Quantum

The glow that fades as one over the time

A glow-in-the-dark sticker charged under a lamp is bright for a minute, dim after an hour and still faintly visible at dawn. It does not fade like a radioactive source, by halves at a steady pace; it fades as one over the time, so that it loses as much brightness between one and ten minutes as between one and ten hours. The reason is that it has no single lifetime. Its light comes from electrons caught in traps of many depths, each emptying at its own exponential rate, and a spread of depths spread evenly adds up to a decay with no time scale at all. Heat the same crystal and it empties its traps in order of depth — which is how a piece of pottery can be dated by how much light it gives off when warmed.

Assumes: A nucleus with no clock · The exponential that is only true in the middle

The stars stuck to a child’s bedroom ceiling glow green when the light is switched off. For the first minute they are bright enough to read a clock by; after half an hour they are a faint smudge; but at four in the morning, for anyone awake to look, they are still there. Exit signs that need no electricity, watch hands, emergency markings on aircraft floors work the same way. The materials in them — the best is strontium aluminate with a little europium and dysprosium added — store light and give it back over hours.

The way they give it back is not the way most things in physics run down. A radioactive source decays exponentially, losing half its activity in each half-life, steadily, for ever. A nucleus with no clock explained why: a nucleus that has not yet decayed has no memory of how long it has waited, so the chance of decaying in the next second is always the same. A glow-in-the-dark sticker has no half-life. Its brightness falls as one over the time since it was charged, roughly, which means it falls by a factor of ten between one minute and ten, by another ten between ten minutes and a hundred, and by another between a hundred minutes and a thousand. That is what makes it useful: a decay with no time scale is visible at every time scale.

Electrons in traps

A phosphor of this kind is a crystal with a band gap, a range of energies in which electrons in a perfect crystal cannot exist, of the kind the hydrogen atom inside a crystal found broken by impurities into a few allowed levels just below the conduction band. Light from a lamp lifts electrons from the activator ions — europium — up into the conduction band, where they move freely. Most fall straight back and emit at once, which is ordinary fluorescence. Some are caught by defects in the crystal: places where an electron can sit at an energy below the conduction band, held there until it is given enough energy to climb out. These are traps, and the depth of a trap is the energy needed to escape it.

Nothing in the dark provides that energy except the crystal’s own heat. Atoms vibrate, and now and then the vibrations concentrate enough energy at a trap to lift its electron back into the conduction band, after which it can find its way to an ionised activator and recombine, emitting a photon. The rate at which this happens is a Boltzmann factor: an attempt frequency ss, of the order of a lattice vibration, about 101210^{12} per second, times e−E/kTe^{-E/kT}, the fraction of attempts that have enough energy. An electron sits in a trap of depth EE for an average time τ=eE/kT/s\tau = e^{E/kT}/s.

Every 58 millielectronvolts of depth is ten times longer. How long an electron stays in a trap before the crystal's heat shakes it out, against the trap's depth, at 20 °C and at 300 °C, for an attempt frequency of 10¹² per second. At room temperature each further 58 meV multiplies the time by ten: traps of about 0.7 eV hold for a second, 0.9 eV for an hour, 1.1 eV for months, 1.5 eV for longer than people have made pottery. Heating to 300 °C shortens every one of them enormously: traps down to about 1.5 eV then empty in seconds.
Fig. 1 How long an electron stays in a trap before the crystal’s heat shakes it out, against the trap’s depth, at 20 °C and at 300 °C, for an attempt frequency of 10¹² per second. At room temperature each further 58 meV multiplies the time by ten: traps of about 0.7 eV hold for a second, 0.9 eV for an hour, 1.1 eV for months, 1.5 eV for longer than people have made pottery. Heating to 300 °C shortens every one of them enormously.

The exponential makes the lifetime extraordinarily sensitive to the depth. At room temperature kTkT is 25 millielectronvolts, and every additional kTln⁡10=58kT\ln 10 = 58 meV of depth multiplies the time by ten. A trap of 0.7 electronvolts holds an electron for about a second; 0.9 eV, about an hour; 1.1 eV, months; 1.5 eV, millions of years. The whole range of times from the blink of an eye to geology is covered by less than one electronvolt of depth. The same exponential sensitivity made the heap that sets itself alight run away when it warmed by a few degrees; here it makes a modest spread of trap depths produce an enormous spread of lifetimes.

A sum of exponentials with no time scale

Each trap empties exponentially, as a radioactive nucleus does. A crystal with traps all of exactly one depth would glow with one lifetime and fade exponentially. Real crystals have traps of many depths, because the defects that make them are not identical, and the glow is the sum of all their exponentials.

