Molar Absorptivity and Transmittance

A spectrophotometer reports a number. That number can lie to you if you don't know its shape. Push the concentration too high and the meter reads a flat, useless 0.003 %T. Dilute too far and it reads 99.1 %T. Both readings carry almost no chemical information. Yet the instrument dutifully displays four digits. Between these extremes lies a narrow window of absorbance. In this window, a spectrophotometric measurement is actually trustworthy. Understanding why that window exists is the real content of this article. Transmittance, absorbance, and molar absorptivity are three descriptions of the same physical event. That event is light passing through a solution and losing intensity to absorbing molecules. But they are not interchangeable in how they behave statistically. Transmittance is what the detector physically measures: a ratio of light powers. Absorbance is a logarithmic transformation of that ratio. It is chosen because it is linear in concentration and path length. Molar absorptivity is the fundamental, wavelength-dependent property of the absorbing species itself. It is stripped of concentration and path length. This lets you compare one chromophore's intrinsic light-absorbing power to another's. It also lets you predict absorbance in an experiment you haven't yet run. This matters because ultraviolet-visible spectrophotometry underlies enormous swaths of chemistry. Examples include protein concentration by A₂₈₀, and DNA purity ratios. They also include drug-content assays in pharmaceutical quality control. Enzyme kinetics followed by a plate reader is another example. So is colorimetric determination of trace metals. Food and environmental testing, done by the million samples per year, is another. Every one of these measurements depends on choosing a concentration, via dilution, and a path length. These must be chosen so the resulting absorbance falls in a good regime. In that regime, instrumental noise is minimized. The calibration curve also stays linear, per the The Beer-Lambert Law. This article builds the full vocabulary — %T, A, ε. It derives how they convert into one another. It also derives the propagation-of-error argument that identifies the optimal absorbance range. That range is traditionally A ≈ 0.4–0.9. There is a sharp minimum in relative error at A = 0.434. The article works through three realistic examples. These span a classroom calibration curve, a real pharmaceutical assay, and a multicomponent mixture problem. It sets up, without yet solving, further questions. These concern nonlinearity and matrix effects. Such effects arise once real instruments and real solutions stop behaving ideally.

Key takeaways

Transmittance $T = P/P_0$ is the fraction of incident light power that exits the sample; percent transmittance is $\%T = 100T$ .
Absorbance $A = -\log_{10}T = \log_{10}(P_0/P)$ is dimensionless and additive over multiple absorbers and multiple path segments.
Molar absorptivity $\varepsilon$ (units L·mol⁻¹·cm⁻¹) is an intrinsic, wavelength-dependent property; $A = \varepsilon b c$ is the The Beer-Lambert Law.
Photometric (shot- and dark-current-limited) error in $c$ is minimized near $A \approx 0.434$ , and the practical usable window is roughly $A = 0.2$ – $1.0$ .
Outside that window, %T readout error is amplified nonlinearly into huge relative concentration error — not because the chemistry changes, but because $\log_{10}$ compresses information at the extremes.
Real instruments deviate from ideal Beer–Lambert behavior at high absorbance for reasons besides statistics (stray light, polychromatic bandwidth), foreshadowing genuine Deviations from Beer's Law.

Key idea: Absorbance is the logarithmically transformed, concentration-linear quantity that both the chemistry (Beer–Lambert) and the statistics (error propagation) agree is the right variable to work in — but only inside a bounded, instrument-defined window of values.

