Acousto-Optic and Magneto-Optic Devices

Send a laser through a transparent crystal while an ultrasonic wave races across it, and the beam splits, bends, and comes out shifted in colour by a few hundred megahertz. Wrap the same crystal in a magnet and a linearly polarized beam emerges with its plane twisted — and, uniquely among optical elements, it refuses to untwist when sent back through. These are the two families of field-controlled optics. Acousto-optics borrows from solid-state physics: the acoustic wave is a stream of phonons, and the photoelastic tensor $p_{ijkl}$ says how their strain reshapes the refractive index. Magneto-optics borrows from atomic physics: an axial field Zeeman-splits the electronic resonances, and that splitting, repackaged as a gyrotropic dielectric tensor, makes the two circular polarizations travel at different speeds. Both devices therefore do the same abstract thing — engineer the dielectric tensor $\varepsilon_{ij}$ with an external field — but with opposite symmetry. The acoustic grating is a real, moving, periodic modulation: reciprocal, and it diffracts, redirecting light and Doppler-shifting it by exactly the acoustic frequency. The magnetic gyration is an imaginary, antisymmetric off-diagonal term $\varepsilon_{xy}=-\varepsilon_{yx}$ : nonreciprocal, and it rotates polarization, breaking the time-reversal symmetry every ordinary lens and mirror obeys. That broken symmetry is the one ingredient that makes a true optical isolator possible — the component that lets a laser fire into a fibre without its own reflection firing back. Together these effects run the modern optics bench: pulse-picking and frequency-shifting beams for cold-atom work, steering and filtering wavelengths, and shielding every high-coherence source from back-reflections.

Key takeaways

An acoustic wave writes a travelling refractive-index grating through the photoelastic effect; that grating diffracts light, enabling beam deflection, spectral analysis, and megahertz frequency shifting.
The Klein–Cook parameter $Q$ sets the regime: the thick-grating Bragg regime ( $Q\gg1$ ) redirects light efficiently into one order; the thin-grating Raman–Nath regime ( $Q\ll1$ ) spreads many symmetric orders.
In the Bragg regime the deflection obeys $\sin\theta_B=\lambda f_a/2nv$ , and the diffracted beam is Doppler-shifted by exactly $\pm f_a$ because the grating moves at the sound speed.
Magneto-optic devices exploit the Faraday effect: an axial field rotates the polarization by $\beta=VBL$ (Verdet constant $V$ ) — a nonreciprocal rotation that adds, rather than cancels, on a return pass.
Faraday rotation builds the optical isolator and circulator, which protect lasers and route light one way in fibre and integrated systems.
Key idea: both devices are dielectric-tensor engineering — acousto-optics writes a moving index grating via strain (reciprocal, diffractive, frequency-shifting), magneto-optics a circular birefringence via a field (nonreciprocal, polarization-rotating) — and both diffraction efficiency and rotation angle scale as interaction length times a material figure of merit.

The core insight

Acousto-optics writes a moving, periodic index grating with strain — reciprocal, diffractive, frequency-shifting — while magneto-optics writes an antisymmetric circular birefringence with a magnetic field — nonreciprocal, polarization-rotating; diffraction efficiency and rotation angle both scale as (interaction length) × (material figure of merit).

Where this lives today

Any cold-atom laboratory is full of acousto-optic modulators: TeO $_2$ cells at 80–200 MHz switch cooling beams in well under a microsecond, shift them by a precise $\pm f_a$ to set the detuning from an atomic resonance, and, double-passed, tune that shift over tens of megahertz without the beam walking off — the standard way to servo a laser onto an atomic line. Sweep the drive instead and the same physics becomes a deflector painting a beam across dozens of spots, or, loaded with a broadband RF signal, a real-time optical spectrum analyzer mapping each RF component to its own angle; tunable filters extend it to hyperspectral imaging with no moving parts. On the magneto-optic side, the Faraday isolator is nearly invisible for how ubiquitous it is: a TGG rotator between crossed polarizers, or garnet films at telecom wavelengths, guards essentially every semiconductor and fibre laser with 30–60 dB of protection against back-reflections. Combine rotators with polarization beam splitters and you get an optical circulator, routing light one way from port 1 to 2 to 3 for full-duplex transmission on one fibre, add/drop with fibre Bragg gratings, and OTDR fault-finding. The open frontier is putting this nonreciprocity on a chip: silicon has no native magneto-optic response, so integrating a compact isolator — via bonded garnet films or magnet-free time-modulation schemes — remains one of the field's most pursued goals.

