And the flash a circling charge sends once a turn showed that a fast electron on such a circle radiates in a narrow cone, the reason synchrotron light comes in pulses. Each of these uses ω=qB/γm\omega = qB/\gamma m; the cyclotron is the device that asked it to have no γ\gamma in it.

How far a hundred turns goes

The scale of the machine follows from the same numbers. A proton gaining 0.2 MeV a turn needs about fifty turns to reach the classical ceiling, and the radius of its orbit at kinetic energy TT is p/qBp/qB, which at ten MeV in a field of 1.5 tesla is about thirty centimetres. The classical cyclotrons that reached the ceiling were therefore a metre or so across, with magnets weighing tens of tonnes; making them bigger made the orbits longer, not the protons faster. A modern isochronous machine for proton therapy reaches 230 MeV in a few metres, and a laser-driven plasma, as the wave an electron rides to a gigaelectronvolt described, does something comparable for electrons in centimetres, by keeping the particle on the crest of a wave that travels with it rather than kicking it at a fixed point once a turn. Every one of these machines is a way of keeping a particle in step with the thing that pushes it.

What the turn-by-turn picture leaves out

The figures follow one particle through a field that is perfectly uniform, crossing a gap of zero width with an instantaneous kick, in a voltage that is a pure sinusoid. A real cyclotron’s field falls with radius for focusing, which makes the particle slower at large radius and the slip larger; its gap has width, and the particle takes time to cross it, which reduces the effective voltage; and the beam is a cloud of particles with a spread of phases and energies, of which only some reach the ceiling. The square-root ceiling is the best case for the classical machine, and real classical cyclotrons approached it only within a factor of order one.

The argument also treats the energy gain per turn as small compared with the particle’s energy, which allows the turns to be smoothed into a continuous variable. That fails in the first few turns and is why the figure is drawn turn by turn; the two agree to within a per cent at the ceiling. Space charge — the particles’ repulsion of each other — is ignored throughout, and in a high-current machine it is often the effect that sets the real limit.

Still open: how much current an isochronous cyclotron can carry

Modern isochronous cyclotrons deliver proton beams of milliamperes at several hundred MeV, and the most powerful of them is limited not by relativity but by the beam’s own space charge and by losses: a tiny fraction of the beam striking the machine’s structure makes it radioactive, and at megawatt beam powers that fraction must be below a part in ten thousand. Whether cyclotrons can be pushed to the ten-megawatt beams some proposed facilities want — for driving subcritical reactors, or producing neutrinos — depends on how the turns can be kept separated at extraction when the beam is that intense, and that is being worked out in simulation and in the machines that exist.

A charge in a magnetic field circles at ω=qB/γm\omega = qB/\gamma m, so a proton’s rate falls 1% for every 9.5 MeV it gains; with a voltage reversing at one fixed rate, its crossing phase slips 2π(γ−1)2\pi(\gamma - 1) a turn until it reaches the edge of the swing, which caps a classical cyclotron at 2mc2qV/π2\sqrt{mc^2qV/\pi} — 10.9 MeV for protons at 100 kV, about 22 MeV with the best detuning, 255 keV for electrons — a ceiling no larger magnet moves; a field that rises with radius to keep step defocuses the beam with index −(γ2−1)-(\gamma^2 - 1) unless it alternates in sectors, and sweeping the frequency down as 1/γ1/\gamma removes the ceiling because the particles’ timing corrects itself. The cyclotron’s coincidence was a non-relativistic one, and getting past it is how accelerators learned to keep step.

Part 20 of 20

This essay is one argument about Relativistic dynamics. The others:

The objects named here

The third axis, after the field and the reading path: the things themselves, and every essay that touches each one.

Cyclotron frequencyField indexIsochronous cyclotronThe Lorentz factorPhase stabilitySynchrocyclotron