Where this lives today

Every UV-Vis spectrophotometer sold today — from a $2,000 teaching-lab Spec-20 successor to a$ 150,000 diode-array research instrument — is built around exactly this triad of transmittance, absorbance, and molar absorptivity, and every protocol that uses one specifies a target absorbance range for precisely the reasons derived above. Molecular biology labs quantify nucleic acids via A₂₆₀ (with the well-known A₂₆₀/A₂₈₀ ratio used as a crude purity check for protein contamination), routinely diluting samples so the reading lands near $A\approx0.5$ – $1.0$ rather than trusting a raw, oversaturated stock reading. Protein concentration by the Bradford or BCA colorimetric assays, and enzyme-linked immunosorbent assays (ELISAs) read on microplate readers, both rely on calibration curves built strictly in absorbance space, with well-designed dilution series specifically targeting the linear, low-error window. Pharmaceutical quality control, as in the acetaminophen example above, uses UV assay as a primary or confirmatory identity and content-uniformity test in nearly every monograph in the USP and European Pharmacopoeia, and analytical method validation protocols explicitly require demonstrating linearity of the calibration curve across a defined absorbance range before a method can be approved for release testing. Environmental and water-quality labs use colorimetric methods (phosphate as the molybdenum blue complex, nitrite via the Griess reaction, chlorine via DPD) where molar absorptivities of the colored complexes, often deliberately chosen or engineered to be very large ( $\varepsilon$ in the tens of thousands), allow trace-level detection while still landing measured absorbances in the workable window through appropriate color-development chemistry and cell path-length choice (some field kits use 5 or 10 cm cells specifically to boost weak absorbances into range). Clinical chemistry analyzers performing thousands of assays per hour — glucose, bilirubin, creatinine — are all, at heart, automated dilution-and-absorbance-reading machines built on the same three quantities defined here, with automatic flagging when a sample's raw absorbance falls outside instrument-validated linear range, triggering an automatic re-run at a different dilution. Even far afield, atmospheric remote sensing (DOAS, differential optical absorption spectroscopy) and Raman/IR quantitation in materials science borrow the identical logic of converting a transmitted-intensity ratio into a linear, additive, concentration-proportional absorbance variable before attempting any quantitative fit.

Ask the tutor: If my measured absorbance keeps drifting slightly above the Beer–Lambert prediction only at high concentrations, is that a statistical error like the one derived here, or something else entirely?

The physical picture

Picture a beam of monochromatic light, intensity $P_0$ , entering a cuvette of path length $b$ containing an absorbing species at molar concentration $c$ . Each absorbing molecule presents a certain effective cross-section to the beam; statistically, in an infinitesimally thin slice $db$ of solution, the fractional loss of power $dP/P$ is proportional to the number of absorbers the beam encounters, which is proportional to $c\,db$ . Integrating this exponential attenuation over the full path length $b$ gives $P = P_0\,10^{-\varepsilon bc}$ (base-10 by convention in analytical chemistry, though the underlying physics is a natural exponential — $\varepsilon$ absorbs the $\ln 10$ conversion). This is exactly the statement of the The Beer-Lambert Law, and it is the reason absorbance, not transmittance, is linear in concentration. The detector, however, does not measure absorbance. A photodiode or photomultiplier tube converts photon flux into a photocurrent, and the instrument's electronics compare the photocurrent with sample in the beam ( $P$ ) to the photocurrent with a reference blank in the beam ( $P_0$ ). The raw, physically-measured quantity is the ratio $T = P/P_0$ , transmittance — a number between 0 and 1, or equivalently between 0% and 100%. The instrument then computes $A = -\log_{10}T$ internally and displays whichever of $T$ , $\%T$ , or $A$ the user has selected. Because the human eye and early instrument dials were calibrated in %T, %T is still what the physical scale on a classic Spec-20 knob shows; %T is also intuitive ("60% of the light got through"). But %T is a poor variable for quantitative work exactly because it is not linear in $c$ — it is an exponential of $c$ , and small errors in reading a compressed, curved scale near either end translate into large concentration errors. The molar absorptivity $\varepsilon$ is what's left over once path length and concentration are divided out: $\varepsilon = A/(bc)$ . It depends on wavelength, on the solvent (through the local electric field a chromophore experiences), on temperature to a much smaller extent, and fundamentally on the transition dipole moment of the electronic transition being excited — but critically it does not depend on how much sample you happen to have or how thick your cuvette is. A molecule with $\varepsilon = 15{,}000\ \text{L}\,\cdot\,\text{mol}^{-1}\text{cm}^{-1}$ at its absorption maximum is simply a far more efficient light absorber than one with $\varepsilon = 150\ \text{L}\,\cdot\,\text{mol}^{-1}\text{cm}^{-1}$ , independent of any particular measurement.

Figure

Transmittance, absorbance, and the nonlinear %T scale

The curve above shows the essential fact: %T falls off exponentially and steeply near $A = 0$ , then flattens dramatically as $A$ grows past about 1. Reading a nearly-flat curve to the same absolute precision (say, ±0.3 %T, typical of a decent single-beam instrument) translates into wildly different relative precision in concentration depending on where on the curve you sit — which is precisely the argument that produces the optimal absorbance range, derived below.

Figure

Additivity of absorbance for a two-component mixture

The core ideas

The three defining relationships, and how they interlock:

T = \frac{P}{P_0}, \qquad \%T = 100\,T

where: $P_0$ is the radiant power incident on (or exiting a reference blank in) the sample, and $P$ is the radiant power that exits the sample cell. Both are measured in the same units (typically as a photocurrent or voltage proportional to photon flux), so $T$ is dimensionless.

