Blazed Gratings

An ordinary ruled grating has a maddening flaw: shine white light on it and most of the power marches straight through, undeviated, into the zeroth order — the one order that carries no spectrum at all. All the colour, all the dispersion, lives in the faint first and second orders off to the side, each holding only a sliver of the light. For a century this was the price of a grating: high resolution bought with terrible efficiency. A blazed grating breaks the bargain. In place of flat, symmetric grooves it carries a sawtooth profile — a staircase of tiny mirror facets, each tilted at a chosen blaze angle γ. The regular spacing of the grooves still fixes where the diffraction orders appear, exactly as the grating equation demands; but the tilt of each little mirror controls how bright those orders are. It aims the specular glint of a single facet away from the useless zeroth order and lands it on a dispersive order, so that at a design wavelength that order lights up — drinking eighty or ninety percent of the light — while the zeroth order goes dark. The physics is a product of two patterns you already know. The many grooves act as an N-slit interferometer, producing a sharp comb of orders; a single facet acts as a small aperture, producing a broad single-slit diffraction envelope. The brightness is the comb times the envelope, and blazing slides the envelope sideways until its peak sits on the comb tooth you want. Push the blaze steep and the order high and you reach the echelle — coarse, steeply cut gratings whose enormous angular dispersion feeds every high-resolution spectrograph on Earth and in orbit.

Key takeaways

A blazed (echelette) grating replaces symmetric grooves with tilted mirror facets, so the single-facet envelope peaks on a chosen dispersive order instead of the undispersed zeroth order.
The grating equation $d(\sin\theta_i+\sin\theta_m)=m\lambda$ fixes where the orders are; the blaze angle γ fixes how energy is shared among them — two independent knobs.
In the Littrow mount incidence, diffraction, and facet normal coincide ( $\theta_i=\theta_m=\gamma$ ), giving $2d\sin\gamma=m\lambda_B$ and blaze wavelengths scaling as $\lambda_B\propto1/m$ .
Steep facets at high order (the echelle, $\tan\gamma=2$ , γ = 63.4°) maximise the angular dispersion $\mathcal D=m/(d\cos\theta)$ and resolving power $R=mN$ — at the cost of a tiny free spectral range.
Key idea: the diffracted intensity is the single-facet envelope times the N-groove comb; blazing tilts the facets to drag the envelope peak off the zeroth order and onto a dispersive order at the design wavelength, where angular dispersion earns its keep.

The core insight

A grating's far-field intensity factorises into a broad single-facet Fraunhofer envelope times a sharp N-groove interference comb. The comb positions are set by the groove spacing $d$ alone; the envelope position by the facet tilt γ alone. Blazing chooses γ so the envelope maximum lands on a non-zero comb tooth, moving light from the undispersed zeroth order into a dispersive one — where the angular dispersion $\mathcal D=m/(d\cos\theta)$ makes it useful.

Where this lives today

Blazed reflection gratings are the dispersing heart of nearly every modern spectrometer. A grating monochromator picks a blaze wavelength near the centre of its band and rotates the grating to scan, the blaze keeping throughput high across roughly $2\lambda_B/3$ to $2\lambda_B$ — which is why catalogue gratings are sold by blaze wavelength. In tunable lasers, a Littrow-mounted blazed grating is the wavelength-selective end mirror, feeding its strong retro-diffracted first order back into the cavity; rotating it tunes the wavelength. At the high-resolution frontier the echelle rules. Cross-dispersed echelle spectrographs — HARPS and ESPRESSO at ESO, HIRES on Keck — pour starlight onto a coarse, steeply blazed grating worked at order 50 to 100, then fan the stacked orders into a two-dimensional format one detector captures at once. The result is resolving power in the tens of thousands to over a hundred thousand, precise enough to weigh planets by the radial-velocity wobble they impose on their stars — a measurement resting entirely on angular dispersion delivered efficiently by a blaze.