A glow with no lifetime fades as one over the time. The light from a phosphor after its charging light is switched off, at 20 °C, on logarithmic scales, for three populations of traps. One depth gives an exponential with a single lifetime, flat on these axes and then a cliff. A narrow spread of depths softens the cliff. Depths spread evenly from 0.7 to 1.1 eV give lifetimes spread evenly in their logarithm, from a second to months, and their sum falls as one over the time — slope −1.00 at a quarter of an hour — with no lifetime anywhere in it. Glow-in-the-dark paint fades this way.
Fig. 2 The light from a phosphor after its charging light is switched off, at 20 °C, on logarithmic scales, for three populations of traps. One depth gives an exponential with a single lifetime, flat on these axes and then a cliff. A narrow spread softens the cliff. Depths spread evenly from 0.7 to 1.1 eV give lifetimes spread evenly in their logarithm, from a second to months, and their sum falls as one over the time — slope −1.00 at a quarter of an hour — with no lifetime anywhere in it.

The arithmetic of the third curve is short. If trap depths are spread evenly over a range, then because the lifetime is exponential in the depth, the lifetimes are spread evenly in their logarithm: as many traps empty between one and ten seconds as between ten and a hundred. At time tt, the traps contributing most light are those whose lifetime is about tt — shorter ones are already empty, longer ones have barely started — and the light from a band of lifetimes around tt is the number of traps in that band divided by their lifetime, which is proportional to 1/t1/t. The integral is exact: a uniform density nn of traps per unit depth gives

L(t)≈n kTtL(t) \approx \frac{n\,kT}{t}

between the lifetimes of the shallowest and deepest traps. Outside that window the decay is faster: before the shallowest traps’ lifetime it is flatter, after the deepest it falls off a cliff. Inside it, the glow has no memory of any time scale at all.

The empirical form of this decay was found long before its explanation. Edmond Becquerel — father of Henri, who found radioactivity — measured phosphorescent materials in the nineteenth century and described their decay by a hyperbolic law, intensity proportional to an inverse power of time near one. The radioactivity his son discovered decayed exponentially and seemed to be the opposite case. The two are the same physics: a Boltzmann factor, or a nucleus’s barrier, giving each emitter its own exponential. The difference is that all nuclei of one isotope have the same barrier, and the traps in a crystal do not.

Equal light in every decade of time

A decay as one over the time has a property that makes it the right shape for a sign that must be visible all night. The light emitted between two times is the integral of nkT/tnkT/t, which is nkTln⁡(t2/t1)nkT\ln(t_2/t_1): it depends only on their ratio. The light given out between one second and ten is the same as between one minute and ten minutes, and the same again between a day and ten days. With trap lifetimes spanning about seven decades, from a second to a few months, the stored light is shared out evenly among the seven: a seventh in the first ten seconds, a seventh between a quarter of an hour and three hours, a seventh between the end of the first day and the end of the eleventh.

An exponential spends its store the other way: most of it in the first lifetime, almost nothing after five. A phosphor with one trap depth tuned to give an hour-long lifetime would be bright for an hour and then dark, invisible at four in the morning however much light it had stored. The spread of trap depths, which looks like a defect of manufacture, is what lets the glow last; a crystal with perfectly uniform traps would make a worse exit sign.

Not the same as the long tail of a single decay

A power-law tail can come from more than one mechanism, and it is worth distinguishing this one from its neighbours. The exponential that is only true in the middle showed that even a single quantum decay departs from an exponential at very short and very long times, the long-time departure being a power law; but that tail sets in only after many lifetimes, when almost nothing is left, and it is far too faint to see. The glow’s power law is not a correction to an exponential; it is the whole visible decay, made of many exponentials of different lengths. Nor is it the behaviour of a sequence of decays, which the chain that runs at its slowest member’s rate found settling down to the pace of its slowest step: a chain is decays in series and has one bottleneck, where a phosphor’s traps empty in parallel and every one of them sets the pace for a while.

A closer relative is the diffusion that slows the longer it is watched, in which a walker that keeps getting stuck in traps spreads anomalously slowly, because its waiting times have no average. That happens when trap depths are distributed exponentially rather than uniformly: deep traps are rarer, but each is so much more sticky that the average wait diverges. A phosphor with an exponential distribution of trap depths glows not as 1/t1/t but as t−1−kT/E0t^{-1-kT/E_0}, with E0E_0 the scale of the distribution, an exponent just above one. The measured exponents of real phosphors, typically a little above one, may reflect this. The glow-in-the-dark sticker and the anomalous walker are reading the same kind of disorder from opposite ends: one counts the electrons escaping, the other the time spent stuck.