A = -\log_{10}T = \log_{10}\!\left(\frac{P_0}{P}\right) = \log_{10}\!\left(\frac{100}{\%T}\right)

where: $A$ is absorbance, also dimensionless, always non-negative for $0 < T \le 1$ . Reading: absorbance is the base-10 logarithm of how many times the light intensity is reduced.

A = \varepsilon b c

where: $\varepsilon$ is the molar absorptivity in L·mol⁻¹·cm⁻¹ (equivalently M⁻¹cm⁻¹), $b$ is the path length in cm, and $c$ is molar concentration in mol/L. This is the The Beer-Lambert Law itself, and it is what makes $A$ — not $T$ — the natural variable for building a calibration curve.

A_{\text{total}} = \sum_i \varepsilon_i b c_i \quad\text{(additivity, non-interacting absorbers, same }\lambda\text{)}

where: each term is the absorbance species $i$ would show alone; this additivity is the basis of multicomponent spectrophotometric analysis and is a direct consequence of $A$ 's logarithmic definition converting multiplicative transmittance losses into additive absorbance losses.

Derivations

Derivation AFrom single-molecule attenuation to A = εbc, and why base 10

Goal. Show how the exponential attenuation of light by discrete absorbers becomes the linear law $A=\varepsilon bc$ , and pin down where the constant $\varepsilon$ and the choice of logarithm base come from.

Step 1 — model the attenuation in a thin slice. Consider a beam traveling through a slab of solution of cross-sectional area $S$ and infinitesimal thickness $dx$ . Each absorbing molecule presents an effective absorption cross section $\sigma$ (units of area) to photons of the given wavelength. The number of absorber molecules in the slice is $n = N_A c\, S\, dx$ , where $N_A$ is Avogadro's number and $c$ is molarity (mol per liter, with unit conversions folded in). The probability that a photon is absorbed while traversing $dx$ is the fraction of the cross-sectional area blocked by absorbers, $n\sigma/S = N_A c\sigma\,dx$ , assuming this fraction is small (true for a thin slice) and that absorbers don't shadow one another.

Step 2 — set up and solve the differential equation. The power lost per unit thickness is proportional to the power present and to this absorption probability per unit length:

dP = -P\,(N_A \sigma c)\,dx

Separating variables and integrating from $x=0$ (power $P_0$ ) to $x=b$ (power $P$ ):

\int_{P_0}^{P}\frac{dP}{P} = -N_A\sigma c\int_0^b dx

\ln\!\left(\frac{P}{P_0}\right) = -N_A\sigma c\, b

Step 3 — convert to base-10 absorbance. Analytical chemistry defines $A \equiv -\log_{10}(P/P_0)$ , not the natural log, purely by historical convention (it makes $\varepsilon$ values come out as tidy tabulated numbers rather than always carrying a stray $\ln 10$ ). Converting,

A = -\log_{10}\left(\frac{P}{P_0}\right) = -\frac{1}{\ln 10}\ln\!\left(\frac{P}{P_0}\right) = \frac{N_A\sigma}{\ln 10}\, b\, c

Define the molar absorptivity as the lumped constant $\varepsilon \equiv N_A\sigma/\ln 10$ (with unit conversions so that with $c$ in mol/L and $b$ in cm, $\varepsilon$ comes out in L·mol⁻¹cm⁻¹). Then:

A = \varepsilon b c

This is exactly the Beer–Lambert law, now with $\varepsilon$ explicitly tied to a microscopic absorption cross-section per molecule — a genuine physical property of the chromophore's electronic transition, not a fitting artifact.

Step 4 — sanity check the limiting case. As $c\to 0$ (no absorber), $A\to 0$ and $T\to 1$ (100% transmittance) — light passes through unattenuated, as it must for a blank. As $b\to 0$ (zero path length), likewise $A\to 0$ regardless of concentration, consistent with there being no material to absorb in. Both limits confirm the algebra: $A=\varepsilon bc$ correctly forces $A=0$ whenever either geometric or chemical "amount of absorber in the path" vanishes.

Go deeper · Derivation B · 2 minThe optimal absorbance range from photometric error propagation

Derivation BThe optimal absorbance range from photometric error propagation

Goal. Given that a real spectrophotometer's transmittance readout carries some absolute uncertainty $\Delta T$ (from shot noise, dark current, and readout digitization, roughly constant for a given instrument over its working range), find the absorbance value that minimizes the relative uncertainty in the derived concentration, $\Delta c/c$ .