The physical picture

Picture a flight of tiny mirror steps. Each step reflects light specularly in a direction set only by its own tilt; the whole periodic staircase, being a grating, is only allowed to send light into the discrete order directions of the grating equation. Blazing orients the mirror steps so their common specular glint points along one of those allowed directions — then that order blazes bright (the origin of the word) while its neighbours starve. Two patterns are multiplied together. One facet is a narrow aperture, throwing a broad single-slit glow — the envelope. The many facets form an N-slit interferometer, throwing a fine picket-fence of razor-sharp orders — the comb. Blazing neither sharpens nor moves the picket-fence — the grating equation still nails each order to its angle — it only slides the broad glow sideways until its bright centre sits on the picket you care about. That envelope-peak steering is the whole trick.

In the Littrow mount you look straight back along your own line of sight: light retro-diffracts into the order you want because the facet faces you squarely — the most efficient way to run a grating, and the reason Littrow blazed gratings serve as the wavelength-selecting mirror in tunable lasers.

The zeroth order is the enemy of spectroscopy — undispersed, every wavelength piled at one angle, and exactly where a flat grating dumps its light. Every design choice in a blazed grating exists to starve it and feed a dispersive order.

The core equations

Because a grating is a periodic array of grooves, the path difference between light scattered from adjacent grooves must be a whole number of wavelengths to interfere constructively. Adding the incident and diffracted path differences $d\sin\theta_i$ and $d\sin\theta_m$ on the same side of the normal gives the grating equation in its reflection form — the law inherited from Grating Equation that fixes every order angle.

d\,(\sin\theta_i + \sin\theta_m) = m\lambda \qquad (1)

where: $d$ — groove spacing (period); $\theta_i$ — angle of incidence from the grating normal; $\theta_m$ — diffraction angle of order $m$ ; $m$ — integer order; $\lambda$ — wavelength. Reading it physically: only the spacing $d$ appears, so the groove pitch alone decides the set of allowed order directions. Nothing here mentions the groove shape — which is why the grating equation is silent about brightness. Brightness is a second, independent statement. In the Littrow mount one sets the incidence angle, the diffraction angle, and the facet normal all equal, $\theta_i=\theta_m=\gamma$ : light strikes each facet at normal incidence, reflects straight back, and that retro-reflected beam must itself be an allowed order. Substituting $\theta_i=\theta_m=\gamma$ into $(1)$ gives the blaze condition.

2d\sin\gamma = m\lambda_B, \qquad \lambda_B(m)=\frac{2d\sin\gamma}{m} \qquad (2)

where: $\gamma$ — blaze angle (facet tilt from the grating plane); $\lambda_B$ — blaze wavelength, the wavelength maximally concentrated into order $m$ ; $d,m$ — spacing and design order. The reading: a single grating has a comb of blaze wavelengths, one per order, in the ratio $\lambda_B(1):\lambda_B(2):\lambda_B(3)=1:\tfrac12:\tfrac13$ . A grating blazed for 500 nm in first order is simultaneously blazed for 250 nm in second — its orders overlap, which is why blazed gratings ride together with the order-sorting ideas of Free Spectral Range. The full pattern is the product of two factors you already own. The N grooves give the array factor of an N-slit interferometer; the finite width of one facet gives a single-slit Fraunhofer envelope — the Single-Slit Diffraction envelope, recalled here as the facet's own diffraction. The measured intensity is comb times envelope.

I(\theta) = I_0\;\underbrace{\left[\frac{\sin(N\delta/2)}{\sin(\delta/2)}\right]^2}_{\text{N-groove comb}}\;\underbrace{\left[\frac{\sin\beta}{\beta}\right]^2}_{\text{single-facet envelope}} \qquad (3)

where: $N$ — illuminated grooves; $\delta=(2\pi d/\lambda)(\sin\theta_i+\sin\theta)$ — phase between adjacent grooves; $\beta=(\pi b/\lambda)(\sin\theta-\sin\theta_\text{spec})$ — half the phase across one facet of width $b$ , from the facet-specular direction $\theta_\text{spec}$ ; $I_0$ — single-facet forward scale. Reading: the comb spikes to height $N^2$ wherever $\delta=2\pi m$ (the grating equation); the envelope multiplies each spike by $[\sin\beta/\beta]^2$ , and blazing sets $\beta=0$ at the wanted order. The reason the dispersive orders are worth chasing is angular dispersion, the portrait of this article. Differentiating the grating equation at fixed incidence shows how far apart two nearby wavelengths are fanned in angle.

