Absorption Coefficient and Beer–Lambert Law

Shine a beam into a block of tinted glass, a cuvette of dye, or a hundred kilometres of optical fibre, and the power that survives always follows the same law: an exponential decay with distance, $I(z)=I_0\,e^{-\alpha z}$ . This is the Beer–Lambert law, and the single number $\alpha$ — the absorption coefficient — measures how strongly the medium drinks the light. It is not an empirical curve fitted to data but a direct consequence of the wave physics of the previous articles: once the refractive index is complex, the wave carries its own decay in its exponent, and the intensity, being the square of the field, falls twice as fast. The whole article rests on one link, recalled by name from aq-optics-complex-refractive-index-extinction: the imaginary part of the index, the extinction coefficient $\kappa$ , sets the absorption coefficient through $\alpha=4\pi\kappa/\lambda_0=2\omega\kappa/c$ . A plane wave with wavevector $k=\tilde n\,\omega/c$ splits into an oscillation times a real decaying envelope $e^{-\kappa\omega z/c}$ ; squaring gives $e^{-\alpha z}$ . Everything else is bookkeeping laid on top of that one fact — and the bookkeeping is where students lose marks, because the same attenuation is quoted three incompatible ways. Physicists write $\alpha$ as a natural-log rate per metre; chemists write an absorbance $A=-\log_{10}(I/I_0)=\varepsilon_M C\ell$ , base ten and proportional to concentration; engineers write decibels per kilometre. The three differ only by fixed factors — $\ln 10=2.3026$ between napierian and decadic, and $1\,\mathrm{Np}=8.686\,\mathrm{dB}$ — yet mixing them is the classic error. Get the factors right and the same coefficient describes a laboratory dye, the optical depth of a star's atmosphere, the skin depth of a metal, and the 0.2 dB/km silica fibre whose transparency — a $1/e$ length of 21.71 km — made the global internet physically possible.

Key takeaways

A beam in an absorbing medium loses intensity exponentially, $I(z)=I_0\,e^{-\alpha z}$ : the Beer–Lambert law, with $\alpha$ the intensity absorption coefficient (units $\mathrm{m^{-1}}$ ).
The exponential is a theorem, not a fit: each thin layer removes a fixed fraction of the passing power, $\mathrm{d}I=-\alpha I\,\mathrm{d}z$ , and that integrates to the exponential.
The coefficient comes from the imaginary index (recalled by name): $\alpha=4\pi\kappa/\lambda_0=2\omega\kappa/c$ , twice the field-amplitude decay rate because intensity is amplitude squared.
Three scales, one physics: napierian $\alpha$ (per metre), decadic absorbance $A=\alpha z/\ln 10=\varepsilon_M C\ell$ (chemistry), and decibels ( $1\,\mathrm{Np}=8.686\,\mathrm{dB}$ , $3.01\,\mathrm{dB}$ = half the power).
For independent absorbers the coefficient is additive and linear in number density, $\alpha=N\sigma$ , defining a per-particle cross-section $\sigma$ that can far exceed an atom's geometric size near resonance.
Anchor: telecom silica at 0.2 dB/km gives $\alpha=4.605\times10^{-5}\ \mathrm{m^{-1}}$ , a $1/e$ attenuation length of 21.71 km, and loses half its power over 15.05 km.
Key idea: the imaginary part of the wavevector turns oscillation into decay — $E\propto e^{-\kappa\omega z/c}$ , so $I\propto e^{-\alpha z}$ with $\alpha=2\omega\kappa/c$ — and every measure of absorption (absorbance, molar absorptivity, decibels, cross-section) is a change of units on that one exponent.

The core insight

The imaginary part of the wavevector turns oscillation into decay: a plane wave with complex index $\tilde n=n+i\kappa$ propagates as $E\propto e^{-\kappa\omega z/c}e^{i(n\omega z/c-\omega t)}$ , so the intensity falls as $I=I_0e^{-\alpha z}$ with $\alpha=2\omega\kappa/c=4\pi\kappa/\lambda_0$ . Cross-sections, molar absorptivity, absorbance and decibels are all bookkeeping on this single decay rate.

