Everything a scatterer removes, from one direction

A particle in a beam removes light from it in two ways: by sending some of it off in other directions, and by absorbing some of it as heat. The two have nothing obvious in common — one is elastic and one is not — and the natural way to find their sum is to compute each separately and add.

There is a much shorter route, and it is exact. The total removed is fixed by what the particle does in one single direction: straight ahead, where the scattered wave travels alongside the incident one and no measurement can separate them.

$$\sigma_{\text{ext}} = \frac{4\pi}{k}\,\mathrm{Im}\,f(0).$$

The scattering in every other direction does not appear. Nor does the absorption, explicitly. Both are already in the forward amplitude.

Why forward, and why the imaginary part #

The whole mechanism is in that picture and it is a statement about adding vectors. A long vector with a small one added along it changes length by the small one’s whole length. The same small vector added at right angles changes the length by an amount second order in its size — the hypotenuse of a long side and a short one is barely longer than the long side.

Energy is the square of the length. So a scatterer whose forward-scattered wave is in phase with the incident one changes the transmitted power only at second order — it is barely removing anything — and one whose forward wave is a quarter turn behind removes energy at first order. The size of the out-of-phase part is the whole extinction.

Two consequences follow that are worth separating.

The first is that the theorem is a conservation law in disguise. Whatever leaves the beam has to have left it somewhere, and the only place a first-order deficit can appear is behind the particle, in the interference between the two waves. So the forward amplitude is forced to know about the total, and a scatterer that sent light off in all directions without a corresponding forward phase shift would be creating energy.

The second is that the theorem is about the beam rather than about the particle. It says nothing about where the removed light went — into other directions, into heat — and it cannot be used to separate them. That separation needs the scattering in every direction, which is the calculation the theorem was avoiding.

The particle that removes twice its shadow #

That a large obstacle removes twice the light its shadow blocks is the extinction paradox, and when the particle is the size of the wave establishes the number from the Mie series — the efficiency of a large sphere reaching two from above. What the theorem adds is the mechanism, and with it the paradox stops being one. A shadow is not a hole in the beam. It is a region where the incident wave has been cancelled by a scattered wave of equal amplitude and opposite sign — which is a scattered wave carrying as much energy as the shadow removes. Half the extinction is the shadow; the other half is that cancelling wave, which spreads out at large distance into a narrow forward diffraction lobe.

The consequence is directly observable. Look at a distant lamp through fog and the light removed from the direct beam is twice what a ray argument gives — and half of it is still travelling almost exactly forward, in a lobe of angular width about the wavelength divided by the droplet size. A detector with a wide acceptance angle collects that lobe and measures half the extinction the theorem predicts; a detector with a narrow one measures all of it. The measured cross-section of an aerosol therefore depends on the instrument’s field of view, which is a standard trap in atmospheric measurement and is a direct consequence of the picture.

The overshoot between the two limits is worth its own sentence. A droplet a few wavelengths across is not opaque: light goes through it as well as round it, and the two emerge with a relative phase set by how much the middle of the drop retarded its share. When they emerge in step the forward amplitude is large and the extinction is large; a little further and they are out of step. That is why the curve rings, and it is the same interference between two paths through one object that gives coloured rings round the Moon and makes a fog of one droplet size look quite different from a fog of another.

The other half of the same amplitude #

The forward amplitude is a complex number and only its imaginary part has been used. The real part is the refractive index.

The argument is the same phasor picture read the other way. Adding a small wave a quarter turn ahead of the incident one, rather than behind it, leaves the length almost unchanged and rotates the sum — which is a phase shift. A phase shift accumulated over a distance is exactly what a refractive index is: the medium retards the wave, and the retardation per unit length is the number of scatterers per unit volume times the real part of the forward amplitude, divided by the wavenumber squared.

