8.6 Nonlinear optical effects

Fundamentals of nonlinear optics

In standard (linear) optics, a material's optical properties stay the same no matter how intense the light is. Nonlinear optics breaks that rule: when the electromagnetic field becomes strong enough, the material response starts to depend on the field amplitude in a nonlinear way. This regime unlocks phenomena like frequency conversion, intensity-dependent refraction, and the generation of non-classical light states, all of which are central to modern photonics and quantum technologies.

Beyond device applications, nonlinear optical measurements serve as a probe of electronic structure and symmetry at the microscopic level, complementing transport and linear spectroscopy techniques in condensed matter physics.

Linear vs nonlinear optical responses

In the linear regime, the polarization of a medium is simply proportional to the applied electric field:

$P = \epsilon_0 \chi^{(1)} E$

Here $\chi^{(1)}$ is the linear susceptibility, and it captures everything you see in ordinary optics: refraction, absorption, reflection.

When the field amplitude grows large (typically requiring laser intensities), higher-order terms in the polarization become non-negligible:

$P = \epsilon_0 \left( \chi^{(1)} E + \chi^{(2)} E^2 + \chi^{(3)} E^3 + \cdots \right)$

Each successive term describes a new class of optical effects. The $\chi^{(2)}$ term drives second-order processes (second harmonic generation, sum/difference frequency generation), while $\chi^{(3)}$ governs third-order processes (Kerr effect, four-wave mixing, two-photon absorption). Before the invention of the laser, fields were too weak for these terms to matter in practice.

Nonlinear susceptibility tensors

The susceptibilities $\chi^{(n)}$ are not simple scalars; they are tensors of rank $n+1$. This means their components connect specific polarization directions of the input and output fields.

• $\chi^{(2)}$ is a third-rank tensor (27 components before symmetry reduction). It governs all second-order processes.
• $\chi^{(3)}$ is a fourth-rank tensor (81 components before symmetry reduction). It governs third-order processes.
• The magnitudes drop rapidly with order. Typical values in a crystal like LiNbO₃: $\chi^{(1)} \sim 1$, $\chi^{(2)} \sim 10^{-12}$ m/V, $\chi^{(3)} \sim 10^{-22}$ m²/V².
• The number of independent, nonzero tensor components is dictated by the point group symmetry of the material, which is why crystal symmetry analysis is essential before choosing a nonlinear medium.

Symmetry considerations in materials

Symmetry is the single most important factor in determining which nonlinear processes a material can support.

• Centrosymmetric materials (those with inversion symmetry, like silicon or NaCl) have all even-order susceptibilities identically zero: $\chi^{(2)} = 0$. This follows directly from the requirement that inverting the field must invert the polarization, which is impossible if $P$ contains an $E^2$ term.
• Non-centrosymmetric materials (like GaAs, LiNbO₃, BBO) can exhibit second-order effects and are the workhorses of frequency conversion.
• Third-order effects ($\chi^{(3)}$) are allowed in all materials, regardless of symmetry.
• Point group analysis using Neumann's principle lets you identify exactly which tensor components are nonzero for a given crystal class. This is a practical first step in any nonlinear optics experiment.

Second-order nonlinear effects

Second-order processes originate from the $\chi^{(2)}$ term in the polarization expansion. They require a non-centrosymmetric medium and involve the mixing of two input field components to produce a new frequency. These effects form the basis of most frequency conversion technology.

Second harmonic generation

Second harmonic generation (SHG) converts two photons at frequency $\omega$ into a single photon at $2\omega$. It was the first nonlinear optical effect ever observed (Franken et al., 1961, using a ruby laser and quartz crystal).

• The process requires phase matching: the fundamental wave at $\omega$ and the generated wave at $2\omega$ must travel at the same phase velocity so that contributions from different points along the crystal add constructively. Without phase matching, the generated signal oscillates with propagation distance and remains weak.
• Conversion efficiency scales with the square of the nonlinear coefficient, the square of the interaction length, and the input intensity.
• Practical applications include green laser pointers (1064 nm → 532 nm via KTP or LiNbO₃) and second harmonic imaging microscopy in biological tissues.