Where this lives today

Beer's law is the quiet engine of quantitative measurement. Every UV–visible and infrared spectrophotometer reads a concentration off an absorbance through $A=\varepsilon_M C\ell$ — a clinical blood assay, a groundwater pollutant, the optical density of a bacterial culture — trusting the optically thin regime $A\lesssim1$ where the exponential is nearly linear. Astronomers correct starlight for atmospheric extinction with the same law, the airmass playing the role of path length and $\tau=\int\alpha\,\mathrm{d}z$ the optical depth; remote-sensing lidars locate methane leaks by differential absorption on and off a molecular line. And the whole architecture of optical telecommunications is a fight to make $\alpha$ small: Kao's 1966 insight that the loss of glass was removable impurity absorption rather than fundamental scattering, cashed out in Corning's low-loss fibre, drove attenuation from thousands of decibels per kilometre to the 0.2 dB/km of the punchline figure — the transparency that lets a pulse cross an ocean with only a handful of amplifiers.

The physical picture

Picture the beam entering a stack of identical thin slabs. Each slab removes the same fraction of whatever power reaches it — not the same absolute amount, because a slab deep inside sees a weaker beam and takes a proportionally smaller bite. Constant fractional loss per unit length is the exact signature of exponential decay: write down that each layer of thickness $\mathrm{d}z$ subtracts $\alpha I\,\mathrm{d}z$ , and integrating $\mathrm{d}I/I=-\alpha\,\mathrm{d}z$ delivers $I=I_0e^{-\alpha z}$ with no further assumptions. The depth at which the intensity has fallen to $1/e=0.368$ of its entry value is the penetration depth $\delta=1/\alpha$ ; after ten such depths only $e^{-10}\approx4.5\times10^{-5}$ of the light remains. The medium responds to the field, not to the intensity, and that is the origin of the most-forgotten factor in the subject. The complex wavevector makes the field envelope shrink at the rate $\kappa\omega/c$ ; a detector, however, reads $|E|^2$ , so the measured power decays at twice that rate, giving $\alpha=2\kappa\omega/c$ . It is exactly the amplitude-versus-power distinction that gives decibels their factor of 10 for power against 20 for amplitude.

Figure

A beam entering an absorbing slab: the intensity falls as I(z) = I₀ e^{−αz}. After one penetration depth δ = 1/α the intensity is 1/e = 0.368 of its entry value; after 2δ it is 0.135, after 3δ only 0.050 — equal steps of depth cost equal fractions of power

Finally, think of the cross-section form $\alpha=N\sigma$ as each absorber casting a shadow of area $\sigma$ across the beam. In a slab of thickness $\mathrm{d}z$ the shadows cover a fraction $N\sigma\,\mathrm{d}z$ of the wavefront, so that fraction of the power is removed — additive, linear in the number density $N$ , and the microscopic reason $\alpha$ is proportional to concentration. The cross-section can vastly exceed the physical size of an atom near a resonance, because it is set by the driven-dipole response, not by geometry. The second figure is the number that changed the world: the transmission of a modern telecom fibre, plotted the way engineers actually read it.

Figure

The punchline — telecom silica fibre at 0.2 dB/km. On a decibel (log) axis the power falls in a straight line: α = 4.605×10⁻⁵ m⁻¹, a 1/e attenuation length of 21.71 km (−4.343 dB = 1 Np), and half the power gone at 15.05 km (−3.01 dB). This transparency is what makes intercontinental fibre possible

The core equations

The law itself is the exponential, and it is worth stating first as the observable — what a power meter reads at the far end of a sample of thickness $z$ . Its exponential form is forced by linearity, as the first derivation shows.