\mathcal D \equiv \frac{\mathrm d\theta_m}{\mathrm d\lambda} = \frac{m}{d\cos\theta_m} \qquad (4)

where: $\mathcal D$ — angular dispersion (angular separation per unit wavelength); $m$ — working order; $d$ — spacing; $\theta_m$ — diffraction angle. The reading: dispersion grows with order $m$ and fineness $1/d$ and diverges as $\theta_m\to90°$ — and a blazed grating is the tool that actually delivers light into that high- $m$ , high-θ corner, which an unblazed grating leaves dark. Finally, the sharpness of each order sets the chromatic resolving power. Because the principal maxima of the comb have angular half-width $\lambda/(Nd\cos\theta)$ , the Rayleigh criterion applied to the grating gives the compact result inherited from Grating Equation and its resolving-power sibling.

R \equiv \frac{\lambda}{\Delta\lambda} = mN \qquad (5)

where: $R$ — resolving power; $\Delta\lambda$ — smallest resolvable wavelength difference; $m$ — order; $N$ — illuminated grooves. The reading: resolution is set by the total interfering groove count $mN$ , so both high order (echelle) and many grooves (long rulings) buy it — and blazing is what makes high- $m$ operation photon-efficient enough to use.

Go deeper · Derivation B · 2 minEnvelope times comb, and how blazing steers the peak

Derivation BEnvelope times comb, and how blazing steers the peak

This card proves the factorisation that underlies everything: the far field of $N$ identical, equally spaced facets is a single-facet Fraunhofer amplitude times an $N$ -source interference sum, so the intensity is one envelope times one comb. Establishing it cleanly is what lets us say precisely — not loosely — that blazing changes brightness without moving a single order, because the two factors depend on wholly separate parameters.

Step 1. Write the total far-field amplitude as a sum over facets. Each facet, displaced by $nd$ , radiates the same complex Fraunhofer amplitude $A(\theta)$ of one facet, carrying an extra phase $n\delta$ from its displacement: $U(\theta)=A(\theta)\sum_{n=0}^{N-1}e^{-in\delta}$ , with $\delta=(2\pi d/\lambda)(\sin\theta_i+\sin\theta)$ . Why: identical grooves means every facet contributes the identical shape $A(\theta)$ ; only the geometric phase $n\delta$ distinguishes them, so the facet amplitude factors cleanly out of the sum.

Step 2. Close the geometric series: $\sum_{n=0}^{N-1}e^{-in\delta}=e^{-i(N-1)\delta/2}\,\dfrac{\sin(N\delta/2)}{\sin(\delta/2)}$ . The leading phase is irrelevant to intensity. Why: this is the standard N-slit array factor; it spikes to magnitude $N$ wherever $\delta=2\pi m$ , which is the grating equation $(1)$ re-derived as the condition for all facets to radiate in phase.

Step 3. Take the squared modulus: $I(\theta)=|A(\theta)|^2\,[\sin(N\delta/2)/\sin(\delta/2)]^2$ , and identify $|A(\theta)|^2=I_0[\sin\beta/\beta]^2$ for a facet of width $b$ , a sinc² centred on the facet-specular direction $\theta_\text{spec}$ where $\beta=0$ . Why: this is equation $(3)$ — the interference term is the comb, sharp because $N$ is large; the facet term is the envelope, broad because one facet is narrow; the comb depends only on $d$ , the envelope only on $b$ and γ.

