Relativity

The mass a stopwatch stops being able to weigh

A magnet measures a charged particle's momentum and a stopwatch its speed, and relativity says the two together give its mass: m = p√(1/β² − 1). At one gigaelectronvolt of momentum a pion and a kaon cross three metres about a nanosecond apart, and a good clock tells them apart easily. At five, they cross 45 picoseconds apart. The mass has not changed. It has moved into the factor γ, which a speed measurement reads through 1 − β ≈ 1/2γ², and every factor of ten in momentum costs a hundredfold in the clock or the path.

Assumes: The speed that stops while the energy keeps coming · The invariant that survives a boost

A collision in a particle accelerator sends out a spray of charged particles, and the first question asked of each is what it is. Most are pions, the lightest particles made of quarks; some are kaons, carrying a strange quark; some are protons; occasionally there is something rarer. They all leave curved tracks in a magnetic field and look, to the tracking detectors, much the same. Telling them apart means weighing them, one at a time, while they are moving at nearly the speed of light, and the first instrument built to do it was a stopwatch.

The idea is simple and depends entirely on relativity. The invariant that survives a boost identified the mass as the one combination of energy and momentum every observer agrees on. A magnet measures the momentum: a particle of charge qq moving across a field BB curves with radius r=p/qBr = p/qB. A stopwatch measures the speed: the time taken to cover a known path. Momentum and speed are related through the mass by

p=γmv=mβc1−β2,p = \gamma m v = \frac{m\beta c}{\sqrt{1-\beta^2}},

so the mass is

m=pc1β2−1.m = \frac{p}{c}\sqrt{\frac{1}{\beta^2} - 1}.

Two measurements, one formula, and the particle is identified. This essay is about why the formula stops working, and how fast.

The stopwatch that found the antiproton

The method has a famous success. In 1955 Owen Chamberlain, Emilio Segrè and their colleagues at Berkeley went looking for the antiproton, which the newly built Bevatron had been designed to make: its beam energy was chosen to clear the threshold that the partner that sets the price of a pair computed for making a proton and an antiproton together on a fixed target. The difficulty was that each antiproton came with tens of thousands of negative pions. The experimenters used magnets to select negative particles of a single momentum, 1.19 GeV/c, and then timed them over a flight path of about twelve metres. At that momentum a pion moves at 0.993 of the speed of light and crosses the path in about 41 nanoseconds; an antiproton, with nearly seven times the mass, moves at 0.785 and takes about 52. Eleven nanoseconds was enormous even for the electronics of 1955, and a small cluster of particles arriving late, at the time the antiproton’s mass predicted, was the discovery.

The numbers show why the experiment was possible and how narrow the window was. The selected momentum had to be high enough for antiprotons to emerge from the target in useful numbers and low enough for their speed to stay well below the pions’. At three times the momentum the delay over the same path would have been a few nanoseconds; at ten times, a few hundred picoseconds, beyond any timing of the day. The discovery was made in the range where a stopwatch can still weigh, and it was designed to be.

Speed sorts a common momentum

At a fixed momentum, heavier particles move more slowly. That is true in Newton’s mechanics, where v=p/mv = p/m, and it stays true in Einstein’s, though the dependence changes shape completely.

Particles of one momentum, sorted by speed. Speed against momentum for five charged particles. At the same momentum a heavier particle moves more slowly, so speed and momentum together fix the mass: m = p√(1/β² − 1). At 0.1 GeV/c the curves are far apart; at 1 GeV/c a pion moves at 0.990 of c and a kaon at 0.897; by 10 GeV/c everything heavier than an electron is within half a per cent of c, and the curves have run together against the line of light.
Fig. 1 Speed against momentum for electrons, pions, kaons, protons and deuterons. At 1 GeV/c a pion moves at 0.990 of the speed of light and a kaon at 0.897; by 10 GeV/c every particle heavier than an electron is within half a per cent of cc, and the curves have merged against the line of light.

At low momentum the curves are spread across the whole range of speeds, and a measurement of speed sorts particles crisply. As the momentum rises, each curve bends towards β=1\beta = 1 and flattens there, and the curves bunch together in the order of their masses. An electron is relativistic already at a few MeV/c; a pion, 273 times heavier, at a few hundred; a proton at a few GeV/c. Beyond that, everything moves at nearly the same speed whatever it is, and a speed measurement, however precise, has very little left to read.

The speed that stops while the energy keeps coming followed the same flattening from the other side: Bertozzi’s electrons, given more and more energy, crossed his flight path in nearly the same time, the speed refusing to rise past cc while the heat they deposited kept growing. There the lesson was that energy and speed part company. Here it is the consequence for anyone trying to use speed as a scale: the particle’s mass, which sets where its curve bends, is written into the speed less and less legibly.