I(z)=I_0\,e^{-\alpha z} \qquad (1)

where: $I(z)$ — time-averaged intensity (irradiance, $\mathrm{W/m^2}$ ) at depth $z$ ; $I_0$ — incident intensity just inside the medium ( $z=0$ ); $\alpha$ — intensity absorption coefficient ( $\mathrm{m^{-1}}$ ); $z$ — path length. The penetration (skin) depth is $\delta=1/\alpha$ , the depth at which $I$ falls to $1/e$ of $I_0$ . The coefficient is not free: it is fixed by the imaginary part of the refractive index, recalled by name from aq-optics-complex-refractive-index-extinction, which introduced $\tilde n=n+i\kappa$ . The factor of two is the intensity-versus-amplitude bookkeeping.

\alpha=\frac{2\omega\kappa}{c}=\frac{4\pi\kappa}{\lambda_0} \qquad (2)

where: $\omega$ — angular frequency; $\kappa$ — extinction coefficient, $\mathrm{Im}\,\tilde n$ ; $c$ — vacuum speed of light; $\lambda_0=2\pi c/\omega$ — vacuum wavelength. The field decays at $\kappa\omega/c$ ; the intensity, being $|E|^2$ , decays at twice that. A telecom fibre's $\kappa$ is astonishingly small — about $5.7\times10^{-12}$ at 1550 nm. Chemists and spectroscopists measure the same attenuation as a base-ten absorbance and tie it to the molar concentration of the absorbing species — the working form of Beer's law in a teaching lab.

A=-\log_{10}\!\frac{I}{I_0}=\frac{\alpha z}{\ln 10}=\varepsilon_M\,C\,\ell \qquad (3)

where: $A$ — absorbance (dimensionless, base ten); $\varepsilon_M$ — molar absorptivity ( $\mathrm{L\,mol^{-1}\,cm^{-1}}$ ); $C$ — molar concentration; $\ell$ — path length. The napierian coefficient and the decadic one differ by $\ln 10$ : $\alpha=\ln 10\,(\varepsilon_M C)$ in consistent units, so an absorbance of 1 means 10% transmitted, of 2 means 1%. Engineers quote attenuation in decibels, a base-ten power ratio, most often per kilometre for a fibre. The neper and the decibel are the natural-log and base-ten members of the same family.

\text{loss (dB)}=10\log_{10}\!\frac{I_0}{I},\qquad 1\ \mathrm{Np}=\frac{20}{\ln 10}\ \mathrm{dB}=8.686\ \mathrm{dB} \qquad (4)

where: a $3.01\,\mathrm{dB}$ loss is exactly a factor of two in power ( $10\log_{10}2=3.01$ ); a $10\,\mathrm{dB}$ loss is a factor of ten. One neper is one $1/e$ in amplitude, hence a factor $e^2$ in power, hence $8.686\,\mathrm{dB}$ . When the absorbers act independently — a dilute gas, a dye in solution, dust along a sightline — the coefficient is additive and linear in their number density, which defines a per-particle cross-section.

\alpha=N\,\sigma(\omega) \qquad (5)

where: $N$ — number density of absorbers ( $\mathrm{m^{-3}}$ ); $\sigma(\omega)$ — absorption cross-section per particle ( $\mathrm{m^2}$ ), the effective shadow each absorber casts. Near a strong resonance $\sigma$ can far exceed the geometric size of the particle.

Derivation ABeer–Lambert from a complex-wavevector plane wave

The goal is to earn the exponential intensity law from first principles rather than assume it, and in doing so to pin down two things students routinely get wrong: that the exponential is a consequence of linearity (not a fitted shape), and that the intensity decay rate $\alpha$ is twice the field-amplitude decay rate because intensity is amplitude squared. We take a monochromatic plane wave in a medium of complex index $\tilde n=n+i\kappa$ , recall the extinction coefficient by name from its owning article, propagate it, and square it.

Step 1. Start from a forward plane wave with the complex wavevector of an absorbing medium, $E(z,t)=E_0\,e^{i(kz-\omega t)}$ with $k=\tilde n\,\omega/c$ , the dispersion relation of the plane-wave article carried over with a complex index. Why: nothing about the structure of the wave changes when the index goes complex — the same solution now simply carries a complex $k$ , and that is the entire origin of absorption.