Step 4. Blaze the grating: choose the tilt γ so the specular direction $\theta_\text{spec}$ hits the order- $m$ comb tooth. Then $\beta=0$ there, the envelope contributes its peak value of unity, and that order drinks the light while its neighbours fall on the envelope's shoulders. Why: the comb positions never moved, so blazing is pure redistribution — it drags the envelope maximum from the zeroth order onto a dispersive one, turning a few-percent order into an eighty-to-ninety-percent one. *Verified: the geometric-series closure $|\sum e^{-in\delta}|^2=[\sin(N\delta/2)/\sin(\delta/2)]^2$ symbolically, the principal height $\to N^2$ , and the envelope peak $[\sin\beta/\beta]^2\to1$ at $\beta=0$ in verify_aq-optics-blazed-gratings.py.*

Limiting cases

Case	Result	Meaning
Unblazed, $\gamma\to0$	envelope peak at $\theta_m=-\theta_i$ (specular = $m=0$ )	flat facets glint into the undispersed zeroth order; the grating disperses but wastes nearly all the light
Littrow blaze	$2d\sin\gamma=m\lambda_B$ , $\theta_i=\theta_m=\gamma$	design order retro-reflected; scalar efficiency approaches unity, real gratings reach 80–90%
Order scaling	$\lambda_B(m)=2d\sin\gamma/m$	one facet blazes $1:\tfrac12:\tfrac13$ across orders; harmonics overlap and need sorting
Echelle, $\tan\gamma=2$ , $m\gg1$	huge $\mathcal D=m/(d\cos\gamma)$ and $R=mN$	coarse, steep grating at high order; enormous dispersion, but $\text{FSR}=\lambda/m$ tiny — cross-disperse
Scalar breakdown, $d\sim\lambda$	comb × envelope fails	groove period near the wavelength needs vector electromagnetic theory; polarisation-dependent Wood's anomalies appear

Common trapBlazing changes where the orders are

The single deadliest confusion is to think the blaze angle moves the diffraction orders. It does not. The grating equation $d(\sin\theta_i+\sin\theta_m)=m\lambda$ depends only on the groove spacing $d$ ; it fixes the order angles rigidly, and a mis-blazed grating obeys exactly the same equation as a perfectly blazed one. What the blaze angle controls is the envelope — the single-facet glow that multiplies the comb — and hence how the fixed orders share the light. Blazing is redistribution among orders already nailed in place, not relocation. The corollary: tilt the facets to change brightness; change $d$ , $m$ , or the mount to change dispersion $\mathcal D=m/(d\cos\theta)$ , which the blaze angle does not touch.

Worked examples

1 · Blazing a 600 l/mm grating for the visible (confidence builder)

A reflection grating is ruled with 600 lines per millimetre. At what blaze angle γ does it concentrate 500 nm light into first order in the Littrow mount, and what wavelength does the same facet blaze in second order?

Step 1. Find the groove spacing: $d=1/600$ mm $=1666.67$ nm. This is the anchor grating for the whole cluster. The blaze condition in Littrow is $2d\sin\gamma=m\lambda_B$ .

Step 2. Solve for the first-order blaze angle at $\lambda_B=500$ nm: $\sin\gamma=\lambda_B/(2d)=500/(2\times1666.67)=0.150$ , so $\gamma=\arcsin(0.150)=\mathbf{8.63^\circ}$ . Each facet is tilted a mere eight and a half degrees — enough to swing the envelope glint from $m=0$ onto $m=1$ .

Step 3. Read off the second-order blaze wavelength. Since $\lambda_B\propto1/m$ , the same facet blazes $\lambda_B(2)=500/2=\mathbf{250\ nm}$ in second order (and $166.7$ nm in third). These orders overlap the first, so a filter must block the unwanted harmonics.

Step 4. Check the forward direction: $2d\sin(8.63°)=500.2$ nm, recovering the design wavelength. A 600 l/mm grating blazes 500 nm in first order at γ = 8.63°, and the same facet blazes 250 nm in second order.

