The Pockels Effect: Switching Light With Voltage in Picoseconds

The core physics: an index ellipsoid that tilts with voltage

Light propagates through a crystal according to its impermeability tensor ηᵢⱼ = ε₀(∂Eᵢ/∂Dⱼ), which defines the optical index ellipsoid (the indicatrix): Σᵢⱼ ηᵢⱼ xᵢxⱼ = 1, where the ellipsoid's semi-axes are the principal refractive indices. In an uniaxial crystal these axes are nₒ (ordinary) and nₑ (extraordinary). The Pockels effect says that applying an electric field E changes the impermeability linearly:

• Δηᵢ = Σⱼ rᵢⱼ Eⱼ, equivalently Δ(1/n²)ᵢ = Σⱼ rᵢⱼ Eⱼ, with i = 1…6 (contracted Voigt notation) and j = x, y, z.
• The coefficients rᵢⱼ form the electro-optic tensor, measured in metres per volt — in practice picometres per volt (pm/V).
• Because Δη is linear in E, reversing the field reverses the index change. This is exactly why the effect is called the linear electro-optic effect.

For the technologically dominant crystal lithium niobate (LiNbO₃, point group 3m), the largest coefficient is r₃₃ ≈ 30.8 pm/V, coupling a z-directed field to light polarized along z. Applying E_z shifts the extraordinary index by Δnₑ = −½ nₑ³ r₃₃ E_z. That cubic-in-n prefactor is the reason high-index crystals win: the response scales as n³.

Why only non-centrosymmetric crystals qualify

The linearity has a strict symmetry consequence. Consider a crystal with a center of inversion (centrosymmetric). Inversion maps E → −E and, by symmetry, leaves the index ellipsoid unchanged. But the Pockels term predicts Δη(−E) = −Δη(E). The only value satisfying both "unchanged" and "sign-flipped" is zero. Therefore:

• In any of the 11 centrosymmetric crystal classes, every rᵢⱼ = 0 identically. No Pockels effect — full stop.
• Only 20 of the 21 non-centrosymmetric classes can host a nonzero electro-optic tensor (class 432 is the lone exception, where symmetry still forces r = 0).
• This is the same symmetry gate that governs piezoelectricity and second-harmonic generation — all three are odd-rank tensor responses that vanish under inversion. A good Pockels crystal is usually piezoelectric and frequency-doubling too.

The physical picture: the index change is really a distortion of the electron cloud's polarizability. In a centrosymmetric lattice, pushing the electrons one way is exactly mirrored by pulling them the other, so the leading response is quadratic (that's the ubiquitous Kerr effect). Break the symmetry — as in the polar LiNbO₃ lattice — and a first-order, direction-dependent response survives.

From index change to a phase shift you can measure

A modulator turns Δn into a controllable optical phase. A beam of vacuum wavelength λ travelling a length L through a region of index change Δn accumulates an extra phase

• Δφ = (2π/λ) · Δn · L.
• Substituting the transverse-field LiNbO₃ result: Δφ = (π/λ) nₑ³ r₃₃ (V/d) L, where V is the applied voltage and d the electrode gap.

The natural figure of merit is the half-wave voltage V_π, the voltage that produces Δφ = π (a half-wavelength of extra path, enough to swing a polarizer or interferometer from full transmission to full extinction):

• V_π = λ d / (nₑ³ r₃₃ L).

Read this formula like a designer: V_π falls if you make the electrode gap d small and the interaction length L long. That single insight is why bulk crystals with d ≈ 5 mm need kilovolts, while an integrated waveguide modulator — confining light to a d ≈ few-μm channel over L ≈ centimetres — reaches V_π of only a few volts, directly drivable by CMOS electronics. The waveguide doesn't change the physics; it changes the geometry by three orders of magnitude in d/L.

The controlling variables and their real numbers

Everything about a Pockels device is a trade among five quantities: the coefficient r, the index n, the wavelength λ, the geometry (d, L), and the crystal's symmetry axis. Some hard numbers:

• Lithium niobate (LiNbO₃): nₑ ≈ 2.20, r₃₃ ≈ 30.8 pm/V. For λ = 1550 nm, d = L (transverse cube) gives V_π ≈ λ/(nₑ³ r₃₃) ≈ 1550×10⁻⁹ / (2.20³ · 30.8×10⁻¹²) ≈ 4.7 kV in bulk — but a folded waveguide brings this to ~3.5 V.
• KDP (KH₂PO₄): the classic longitudinal Pockels cell, r₆₃ ≈ 10.5 pm/V, nₒ ≈ 1.51. Longitudinal V_π = λ/(2 nₒ³ r₆₃) ≈ 8.8 kV at 633 nm — independent of length, because field and light run the same way.
• β-BBO and RTP/KTP: chosen for high damage threshold (GW/cm² pulses) in laser Q-switches; RTP handles high repetition rates with modest V_π.