The physical picture

Picture the sound wave as a stack of moving mirrors — sheets of compressed, higher-index material spaced by one acoustic wavelength $\Lambda$ . Light reflects a little from each sheet, and when the path difference between successive sheets equals one optical wavelength the reflections add in phase and the beam is efficiently redirected. That is the Bragg condition, identical in form to X-ray diffraction from crystal planes, with $\Lambda$ as the lattice spacing. Because the mirrors march at the sound speed $v$ , every reflection is a Doppler bounce: the photon scatters off a phonon, absorbing energy $hf_a$ and momentum $\hbar\mathbf{K}$ , so it comes out both redirected and shifted in frequency by exactly $f_a$ — energy and momentum bookkeeping deliver the deflection and the frequency shift in one stroke.

Figure

An acoustic wave of frequency f_a = 200 MHz in fused silica (v ≈ 6×10³ m/s) writes a moving index grating of period Λ = v/f_a = 30 μm; a 633 nm beam incident at the Bragg angle θ_B ≈ 0.41° is diffracted into a single order and Doppler-shifted from f_i to f_i + f_a, while the undiffracted order passes straight through

For Faraday rotation, decompose the linear input into left- and right-circular light. The axial field makes the two hands see slightly different indices — it pushes the electronic resonances up for one sense and down for the other, the Zeeman shift seen through the dielectric response. The hands accumulate different phases, so their sum re-forms into linear light whose plane has rotated by half the phase difference. Reverse the beam and $\mathbf{B}$ still points the same way, so the fast and slow hands swap roles relative to propagation and the rotation adds instead of cancelling. That nonreciprocity is exactly what lets a rotator pass light one way and block its own reflection — the isolator drawn below.

Figure

A 45° Faraday rotator (TGG, ≈ 5.9 mm at B ≈ 1 T) sits between polarizers at 0° and 45°: forward light rotates 0° → 45° and passes the output polarizer, while a back-reflection rotates a further 45° → 90° and is extinguished by the input polarizer, giving ≈ 35 dB isolation — the rotation adds because B is fixed in the lab, it does not cancel

The unifying theme is that a small perturbation of the dielectric tensor redirects energy between optical modes. In acousto-optics it is a real, periodic modulation coupling different wavevectors $\mathbf{k}$ ; in magneto-optics an imaginary, antisymmetric off-diagonal term coupling the two polarizations. The first breaks translational periodicity and moves; the second breaks time-reversal symmetry and stands still.

The core equations

The acoustic wave sets the stage. A transducer driven at radio frequency $f_a$ launches sound of velocity $v$ , and the periodic strain writes an index grating of period one acoustic wavelength, whose grating vector $K$ the diffracting photon must borrow.

\Lambda=\frac{v}{f_a},\qquad K=\frac{2\pi}{\Lambda}=\frac{2\pi f_a}{v} \qquad (1)

where: $\Lambda$ — acoustic (grating) wavelength; $v$ — acoustic phase velocity; $f_a$ — acoustic (RF drive) frequency; $K$ — acoustic wavenumber. Raising $f_a$ makes a finer grating and a larger deflection. Efficient single-order deflection happens only at the Bragg angle, where reflections from successive wavefronts add in phase. Inserting equation (1) puts the internal condition $2\Lambda\sin\theta_B=\lambda/n$ entirely in terms of the drive frequency.