Heat the crystal and it reads out its traps

Holding a phosphor at room temperature lets its traps empty in order of depth, shallowest first, over a span of time. Heating it does the same in a span of temperature, and much faster. As the temperature rises steadily, each depth reaches the temperature at which its escape rate becomes comparable to the heating rate, empties in a burst of light, and is done. The light emitted against temperature is a glow curve.

Heating a crystal reads its traps out in order of depth. Thermoluminescence: the light from a crystal heated at 2 kelvin a second, for equal numbers of electrons in traps 1.0, 1.3 and 1.6 eV deep. Each depth empties over a narrow range of temperature, where its escape rate first becomes comparable to the heating rate, and gives a peak: at 120, 234, 347 °C. The deeper peaks are lower and wider, because the same number of electrons leaves over a wider span of temperature. A glow curve is a list of the trap depths and how full each one was.
Fig. 3 Thermoluminescence: the light from a crystal heated at 2 kelvin a second, for equal numbers of electrons in traps 1.0, 1.3 and 1.6 eV deep. Each depth empties over a narrow range of temperature and gives a peak, at 120, 234 and 347 °C. The deeper peaks are lower and wider, because the same number of electrons leaves over a wider span of temperature.

Each peak’s temperature is set by its depth: roughly, the temperature at which the trap’s lifetime equals the time the heating spends near that temperature. For an attempt frequency of 101210^{12} per second and a heating rate of two kelvin a second, traps of 1.0, 1.3 and 1.6 eV glow at about 120, 234 and 347 °C. The peak’s shape is computed from first-order kinetics — each electron escaping independently and not being caught again — in the form given by Randall and Wilkins in 1945. A glow curve is therefore a spectrum of the crystal’s traps: where the peaks are says how deep the traps are, and how much light each gives says how full they were.

This is thermoluminescence, and it is a measuring instrument. Crystals that have been exposed to ionising radiation have their traps filled in proportion to the dose, since radiation does what the lamp did — lifts electrons into the conduction band to be caught. Lithium fluoride chips worn as badges by people who work with radiation are read this way: heated, they give out light in proportion to the dose they received since they were last read.

Dating by what accumulated

The most striking use turns the measurement round. Pottery is made of clay containing quartz and feldspar grains, crystals with traps. Firing a pot heats it to hundreds of degrees and empties every trap. After that, the grains sit in the pot, and the pot sits in the ground, receiving a slow, steady dose of radiation from uranium, thorium and potassium in the clay and the soil and from cosmic rays. Their traps fill. Heated in a laboratory centuries later, they give out light in proportion to the dose received since firing, and the dose divided by the dose rate is the time since the pot was fired.

Two thousand years empty the shallow traps and spare the deep. Glow curves from equal doses in traps 1.0, 1.3 and 1.6 eV deep, measured just after the dose (dashed) and after the dose was received 2,000 years ago at 15 °C (solid). In twenty centuries the 1.0 eV traps have emptied entirely, the 1.3 eV traps have kept 32% and the 1.6 eV traps all of it. Dating by thermoluminescence uses only the part of the curve where the old sample and the fresh one agree in shape: there the traps are deep enough to have kept every electron the years put into them.
Fig. 4 Glow curves from equal doses in traps 1.0, 1.3 and 1.6 eV deep, measured just after the dose (dashed) and after the dose was received 2,000 years ago at 15 °C (solid). In twenty centuries the 1.0 eV traps have emptied entirely, the 1.3 eV traps have kept 32% and the 1.6 eV traps all of it.

The method works only with traps deep enough to keep what they collect. Over two thousand years at fifteen degrees, the shallow traps have emptied many times over and remember nothing of the pot’s age; traps 1.3 eV deep have kept a third of their electrons and give an age too young by that factor; only traps of 1.6 eV and deeper have kept everything. So a dating laboratory compares the glow curve of the ancient sample with that of a portion of the same sample given a known dose in the laboratory, and uses only the temperature range — the plateau — where the two have the same shape. There the traps are deep enough that their memory has not leaked. The comparison is the same logic as the clock read by counting atoms, which reads age from a radioactive parent and its daughter; here the clock is not a decay but an accumulation, and the trap depth decides how long it can run before it starts to forget.

