Astrophysics

The tube that turns its beam into heat

An X-ray tube is an electron gun aimed at a block of tungsten, and the X-rays are the light the electrons give off as they are stopped. Fewer than one electron-volt in a hundred comes out that way. The rest heats the target, so violently that a hospital CT tube's focal spot would melt in thirty-six microseconds if the tungsten under it stood still. The fraction radiated is set by one ratio — how hard the nucleus deflects an electron compared with how often the atom's own electrons nudge it — and the same ratio, read a thousand times higher up the energy scale, is where an electromagnetic shower begins.

Assumes: The bill that arrives when the pushing stops · A charge that turns must glow

A dental X-ray set draws a few hundred watts from the wall for a fraction of a second, and a CT scanner draws tens of kilowatts for several seconds at a time. In both, the electrical power goes into an electron beam, accelerated through a potential of a few tens to a hundred and fifty kilovolts and driven into a metal target. The X-rays that leave the target are light emitted by those electrons as they slow down — braking radiation, bremsstrahlung — plus a few sharp lines from atoms the beam has ionised. And almost none of the beam’s energy becomes either. Of every hundred joules a tungsten target receives at 100 kilovolts, about six-tenths of a joule leaves as X-rays.

That number is not an engineering shortfall that better tubes have failed to close. It follows from a single ratio, the rate at which an electron radiates compared with the rate at which it loses energy to the atom’s electrons, and the ratio is fixed by the target’s atomic number and the electron’s energy and by nothing else. This essay follows it from the shape of a tube’s spectrum to the heat that forces its target to spin, and then up the energy scale to the point where it reaches one and radiation stops being the minor partner.

A triangle for every voltage

The bill that arrives when the pushing stops followed what a charge radiates when it is brought to rest suddenly: a broad spectrum, flat up to a cut-off set by how sudden the stop is. A thick target does not stop an electron in one event. It slows it through millions of gentle encounters with atomic electrons and occasionally deflects it hard in the field of a nucleus, and the hard deflections are where the radiation comes from. Each one emits a flat spectrum up to the electron’s energy at that moment — the shape the field that points where the charge is now derived for a sudden deflection, a number of photons per unit frequency falling as one over the frequency, which is a flat spectrum of energy; summing over the whole slowing-down gives the thick-target spectrum Hendrik Kramers derived in 1923, in which the energy emitted per unit of photon energy falls in a straight line from the low end to zero at the full tube voltage.

Every tube voltage draws the same triangle, scaled. The unfiltered thick-target spectrum of a tungsten anode at 60, 80, 100 and 120 kV, in Kramers' form: the energy emitted per keV of photon energy falls in a straight line from the low-energy end to zero at the full tube voltage. The radiated energy is the triangle's area, which grows as the square of the voltage while the beam's energy grows only as the voltage — so the fraction radiated grows in proportion to the voltage. Tungsten's characteristic K lines at 59.3 and 67.2 keV appear only above its 69.5 kV K edge and carry a few per cent of the output at 100 kV; the window and the patient remove the low-energy end in practice.
Fig. 1 The unfiltered thick-target spectrum of a tungsten anode at 60, 80, 100 and 120 kV, in Kramers’ form: the energy emitted per keV of photon energy falls in a straight line to zero at the full tube voltage. The radiated energy is each triangle’s area, which grows as the square of the voltage while the beam’s energy grows only as the voltage. Tungsten’s K lines at 59.3 and 67.2 keV appear only above its 69.5 kV K edge.

The upper end of each triangle is the Duane–Hunt limit: no photon can carry more energy than one electron brought to the target, eVeV, which is how Planck’s constant was first measured to three figures from X-rays. The lower end in practice is cut away by the tube’s window and by whatever filter sits in front of the patient, because low-energy photons are absorbed in the first centimetres of tissue and add dose without adding to the picture — the steps in an absorption curve followed how steeply absorption rises as the photon energy falls. What matters here is the area. Kramers’ triangle has a height proportional to the voltage and a base equal to it, so its area — the energy radiated per electron — grows as the square of the voltage. The electron brought only the voltage. The fraction radiated therefore grows in proportion to the voltage, and doubling the voltage doubles the efficiency while quadrupling the output per electron.

