Monochromators and Wavelength Selection

Every double-beam UV-Vis spectrophotometer in a modern teaching or QC lab uses a grating monochromator, almost always in the Czerny–Turner layout, with a motor-driven sine-bar mechanism that converts uniform rotation into a linear wavelength scan — the mechanical descendant of Rowland's original ruling precision. Atomic absorption and atomic emission spectrometers depend on a carefully chosen bandpass (commonly selectable in steps of 0.2, 0.5, 1.0, and 2.0 nm) to isolate one resonance line from neighboring lamp emission lines, exactly the calculation worked in Example 2; getting this wrong is a classic cause of nonlinear calibration curves in clinical and environmental trace-metal analysis. High-resolution Raman spectrometers and echelle-grating array spectrographs push resolving power into the tens of thousands by using high diffraction orders ($m > 1$) combined with cross-dispersing prisms to separate overlapping orders, letting a single compact instrument cover the full visible-to-near-IR range simultaneously — a direct scaling-up of the $R = mN$ relation derived here. Interference filters remain standard in point-of-care clinical photometers, portable colorimeters, and fluorescence microscopy filter cubes, where only one or two fixed wavelengths are ever needed and a full scanning monochromator would be needless expense. In fiber-optic and array-detector spectrometers, the "monochromator" concept survives even without a moving slit: a fixed grating disperses light across a linear detector array (a photodiode array or CCD), and each detector pixel effectively behaves as its own tiny exit slit, with the pixel pitch and dispersion together setting the instrument's practical bandpass — the same reciprocal linear dispersion relationship derived above, applied pixel by pixel. Understanding this wavelength-selection machinery is the necessary bridge to the full instrument architecture treated in UV-Vis spectrophotometer instrumentation, and to understanding the instrumental causes of nonlinearity catalogued under deviations from Beer's law.

Ask the tutor: If I halve my monochromator's exit slit width to improve resolution, roughly how much do I expect my signal-to-noise ratio to change, and why doesn't resolving power simply double?

A monochromator converts polychromatic radiation into a beam whose wavelength content is confined to a narrow, continuously tunable band. Every monochromator, however built, has the same skeleton: an entrance slit that defines a point (line) source, a collimating mirror or lens that turns the diverging light from that slit into a parallel beam, a dispersing element (prism or grating) that spatially separates wavelengths by angle, a focusing mirror or lens that reimages the dispersed beam, and an exit slit that physically selects which small range of angles — hence which small range of wavelengths — is allowed to reach the detector. Rotating the dispersing element sweeps different wavelengths across the fixed exit slit; this rotation is the "wavelength dial."

Filters are the simplest wavelength selectors and require no moving parts. An absorption filter is a piece of colored glass or a dye solution that absorbs unwanted wavelengths and transmits a band, typically 30–250 nm wide (FWHM) — cheap, robust, but broad and lossy at the band edges. An interference filter exploits constructive interference in a thin dielectric spacer (often CaF₂ or MgF₂) sandwiched between semi-transparent metal or dielectric mirror layers, analogous to a miniature Fabry–Pérot cavity; light reflects back and forth, and only wavelengths satisfying

survive constructive reinterference and exit the far side, where $n$ is the spacer's refractive index, $d$ its thickness, $\theta$ the internal angle of incidence, and $m$ an integer order. Interference filters achieve bandpasses of 1–15 nm, are wavelength-fixed by construction, and are the workhorses of simple photometers and clinical colorimeters where only one or two wavelengths are ever needed.

Prism monochromators use the fact that a transparent medium's refractive index $n(\lambda)$ decreases with increasing wavelength (normal dispersion): blue light bends more than red on passing through a prism, fanning white light into a spectrum. The angular separation between two wavelengths depends on $dn/d\lambda$, which is largest in the ultraviolet and smallest in the near infrared — so a prism instrument's resolution is wavelength-dependent, typically far better in the UV/blue than in the red, and its dispersion is nonlinear across the spectrum (wavelengths are not evenly spaced along the focal plane). Prisms have essentially 100% throughput (no diffraction orders competing for light) and no higher-order overlap problems, but they are bulkier and costlier to manufacture with high optical quality than gratings for a given resolving power.

Diffraction gratings are now nearly universal in commercial UV-Vis, IR, and atomic spectrometers. A grating is a surface ruled (or holographically recorded) with thousands of closely, evenly spaced parallel grooves — typically 300 to 3600 grooves per millimeter. Each groove acts as a secondary source; light reflected (reflection grating) or transmitted (transmission grating) from adjacent grooves interferes, and constructive interference occurs only at angles satisfying the grating equation (derived below). Because the groove spacing $d$ is fixed by manufacture, a grating's angular dispersion is far more linear across wavelength than a prism's, and gratings can be blazed — grooves cut at a controlled facet angle — to concentrate diffracted intensity into one particular order at one particular wavelength, dramatically improving efficiency where it is needed.