Step 2. Split the wavevector into real and imaginary parts, $kz=\dfrac{n\omega}{c}z+i\dfrac{\kappa\omega}{c}z$ , so the field factorises into an oscillation and a real, decaying envelope, $E=E_0\,e^{-\kappa\omega z/c}\,e^{i(n\omega z/c-\omega t)}$ . Why: the real part of $k$ advances the phase (phase velocity $c/n$ ); the imaginary part multiplies $i\cdot i=-1$ into a real negative exponent — refraction and absorption are the two halves of one complex number.

Step 3. Form the time-averaged intensity, which goes as the square of the field amplitude: $I(z)=\tfrac12 c\varepsilon_0 n\,|E|^2\propto e^{-2\kappa\omega z/c}\equiv e^{-\alpha z}$ , so $\alpha=\dfrac{2\kappa\omega}{c}=\dfrac{4\pi\kappa}{\lambda_0}$ using $\omega=2\pi c/\lambda_0$ . Why: this is the factor-of-two trap made concrete — the amplitude $1/e$ length is $2/\alpha$ , but the intensity $1/e$ length, the penetration depth actually measured, is $\delta=1/\alpha$ .

Step 4. Differentiate the result to recover the microscopic loss law: $\mathrm{d}I/\mathrm{d}z=-\alpha I$ , i.e. $\mathrm{d}I=-\alpha I\,\mathrm{d}z$ . Why: this closes the loop with the physical-picture argument — constant fractional loss per unit length is both the cause (each layer takes a fixed fraction) and the consequence (the exponential differentiates back to it), so the exponential is forced by linearity alone. *Verified: $\alpha=2\omega\kappa/c=4\pi\kappa/\lambda_0$ (symbolic identity via $\omega=2\pi c/\lambda_0$ ); $\mathrm{d}I/\mathrm{d}z=-\alpha I$ by differentiation; $\kappa=5.68\times10^{-12}$ at 1550 nm reproduces $\alpha=4.605\times10^{-5}\ \mathrm{m^{-1}}$ in verify_aq-optics-absorption-coefficient-beer-lambert.py.*

Go deeper · Derivation B · 1 minThree scales for one attenuation: napierian, decadic, and decibel

Derivation BThree scales for one attenuation: napierian, decadic, and decibel

The goal is to make the unit conversions between the physicist's $\alpha$ , the chemist's absorbance and molar absorptivity, and the engineer's decibel completely unambiguous, because these three communities describe the same exponential decay with numbers that differ by fixed but easily forgotten factors — $\ln 10$ and the $8.686$ of the neper-to-decibel step. We start from the one physical exponent and translate it into each language in turn.

Step 1. Take the napierian (natural-log) law $I=I_0e^{-\alpha z}$ as the reference. Its exponent $\alpha z$ is the optical depth $\tau$ , dimensionless, and $\alpha$ is measured in $\mathrm{m^{-1}}$ . Why: the natural log is the form that falls straight out of the wave physics, so it is the honest baseline against which the human-convenient scales are defined.

Step 2. Convert to base ten. Since $e^{-\alpha z}=10^{-\alpha z/\ln 10}$ , the absorbance is $A=-\log_{10}(I/I_0)=\alpha z/\ln 10$ , and with independent absorbers of molar concentration $C$ this is $A=\varepsilon_M C\ell$ . Hence $\alpha=\ln 10\,(\varepsilon_M C)$ with $\ln 10=2.3026$ . Why: the base of the logarithm is a human choice, so a decadic coefficient is always a napierian one divided by $\ln 10$ — the single most common numerical slip between a physics text and a chemistry text.

Step 3. Convert to decibels, a base-ten power-ratio scale: loss(dB) $=10\log_{10}(I_0/I)=10\,A$ . So a decibel is ten times an absorbance; a $3.01\,\mathrm{dB}$ loss halves the power ( $10\log_{10}2=3.01$ ) and a $10\,\mathrm{dB}$ loss cuts it to a tenth. Why: engineers count powers of ten in tenths of a bel, so a fibre's 0.2 dB/km is just $A=0.02$ per km, i.e. $\alpha=\ln 10\times0.02/1000=4.605\times10^{-5}\ \mathrm{m^{-1}}$ .