Step 1 — express c in terms of T. From $A=\varepsilon bc$ and $A = -\log_{10}T$ ,

c = \frac{-\log_{10}T}{\varepsilon b} = \frac{-\ln T}{\varepsilon b\,\ln 10}

Step 2 — propagate the fixed instrumental uncertainty ΔT through to Δc. Treat $\Delta T$ as the (roughly $T$ -independent) standard uncertainty in the transmittance reading. Differentiate $c(T)$ :

\frac{dc}{dT} = \frac{-1}{\varepsilon b \,\ln 10}\cdot\frac{1}{T}

so, propagating errors (first-order, $\Delta c \approx |dc/dT|\,\Delta T$ ),

\Delta c = \frac{\Delta T}{\varepsilon b\,\ln 10\, T}

Step 3 — form the relative error and substitute T in terms of A. Divide by $c$ from Step 1:

\frac{\Delta c}{c} = \frac{\Delta T/(\varepsilon b\ln10\,T)}{-\ln T/(\varepsilon b\ln10)} = \frac{-\Delta T}{T\ln T}

Since $\ln T = -A\ln 10$ and $T = 10^{-A}$ , substitute to express relative error purely as a function of absorbance:

\frac{\Delta c}{c} = \frac{\Delta T}{\,T\ln T\,}\cdot(-1)= \frac{0.4343\,\Delta T}{10^{-A}\,A}

(using $-1/\ln 10 = -0.4343$ and absorbing signs so the right side is positive).

Step 4 — minimize over A. Let $f(A) = A\,10^{-A}$ appear in the denominator; relative error is minimized where $f(A)$ is maximized. Differentiate $f$ and set to zero:

\frac{df}{dA} = 10^{-A} - A\,(\ln 10)\,10^{-A} = 10^{-A}\left(1 - A\ln 10\right) = 0

Since $10^{-A}\neq0$ , the extremum requires

1 - A\ln 10 = 0 \quad\Longrightarrow\quad A^* = \frac{1}{\ln 10} = 0.4343

The second derivative is negative there (this is a maximum of $f$ , hence a minimum of relative error), so $A^* \approx 0.434$ is the theoretically optimal absorbance for minimizing relative concentration error under a constant- $\Delta T$ noise model.

Step 5 — quantify how sharp the minimum is (why a range, not one number, is used in practice). Evaluating relative error (in percent, taking a typical $\Delta T = 0.0033$ , i.e. ±0.33 %T, characteristic of a mid-range instrument) at several $A$ values:

$A$	$T$	$\Delta c/c$ (%)
0.10	0.794	18.2
0.20	0.631	11.5
0.434	0.368	9.1
0.70	0.200	10.3
1.00	0.100	14.5
2.00	0.010	108.6

The curve is shallow near the minimum (9.1% at the true optimum vs. 10.3–11.5% a fair distance away at $A=0.2$ – $0.7$ ) but explodes at either extreme — at $A=2.00$ the relative error has ballooned past 100%, meaning the "concentration" extracted from that reading is not trustworthy at all. This is exactly why working guidance is a band, conventionally $A\approx0.2$ – $1.0$ (some texts tighten this to $0.4$ – $0.9$ for the most demanding work), rather than insisting on hitting $A=0.434$ exactly.

Limiting check. As $A\to0$ ( $T\to1$ , essentially no absorption), $\Delta c/c\to\infty$ because $A$ itself appears in the denominator of Step 3's formula — you are trying to detect a signal that isn't there, so any fixed $\Delta T$ is a huge fraction of an infinitesimal true signal. As $A\to\infty$ ( $T\to0$ , essentially opaque), $10^{-A}\to0$ even faster than $A$ grows, so $\Delta c/c\to\infty$ again — almost no photons reach the detector, so shot noise dominates and the reading is unreliable. Both boundary behaviors are physically sensible and confirm the derivation captures a real U-shaped error curve, not an artifact of the algebra.