A warm glow is brighter and shorter

The same physics says how a phosphor’s afterglow depends on its temperature.

A warm glow is brighter and shorter. The afterglow of a phosphor with trap depths spread evenly from 0.7 to 1.1 eV, held at 0, 20 and 40 °C. In the middle of its range the glow is close to kT/t for each temperature, nearly the same, but warmth shifts the whole range of lifetimes earlier: at 40 °C the shallow traps have already emptied before the first second and the deepest are exhausted within days, so the glow both starts brighter and ends sooner. The light stored is fixed; temperature only decides how quickly it is spent.
Fig. 5 The afterglow of a phosphor with trap depths spread evenly from 0.7 to 1.1 eV, held at 0, 20 and 40 °C. In the middle of its range the glow is close to kT/tkT/t at each temperature, nearly the same, but warmth shifts the whole range of lifetimes earlier: at 40 °C the shallow traps have already emptied before the first second and the deepest are exhausted within days, so the glow both starts brighter and ends sooner.

Warmth shortens every lifetime by the same exponential factor, so it slides the window of lifetimes earlier without changing its shape. Within the window the glow is nkT/tnkT/t, barely dependent on temperature; at its edges, temperature decides everything. A warm room spends a phosphor’s stored light faster: brighter in the first minutes, exhausted sooner. A cold one rations it. A sticker that is still visible at dawn in a cool bedroom may be gone by midnight on a warm night. The total light is the same in both rooms, because every trapped electron eventually escapes and recombines; temperature decides only when. It is the same accounting as the ball that forgets how rough the floor was, whose skid makes the same heat on any surface: the store is fixed by how the phosphor was charged, and the surroundings only set the schedule on which it is paid out. And a phosphor cooled far enough holds its electrons for so long that it stores light indefinitely, to be released only when it is warmed — the principle of the infrared detection cards and the photostimulable plates once used in place of X-ray film, which hold an image as filled traps until a laser or heat reads them out.

What the picture leaves out

The figures use one attempt frequency for every trap, first-order kinetics in which no released electron is ever recaptured, and idealised distributions of depth. Real phosphors have attempt frequencies that differ between trap types, and their released electrons are often recaptured before recombining, which changes the shapes of glow peaks and the exponent of the afterglow. Real trap distributions are measured, not assumed, and the best long-afterglow phosphors were developed empirically before their traps were understood. Some emission comes not from thermal escape at all but from tunnelling between a trapped electron and a nearby recombination centre, which is independent of temperature and gives its own slow decay; in some dating minerals it causes a fading that the simple model does not predict and that laboratories must correct for.

The quantum part of the story — what a trap is, why an electron caught in one is bound by a particular energy, how its escape couples to lattice vibrations — is compressed here into a single depth and a Boltzmann factor. It is not the same as the slow glow of molecules whose excited electrons are stuck in a triplet state, the phosphorescence four states and one of them is odd explained through spin, which fades exponentially over milliseconds to seconds; the hours-long afterglow of a glow-in-the-dark sticker is trap physics, not spin physics, although both are called phosphorescence.

Still open: what the deepest traps are

The persistent phosphors used in signs and watches were found by trial in the 1990s, and the identity of their traps — which defects, created by which dopants, at which depths — is still argued over, with competing models placing the electrons on dysprosium ions, on oxygen vacancies or on both. In dating, the question is how long the deepest traps really keep their electrons. Some feldspars lose part of their stored signal over geological times by a slow, temperature-independent fading attributed to tunnelling, and correcting for it limits the ages that can be measured; whether every apparently stable trap is stable enough for dates of a hundred thousand years or more is decided mineral by mineral.

An electron in a crystal trap of depth EE escapes at rate s e−E/kTs\,e^{-E/kT}, so at room temperature each 58 meV of depth multiplies its stay by ten, and a phosphor whose trap depths are spread evenly from 0.7 to 1.1 eV holds lifetimes spread evenly in their logarithm from a second to months, whose sum glows as nkT/tnkT/t — one over the time, with no half-life anywhere in it; heated at 2 K/s, the same traps empty in order of depth, as peaks at 120, 234 and 347 °C for 1.0, 1.3 and 1.6 eV, which is how a dose of radiation, and the age of a fired pot, are read from the deep traps that have not forgotten it. A decay with no time scale is a crowd of decays with every time scale at once.

Part 9 of 9

This essay is one argument about Decay. 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.

Activation energyThe Boltzmann factorExponential decayLuminescence datingPower lawThermoluminescence