Step 4. Relate the neper to the decibel. One neper is a factor $e$ in amplitude, hence $e^2$ in power, so in decibels $1\,\mathrm{Np}=10\log_{10}e^2=20/\ln 10=8.686\,\mathrm{dB}$ . Why: this is the amplitude-versus-power factor of two wearing different clothes — the neper counts $1/e$ lengths of the field, the decibel counts power ratios, and the bridge between them carries the same 2 that put the factor into $\alpha=2\kappa\omega/c$ . *Verified: $\ln 10=2.302585$ ; $A=\alpha z/\ln 10$ gives $A=0.02$ for 1 km of 0.2 dB/km fibre; $\alpha=\ln 10\times a_{\text{decadic}}$ ; $1\,\mathrm{Np}=8.686\,\mathrm{dB}$ ; molar $A=\varepsilon_M C\ell=1\Rightarrow T=10\%$ in verify_aq-optics-absorption-coefficient-beer-lambert.py.*

Limiting cases

Case	Result	Meaning
Transparent window	$\kappa\to0\Rightarrow\alpha\to0,\ I\to I_0$	far from every resonance the medium is lossless; only the real index $n$ matters (glass in the visible)
Optically thin sample	$\alpha z\ll1\Rightarrow(I_0-I)/I_0\approx\alpha z$	fractional absorption is linear in path length and concentration — the regime where Beer's law quantitation is most reliable
Optically thick	$\alpha z\gg1\Rightarrow I\to I_0e^{-\alpha z}$ collapses fast	ten penetration depths leave $4.5\times10^{-5}$ of the power; a "black" absorber is just many $\delta$
Strong absorber / metal	$\kappa\gg n\Rightarrow\delta=1/\alpha\ll\lambda_0$	the field is evanescent within a fraction of a wavelength — the optical skin depth of a conductor
Gain medium (inverted)	$\alpha<0\Rightarrow I=I_0e^{+\|\alpha\|z}$	population inversion flips the sign: the beam is amplified — the physical basis of the laser

Common trap“The absorption coefficient α is just the field decay rate, and its three unit systems are interchangeable numbers.”

Two traps, both about factors that look cosmetic and are not. First, the factor of two: the field amplitude decays at $\kappa\omega/c$ , but the intensity — the thing a detector measures — decays at twice that, so $\alpha=2\kappa\omega/c$ , not $\kappa\omega/c$ . The amplitude $1/e$ length is $2/\alpha$ ; the intensity $1/e$ length, the penetration depth $\delta$ , is $1/\alpha$ . Quote the wrong one and every attenuation length is off by two. Second, the three scales are not interchangeable numbers: the napierian $\alpha$ (per metre, natural log) and the decadic molar absorptivity $\varepsilon_M$ (per centimetre, base ten) differ by $\ln 10=2.303$ and by using number density versus molar concentration, while the decibel is ten times the base-ten absorbance and the neper is $8.686\,\mathrm{dB}$ . Reporting a chemist's $\varepsilon_M$ as if it were a physicist's $\alpha$ , or a "3 dB point" as " $e^{-3}$ ", is the classic Beer–Lambert error. The exponent is one physical quantity; the number you write depends entirely on the logarithm base and on whether you are counting amplitude or power.

Worked examples

1 · The fibre that carries the internet (the real measurement)

A single-mode silica fibre is specified at 0.2 dB/km near 1550 nm. Find the absorption coefficient $\alpha$ , the $1/e$ attenuation length, and the distance over which the signal loses half its power.

Step 1. Convert the decibel rate to a napierian coefficient. A loss of 0.2 dB/km is an absorbance of $A=0.2/10=0.02$ per kilometre, so $\alpha=\ln 10\times A/z=2.3026\times0.02/1000=\mathbf{4.605\times10^{-5}\ m^{-1}}$ . The extinction coefficient behind this is a staggeringly small $\kappa=\alpha\lambda_0/4\pi=5.7\times10^{-12}$ .