2\Lambda\sin\theta_B=\frac{\lambda}{n}\;\Longrightarrow\;\sin\theta_B=\frac{\lambda f_a}{2nv} \qquad (2)

where: $\theta_B$ — Bragg angle inside the medium; $\lambda$ — optical vacuum wavelength; $n$ — refractive index. The deflection is small and tunable: sweeping $f_a$ sweeps $\theta_B$ , steering the beam across many angular positions. Because the grating is not merely periodic but racing along at $v$ , each diffraction event is a Doppler bounce. Equivalently the photon absorbs or emits one phonon, so its frequency comes out shifted by exactly the acoustic frequency.

f_d=f_i\pm f_a \qquad (3)

where: $f_i,\;f_d$ — incident and diffracted optical frequencies; $+$ — deflection along the acoustic propagation (up-shift); $-$ — the opposite geometry (down-shift). The shift is exact and phase-coherent, making the modulator a precision frequency shifter. Whether the device gives one clean Bragg order or a fan of Raman–Nath orders is decided by a single dimensionless number. It compares the grating thickness to $\Lambda^2/\lambda$ and is called the Klein–Cook parameter.

Q=\frac{2\pi\lambda L}{n\Lambda^2} \qquad (4)

where: $L$ — interaction length (acoustic beam width traversed); $Q\gg1$ — thick-grating Bragg regime (one order); $Q\ll1$ — thin-grating Raman–Nath regime (many orders). Practical devices sit deep in the Bragg regime by making $L$ large. How strongly a crystal diffracts per watt of sound is captured by an acousto-optic figure of merit. It is a specific grouping of material constants, the number read off a table when choosing a crystal.

M_2=\frac{n^6 p^2}{\rho v^3} \qquad (5)

where: $p$ — the relevant elasto-optic (photoelastic) coefficient; $\rho$ — mass density; $v$ — acoustic velocity; $M_2$ — diffraction efficiency per unit acoustic power. The slow shear mode of TeO $_2$ has a very large $M_2$ (its low $v$ enters as $v^{-3}$ ), which is why it dominates commercial deflectors. Turning to magneto-optics, the defining relation is disarmingly simple. The polarization plane rotates by an angle proportional to the field and the path length, with the constant of proportionality a material property.

\beta=V\,B\,L \qquad (6)

where: $\beta$ — rotation angle of the polarization plane; $V$ — Verdet constant (rad T $^{-1}$ m $^{-1}$ ), material- and wavelength-dependent; $B$ — flux density along propagation; $L$ — path length. Terbium gallium garnet (TGG) has a large $V$ in the visible and near-infrared, the reason it fills isolators. Microscopically the rotation is circular birefringence. The two circular polarizations travel with different indices, and the linear plane rotates by half of their accumulated phase difference.

\beta=\frac{\pi L}{\lambda}\,(n_+-n_-) \qquad (7)

where: $n_+,\;n_-$ — indices for right- and left-circular polarization; $n_+-n_-$ — the magnetically induced circular birefringence. Equations (6) and (7) are one statement: $V$ packages the field dependence of $n_+-n_-$ into a measured number. The property that makes magneto-optics irreplaceable is nonreciprocity. Retracing the path does not undo the rotation but doubles it, because the axial field is fixed in the lab while the propagation direction flips.

\beta_{\text{double pass}}=2VBL\quad(\text{not }0) \qquad (8)

where: $\beta_{\text{double pass}}$ — net rotation after light retraces its path; $2VBL$ — the doubling, versus the exact cancellation of reciprocal optical activity. This factor of two, versus zero, is the whole reason an isolator works. For a deflector the useful figure is how many distinct spots it can resolve. It works out to a time–bandwidth product: the RF bandwidth times the acoustic transit time across the beam.

\delta\theta\sim\frac{\lambda}{D},\qquad N_{\text{spots}}=\frac{\Delta\theta}{\delta\theta}=\frac{\Delta f_a\,D}{v}=\Delta f_a\,\tau \qquad (9)

where: $\Delta f_a$ — RF bandwidth; $D$ — optical beam diameter across the aperture; $\tau=D/v$ — transit time, which also sets the access/switching time; $N_{\text{spots}}$ — resolvable angular positions. A big aperture buys more spots but a slower response.