The two spikes on the triangles above 69.5 kilovolts are a different process with its own bookkeeping. An electron in the beam knocks a tungsten atom’s innermost electron out altogether, and an electron from the next shell drops into the hole and emits the difference, at an energy that belongs to tungsten and not to the tube: the X-ray line that counts the protons followed how Moseley read the atomic number off exactly these lines. The ionisation that makes them is a collision with an atomic electron, the kind of event that does most of the stopping, and only a small share of the vacancies it leaves end in an X-ray rather than in an ejected Auger electron. At 100 kilovolts the lines carry a few per cent of a tungsten tube’s output. They do not change the conclusion; they are a second small channel beside the first.

That is the first sign that the efficiency is a ratio which scales with energy rather than a fixed property of the tube. To see why it scales, the two ways an electron loses energy in matter have to be compared directly.

Two ways to slow down

An electron crossing a solid loses energy almost continuously to the electrons of the atoms it passes, kicking each a little and ionising some. This collisional loss is what stops it, and it is large: a 100 keV electron in tungsten travels only about twenty micrometres before it is spent. Occasionally the electron passes close to a nucleus, is bent sharply, and — as a charge that turns must glow established for any accelerated charge — radiates. The ratio of the two rates of energy loss is, to the accuracy that matters for the argument, proportional to the atomic number and to the electron’s energy:

(dE/dx)rad(dE/dx)col≈(Z+1.24) E610 MeV.\frac{(dE/dx)_\text{rad}}{(dE/dx)_\text{col}} \approx \frac{(Z + 1.24)\,E}{610\ \text{MeV}}.

One ratio, read at two ends of the energy scale. The rate at which a fast electron loses energy by radiating, divided by the rate at which it loses it to atomic electrons, against its energy, for tungsten, copper and water, using the rule (Z + 1.24)E/610 MeV. The ratio grows in proportion to both the atomic number and the energy. In an X-ray tube's range it is about a hundredth: at 100 keV in tungsten 1.2%. It reaches one — the critical energy, above which an electron sheds energy mainly as photons and an electromagnetic shower develops — at 8.1 MeV in tungsten and about 71 MeV in water.
Fig. 2 The rate at which a fast electron loses energy by radiating, divided by the rate at which it loses it to atomic electrons, against its energy, for tungsten, copper and water. The ratio grows in proportion to both the atomic number and the energy. At 100 keV in tungsten it is 1.2%. It reaches one — the critical energy — at 8.1 MeV in tungsten and about 71 MeV in water.

The factor of ZZ has a clean origin. The power a charge radiates goes as the square of its acceleration, and the acceleration in a nucleus’s field is proportional to the nuclear charge, so the radiation from one close encounter goes as Z2Z^2. The collisional loss is shared among the atom’s ZZ electrons and goes as ZZ. The ratio is left with one power of ZZ, which is why tubes use tungsten, at 74, and not aluminium, at 13; it would be a still better choice if tungsten’s other virtue, a melting point of 3,422 °C, were not the deciding one.

There is a quieter reason the atom’s electrons radiate so little themselves, even though they are the ones doing almost all the stopping. When an electron collides with another electron, the pair’s centre of charge and centre of mass coincide — the two have the same charge-to-mass ratio — so the collision does not change their combined electric dipole moment at all, and the dipole radiation that dominates every other case is absent. What is left is far weaker. A nucleus is thousands of times heavier than the electron and barely recoils, so in an electron–nucleus encounter the dipole moment swings fully and radiates fully. The atom’s electrons take the energy and give back almost no light; the nucleus takes almost no energy and gives back the light. It is the same separation of roles that makes the cross-section that forgets the colour so small for a heavy particle: how strongly a charge radiates when it is shaken depends on how easily it is shaken, and a nucleus thousands of times heavier than an electron is the anvil, not the bell.