Worked examples

Example 1 (foundational) — converting %T to A and back, and reading ε off a calibration point. A solution of Cu(NH₃)₄²⁺ is placed in a 1.00 cm quartz cuvette and measured at 620 nm against a water blank. The instrument reads $\%T = 36.8\%$ . Step 1. Convert to transmittance: $T = 36.8/100 = 0.368$ . Step 2. Convert to absorbance:

A = -\log_{10}(0.368) = 0.434

(This is, not coincidentally, the optimum absorbance from the derivation above — chosen deliberately for this example.) Step 3. If the solution's concentration is known independently to be $c = 2.50\times10^{-3}$ M, find $\varepsilon$ at 620 nm using $A=\varepsilon bc$ with $b=1.00$ cm:

\varepsilon = \frac{A}{bc} = \frac{0.434}{(1.00\ \text{cm})(2.50\times10^{-3}\ \text{mol/L})} = 173.6\ \text{L}\,\cdot\,\text{mol}^{-1}\text{cm}^{-1}

Reporting to 3 sig figs (limited by the %T reading, typically good to ±0.1–0.3%T on a decent instrument): $\varepsilon = 174\ \text{L}\,\cdot\,\text{mol}^{-1}\text{cm}^{-1}$ . Step 4 — limiting-case sanity check. If instead the same solution were diluted to $c\to0$ , $A\to0$ and $\%T\to100\%$ ; the formula correctly predicts a blank-like reading with no information about $\varepsilon$ extractable (division by $c\approx0$ blows up any attempted calculation, as it must — you cannot determine a chromophore's molar absorptivity from a solution containing essentially none of it).

Example 2 (applied) — a real pharmaceutical UV assay, dilution planning, and the optimal-A window. A quality-control lab must assay acetaminophen tablets by UV spectrophotometry at $\lambda_{max} = 243$ nm, where the literature molar absorptivity in dilute aqueous/methanolic buffer is $\varepsilon = 9.85\times10^{3}$ L·mol⁻¹cm⁻¹ (a typical order of magnitude for an aromatic carbonyl π→π transition). A tablet is dissolved and diluted to give a stock solution calculated (from label claim) to contain acetaminophen at $c_{\text{stock}} = 4.00\times10^{-3}$ M. The analyst wants the final, measured solution to land at $A\approx0.50$ , comfortably inside the optimal window, using a standard 1.00 cm cell. Step 1 — find the concentration that gives A = 0.50.*

c_{\text{target}} = \frac{A}{\varepsilon b} = \frac{0.50}{(9.85\times10^3\ \text{L}\,\cdot\,\text{mol}^{-1}\text{cm}^{-1})(1.00\ \text{cm})} = 5.08\times10^{-5}\ \text{M}

Step 2 — determine the dilution factor needed from the stock.

\text{DF} = \frac{c_{\text{stock}}}{c_{\text{target}}} = \frac{4.00\times10^{-3}}{5.08\times10^{-5}} = 78.7

So the analyst must dilute roughly 1-in-79 — practically, take, e.g., 1.00 mL of stock to a final volume of 78.7 mL (or a convenient two-stage dilution such as 1.00 mL → 10.00 mL, then 5.00 mL of that → 39.4 mL, giving the same overall factor with better volumetric precision than one huge single dilution). Step 3 — check the measurement and back-calculate an unknown tablet's content. Suppose the diluted unknown tablet solution is measured and gives $A_{\text{obs}} = 0.462$ under the identical dilution scheme. Concentration in the cuvette:

c_{\text{cuvette}} = \frac{0.462}{9.85\times10^{3}\times1.00} = 4.69\times10^{-5}\ \text{M}

Multiply back through the dilution factor (78.7) to find the original stock concentration:

c_{\text{stock,found}} = 4.69\times10^{-5}\ \text{M}\times78.7 = 3.69\times10^{-3}\ \text{M}

Compare to the label-claim stock of $4.00\times10^{-3}$ M: the tablet assays at

\frac{3.69\times10^{-3}}{4.00\times10^{-3}}\times100\% = 92.3\%

of label claim — a result a QC chemist would flag, since USP-type acceptance windows are typically 90.0–110.0% of label claim, so this would pass but sit uncomfortably close to the lower limit and likely trigger a retest. Step 4 — why this dilution scheme was worth planning. Had the analyst instead measured the undiluted stock directly ( $c = 4.00\times10^{-3}$ M) in the same 1.00 cm cell, the absorbance would have been

A = (9.85\times10^3)(1.00)(4.00\times10^{-3}) = 39.4

which is physically meaningless as a transmittance reading (it would correspond to $T = 10^{-39.4}$ , i.e., essentially zero photons reaching the detector — deep into the noise floor, total instrumental darkness). This underlines concretely why the dilution step, targeting $A\approx0.5$ , is not optional bookkeeping but the difference between a real measurement and pure noise.