Derivation ABragg diffraction from photon phonon momentum conservation

The goal is to obtain the Bragg deflection condition of equation (2) from conservation laws rather than geometry, so the deflection angle and the Doppler shift emerge from one diagram. We treat the acoustic wave as phonons of energy $hf_a$ and momentum $\hbar\mathbf{K}$ and read the angle off an isosceles momentum triangle.

Step 1. Impose momentum conservation, $\mathbf{k}_d=\mathbf{k}_i\pm\mathbf{K}$ . Why: diffraction from a periodic structure is momentum exchange in units of the grating vector, and the moving grating supplies exactly $\pm\mathbf{K}$ per scattered photon.

Step 2. Impose energy conservation; since the photon gained only $hf_a$ , $|\mathbf{k}_d|\approx|\mathbf{k}_i|=2\pi n/\lambda$ . Why: $f_a$ is at most a gigahertz against hundreds of terahertz, so the wavevector length changes by about $10^{-7}$ and the two photon vectors are equal to superb accuracy.

Step 3. Two equal-length vectors differing by $\mathbf{K}$ form an isosceles triangle, so $K=2|\mathbf{k}_i|\sin\theta_B$ . Why: the perpendicular from the apex onto $\mathbf{K}$ splits it into right triangles with common side $|\mathbf{k}_i|\sin\theta_B$ .

Step 4. Substitute $K=2\pi/\Lambda$ and $|\mathbf{k}_i|=2\pi n/\lambda$ , cancel $2\pi$ , and get $\sin\theta_B=\lambda f_a/2nv$ , while Step 2's energy balance is exactly the Doppler shift $f_d=f_i\pm f_a$ . Why: one triangle carries both the angle and the frequency shift, so deflection and colour change are two faces of one conservation law.

Go deeper · Derivation B · 1 minDiffraction efficiency by coupled mode theory

Derivation BDiffraction efficiency by coupled mode theory

The aim is to compute not just where the light goes but how much, deriving the power exchange between the undiffracted and first-order beams in a thick grating and showing the efficiency oscillates as $\sin^2$ of a coupling that grows as the square root of acoustic power. We keep only the two phase-matched waves and reduce Maxwell's equations to a pair of coupled envelope equations. The result $\eta=\sin^2(\kappa L)$ is the design equation for every modulator: full switching is reachable, and it fixes the drive power to get there.

Step 1. Write the permittivity as a static part plus the grating, $\varepsilon(\mathbf{r},t)=\varepsilon_0 n^2+\Delta\varepsilon\cos(Kz-\Omega t)$ , with $\Delta\varepsilon=2\varepsilon_0 n\,\Delta n$ . Why: to first order in the small strain the modulation is a single sinusoid, so only its fundamental $K$ matters for coupling.

Step 2. Expand the field in two plane waves, the incident $A_0$ and first order $A_1$ with $\mathbf{k}_1=\mathbf{k}_0+\mathbf{K}$ , dropping all others. Why: in the thick ( $Q\gg1$ ) grating only the Bragg-matched order is synchronous; every other order accumulates a rapid phase mismatch and averages away.

Step 3. Insert the ansatz and keep synchronous terms to get $\mathrm{d}A_0/\mathrm{d}x=-i\kappa A_1$ , $\mathrm{d}A_1/\mathrm{d}x=-i\kappa A_0$ , with $\kappa=\pi\Delta n/\lambda\cos\theta_B$ . Why: the grating couples the two amplitudes symmetrically — two resonantly coupled modes exchanging energy, exactly like a directional coupler.