How far apart they arrive

A detector does not measure speed directly. It measures the time a particle takes to cross from one counter to another, a few metres apart, and compares particles of the same momentum. The time for a path LL is

t=Lβc=Lc1+m2c2p2,t = \frac{L}{\beta c} = \frac{L}{c}\sqrt{1 + \frac{m^2c^2}{p^2}},

and for two masses at the same high momentum the difference is approximately

Δt≈Lc (m22−m12) c22p2.\Delta t \approx \frac{L}{c}\,\frac{(m_2^2 - m_1^2)\,c^2}{2p^2}.

How far apart in time particles of one momentum arrive. The difference in arrival time over a 3 m flight between particles of equal momentum, on a logarithmic scale. At high momentum it falls as (m₂² − m₁²)L/2p²c, a factor of a hundred for every factor of ten in momentum. With a clock good to 100 ps, three standard deviations of separation last up to 0.57 GeV/c for electrons from pions, 1.92 for pions from kaons and 3.22 for kaons from protons. Above that the particles arrive together to within the clock's own blur.
Fig. 2 The difference in arrival time over a 3 m flight for particles of the same momentum, on a logarithmic scale. Above a few hundred MeV/c it falls as 1/p21/p^2, a hundredfold for each tenfold rise in momentum. With a 100 ps clock, three standard deviations of separation reach 0.57 GeV/c for electrons and pions, 1.92 for pions and kaons and 3.22 for kaons and protons.

The inverse square is the whole story. A pion and a kaon of 1 GeV/c, flying three metres, arrive about 1.1 nanoseconds apart, which a clock good to a tenth of a nanosecond resolves easily. At 2 GeV/c they arrive a quarter of that apart, and at 10 GeV/c eleven picoseconds apart, well below any timing that a large detector can maintain across thousands of channels. The clock’s resolution draws a horizontal line on the figure, three times its standard deviation for a confident separation, and each pair of particles can be told apart only to the left of where its curve crosses the line.

The difference in time comes from the difference in mass squared. That is what relativity puts into the formula and Newton’s mechanics does not: in Newton’s world, two particles with the same momentum and masses differing by a factor of three would have speeds differing by a factor of three at any momentum, and a stopwatch would separate them forever. In relativity the speeds converge, and the convergence is controlled by m2/p2m^2/p^2, the same combination that appears in the invariant E2−p2c2=m2c4E^2 - p^2c^2 = m^2c^4. The mass is still there; it has moved from the speed into the energy, where a stopwatch cannot reach it.

What a clock can still weigh

The time difference sets whether two species can be separated. The mass itself is reconstructed particle by particle, and it is cleanest to work with the mass squared, which the measurement gives directly:

m2=p2c2(c2t2L2−1).m^2 = \frac{p^2}{c^2}\left(\frac{c^2t^2}{L^2} - 1\right).

Each particle yields one value. With a perfect clock and a perfect magnet, the values would stack into sharp spikes at the squares of the true masses. Real measurements spread them, and the spread is set by the uncertainties in time and momentum propagated through the formula.

The masses a stopwatch can still tell apart. Masses reconstructed from a 1% momentum measurement and a 100 ps flight time over 3 m, for equal numbers of pions, kaons and protons, at two momenta. At 1 GeV/c three clean peaks stand at the squares of the three masses, the kaon's 0.023 (GeV/c²)² wide. At 2.5 GeV/c the same clock gives the kaon a width of 0.127, the pion and kaon peaks have run together into one lumpy hump, and only the proton, five widths further on, still stands clear. Nothing about the particles changed; their speeds moved closer to c.
Fig. 3 The mass squared reconstructed from a 1 per cent momentum measurement and a 100 ps flight time over 3 m, for equal numbers of pions, kaons and protons. At 1 GeV/c three separate peaks stand at the true masses squared; at 2.5 GeV/c the same instruments give peaks six times wider, the pion and kaon run together, and only the proton stands clear.

At 1 GeV/c the three peaks are well separated, the kaon’s 0.023 square gigaelectronvolts wide against a spacing of 0.22 from the pion. At 2.5 GeV/c the same clock and the same magnet give peaks 0.13 wide, the pion and the kaon have run into one hump, and only the proton, five widths further on, still stands apart. The broadening comes almost entirely from the clock. A 1 per cent momentum error contributes 2m22m^2 times 1 per cent to the mass squared, which for a kaon is a hundredth of a square gigaelectronvolt and does not grow with momentum; the timing error contributes 2p2/β22p^2/\beta^2 times the relative timing error, which grows as the square of the momentum.

Where the mass goes

The relative uncertainty in the mass, from the timing alone, has a compact form. Differentiating the mass formula and keeping the momentum fixed gives

σmm=γ2 σtt,\frac{\sigma_m}{m} = \gamma^2\,\frac{\sigma_t}{t},

with γ=1+p2/m2c2\gamma = \sqrt{1 + p^2/m^2c^2} the particle’s Lorentz factor.