Two regimes matter. In the longitudinal configuration (E ∥ beam, KDP) V_π is set purely by material constants — geometry cancels — so you cannot trade length for voltage. In the transverse configuration (E ⊥ beam, LiNbO₃) the L/d ratio buys you low voltage, which is why every high-speed telecom modulator is transverse. The price of the transverse geometry is sensitivity to crystal length variation and residual natural birefringence, usually cancelled with a temperature-compensated dual-crystal pair.

Real devices: modulators, Q-switches, and Pockels cells

The Pockels effect is not a laboratory curiosity — it is embedded in the global optical infrastructure.

• Telecom Mach–Zehnder modulators: split light into two waveguide arms, apply the electro-optic phase to one, and recombine. A π phase difference switches from constructive to destructive interference — an optical on/off. Titanium-diffused LiNbO₃ modulators ran the internet's backbone at 10–40 Gbit/s for decades; thin-film lithium-niobate (TFLN) devices now demonstrate bandwidths beyond 100 GHz with V_π·L ≈ 2 V·cm.
• Laser Q-switching: a Pockels cell inside a laser cavity acts as a voltage-controlled shutter. Hold the cavity lossy (Q low) while the gain medium stores energy, then drop the voltage in <10 ns to open the cavity — the stored population inverts into a giant pulse of megawatts peak power, nanoseconds long. This is how Nd:YAG surgical and machining lasers make their pulses.
• Regenerative amplifiers and pulse pickers: Pockels cells select single pulses from a mode-locked train at MHz–kHz rates, essential for ultrafast (femtosecond) amplification chains.
• Quantum optics and metrology: fast, low-jitter phase and polarization control for single photons, squeezed light, and cavity locking.

The reason the Pockels effect beats mechanical or thermal alternatives everywhere is speed: the electronic polarizability responds within an optical cycle, so device bandwidth is limited only by RC electrical time constants and the microwave/optical velocity match — not by any physical inertia.

Subtleties and misconceptions

A few points trip up newcomers:

• It is not birefringence caused by "squeezing" the crystal. The Pockels effect has two contributions: a true clamped (constant-strain) electronic part, and a secondary piezoelectric part where the field strains the lattice, which then changes the index elasto-optically. At high modulation frequency (above mechanical resonances, typically >few MHz) the crystal cannot deform in time, so only the fast electronic (clamped, r^S) coefficient contributes — usually smaller than the low-frequency (unclamped, r^T) value quoted in handbooks.
• Half-wave voltage is not fixed for a material. V_π depends on wavelength, geometry, and which coefficient you couple to. Quoting "V_π of LiNbO₃" without the configuration is meaningless.
• Sign matters. Because Δn ∝ +E, reversing polarity swings the phase the other way. This bipolar response is what lets Mach–Zehnder modulators encode both amplitude and phase (used in coherent QAM optical links).
• Pockels vs Kerr is a symmetry statement, not a strength statement. In a non-centrosymmetric crystal both effects exist, but the linear Pockels term dominates at ordinary fields; only when Pockels vanishes by symmetry (glasses, centrosymmetric crystals, liquids) does the quadratic Kerr term become the leading electro-optic response.

Pockels (linear) vs Kerr (quadratic) electro-optic effects

Property	Pockels effect	Kerr effect
Index change vs field	Δn ∝ E (linear)	Δn ∝ E² (quadratic)
Coefficient	rᵢⱼ ≈ 1–300 pm/V	K or s ≈ 10⁻¹⁴–10⁻¹⁸ m²/V²
Symmetry requirement	non-centrosymmetric only	any material (incl. liquids, glass)
Sign flips with −E?	yes (Δn reverses)	no (E² even)
Typical materials	LiNbO₃, KDP, BBO, KTP, β-BaB₂O₄	nitrobenzene, CS₂, glass fibers
Drive voltage for π shift	~kV bulk / few V waveguide	tens of kV