Step 4. Solve with $A_0(0)=1,\,A_1(0)=0$ : $A_0=\cos\kappa x$ , $A_1=-i\sin\kappa x$ , so $\eta=|A_1(L)|^2=\sin^2(\kappa L)$ , and since $\Delta n\propto\sqrt{M_2 P_a/H}$ this is $\eta=\sin^2\!\big(\tfrac{\pi}{\lambda}\sqrt{M_2 L P_a/2H}\big)$ . Why: energy sloshes fully between the beams like a Rabi oscillation, reaching 100% at $\kappa L=\pi/2$ and rising linearly in $P_a$ at low drive.

Go deeper · Derivation C · 1 minFaraday rotation from the gyrotropic dielectric tensor

Derivation CFaraday rotation from the gyrotropic dielectric tensor

The purpose is to show that an axial field forces an antisymmetric imaginary off-diagonal element into the dielectric tensor, that its eigenmodes are the two circular polarizations with split indices, and that the resulting circular birefringence rotates a linear input by $\beta=VBL$ . We start from the classical electron driven by the optical field and the magnetic Lorentz force, extract the gyration term, diagonalize, and combine with equation (7). The route also exposes the nonreciprocity: the gyration is odd in $\mathbf{B}$ , not in the propagation direction.

Step 1. Drive a bound electron with the optical field and the static field, $m\ddot{\mathbf{r}}+m\omega_0^2\mathbf{r}=-e\mathbf{E}-e\dot{\mathbf{r}}\times\mathbf{B}$ , $\mathbf{B}=B\hat z$ . Why: the $\dot{\mathbf{r}}\times\mathbf{B}$ term couples the $x$ and $y$ motions, and this cross-coupling becomes the off-diagonal susceptibility.

Step 2. Solve to first order in $B$ and read off $\varepsilon_{xy}=-\varepsilon_{yx}=ig$ , gyration constant $g\propto B$ . Why: the magnetic coupling is linear in $B$ and, being velocity-dependent, enters as a purely imaginary antisymmetric term — the signature of a nonreciprocal medium.

Step 3. Diagonalize: the eigenvectors are the right- and left-circular polarizations with $n_\pm^2=\varepsilon_r\pm g$ , so $n_\pm\approx n\pm g/2n$ . Why: a tensor with equal diagonals and $\pm ig$ off-diagonals is diagonal in the circular basis, which is why circular polarizations are the natural modes in a magnetized medium.

Step 4. Feed the split indices into equation (7): $\beta=(\pi L/\lambda)(n_+-n_-)=\pi g L/\lambda n$ , and since $g\propto B$ this is $\beta=VBL$ . Why: Faraday rotation is the circular-birefringence rotation of equation (7) with the birefringence supplied by magnetic gyration, and because $g$ tracks $\mathbf{B}$ not the propagation direction, reversing the beam does not reverse $\beta$ — nonreciprocity.

Limiting cases

The behaviour splits along three axes: grating thickness, drive strength, and how close the field is to a magneto-optic resonance.

Case	Result	Meaning
Raman–Nath (thin grating)	$Q\ll1$	Light crosses only a few periods; many symmetric orders appear with amplitudes $\propto J_m$ (Bessel functions). Low-frequency, short-interaction modulators.
Bragg (thick grating)	$Q\gg1$	Only the zeroth and one first order survive, efficiency up to 100%. Practical deflectors, shifters, and tunable filters.
Weak-drive efficiency	$\eta\approx\dfrac{\pi^2 M_2 L P_a}{2\lambda^2 H}$	Below saturation the diffracted power is linear in acoustic power $P_a$ ( $H$ = transducer height) — a near-linear intensity control.
Faraday, far from resonance	$\beta=VBL$	$V$ is smooth and roughly frequency-independent; rotation strictly linear in $B$ and $L$ — the isolator regime.
Faraday, near resonance	$V\propto\dfrac{\mathrm{d}n}{\mathrm{d}\omega}$	Near an atomic line $V$ grows large and dispersive, enabling ultra-narrow atomic (FADOF) filters at the cost of steep wavelength dependence.