The energy dependence comes from the other side of the ratio. A slow electron spends longer near each atomic electron it passes and transfers more energy to it, so the collisional loss per unit path falls as the electron speeds up, roughly as the inverse of its energy until it nears the speed of light. The radiative loss per unit path does not fall: a faster electron is bent less by each nucleus, but it carries more energy to radiate. So the ratio rises, and at non-relativistic energies it rises in proportion to the energy.

Less than one per cent, and why the rule of thumb is linear

The target is thick, so each electron passes through every energy from the tube voltage down to zero, radiating at the local ratio as it goes. Averaging the ratio over the slowing-down gives the fraction of the beam that leaves as radiation, the thick-target yield:

Y(T)=1−ln⁡(1+aT)aT,a=Z+1.24610 MeV.Y(T) = 1 - \frac{\ln(1 + aT)}{aT}, \qquad a = \frac{Z + 1.24}{610\ \text{MeV}}.

When aTaT is small, as it is in every diagnostic tube, this is aT/2aT/2 — half the ratio at the starting energy, because the electron spends its slowing-down passing evenly through every lower energy, where the ratio is smaller.

A tube radiates less than one per cent of what it is fed. The fraction of an electron beam's energy that comes out as X-rays when the beam is stopped in a thick target, against tube voltage, for tungsten, molybdenum and copper anodes. It grows in proportion to the voltage and to the atomic number: 0.61% for tungsten at 100 kV and 0.10% for a molybdenum mammography tube at 28 kV. The dashed line is the radiographer's rule of thumb for tungsten, 0.9 × 10⁻⁹ times Z times the voltage in volts. The rest of the beam's energy, more than ninety-nine parts in a hundred, becomes heat in the anode.
Fig. 3 The fraction of an electron beam’s energy that comes out as X-rays when the beam is stopped in a thick target, against tube voltage, for tungsten, molybdenum and copper. It grows in proportion to the voltage and to the atomic number: 0.61% for tungsten at 100 kV and 0.10% for a molybdenum mammography tube at 28 kV. The dashed line is the radiographer’s rule of thumb, 0.9 × 10⁻⁹ times Z times the voltage in volts.

The radiographer’s rule of thumb — efficiency equals roughly 10−910^{-9} times the atomic number times the voltage in volts — is this formula with the constants collected. The rule and the formula differ by about ten per cent at 100 kilovolts, which is a fair statement of how well either is known: the 610 MeV constant is fitted at high energies, where showers are studied, and is being used here three decades below them. The point is not the second figure. It is that the efficiency is linear in two quantities a tube designer controls only weakly. The voltage is set by what has to be imaged: enough to get through the patient, not so much that soft tissues stop being told apart. The atomic number is set by what can survive the heat. A mammography tube, which needs low-energy photons to see the faint differences between breast tissues, runs a molybdenum target at 28 kilovolts and turns a tenth of a per cent of its beam into X-rays. The physics that makes it a good picture makes it a poor converter.

Calling the tube inefficient is in one sense unfair, because nobody wants most of the beam’s energy in the radiation. A radiograph needs a modest number of photons; too many simply increase the dose. The cost of the low yield is not wasted electricity, which is small, but the heat it leaves behind, which is concentrated in a spot a millimetre across and has to go somewhere in the fraction of a second the exposure lasts.

Where the other ninety-nine per cent goes

A CT scanner’s tube runs at 120 kilovolts and several hundred milliamperes: 60 kilowatts in the beam, of which about 400 watts emerges as X-rays and the rest arrives as heat on a small patch of tungsten. The patch is made long and narrow — a millimetre across and several long — and the target face is tilted so that, seen from the patient, the long patch looks square: the line-focus principle, which buys area for heat without blurring the image. Even so, 60 kilowatts on seven square millimetres is a flux of eight and a half gigawatts per square metre, about ten thousand times the flux at the surface of the Sun.