Example 3 (exam/competition level) — resolving a two-component mixture using additivity of absorbance. A mixture contains two absorbing species, X and Y, in the same 1.00 cm cell. Their molar absorptivities at two wavelengths are known from pure-standard measurements:

	400 nm	500 nm
$\varepsilon_X$ (L·mol⁻¹cm⁻¹)	1.20×10⁴	2.10×10³
$\varepsilon_Y$ (L·mol⁻¹cm⁻¹)	3.50×10³	8.40×10³

The mixture, in a 1.00 cm cell, gives measured absorbances $A_{400} = 0.842$ and $A_{500} = 0.615$ . Find $c_X$ and $c_Y$ . Step 1 — write the additivity equations (b = 1.00 cm throughout, so it drops out of every term).

0.842 = (1.20\times10^4)\,c_X + (3.50\times10^3)\,c_Y

0.615 = (2.10\times10^3)\,c_X + (8.40\times10^3)\,c_Y

Step 2 — solve the linear system. From the first equation,

c_X = \frac{0.842 - 3.50\times10^3\,c_Y}{1.20\times10^4}

Substitute into the second:

0.615 = 2.10\times10^3\left(\frac{0.842 - 3.50\times10^3\,c_Y}{1.20\times10^4}\right) + 8.40\times10^3\,c_Y

0.615 = 175.0\,(0.842) - 175.0(3.50\times10^3/1.20\times10^4)^{-1}\ldots

More carefully, compute the coefficient: $2.10\times10^3/1.20\times10^4 = 0.1750$ . So

0.615 = 0.1750(0.842) + \left(8.40\times10^3 - 0.1750\times3.50\times10^3\right)c_Y

0.615 = 0.1474 + (8.40\times10^3 - 612.5)\,c_Y = 0.1474 + 7.788\times10^3\,c_Y

c_Y = \frac{0.615 - 0.1474}{7.788\times10^{3}} = \frac{0.4676}{7.788\times10^3} = 6.005\times10^{-5}\ \text{M}

Step 3 — back-substitute for c_X.

c_X = \frac{0.842 - (3.50\times10^3)(6.005\times10^{-5})}{1.20\times10^4} = \frac{0.842 - 0.2102}{1.20\times10^4} = \frac{0.6318}{1.20\times10^4} = 5.265\times10^{-5}\ \text{M}

Step 4 — check by recomputing both absorbances.

A_{400} = (1.20\times10^4)(5.265\times10^{-5}) + (3.50\times10^3)(6.005\times10^{-5}) = 0.6318+0.2102 = 0.842\ \checkmark

A_{500} = (2.10\times10^3)(5.265\times10^{-5}) + (8.40\times10^3)(6.005\times10^{-5}) = 0.1106+0.5044=0.615\ \checkmark

Both recovered exactly, confirming the algebra; $c_X = 5.27\times10^{-5}$ M and $c_Y = 6.01\times10^{-5}$ M to three sig figs. Step 5 — limiting-case check. If X and Y had identical spectra ( $\varepsilon_X(\lambda)=\varepsilon_Y(\lambda)$ at both wavelengths), the two equations would become linearly dependent (same equation twice), the determinant of the coefficient matrix would vanish, and the system would be unsolvable for individual concentrations — only the sum $c_X+c_Y$ could be recovered. This is the correct physical limit: spectrally identical species cannot be resolved by absorbance alone, no matter how many (dependent) wavelengths you throw at the problem; you would need a third, independent piece of information (a different technique, or a chemical separation) to split them.

Limiting cases, scope & common traps

Situation	What happens	Correct handling
$A \to 0$ (very dilute, or very short path)	$T\to1$ ; %T reading pinned near 100%, tiny absolute signal	Increase path length (longer cell) or concentrate sample; relative error in $c$ diverges as derived above
$A \to \infty$ (very concentrated or long path)	$T\to0$ ; detector near dark-current floor, dominated by shot noise and stray light	Dilute the sample or shorten path; never trust a reading near 0 %T
Two absorbers, same $\lambda_{max}$ shape	Additivity equations become linearly dependent	Only total concentration recoverable from absorbance alone; need a separative or spectrally-orthogonal method
Polychromatic source / wide bandpass	Apparent $\varepsilon$ becomes concentration-dependent, curve bends downward at high $A$	This is a genuine instrumental deviation from Beer's law, distinct from the statistical error argument here
Stray light reaching detector	%T reading has a nonzero floor even at "infinite" absorbance	Sets a hard instrumental ceiling on usable $A$ , independent of the ideal derivation
Reporting $\varepsilon$ without stating $\lambda$ and solvent	Number is meaningless out of context	Always quote $\varepsilon_\lambda$ in a stated solvent; $\varepsilon$ shifts with solvatochromic effects