Worked examples

1 · A fused-silica Bragg cell — angle, colour, and regime

An acousto-optic modulator in fused silica ( $v\approx6\times10^3$ m/s, $n\approx1.46$ ) is driven at $f_a=200$ MHz and used with a 633 nm helium–neon beam. Find the grating period, the internal Bragg angle, the deflection seen in air, the frequency shift, and confirm it operates in the Bragg regime for a 2 mm interaction length.

Step 1. Grating period from equation (1): $\Lambda=v/f_a=6000/(2\times10^8)=3.0\times10^{-5}$ m $=\mathbf{30\ \mu m}$ .

Step 2. Internal Bragg angle from equation (2): $\sin\theta_B=\lambda f_a/2nv=(633\times10^{-9})(2\times10^8)/(2\times1.46\times6000)=7.23\times10^{-3}$ , so $\theta_B=\mathbf{0.41^\circ}$ inside. Refraction out of the crystal magnifies angles by roughly $n$ , so the separation between the two beams in air is about $2n\theta_B\approx\mathbf{1.2^\circ}$ — small, as always for these devices.

Step 3. Frequency shift from equation (3): $f_d=f_i\pm200$ MHz. Against the carrier $f_i=c/\lambda=4.74\times10^{14}$ Hz this is a fractional shift of only $4.2\times10^{-7}$ — tiny in relative terms, but a rock-stable, phase-coherent 200 MHz offset that atomic-physics experiments rely on.

Step 4. Regime check from equation (4) with $L=2$ mm: $Q=2\pi(633\times10^{-9})(2\times10^{-3})/[1.46\,(30\times10^{-6})^2]\approx\mathbf{6.1}$ . Since $Q\gg1$ this is comfortably Bragg — essentially all diffracted light lands in one order. Λ = 30 μm, θ_B = 0.41° inside (≈1.2° in air), shift ±200 MHz, Q ≈ 6.

2 · Designing a TGG Faraday isolator

You need a $45^\circ$ Faraday rotator at 633 nm using terbium gallium garnet, whose Verdet constant there is $V\approx134$ rad T $^{-1}$ m $^{-1}$ , in a permanent-magnet field of $B\approx1.0$ T. How long must the crystal be, what does a back-reflection do, and how does the length change at 1064 nm where $V\approx40$ rad T $^{-1}$ m $^{-1}$ ?

Step 1. Required rotation $\beta=45^\circ=\pi/4=0.785$ rad. From equation (6), $L=\beta/(VB)=0.785/(134\times1.0)=5.86\times10^{-3}$ m, so $L\approx\mathbf{5.9\ mm}$ — the scale of a real commercial visible-wavelength isolator.

Step 2. Forward, the input polarizer sets $0^\circ$ , the rotator adds $+45^\circ$ , and the output polarizer at $45^\circ$ transmits fully. A back-reflection re-enters at $45^\circ$ ; by nonreciprocity the rotator adds another $+45^\circ$ (equation 8), bringing it to $90^\circ$ , and the input polarizer at $0^\circ$ rejects it: $T_{\text{back}}=\cos^2(90^\circ)=\mathbf{0}$ ideally, 30–40 dB in practice per stage.

Step 3. At 1064 nm the Verdet constant is smaller, so the crystal must be longer: $L=0.785/(40\times1.0)=1.96\times10^{-2}$ m $\approx\mathbf{20\ mm}$ . A 45° isolator needs ≈ 5.9 mm of TGG at 633 nm but ≈ 20 mm at 1064 nm — the length tracks $1/V$ , and $V$ falls toward the infrared.

3 · Resolvable spots of a TeO₂ deflector

A tellurium dioxide deflector uses the slow shear mode ( $v\approx4200$ m/s), a 5 mm optical aperture, and a 50 MHz RF bandwidth. Estimate the acoustic transit time and the number of resolvable spots.

Step 1. Transit (access) time: $\tau=D/v=5\times10^{-3}/4200=1.19\times10^{-6}$ s $\approx\mathbf{1.2\ \mu s}$ — how long the pattern takes to cross the beam, hence the fastest the deflector can be repositioned.