2 · An echelle for high-resolution spectroscopy (exam level)

A coarse echelle grating has 79 lines per millimetre and a steep R2 blaze, $\tan\gamma=2$ . Working near 500 nm, find the blaze order, the angular dispersion, and compare it with the 600 l/mm grating of Example 1. Why does such a grating need a cross-disperser?

Step 1. Geometry first: $\tan\gamma=2\Rightarrow\gamma=\arctan(2)=\mathbf{63.43^\circ}$ , and $d=1/79$ mm $=12.66$ µm — coarse grooves, steep facets, the R2 echelle signature.

Step 2. Find the Littrow blaze order at 500 nm from $2d\sin\gamma=m\lambda_B$ : $m=2d\sin\gamma/\lambda=2\times12658\times0.8944/500=\mathbf{45}$ (rounding to the nearest whole order, which blazes 503 nm exactly). The echelle lives at very high order — the opposite of a first-order grating.

Step 3. Compute the angular dispersion at $\theta_m=\gamma$ : $\mathcal D=m/(d\cos\gamma)=45/(12.66\times10^{-6}\times0.4472)=7.95\times10^{6}$ rad/m $=\mathbf{7.95\ mrad/nm}$ . Against Example 1's $0.649$ mrad/nm this is a factor of 12.3 more dispersion — the echelle's whole reason for being. Its resolving power $R=mN$ climbs by the same factor of $m$ .

Step 4. See the catch: the free spectral range is $\Delta\lambda_\text{FSR}=\lambda/m=500/45=\mathbf{11.1\ nm}$ , a tiny clean window before order 45 overlaps 44 and 46. The R2 echelle reaches 7.95 mrad/nm and enormous $R=mN$ at $m=45$ , but its 11 nm free spectral range forces a cross-disperser to separate the stacked orders — the architecture of every echelle spectrograph.

Physics of the portrait

Angular dispersion is the quantity a spectroscopist actually spends. It is the lever arm that converts a wavelength difference into an angular one, and a blazed grating is the machine built to make that lever both long and bright.

Origins · 1 minangular dispersion and how this was discovered

angular dispersion

the spectroscopist's lever — dθ/dλ = m ⁄ (d cosθ) · dispersive optics & spectroscopy

Angular dispersion, $\mathcal D=\mathrm d\theta_m/\mathrm d\lambda=m/(d\cos\theta_m)$ , measures how widely an instrument fans neighbouring colours. A prism disperses by the slow variation of refractive index and rarely beats fractions of a degree across the visible; a grating disperses by geometry, and by climbing to high order $m$ and steep angle θ it does far more. The blazed grating made this practical: dispersion lives in the high-order corner where $\mathcal D$ is large, but that corner is dark on an ordinary grating whose glint sits on the useless zeroth order. Robert W. Wood, at Johns Hopkins around 1910, was the first to rule controlled sawtooth "echelette" gratings whose tilted facets deliberately threw light into a chosen dispersive order — turning angular dispersion from a faint possibility into a bright, usable one. George Harrison's ruling engines at MIT in the late 1940s pushed the idea to its extreme in the echelle. Every high-resolution spectrograph now earns its resolution from angular dispersion delivered by a blaze.

History

1898

— Albert Michelson builds the echelon, a stack of glass plates reaching very high interference order

1910

— Robert W. Wood rules the first controlled blazed (echelette) gratings, concentrating light into a chosen order

1949

— George Harrison introduces the echelle grating, coarse and steeply blazed for very high order

1955

— Harrison and George Stroke bring the MIT ruling engine under interferometric servo control

Going further

Texts. Hecht, Optics, 5th ed., §10.2 (diffraction gratings, blazing) · Saleh & Teich, Fundamentals of Photonics, 3rd ed. (diffraction gratings) · Demtröder, Laser Spectroscopy, Vol. 1, ch. 4 (spectrometers, blazed gratings) · Foot, Atomic Physics, appendix on gratings and resolving power · Loudon, The Quantum Theory of Light (cavity gratings, mode selection).