Step 2. The $1/e$ attenuation length is the penetration depth, $\delta=1/\alpha=1/(4.605\times10^{-5})=2.171\times10^{4}\ \mathrm{m}=\mathbf{21.71\ km}$ . After 21.71 km the power is down to $1/e=37\%$ ; this is one neper, equivalently $-4.343\,\mathrm{dB}$ .

Step 3. Half power is a 3.01 dB loss ( $10\log_{10}2=3.01$ ), reached at $z=\ln 2/\alpha=0.693/(4.605\times10^{-5})=\mathbf{15.05\ km}$ — or directly, $3.01\,\mathrm{dB}/0.2\,\mathrm{dB\,km^{-1}}=15.05$ km. α = 4.605×10⁻⁵ m⁻¹, δ = 21.71 km, half-power at 15.05 km. That a beam can travel more than twenty kilometres before fading to $1/e$ is why long-haul fibre needs a repeater only every few tens of kilometres, and it is the single number that made optical telecommunications economic.

2 · Molar absorptivity, a neutral-density filter, and cross-sections (exam level)

(a) A dye solution has molar absorptivity $\varepsilon_M=1000\ \mathrm{L\,mol^{-1}\,cm^{-1}}$ ; a beam crosses $\ell=1\ \mathrm{cm}$ of a $C=10^{-3}\ \mathrm{mol/L}$ sample. What fraction is transmitted? (b) A neutral-density filter is stamped "OD 2". How much light does it pass? (c) A gas has number density $N=10^{23}\ \mathrm{m^{-3}}$ and per-molecule cross-section $\sigma=10^{-22}\ \mathrm{m^2}$ ; find $\alpha$ and the penetration depth.

Step 1. Absorbance of the dye: $A=\varepsilon_M C\ell=1000\times10^{-3}\times1=1$ . Transmission is $T=10^{-A}=10^{-1}=\mathbf{10\%}$ (90% absorbed). Note the base ten: an absorbance of 1 is not $e^{-1}=37\%$ but $10^{-1}=10\%$ — the chemist's log is decadic.

Step 2. Optical density is just absorbance, so "OD 2" means $A=2$ and $T=10^{-2}=\mathbf{1\%}$ . Stacking two OD-2 filters gives $A=4$ , i.e. $T=10^{-4}$ — absorbances add, transmissions multiply.

Step 3. For the gas, the coefficient is additive over independent absorbers: $\alpha=N\sigma=10^{23}\times10^{-22}=\mathbf{10\ m^{-1}}$ , so the penetration depth is $\delta=1/\alpha=\mathbf{0.10\ m}$ . T = 10% (dye), 1% (OD 2), α = 10 m⁻¹ with δ = 10 cm (gas). The same coefficient wears three faces — a molar $\varepsilon_M$ , a filter's OD, a microscopic $N\sigma$ — and all three feed the one exponent of Beer's law.

Physics of the portrait

The exponential attenuation law is one of the oldest quantitative results in optics, and its name records three people across a century and a quarter who each added one clause to the same sentence.

Origins · 1 minbeer-lambert law and how this was discovered

beer-lambert law

the exponential attenuation law I = I₀e^(−αz), additive over absorbers and linear in concentration (Bouguer–Lambert–Beer) · photometry and the optics of absorbing media