How fast that heats the metal depends on how long the beam rests on any one point. A sudden flux on the surface of a solid raises the surface temperature in proportion to the square root of the time, because the heat diffuses inward a distance that grows as the square root of the time and the heated layer thickens as it goes; this is the same arithmetic as the metal that feels colder than the wood, where a contact temperature is set by how fast each material lets heat into its depth. For tungsten, which conducts well and stores a lot of heat per volume, the rise is

ΔT=2qtπkρc.\Delta T = 2q\sqrt{\frac{t}{\pi k \rho c}}.

The anode spins because the heat cannot be allowed to stay. The rise in surface temperature of tungsten under a CT tube's 60 kW beam focused on a 1 mm by 7 mm track, against how long the beam rests on one spot, from the solution for a flux suddenly applied to a solid surface: the rise grows as the square root of the time. A stationary target would reach tungsten's melting point after 36 microseconds. A track of 50 mm radius turning at 9,000 rpm passes under the beam at 47 m/s, so each point is heated for 21 µs and rises about 2522 K before it moves on and cools for the rest of the turn — a margin of less than a factor of two in time, which is why the track runs white-hot.
Fig. 4 The rise in surface temperature of tungsten under a CT tube’s 60 kW beam focused on a 1 mm by 7 mm track, against how long the beam rests on one spot. A stationary target would reach tungsten’s melting point after 36 microseconds. A track of 50 mm radius turning at 9,000 rpm passes under the beam at 47 m/s, so each point is heated for 21 µs and rises about 2,500 K before it moves on.

A target standing still would melt in thirty-six microseconds. So the target does not stand still: it is a disc several centimetres across, spun at up to about ten thousand revolutions a minute on bearings inside the vacuum, and the beam strikes a track near its rim. Each point of the track is under the beam for only twenty-one microseconds before it is carried away to cool for the rest of the turn, spreading the heat over a ring three hundred times longer than the spot. The arithmetic leaves a margin of less than a factor of two in time, which is why the focal track glows white during an exposure, why tube makers rate a tube by the heat its anode can store and shed between scans, and why CT departments schedule around tube cooling.

The heat that leaves the anode must leave it by radiation, because the anode sits in a vacuum and its bearings cannot carry much. A disc at well over a thousand degrees radiates in proportion to the fourth power of its temperature, which is the only reason it can shed kilowatts at all, and the glass or metal envelope and the oil around it carry the heat out of the housing. A large CT tube’s anode stores several megajoules. The whole structure — rotating disc, graphite backing to add heat capacity, oil bath, heat exchanger — exists to dispose of the 99.4 per cent that the ratio says cannot be light.

A thousand times further up

Follow the yield to higher energies and it changes character. At 1 MeV in tungsten an electron radiates about a tenth of its energy; at 6 MeV, the energy of a radiotherapy linac, a quarter.

Raise the energy far enough and radiating becomes the main way to stop. The share of an electron's energy radiated as it is brought to rest in a thick target, against its starting energy, for tungsten, copper and water. At diagnostic energies it is under one per cent; a radiotherapy linac's 6 MeV electrons on a tungsten target radiate about 25%; at a few hundred MeV almost everything goes into photons, which make pairs, which radiate again — an electromagnetic shower. The curve is one function of Z times the energy, which is why water's lies to the right of tungsten's by the ratio of their atomic numbers.
Fig. 5 The share of an electron’s energy radiated as it is brought to rest in a thick target, against its starting energy, for tungsten, copper and water. At diagnostic energies it is under one per cent; a 6 MeV therapy linac’s electrons on a tungsten target radiate about 25%; at a few hundred MeV almost everything goes into photons. The curve is one function of Z times the energy, which is why water’s lies to the right of tungsten’s by the ratio of their atomic numbers.