Provenance. Equations are standard: the grating equation, the Littrow blaze condition $2d\sin\gamma=m\lambda_B$ , the angular dispersion $m/(d\cos\theta_m)$ , the resolving power $R=mN$ , and the comb × envelope factorisation appear in Hecht (Optics, 5th ed., §10.2), Saleh & Teich, and Demtröder (Laser Spectroscopy, Vol. 1, ch. 4). A verifier should recompute the worked check ( $\sin\gamma=\lambda_B/2d\Rightarrow\gamma\approx8.6°$ for 600 lines/mm, 500 nm) and confirm the harmonic order scaling. No non-standard numerical values are asserted. Every number in this article is reproduced from scratch in verify_aq-optics-blazed-gratings.py, including the echelle order $m=45$ and its $7.95$ mrad/nm dispersion.

Prerequisites. Fraunhofer Diffraction Regime · Single-Slit Diffraction · Huygens–Fresnel Principle · Rayleigh Resolution Criterion. Leads to. Grating Spectrometers and Resolving Power · Littrow Laser Tuning · Echelle Gratings.

Does blazing change the angles at which the orders appear?

No — and this is the point students most often miss. The order angles come entirely from the grating equation $d(\sin\theta_i+\sin\theta_m)=m\lambda$ , which contains only the groove spacing $d$ . Tilt the facets all you like and every order stays exactly where it was. What the facet tilt changes is the single-facet diffraction envelope that multiplies the sharp order comb — it decides how the fixed orders share the light. Blazing is redistribution of brightness among stationary orders, never relocation of them.

Why the Littrow mount specifically?

Because it makes the facet do double duty. In Littrow the incidence angle, the diffraction angle, and the facet normal all coincide, so light hits each facet at normal incidence and retro-reflects straight back into the very order you are collecting. That is the geometry of maximum efficiency — the facet faces the light squarely — and it also gives the cleanest blaze condition, $2d\sin\gamma=m\lambda_B$ . It is exactly why a Littrow-mounted blazed grating makes an ideal wavelength-selective end mirror for a tunable laser cavity.

Can one grating really be efficient across the whole visible?

Only over a band, not the whole range at once. Efficiency follows the single-facet envelope, which peaks at $\lambda_B$ and falls off to either side; a good rule of thumb is that first-order efficiency stays high from roughly $2\lambda_B/3$ to $2\lambda_B$ . A grating blazed at 500 nm thus serves well from about 330 nm to 1 µm in first order, but no single grating is optimal everywhere — hence the catalogue of blaze wavelengths.

What actually makes an echelle special?

Two choices pushed to the extreme: coarse grooves and a steep blaze, worked at very high order. A steep R2 facet ( $\tan\gamma=2$ , $\gamma=63.4°$ ) concentrates light into orders of 50 to 100, and at such high $m$ the angular dispersion $m/(d\cos\theta)$ and resolving power $mN$ are enormous. The penalty is a free spectral range $\lambda/m$ so small that dozens of orders overlap, so an echelle is always paired with a cross-disperser that fans them apart onto a two-dimensional detector.

When does the comb-times-envelope picture break down?

When the groove period $d$ shrinks toward the wavelength. The scalar factorisation assumes the facet is many wavelengths wide and that light scatters like a simple aperture; once $d\sim\lambda$ that assumption fails. The efficiency then becomes strongly polarisation-dependent and shows sharp features — Wood's anomalies — near the angles where an order becomes evanescent. Predicting efficiency there requires full vector electromagnetic grating theory, not the scalar envelope. For ordinary visible gratings with $d$ well above a wavelength, the scalar picture is excellent.

How is angular dispersion different from resolving power?

They answer different questions. Angular dispersion $\mathcal D=m/(d\cos\theta)$ says how far apart in angle two given wavelengths land — the plate scale, millimetres per nanometre on the detector. Resolving power $R=mN$ says how close two wavelengths can be and still be told apart, set by the sharpness of each order and hence the total groove count. You can have high dispersion with poor resolving power or the reverse; a good spectrograph needs both, and a blaze to make them bright.