The law is properly the Bouguer–Lambert–Beer law, and each name marks a distinct discovery. Pierre Bouguer, a French hydrographer and the founder of photometry, published in his 1729 Essai d'optique sur la gradation de la lumière the crucial observation that equal thicknesses of an absorbing medium remove equal fractions of the light — the statement of constant fractional loss from which the exponential follows. Johann Heinrich Lambert, in his 1760 Photometria, made that observation exact, writing the attenuation as an explicit exponential in the path length and grounding it in the differential law $\mathrm{d}I=-\alpha I\,\mathrm{d}z$ ; he credited Bouguer generously. The third clause waited for chemistry: August Beer, in 1852, studying coloured solutions, showed that for a dilute absorber the coefficient is proportional to the concentration, $A=\varepsilon_M C\ell$ , turning the law into the quantitative tool of analytical chemistry. The three statements are logically independent — a fractional-loss geometry (Bouguer), an exponential in distance (Lambert), and a linearity in concentration (Beer, valid only while the absorbers act independently) — and only together do they make the modern law. That the same exponential also governs a metal's skin depth, a star's atmospheric extinction, and a fibre's decibels per kilometre is the reason it is invoked across so many fields under so many names.

History

1729

— Pierre Bouguer, in his Essai d'optique, finds that equal layers of a medium remove equal fractions of the light

1760

— Johann Heinrich Lambert, in Photometria, casts that fractional loss as the exponential attenuation law

1852

— August Beer shows the absorbance of a dilute solution is proportional to the concentration of absorber

1966

— Kao and Hockham argue that residual absorption, not scattering, is the removable barrier to glass-fibre communication

1970

— Corning demonstrates 20 dB/km silica fibre; modern telecom fibre reaches 0.2 dB/km near 1550 nm

Going further

Texts. Hecht, Optics, 5th ed., §4.8 (absorption and the complex index) · Saleh & Teich, Fundamentals of Photonics, 3rd ed., §5.5 (absorption and dispersion) · Foot, Atomic Physics, §7.1–7.3 (absorption cross-section and lineshapes) · Loudon, The Quantum Theory of Light, 3rd ed., Ch. 1 · Demtröder, Laser Spectroscopy, Vol. 1, §2.7 (absorption coefficients and cross-sections).

Provenance. The coefficient $\alpha=2\omega\kappa/c=4\pi\kappa/\lambda_0$ is recalled by name from aq-optics-complex-refractive-index-extinction, which introduced $\tilde n=n+i\kappa$ ; the plane-wave propagation factor is recalled from aq-optics-plane-waves-isotropic-media. A verifier should confirm: (1) $\alpha=4\pi\kappa/\lambda_0=2\omega\kappa/c$ from $2\,\mathrm{Im}(k)$ with $k=\tilde n\omega/c$ (Hecht §4.8, Saleh & Teich §5.5); (2) $\mathrm{d}I=-\alpha I\,\mathrm{d}z$ integrates to $I_0e^{-\alpha z}$ ; (3) the unit chain $A=\alpha z/\ln 10=\varepsilon_M C\ell$ , $1\,\mathrm{Np}=8.686\,\mathrm{dB}$ , $3.01\,\mathrm{dB}$ = factor 2 in power, with $\ln 10=2.302585$ ; (4) the fibre anchors $\alpha=4.605\times10^{-5}\ \mathrm{m^{-1}}$ , $1/e$ length 21.71 km, half-power 15.05 km from 0.2 dB/km; (5) the cross-section form $\alpha=N\sigma$ . All molar and neutral-density numbers use standard textbook definitions; no constants are fabricated.

Prerequisites. Plane Waves in Homogeneous Isotropic Media · Complex Refractive Index and the Extinction Coefficient · The Refractive Index · Dielectric Susceptibility and Permittivity. Leads to. The Lorentz Oscillator Model · Kramers–Kronig Relations · Dispersion: Normal and Anomalous · Sellmeier and Cauchy Dispersion Equations.

Why is there a factor of 2 between α and the field decay rate κω/c?

Because absorption is defined on the intensity, and intensity is the square of the field. A plane wave in an absorbing medium is $E\propto e^{-\kappa\omega z/c}$ , so the field amplitude falls with a $1/e$ length of $c/(\kappa\omega)$ . A detector, however, responds to power, $I\propto|E|^2\propto e^{-2\kappa\omega z/c}$ , which falls twice as fast — hence $\alpha=2\kappa\omega/c$ and the intensity $1/e$ length (the penetration depth) is $\delta=1/\alpha$ , half the amplitude length. The same factor of two reappears everywhere the amplitude/power distinction does: it is why decibels count 20 for amplitude ratios but 10 for power ratios, and why a neper is $8.686$ dB rather than $4.343$ . Whenever a length or a coefficient is off by exactly two, this is the first thing to check.