The mass hides in γ as the speed runs out. The relative uncertainty in a particle's mass found from a 100 ps flight time over 3 m, against momentum, with the momentum taken as exact. It is γ² times the relative timing error: the mass sits in the factor γ = √(1 + p²/m²c²), and a clock measures β, whose dependence on γ flattens as 1 − 1/2γ². A kaon at 1 GeV/c is weighed to 5%; at 5 GeV/c, γ is ten and the uncertainty exceeds the mass. A proton, with smaller γ at every momentum, is weighed best.
Fig. 4 The relative uncertainty in mass from a 100 ps flight time over 3 m, against momentum, with the momentum known exactly: γ2\gamma^2 times the clock’s relative error. A kaon is weighed to 5 per cent at 1 GeV/c; at 5 GeV/c, where its γ\gamma is ten, the uncertainty exceeds the mass.

The factor γ2\gamma^2 is the price of measuring a quantity through something that saturates. The speed is β=1−1/γ2\beta = \sqrt{1 - 1/\gamma^2}, and for large γ\gamma, 1−β≈1/2γ21 - \beta \approx 1/2\gamma^2: the information about γ\gamma, and through it about the mass, lives in the distance of β\beta from one, which shrinks as the inverse square. A clock that measures β\beta to one part in a hundred is measuring 1−β1 - \beta to one part in a hundred only while 1−β1 - \beta is close to one; when 1−β1 - \beta is a hundredth, the same clock measures it to no better than its own size. A kaon at 1 GeV/c has γ\gamma of 2.2 and is weighed to five per cent. At 5 GeV/c its γ\gamma is ten, and the measurement no longer tells a kaon from nothing.

This is the reverse of a familiar statement about relativity. It is usually said that as a particle approaches the speed of light its energy grows without limit while its speed barely changes, and that is the energy side of the same fact. On the measurement side, a scale that reads speed loses its sensitivity at exactly the rate the energy gains. The two sides are tied by γ\gamma: whatever γ\gamma multiplies, energy or momentum, grows; whatever it hides in, 1−β1 - \beta, shrinks.

The path a clock would need

Faced with the inverse square, a detector designer has two levers: a better clock or a longer path. The time difference grows only in proportion to the path, so both levers are linear, and the problem is quadratic.

The path a stopwatch needs grows as the square of the momentum. The flight path needed to separate pions from kaons of the same momentum by three standard deviations of the clock, for clocks good to 20, 50 and 100 ps. The time difference grows only in proportion to the path and falls as the square of the momentum, so each factor of ten in momentum demands a hundred times the path. Even with a 20 ps clock, separating them at 10 GeV/c takes 16 m. That is why fast particles are identified by the light they emit, or the energy they lose, rather than by a stopwatch.
Fig. 5 The flight path needed to separate pions from kaons by three standard deviations, against momentum, for clocks good to 20, 50 and 100 ps. Each factor of ten in momentum multiplies the path a hundredfold; with a 20 ps clock, 10 GeV/c needs 16 m.

A detector a few metres across, with timing of fifty to a hundred picoseconds, identifies pions and kaons up to one or two GeV/c and protons to three or four. That covers most of the particles produced in heavy-ion collisions, where thousands of slow particles emerge from each event, and time-of-flight walls are a standard part of the detectors built for them. But to reach ten GeV/c with the same separation needs a path of tens of metres and timing of tens of picoseconds across a large area, and no collider detector has room for a flight path that long inside its magnet.

So fast particles are identified by quantities that do not saturate the same way. The cone of light a fast charge emits in a transparent medium, which why the glow of a fast charge is blue followed, opens at an angle set by cos⁡θ=1/nβ\cos\theta = 1/n\beta; a medium with an index very close to one, a gas, keeps that angle sensitive to β\beta up to much higher momenta, because what is measured is β\beta against 1/n1/n rather than against 1. The energy a particle loses by ionisation as it crosses matter, the subject of the kick a fast particle can give an electron, depends on βγ\beta\gamma and rises slowly with it in a way that differs between masses, and the radiation a particle emits crossing a boundary between materials depends on γ\gamma itself and grows with it, which favours exactly the light, high-γ\gamma particles a stopwatch gives up on. Each method has a momentum range where it works, and a detector carries several to cover the whole spray.

The magnet’s half of the measurement

The figures above treat the momentum as known to one per cent, and at the momenta where timing works that is roughly right; but the magnet has its own version of the problem. A track’s momentum is measured from its curvature, and the curvature of a fast particle is small: the turns a muon makes at any energy used the fact that the radius of a charged particle’s orbit grows in proportion to its momentum, γ\gamma included, and a straighter track is a harder track to measure. The sagitta, the depth of the arc across a tracker of given length, falls as one over the momentum, so a tracker with a fixed position resolution measures the momentum with a relative error that grows in proportion to it. Unlike the clock, though, the magnet loses information linearly rather than quadratically, and its contribution to the mass squared is 2m22m^2 times its relative error, with no factor of γ\gamma. The magnet keeps working long after the clock has given up; its information is about momentum, and momentum alone says nothing about which particle carries it.