Where the ratio reaches one, at the energy physicists call the critical energy, a fast electron loses as much by radiating as by ionising, and above it the radiation wins. The photons it emits are themselves above the threshold the partner that sets the price of a pair priced at 1.022 MeV, so they turn into electron–positron pairs in the next centimetre of material, and the pairs radiate again. The energy cascades through a growing population of particles, each generation with less energy than the last, until the individual particles fall below the critical energy and the cascade dies away into ionisation. This is an electromagnetic shower, and its depth, its width and the number of particles at its peak are all set by the critical energy and by the distance over which an electron radiates most of its energy. Calorimeters at particle colliders are built from lead or tungsten for the reason a tube’s target is: the large ZZ makes the radiation win early, so a shower of a hundred gigaelectronvolts is contained in twenty-odd centimetres of metal.

The surprising thing is that these are the same ratio. The radiographer’s 10−9 ZV10^{-9}\,Z V and the calorimeter builder’s critical energy, 7 MeV in lead, are one formula read at two energies a hundred thousand times apart. The tube sits at the bottom of the curve, where radiation is a hundredth of the story and heat is the problem; the calorimeter sits above the crossing, where radiation is the story and the heat it finally makes is the measurement. Nothing changes about the electron or the nucleus between the two. Only the electron’s energy has moved along the axis.

What a factor of two hides

The formula used throughout is a fit, and it carries its domain with it. The constant 610 MeV and the offset 1.24 come from matching the critical energies of many materials at high energy, where the radiative and collisional losses are computed from relativistic quantum theory and measured well. Below about a hundred kiloelectronvolts neither loss is exactly proportional to anything; the collisional loss depends on the target’s mean excitation energy, which differs between materials in ways the ZZ alone does not capture, and the radiative loss at low energy has its own corrections from screening by the atom’s electrons. That the fit lands within ten to twenty per cent of measured tube efficiencies is useful agreement, not a derivation, and the efficiencies here should be read as good to that.

The thick-target yield also assumes every electron is stopped in the target. In reality some fraction is scattered back out of the surface without being stopped — about half in tungsten, by number, though they leave with part of their energy — and both their energy and their radiation are lost from the count. The spectrum figure shows the unfiltered output; what reaches a patient is hardened by the window and filter, and its shape is no longer a triangle. And the focal-spot heating uses constant properties for tungsten and a uniform flux, where a real track has a temperature-dependent conductivity, a beam with a profile and a starting temperature that rises through a long scan. The surface temperature in the figure is the rise during one pass; the bulk of the anode adds its own, which is what the tube’s heat rating is about.

Still open: how far a target can be pushed

The limits on X-ray tubes are now almost entirely thermal, and the ways around them are inventive: anodes spun faster on magnetic or liquid-metal bearings, targets of liquid metal itself — a jet of gallium alloy that cannot be damaged because it is replaced continuously — and, for the highest brightness, the abandonment of the tube altogether for a synchrotron, where the flash a circling charge sends once a turn is the source: electrons radiate in magnetic fields instead of in nuclei, keep their energy between flashes, and nothing has to absorb the beam. A storage ring is the opposite bargain to a tube. The tube throws away more than ninety-nine per cent of every electron to get a little light from it once; the ring takes a little light from each electron on every turn and keeps the electron for hours. How small and how bright a laboratory source can be made, before the heat in the focal spot sets a limit no target material survives, is a live engineering question; the liquid-metal-jet sources are its current frontier, and their ceiling is set by how fast the jet can be made to flow while staying smooth.

An electron stopped in a target loses energy to the atom’s electrons, which barely radiate because a pair of electrons has no changing dipole, and to the nucleus, which radiates fully; the ratio of the two grows as the atomic number times the energy — 1.2% at 100 keV in tungsten — so a thick target radiates half that, 0.61% at 100 kV and 0.10% for molybdenum at 28 kV, and the rest is heat that would melt a stationary CT focal spot in 36 microseconds and so has to be carried away on a spinning disc. The same ratio reaches one at 8.1 MeV in tungsten, and above that an electron sheds its energy mainly as light and a shower begins: the radiographer’s inefficiency and the physicist’s calorimeter are one formula read at opposite ends of the energy scale.

Part 12 of 12

This essay is one argument about Radiating charge. 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.

BremsstrahlungCritical energyElectromagnetic showerHeat conductionLarmor formulaStopping power