So an index is not a property a medium has in addition to scattering. A transparent medium is a medium whose forward amplitude is nearly real, and it slows light because each molecule sends a tiny wave forward that adds to the incident one with a phase lag. The wave that arrives is not the original wave still travelling at $c$; it is the sum of the original and a great many small scattered contributions, and the sum has crests where the original would not have had them.

That reading resolves an old confusion about what “light travels slower in glass” means. Nothing travels slower than $c$: every scattered wavelet travels at $c$, and so does the original. What is slower is the position of the crest of the sum, which is a property of the addition rather than of any wave in it. The same statement explains why the index can be less than one — as it is for X-rays, where the forward amplitude has the other sign — with nothing exceeding $c$ anywhere.

The two halves are also connected by more than sharing an amplitude. The real and imaginary parts of a response that respects causality are not independent: given one at all frequencies, the other follows. That relation between what a medium absorbs and how it delays is what makes it impossible to have a transparent medium with an interesting index and no absorption anywhere.

The shadow that is made of light #

The extinction paradox deserves being turned over once more, because the usual objection to it is a good one and has a good answer.

The objection: a large opaque disc casts a shadow, the shadow is dark, and no light is missing beyond the geometrical shadow — so how can the disc remove twice what it blocks? The answer is that the shadow at any finite distance is not dark at its edges and does not stay the same size. What has actually happened is that the disc launched a wave equal and opposite to the incident one over its own area, and that wave spreads: near the disc it cancels the beam exactly and the shadow is sharp, and far away it has spread into a lobe of angular width the wavelength over the disc’s radius.

So there is a distance beyond which the shadow has filled in and the light that made it is travelling almost forward as a diffraction lobe. Before that distance a detector sees a sharp shadow and measures the disc’s geometrical area; after it, a small detector on the axis sees light again. Whether a measurement records the extinction as one shadow’s worth or two is decided by where the detector is and how large it is — the near field and the far field give different answers, and both are correct answers to different questions.

The distance where the changeover happens is the Fresnel distance, the disc’s radius squared over the wavelength, and it is a metre or so for a millimetre obstacle at optical wavelengths. That is why the paradox is not visible with everyday objects and is unavoidable in atmospheric optics, where the obstacles are micrometres across and every measurement is in the far field.

Where the theorem earns its keep #

The reason the forward direction is privileged at all is in that figure. Scatterers at random positions send waves that arrive at a detector with random relative phases, so their intensities add and the total goes as their number — which is why the sky is bright at all rather than dark. Straight ahead, the extra path from source to scatterer to detector is the same for every scatterer, so their amplitudes add and the total goes as the square of their number.

That is why a medium has a coherent forward beam and an incoherent halo, why the forward beam has a well-defined phase and the halo does not, and why the forward direction is where a statement about a whole medium can be made from a statement about one particle. Everywhere else, the answer depends on where the particles are.

The theorem’s practical value is that it turns a hard measurement into an easy one. Measuring a total cross-section by collecting everything scattered in every direction is a nightmare; measuring the attenuation of a beam is straightforward. And in particle physics the same theorem, in the same form, relates the total cross-section of a collision to the imaginary part of the forward elastic amplitude — a relation used constantly, derived by exactly the argument above, and one of the very few statements that holds without any model of what is happening.

What a measurement of attenuation is measuring #

Reading a transmission measurement through the theorem changes what it is a measurement of, and the change is useful.

The direct beam falls exponentially with path length, and the exponent is the number density times the extinction cross-section. That is the ordinary law of attenuation and it is usually presented as a definition of the cross-section. The theorem makes it something else: a measurement of the imaginary part of one molecule’s forward scattering amplitude, made on a beam a hundred kilometres long.

Two things follow. The measurement is enormously more sensitive than any laboratory scattering experiment, because the path length does the amplification — an atmospheric transmission measurement determines a molecular cross-section of order $10^{-31}$ square metres, which nothing bench-sized could reach. And it is a measurement of a sum: the extinction includes absorption by every species present, so a transmission spectrum is a superposition, and separating it into constituents is an inverse problem rather than a reading.