The exit slit is the last and, for quantitative purposes, the most important element: it physically limits which range of diffraction angles — and therefore which range of wavelengths — reaches the detector. Widening the exit slit lets more light through (higher signal) but also lets a wider spread of wavelengths through (worse resolution). This single trade-off, throughput versus resolution, governs every monochromator design decision.

The wavelength at which constructive interference occurs from a reflection grating illuminated at incidence angle $i$ and observed at diffraction angle $r$ (both measured from the grating normal, on the same or opposite sides by sign convention) is

where:

• $m$ = diffraction order (integer: 1, 2, 3, …)
• $\lambda$ = wavelength (nm)
• $d$ = groove spacing = $1/(\text{grooves per mm})$
• $i$ = angle of incidence from the grating normal
• $r$ = angle of diffraction from the grating normal

Reading it: for fixed $i$, each wavelength diffracts into its own angle $r$ — this is dispersion. Rotating the grating (changing $i$ and $r$ together, mechanically linked) sweeps a chosen wavelength across a fixed exit slit; that is how the "wavelength dial" works. Differentiating at fixed incidence angle gives the angular dispersion,

where a larger $m$ or a finer groove spacing $d$ (more grooves/mm) spreads wavelengths over a wider angular range — better dispersion. The reciprocal linear dispersion, the practical quantity engineers quote in nm per mm of exit-slit travel, follows from the focal length $f$ of the focusing mirror:

where $x$ is position along the focal plane. A monochromator's reciprocal linear dispersion (often 1–4 nm/mm for compact instruments) multiplied by the exit slit width $w$ gives the effective bandpass,

The theoretical diffraction-limited resolving power of a grating — the finest wavelength difference it can in principle separate, ignoring aberrations and slit width — is

where:

• $R$ = resolving power (dimensionless)
• $\lambda$ = wavelength of interest
• $\Delta\lambda_{\min}$ = smallest resolvable wavelength difference
• $m$ = diffraction order
• $N$ = total number of illuminated grooves

Reading it: resolving power depends only on how many grooves are illuminated and in which order you work — not directly on groove density alone, since a wider grating illuminates proportionally more grooves at the same density. This is why spectroscopists quote grating size (mm illuminated) together with groove density.

Example 1 (Foundational): Where does 486.1 nm land, and how finely can a classroom grating resolve it? A student grating spectrometer uses a reflection grating ruled at 1200 grooves/mm, illuminated at normal incidence ($i = 0°$). The student wants to find the first-order diffraction angle for the hydrogen Balmer-β line at $\lambda = 486.1$ nm, and estimate the resolving power if 3.5 mm of the grating (measured along the ruled direction) is illuminated by the collimated beam. Groove spacing:

Diffraction angle, first order, normal incidence ($i=0$, so $\sin i = 0$):

Number of illuminated grooves:

Theoretical resolving power in first order:

So in principle this small grating, in first order, could resolve two lines only 0.12 nm apart near 486 nm — easily good enough to separate the H-β line from nearby solar Fraunhofer features, though whether the instrument actually achieves this depends on whether the exit slit is narrow enough not to swamp this diffraction-limited value (Example 3 addresses that trade-off explicitly). As a limiting check: if the illuminated width were doubled to 7.0 mm (all else fixed), $N$ doubles to 8400, $R$ doubles to 8400, and $\Delta\lambda_{\min}$ halves to about 0.058 nm — resolving power scales linearly with illuminated grating width, exactly as $R = mN$ predicts.

Example 2 (Applied): Choosing a spectral bandpass for a lead atomic absorption line Atomic absorption spectrometry (AAS) for lead uses the resonance line at 283.31 nm from a lead hollow-cathode lamp. Nearby, an unwanted lamp/background emission line sits at 283.31 nm ± interference from a neighboring line at 282.6 nm is not actually a concern here (lead's isolated resonance line has no strong neighbor within 1 nm), but the monochromator must still resolve the 283.31 nm line cleanly from a much weaker non-absorbing neon filler-gas line at 283.85 nm, a separation of 0.54 nm, to avoid a nonlinear calibration curve caused by co-measuring two lines with different absorption behavior. The instrument's grating has a reciprocal linear dispersion of $1.6$ nm/mm at 283 nm. What is the maximum exit slit width that keeps the effective bandpass at or below one-third of the 0.54 nm line separation — a common rule of thumb ensuring the unwanted line contributes negligible signal? Target bandpass:

Required slit width, from $\Delta\lambda \approx (d\lambda/dx)\,w$:

The analyst should set the slit no wider than about 0.11 mm (110 μm). Most commercial AAS instruments offer discrete slit settings (e.g., 0.2, 0.5, 1.0, 2.0 nm spectral bandpass rather than physical mm), so in practice the analyst would select the nearest available spectral bandpass setting at or below 0.18 nm — commonly the instrument's 0.2 nm setting is used and checked against a calibration curve for linearity, or a tighter 0.1–0.2 nm slit is chosen if the manufacturer's table lists it. As a limiting check: if the two lines were much farther apart — say 5.0 nm instead of 0.54 nm, as for well-isolated resonance lines in simpler spectra — the target bandpass would relax to $5.0/3 \approx 1.67$ nm, permitting $w \approx 1.04$ mm, a far more generous (higher-throughput, better signal-to-noise) slit setting. This illustrates the general principle: closely spaced spectral features force narrow slits and lower throughput, while isolated lines permit wide-open slits and higher sensitivity.