Step 2. Resolvable spots from equation (9): $N=\Delta f_a\,\tau=(5\times10^7)(1.19\times10^{-6})\approx\mathbf{60}$ . About 60 spots with a 1.2 μs access time; a bigger aperture raises $N$ but slows the response, since $N=\Delta f_a\tau$ is fixed by the RF bandwidth.

Physics of the portrait

Two names dominate this material. Léon Brillouin foresaw that light would scatter from sound; but the older surprise — that magnetism could touch light at all — belongs to Michael Faraday, whose 1845 experiment first tied electromagnetism to optics and gave nonreciprocity its founding demonstration.

Origins · 1 minMichael Faraday and how this was discovered

Michael Faraday

1791–1867, London · classical electromagnetism and magneto-optics

Michael Faraday came to science with almost no formal mathematics — apprenticed as a bookbinder before becoming Davy's assistant at the Royal Institution — and reasoned through physical pictures of lines of force rather than equations, which made him the century's greatest experimentalist. By 1845 he was convinced light, electricity, and magnetism were one family and set out to find a magnetic effect on light. After many failures he passed polarized light through a piece of dense lead-borate "heavy glass" between the poles of a strong electromagnet, field along the beam. The plane of polarization rotated, and reversed when he reversed the field. This magneto-optical rotation, now the Faraday effect, was the first hard evidence that a magnetic field acts on light, and it carried the seed of nonreciprocity: because the rotation follows the field, not the beam direction, retracing the path doubles rather than undoes it. He could not explain it quantitatively — that waited for Maxwell and the electron — but the same effect he saw by eye now sits, as TGG between crossed polarizers, at the output of nearly every laser.

History

1845

— Michael Faraday observes magnetic rotation of the plane of polarization in heavy lead-borate glass, the first magneto-optical effect

1854

— Émile Verdet begins the systematic measurements that define the constant now bearing his name

1922

— Léon Brillouin predicts the scattering of light by thermally excited acoustic waves

1932

— Debye and Sears, and independently Lucas and Biquard, observe light diffracted by an ultrasonic wave

1967

— Klein and Cook classify the Bragg and Raman–Nath diffraction regimes through the parameter Q

Going further

Texts. Hecht, Optics, 5th ed., §8.10 (acousto-optic and magneto-optic effects) · Saleh & Teich, Fundamentals of Photonics, 3rd ed., Ch. 20 (Acousto-Optics) and Ch. 6 (Polarization Optics) · Foot, Atomic Physics, §5.5 and §7.6 (Zeeman effect and the Faraday-rotation context) · Loudon, The Quantum Theory of Light, 3rd ed. (photon–phonon scattering) · Demtröder, Laser Spectroscopy, Vol. 1 (acousto-optic modulators for laser control).

Provenance. The Bragg condition, Doppler shift, Klein–Cook parameter, $M_2$ , and the coupled-mode efficiency $\eta=\sin^2\kappa L$ are standard, checkable against Saleh & Teich Ch. 20 and Hecht §8.10; Faraday rotation $\beta=VBL$ , the gyrotropic tensor, and the $45^\circ$ -rotator isolator follow Hecht §8.10 and Foot. A verifier should recompute: (1) $\Lambda=v/f_a$ and $\sin\theta_B=\lambda f_a/2nv$ for the fused-silica cell (30 μm, 0.41°); (2) $Q\approx6$ at $L=2$ mm; (3) the TGG isolator length $L=\beta/VB\approx5.9$ mm at 633 nm, ≈20 mm at 1064 nm; (4) the double-pass factor of 2 for Faraday versus 0 for optical activity; (5) $N=\Delta f_a\tau\approx60$ . The TGG Verdet values (≈134 and ≈40 rad T $^{-1}$ m $^{-1}$ ) are the widely quoted room-temperature figures, used only illustratively.

Prerequisites. Waveguide Modes and the Wave Equation · Coupled-Mode Theory and Directional Couplers · Polarization-Mode Dispersion and Birefringent Fibers. Leads to. Integrated Photonic Waveguides and Circuits · Microring Resonators and Optical Cavities on Chip.