That asymmetry is itself a statement about relativity. Momentum and energy grow without limit as a particle is accelerated, and mass is a form of energy is the reason energy can carry the mass’s signature at any speed: E2=p2c2+m2c4E^2 = p^2c^2 + m^2c^4. Measuring energy and momentum together, in a calorimeter and a tracker, would weigh a particle at any momentum in principle. In practice the energy of a single hadron is measured far less precisely than its momentum, to tens of per cent, and the mass is buried in the difference of two large and nearly equal numbers, which is the same subtraction the stopwatch faces, moved from the speed to the energy.

A clock that runs at the particles’ own speed

There is a further twist, which shows what is really being measured. A flight-time measurement compares the particle with light: the start signal is usually taken from the collision itself, via photons or very fast particles arriving at a counter near the beam, so what is timed is how far the particle lags behind something moving at nearly cc. The lag over the path is

t−Lc=Lc(1β−1)≈Lc m2c22p2,t - \frac{L}{c} = \frac{L}{c}\left(\frac{1}{\beta} - 1\right) \approx \frac{L}{c}\,\frac{m^2c^2}{2p^2},

which is the same m2/2p2m^2/2p^2 as the separation. A stopwatch at high momentum is really measuring how far behind light a particle falls, and that distance is the mass squared divided by twice the momentum squared, times the path. The speed that stops while the energy keeps coming computed it for an electron at the Large Electron–Positron collider, a few millimetres behind light in a second of flight; the same arithmetic puts a 1 GeV/c kaon 35 centimetres behind light after three metres, and a 10 GeV/c kaon less than four millimetres behind.

The lag also explains why heavier particles are always easier to weigh. A proton at a given momentum has a smaller γ\gamma than a kaon, so its speed is further from cc and its lag behind light longer. The resolution curves in the fourth figure are ordered by mass for that reason, and a deuteron, twice the proton’s mass, can be picked out of a crowd of pions at momenta where pions and kaons have long since merged.

What the figures leave out

The figures take the flight path as straight and known. In a magnetic field the path is a helix, and its length depends on the momentum and the angle; a real time-of-flight measurement reconstructs each track’s path length from the tracker, with its own uncertainty, which adds to the timing error and grows with the path. They take the start time as perfect. In a collider, the moment of the collision is known only to the spread of the colliding bunches, tens of picoseconds or more, and is estimated event by event from the earliest particles, which adds a correlated error to every particle in an event. They take the particles as stable over the flight. Pions and kaons decay, with lifetimes of 26 and 12 nanoseconds in their own frames, and a kaon of 1 GeV/c travels 7.5 metres on average before decaying, so a fraction of them decay before reaching the outer counters and are lost or misidentified. And they take the populations as equal, while in a real collision pions outnumber kaons several times and kaons outnumber protons, so a kaon peak near a pion peak is a small hump on the shoulder of a large one, harder to separate than equal peaks would be.

Within those limits the conclusion is exact, because it rests only on the relation between momentum, speed and mass that relativity fixes: separation in time falls as the inverse square of the momentum, and the mass uncertainty from timing grows as γ2\gamma^2.

Still open: how fine a clock can be made over a large area

Timing a single charged particle to ten or twenty picoseconds is possible with small detectors; doing it across tens of square metres, in millions of channels, in a detector bathed in radiation, is an engineering frontier. Detectors built for the next generation of colliders aim at a few tens of picoseconds, mostly to untangle the hundreds of collisions that happen in each crossing of the beams rather than to identify particles. Whether timing can be pushed to the few-picosecond level over a whole detector, which would extend identification by flight time to several times higher momenta, depends on sensors and electronics whose limits are not yet known.

The habit worth carrying away is to ask where a quantity is written. Mass is p1/β2−1p\sqrt{1/\beta^2-1}, and as the momentum rises it moves out of the speed into γ\gamma, which a stopwatch reads through 1−β≈1/2γ21-\beta \approx 1/2\gamma^2: two particles of equal momentum differ in flight time by (m22−m12)L/2p2c(m_2^2-m_1^2)L/2p^2c, so a 100 ps clock over 3 m tells pions from kaons only up to 1.92 GeV/c, and the relative error in a mass from timing is γ2\gamma^2 times the clock’s. A stopwatch weighs a particle by how far it lags behind light. At high enough momentum, nothing lags far enough to weigh.

Part 19 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.

Invariant massThe Lorentz factorMeasurement uncertaintyParticle identificationRelativistic momentumTime of flight