Example 3 (Exam/competition level): How much does bandpass distort a Beer's-law absorbance reading, and what slit avoids it? A UV-Vis instrument measures an analyte with a Gaussian absorption band centered at $\lambda_0 = 400.0$ nm, molar absorptivity peak $\varepsilon_0 = 1.20\times10^{4}$ M⁻¹cm⁻¹, and a natural bandwidth (FWHM of the absorption band) of $\Delta\lambda_{1/2} = 8.0$ nm. The sample is $2.50\times10^{-5}$ M in a 1.00 cm cell. The monochromator has a reciprocal linear dispersion of 3.2 nm/mm; the analyst can choose an exit slit width $w$.

(a) Compute the "true" (infinitely narrow bandpass) absorbance at the peak.

(b) Estimate the measured absorbance if the analyst carelessly opens the slit to $w = 2.0$ mm, giving an effective spectral bandpass of

Because $\Delta\lambda$ (6.4 nm) is comparable to the absorption band's natural FWHM (8.0 nm) rather than negligible compared to it, the transmittance measured is really an average of $T(\lambda) = 10^{-\varepsilon(\lambda)bc}$ across the passed bandpass, not the peak value alone — and since transmittance (not absorbance) is what physically averages linearly on the detector, the correct treatment is to average $T(\lambda)$, then convert back to absorbance:

where $S(\lambda)$ is the triangular slit function. Using a simplifying two-point estimate (the standard classroom approximation): approximate the band as effectively sampling $\varepsilon(\lambda)$ at $\lambda_0$ and at the bandpass half-width $\lambda_0 \pm \Delta\lambda/2 = 400.0 \pm 3.2$ nm, and treat the measured transmittance as the average of the three point transmittances weighted 1:2:1 (triangular slit approximation), since a monochromator with equal entrance and exit slit widths produces a triangular (not rectangular) effective slit function. For a Gaussian absorption band, $\varepsilon(\lambda) = \varepsilon_0 \exp\!\left[-4\ln 2 \left(\frac{\lambda-\lambda_0}{\Delta\lambda_{1/2}}\right)^2\right]$. At the wings, $\lambda - \lambda_0 = \pm 3.2$ nm:

giving absorbances $A_{\text{wing}} = (7.70\times10^3)(1.00)(2.50\times10^{-5}) = 0.1925$ and corresponding transmittances:

Triangular-weighted average transmittance (weights 1:2:1, total weight 4):

The wide-slit measurement reads $A = 0.243$ against a true peak absorbance of $0.300$ — an error of about $-19\%$, entirely an instrumental artifact of using too coarse a bandpass relative to the natural linewidth, and precisely the kind of systematic negative deviation from Beer's law addressed further when treating instrumental causes of nonlinearity.

(c) Find the maximum slit width keeping this error under 2%. A widely used engineering rule of thumb states that instrumental bandpass-induced absorbance error stays below about 1–2% when $\Delta\lambda \le 0.1 \times \Delta\lambda_{1/2}$. Applying it:

A slit at or below 0.25 mm (spectral bandpass ≤ 0.8 nm) keeps the deviation from true Beer's-law absorbance below the 2% engineering tolerance. As a check on the limiting case: if the natural band were far broader — say $\Delta\lambda_{1/2} = 80$ nm, typical of a broad charge-transfer band — the same 6.4 nm bandpass from part (b) would be less than 10% of the linewidth, and the wide-slit error would shrink to only a percent or two, confirming that bandpass distortion matters most for narrow spectral features (atomic lines, sharp vibronic bands) and is nearly negligible for broad, featureless absorption bands.

Henry Rowland (1848–1901), an American physicist at Johns Hopkins University, transformed grating spectroscopy from a delicate laboratory curiosity into a precision instrument industry. Rowland designed and built ruling engines capable of cutting gratings with thousands of grooves per millimeter across surfaces many centimeters wide, with spacing errors small enough to suppress "ghost" lines (spurious diffraction artifacts from periodic ruling errors) that had plagued earlier gratings. He also discovered the concave grating, a ruled grating on a spherical mirror surface that simultaneously disperses and focuses light without any additional lenses — eliminating chromatic aberration entirely and enabling the compact "Rowland circle" mounting still used in many modern spectrographs. Rowland's gratings, sold internationally from his Baltimore workshop, set the standard of dispersive precision for half a century and underpinned the era's most accurate measurements of atomic spectral line wavelengths, which in turn anchored the developing quantum theory of atomic structure.