Common trapCommon misconceptions

"Absorbance is just a rescaled way of writing transmittance, so it shouldn't matter which one I use for a calibration curve." False in practice: $A$ is linear in $c$ (Beer–Lambert), while $T$ is exponential in $c$ . A calibration curve built in %T is curved and awkward to fit; built in $A$ , it is a straight line through the origin (ideally), which is why every real calibration protocol works in absorbance units, never raw %T. "A higher absorbance reading always means a more precise measurement, since the signal is stronger." False — beyond roughly $A\approx1$ , the amount of light reaching the detector becomes so small that shot noise and the instrument's dark-current floor dominate, and relative error in concentration rises steeply again, exactly as the error-propagation derivation shows. Precision is optimized in a middle band, not at either extreme.

The chemist behind it

The modern habit of thinking in absorbance rather than transmittance descends from two independent 19th-century observations bearing the names now fused into the "Beer–Lambert law." Pierre Bouguer first noted (1729) that successive equal thicknesses of an absorbing medium each transmit the same fraction of incident light — the origin of the exponential attenuation law, later rediscovered and popularized by Johann Heinrich Lambert (1760), who framed it in terms of a medium's thickness. August Beer, working in Bonn in 1852, extended the same logic to the concentration of an absorbing solute dissolved in a nonabsorbing solvent, showing that absorbance scales linearly with the amount of chromophore present, not merely with the geometric path.

Origins · 1 minAugust Beer and how this was discovered

August Beer

1825–1863 · Bonn, mid-19th-century physical and analytical chemistry

Beer was a German physicist and mathematician who worked at the University of Bonn, where he studied the transmission of light through absorbing solutions using early photometric methods — visually comparing the brightness of two light paths through a null-balance instrument, since no photoelectric detectors yet existed. His 1852 paper established that absorption depends multiplicatively (in transmittance) or additively (in the logarithmic measure) on the concentration of dissolved absorbing substance, generalizing Lambert's path-length result to a genuinely chemical variable. Beer died young, at 38, but the combined Bouguer–Lambert–Beer relationship became, over the following century, the single most-used quantitative relationship in all of analytical spectroscopy — underpinning every UV-Vis, IR, and much of atomic absorption instrumentation built since.

History

- 1729 — Pierre Bouguer observes that successive equal-thickness layers of an absorbing medium each transmit an equal fraction of incident light.

- 1760 — Johann Heinrich Lambert publishes Photometria, formalizing the exponential dependence of transmitted light on path length.

- 1852 — August Beer extends the law to solute concentration, establishing the linear dependence of absorbance on the amount of dissolved chromophore.

- 1940s — Commercial photoelectric spectrophotometers (e.g., the Beckman DU, 1941) make routine, precise transmittance measurement practical outside specialized optics labs.

- 1950s–60s — Ringbom and others formalize the photometric error-propagation analysis, establishing the now-standard optimal absorbance window (A ≈ 0.2–1.0) taught in every quantitative analysis course.

Why is absorbance defined with a base-10 logarithm instead of a natural logarithm?

Purely historical and practical convention. The underlying physics (Step 2 of the first derivation) naturally produces a natural-log relationship, $\ln(P/P_0) = -N_A\sigma bc$ . Early spectroscopists, working with slide rules and log tables, found base-10 logarithms more convenient to tabulate and compute by hand, and the resulting molar absorptivity values came out as "nicer" numbers in that convention. The base-10 definition, $A=-\log_{10}T$ , has simply persisted as the analytical-chemistry standard ever since, and all tabulated $\varepsilon$ values in the literature assume it. If you ever see a "natural absorbance" or use $\ln$ instead, the numeric value of $\varepsilon$ you compute will differ by a factor of $\ln 10 \approx 2.303$ from the standard tabulated value, so consistency matters more than which convention you pick.

Does molar absorptivity ever depend on concentration?

Ideally, no — $\varepsilon$ is defined as an intrinsic molecular property, independent of $c$ , which is exactly why the Beer–Lambert law $A=\varepsilon bc$ predicts a straight line through the origin on an absorbance-vs-concentration plot. In practice, apparent $\varepsilon$ can appear to drift with concentration at high $c$ due to several real effects: chemical equilibria that shift with concentration (e.g., dimerization or aggregation of the chromophore, protonation equilibria whose position depends on total concentration), changes in the solution's refractive index at high solute loading, and instrumental polychromatic-light effects that become worse at high absorbance. These are all genuine physical or instrumental complications, not violations of the underlying single-molecule law, and are the subject of the deviations-from-Beer's-law analysis that follows on from this foundation.