The same logic runs through every attenuation measurement there is — the opacity of a stellar interior, the absorption of a solution, the loss of a fibre. In each case what is measured is a forward amplitude and what is wanted is usually something else, and the theorem is what connects them.

Who found it, and in what subject #

The theorem has been discovered at least three times in three subjects, and the history is a fair illustration of how far a piece of arithmetic can travel.

It appears first in the 1870s in the theory of light scattering, where Rayleigh’s work on the blue sky needed exactly this relation between what a molecule scatters and what a column of air transmits. It appears again in the 1940s in quantum mechanics, where the same statement relates a total cross-section to the imaginary part of a forward scattering amplitude, derived from the requirement that probability be conserved. And it appears in the 1950s in high-energy physics, where it constrains the growth of total cross-sections with energy and is one of the few statements in the subject that does not depend on any model of what particles are made of.

The three derivations look nothing alike — one is about interference behind an obstacle, one about the unitarity of an operator, one about the analytic structure of an amplitude — and they establish the same equation. What they share is the underlying fact: something is conserved, and a wave that removes some of it has to be interfering with what it removed it from.

That recurrence is the reason to learn the argument in the simplest setting. The phasor picture is the whole of it, and everything else is that picture in a language where drawing it is harder.

Where the model stops #

The theorem assumes a single scatterer and a plane wave. In a dense medium the wave arriving at each scatterer has already been modified by the others, and the extinction is no longer the single-particle value times the number density. That failure begins when the particles are separated by less than a wavelength, and it is why a dense fog is not simply a thin fog with more drops in it.

The extinction curve is an approximation valid for particles that bend light only slightly. Anomalous diffraction assumes rays pass through the particle undeviated and only pick up a phase, which is good for water droplets in air and poor for metal particles or anything with a large index. The exact Mie solution shows finer structure — sharp resonances where a wave circulates round the particle — that the smooth curve here does not contain.

The refractive index calculation is the dilute limit. It ignores the field a molecule sees from its neighbours, which for a gas is a correction of parts per million and for a liquid is the difference between the Lorentz–Lorenz relation and the naive sum. The essay’s claim is about the mechanism, and the mechanism survives; the coefficient does not.

And the theorem is about a beam, not about a photon. It relates powers, and every step in the argument is about adding amplitudes and squaring. Nothing in it says what happens to any individual quantum, and the question of which photon was removed has no answer within the argument.

What the pictures cannot show #

The phasor figure draws two waves adding at one point behind the scatterer, and the real situation is an integral over a plane. The forward-scattered wave is spread over an angular lobe rather than confined to a direction, and the interference that removes energy happens across the whole of the beam’s cross-section. What the single phasor gets right is the sign and the order of the effect; what it cannot show is that “forward” means an angular region whose width is set by the size of the beam and the size of the scatterer.

The extinction curve draws an efficiency and hides the angular distribution entirely. Two particles with the same extinction can send their scattered light in completely different patterns — one mostly forward, one nearly isotropic — and everything about how a medium looks depends on which. That distribution is the part the theorem deliberately does not compute, and it is the part a haze’s appearance is made of.

Where the ladder goes next #

The scattering ladder began with why the sky is blue and the sunset is not, went through what happens when the particle is the size of the wave and why a litre of water is not blue for the same reason. This rung asks how much is removed altogether, and finds it in one direction. The rungs after it: the phase function, which is the angular distribution the theorem refuses to give and which decides how a cloud looks; multiple scattering, where the single-particle account stops and a transport equation begins; and inelastic scattering, where the removed light returns at a different frequency and the accounting needs a second beam.

The habit worth carrying away is to look for the direction in which nothing can be separated. Where a scattered wave and an incident wave are indistinguishable, they interfere, and the interference is a first-order effect where scattering itself is second order. That is a general shape of argument, it is why forward scattering carries so much more information than any other angle, and it is worth checking for whenever a small perturbation is supposed to have a large consequence.