Ask the tutor: Ask me to work through why the same crystal that deflects a beam also shifts its frequency by exactly the RF drive, or how a 45° Faraday rotator turns a back-reflection into a blocked 90° beam.

Why does the diffracted beam change colour at all, and by exactly how much?

Because the index grating is racing along at the acoustic velocity, so reflecting off it is a Doppler process — equivalently, the photon absorbs or emits one phonon of energy $hf_a$ . Its frequency therefore shifts by precisely $f_a$ , the sign set by whether the beam is deflected along or against the sound. The shift equals the RF drive exactly and stays phase-locked to it, which is why modulators double as extremely stable frequency shifters. Relative to the optical carrier it is only about $10^{-7}$ , but in absolute, coherent terms it is a precision offset that atomic and metrology experiments depend on.

What is the difference between the Bragg and Raman–Nath regimes, practically?

Thick grating versus thin grating, sorted by the Klein–Cook parameter $Q=2\pi\lambda L/n\Lambda^2$ . When $Q\gg1$ the light crosses many acoustic periods and only one order is phase-matched, so essentially all the diffracted power lands in a single beam that can reach 100% efficiency — the Bragg regime, where every practical deflector, shifter, and tunable filter lives. When $Q\ll1$ the grating acts like a thin phase mask, giving a symmetric fan of orders with Bessel-function amplitudes — the Raman–Nath regime. Designers push $Q$ large with a long interaction length and a high drive frequency (small $\Lambda$ ).

Why is nonreciprocity such a big deal — can't I just use two polarizers?

No arrangement of reciprocal optics can isolate, because an isolator must do two contradictory things: pass a beam one way and block the same beam returning along the same path. With only reciprocal elements — lenses, waveplates, even optically active sugar cells — any back-reflection retraces its steps and comes out where it started. The Faraday rotator escapes this because its rotation is tied to a lab-fixed magnetic field, not to the propagation direction, so forward and backward passes rotate the same way and the return is turned to $90^\circ$ and rejected. Breaking time-reversal symmetry is the non-negotiable ingredient.

Why does tellurium dioxide dominate acousto-optic devices?

Because of how the figure of merit $M_2=n^6p^2/\rho v^3$ is built: it rewards a high index (sixth power) and a large photoelastic coefficient, but above all a low acoustic velocity, which enters as $v^{-3}$ . TeO $_2$ has a slow shear mode far slower than fused silica's 6000 m/s and an index near 2.2, so its $M_2$ is enormous, giving strong diffraction for modest RF power. The same slow velocity means slower switching, so TeO $_2$ is chosen where efficiency matters more than speed, and faster materials where the reverse holds.

Does the Verdet constant depend on wavelength, and why does it matter?

Yes, strongly. Away from resonances $V\propto\mathrm{d}n/\mathrm{d}\omega$ , so it grows toward the blue and falls toward the infrared — for TGG about 134 rad T $^{-1}$ m $^{-1}$ at 633 nm but only about 40 at 1064 nm. That is why the same $45^\circ$ isolator needs roughly 5.9 mm of crystal in the visible but around 20 mm at 1064 nm: with $\beta=VBL$ , a smaller $V$ must be paid for in length or field. Near an atomic resonance $V$ becomes very large and sharply dispersive — a nuisance for broadband isolators but the basis of ultra-narrow atomic (FADOF) filters.

Can a single Faraday isolator give arbitrarily high isolation?

Not really — one stage typically delivers 30–40 dB, limited by finite polarizer extinction, imperfect $45^\circ$ rotation, thermal drift of $V$ under high power, and stress birefringence. Because the rotation drifts with temperature and wavelength, the return beam never sits at exactly $90^\circ$ , and a little light leaks through $\cos^2$ of the residual angle. When more is needed — for narrow-linewidth or high-power lasers acutely sensitive to feedback — two rotator stages are cascaded to reach 60 dB or more, with heat load a real concern since absorbed power both shifts $V$ and distorts the crystal.