Why can't I just measure %T directly and skip computing absorbance?

You can record %T, but you should not build a calibration curve or report concentration directly from %T without converting to absorbance first, because %T is not linear in concentration. A plot of %T vs. c is a decaying exponential curve, not a straight line, so ordinary linear regression is the wrong tool and will give a biased fit if applied naively. Converting to $A=-\log_{10}(\%T/100)$ restores linearity in $c$ (per Beer–Lambert), letting you use simple, well-understood linear least-squares regression with reliable confidence intervals — which is exactly why every quantitative UV-Vis protocol works in absorbance space even though the raw detector signal is transmittance.

What's a "typical" molar absorptivity, and how do I know if a chromophore is a strong or weak absorber?

Rough benchmarks used across the field: fully allowed π→π* transitions in conjugated organic chromophores (like the acetaminophen example, or many dyes) typically show $\varepsilon$ in the range $10^3$ – $10^5$ L·mol⁻¹cm⁻¹. Symmetry-forbidden or spin-forbidden transitions, common in many d-d transitions of transition-metal complexes (like the Cu(NH₃)₄²⁺ example, $\varepsilon\approx170$ ), are far weaker, typically $\varepsilon\sim 1$ – $100$ L·mol⁻¹cm⁻¹, because the transition dipole moment is small. Charge-transfer transitions and strongly conjugated dyes (phenolphthalein in base, many indicator dyes, metal-ligand charge-transfer complexes used in colorimetric assays) can reach $\varepsilon > 10^4$ – $10^5$ , which is precisely why such complexes are deliberately chosen for trace-level colorimetric determinations — a large $\varepsilon$ lets you reach a workable absorbance at very low concentration.

How does path length choice interact with the optimal absorbance window?

Because $A=\varepsilon bc$ , path length $b$ and concentration $c$ enter absorbance in exactly the same multiplicative way — doubling either one doubles $A$ . This gives you two independent levers to bring a measurement into the optimal window (roughly $A=0.2$ – $1.0$ ): if a sample is too dilute to hit a good absorbance in a standard 1.00 cm cell, you can switch to a longer-path cell (5 cm and 10 cm cells are common for trace analysis) instead of trying to concentrate the sample, which might not be practical or might risk shifting an equilibrium. Conversely, for a very strongly absorbing or highly concentrated sample, a short-path "micro" cell (as short as 0.1 cm, or even purpose-built micro-volume cells with sub-millimeter effective path lengths) can bring an otherwise off-scale absorbance back down into range without diluting the actual chemical sample, which matters when sample volume is precious (as in many biological assays).

Is there a difference between "absorbance" and "optical density"?

In practice, essentially no difference in ideal cases — both are defined as $-\log_{10}(P/P_0)$ and both are dimensionless. "Optical density" (OD) is the older, more general term (often used in physics and in contexts involving scattering as well as true absorption, such as reporting bacterial culture turbidity as "OD600"), while "absorbance" is the analytical-chemistry term reserved specifically for light loss due to true molecular absorption governed by the Beer–Lambert law. The subtlety is that OD600 readings of a bacterial culture are not really governed by molar absorptivity of a chromophore at all — the signal loss is mostly light scattering by cells, not electronic absorption — so while the arithmetic ( $-\log_{10}T$ ) is identical, treating a turbidity "OD" as if it obeyed a true molar Beer–Lambert relationship with a fixed $\varepsilon$ is not strictly justified.

What goes wrong experimentally if I ignore the optimal absorbance range and just report whatever reading I get?

Nothing goes wrong with the arithmetic — $A=\varepsilon bc$ still holds as a chemical relationship regardless of what $A$ happens to be — but the precision of your reported concentration silently degrades at the extremes, often without any obvious warning sign on the instrument display. A measurement at $A=0.02$ or $A=2.5$ can look perfectly normal on the readout, complete with several decimal digits of apparent precision, while actually carrying 30–100%+ relative uncertainty once you propagate the instrument's real photometric noise, as the error-propagation derivation shows explicitly. This is why validated analytical methods specify an acceptable absorbance (or working concentration) range as part of the method itself, and why a competent analyst dilutes or concentrates a sample rather than trusting an out-of-range reading just because the display shows a number.