Why is an absorbance of 1 a transmission of 10%, not 37%?

Because the chemist's absorbance uses a base-ten logarithm, while the physicist's optical depth uses the natural one. Absorbance is defined as $A=-\log_{10}(I/I_0)$ , so $A=1$ means $I/I_0=10^{-1}=10\%$ . The napierian optical depth $\tau=\alpha z=-\ln(I/I_0)$ instead gives $\tau=1\Rightarrow I/I_0=e^{-1}=37\%$ . The two are the same physical attenuation written in different log bases, related by $\tau=A\ln 10=2.303\,A$ . This is the single most common slip when moving between a chemistry lab (absorbance, base ten, per centimetre) and an optics text (coefficient $\alpha$ , natural log, per metre). Always ask which logarithm before trusting a number.

Can the absorption coefficient be negative?

Yes, and it is one of the most important sign changes in physics. If a medium is prepared with a population inversion — more absorbers in the excited state than the ground state — then stimulated emission outpaces absorption, the imaginary part of the susceptibility flips sign, and $\alpha<0$ . Beer's law then reads $I(z)=I_0e^{+|\alpha|z}$ : the beam is amplified as it propagates. This is exactly the small-signal gain of a laser amplifier, and the same exponential bookkeeping, with a gain coefficient replacing $\alpha$ , sets the threshold when gain is balanced against cavity loss. Nothing about the derivation changes; only the sign of $\kappa$ does, which requires the medium to be out of thermal equilibrium.

When does Beer's law stop being linear in concentration?

The linearity $A=\varepsilon_M C\ell$ is an approximation valid only while the absorbers act independently and the light is monochromatic. It fails on several fronts. At high concentration the absorbers interact — they shift each other's energy levels, aggregate, or change $\varepsilon_M$ — and the calibration bends. If the source is not monochromatic, the detector averages transmission over a band where $\varepsilon_M$ varies, and the logarithm of an average is not the average of logarithms, curving the plot at high absorbance. Stray light sets a ceiling: once the transmitted beam falls near the instrument's stray-light floor, apparent absorbance saturates. And scattering (turbid samples) removes light without true absorption. In practice, quantitation is trusted in the optically thin window $A\lesssim1$ , where every one of these effects is small.

How can a cross-section σ be larger than the physical size of the atom?

Because the cross-section measures how effectively the atom couples to the field, not how big it is. Near a resonance the bound electron is driven at its natural frequency and responds with an amplitude far larger than its static displacement, so the atom scatters and absorbs from a region of the wavefront much wider than its own diameter — the resonant cross-section can reach $\sim\lambda^2/2\pi$ , thousands of times the geometric area. The "shadow" picture behind $\alpha=N\sigma$ still holds; the shadow is just set by the driven-dipole response, which peaks on resonance and collapses far from it. This is why a trace gas with a strong, isolated line can block light out of all proportion to its density, and why resonant absorption is such a sensitive probe of dilute species.

Why do telecom fibres run at 1550 nm specifically?

Because that wavelength sits in the deepest transparency window of silica glass, where $\alpha$ is smallest. Glass absorbs in the ultraviolet (electronic transitions) and the infrared (Si–O vibrational overtones), and scatters more strongly toward the blue (Rayleigh scattering falls as $1/\lambda^4$ ). Between these, near 1550 nm, the total attenuation bottoms out at about 0.2 dB/km — the $1/e$ length of 21.71 km in the figure. A second, slightly lossier window near 1310 nm coincides with silica's zero material dispersion, so systems trade off the two. The historic push, from Kao's 1966 argument that the loss was removable to Corning's low-loss fibre, was precisely a campaign to purify the glass until its residual absorption reached this fundamental floor — after which the limit becomes Rayleigh scattering, not chemistry.