Sum and difference frequency generation

These are generalizations of SHG where two different input frequencies mix:

• Sum frequency generation (SFG): $\omega_3 = \omega_1 + \omega_2$. Useful for reaching shorter wavelengths and for surface-sensitive vibrational spectroscopy (SFG spectroscopy probes interfaces because bulk centrosymmetric media cannot contribute).
• Difference frequency generation (DFG): $\omega_3 = \omega_1 - \omega_2$. Used to generate mid-infrared and terahertz radiation.
• Both processes obey energy conservation ($\hbar\omega_3 = \hbar\omega_1 \pm \hbar\omega_2$) and momentum conservation (phase matching).
• Optical parametric amplifiers (OPAs) and oscillators (OPOs) are built on DFG, providing widely tunable coherent light sources.

Optical rectification

When an intense optical pulse passes through a $\chi^{(2)}$ material, it can induce a quasi-DC polarization. This is the zero-frequency limit of difference frequency generation (the pulse mixes with itself).

• In practice, a femtosecond laser pulse in a crystal like ZnTe or GaP generates a broadband terahertz pulse, because the DC polarization follows the pulse envelope and thus contains frequency components in the THz range.
• This is one of the primary methods for generating and detecting terahertz radiation, used in THz time-domain spectroscopy for imaging and materials characterization.

Pockels effect

The Pockels (linear electro-optic) effect is the change in refractive index that is linearly proportional to an applied static or low-frequency electric field. It's technically a second-order process where one of the "input fields" is at DC.

• The index change is $\Delta n = -\frac{1}{2} n^3 r E_{\text{DC}}$, where $r$ is the relevant electro-optic coefficient.
• This effect is the basis for electro-optic modulators (Mach-Zehnder modulators in LiNbO₃) used in fiber-optic telecommunications, where data is encoded onto light at rates exceeding 100 Gbit/s.
• It also enables Q-switching and cavity dumping in pulsed laser systems.
• Only non-centrosymmetric materials exhibit the Pockels effect. The analogous effect in centrosymmetric media is the Kerr (quadratic) electro-optic effect, which is much weaker.

Third-order nonlinear effects

Third-order processes arise from $\chi^{(3)}$ and are present in every material, including glasses, liquids, and gases. This universality makes them especially important in optical fibers and silicon photonics, where the medium is centrosymmetric and $\chi^{(2)}$ vanishes.

Third harmonic generation

Third harmonic generation (THG) converts three photons at $\omega$ into one photon at $3\omega$.

• Phase matching is more difficult than for SHG because the dispersion between $\omega$ and $3\omega$ is typically larger.
• THG is generally much less efficient than SHG. In practice, generating the third harmonic is often done in two steps: SHG ($\omega \to 2\omega$) followed by SFG ($\omega + 2\omega \to 3\omega$).
• Applications include generating UV light from near-infrared sources (e.g., 1064 nm → 355 nm) for spectroscopy and micromachining.

Four-wave mixing

Four-wave mixing (FWM) is the general $\chi^{(3)}$ process involving the interaction of four photons. Energy conservation requires $\omega_1 + \omega_2 = \omega_3 + \omega_4$ (in the degenerate case, two pump photons create a signal and idler pair).

• FWM is the dominant parametric process in optical fibers and silicon waveguides.
• It enables wavelength conversion, parametric amplification, and phase conjugation in telecom systems.
• In quantum optics, spontaneous FWM in fibers generates correlated (entangled) photon pairs, serving as a source for quantum communication experiments.

Optical Kerr effect

The optical Kerr effect is the intensity-dependent refractive index:

$n = n_0 + n_2 I$

where $n_0$ is the linear index, $n_2$ is the nonlinear refractive index (related to $\text{Re}[\chi^{(3)}]$), and $I$ is the optical intensity.

• Self-focusing: A beam with a Gaussian intensity profile sees a higher index at its center than at its edges, effectively creating a lens that focuses the beam further. Above a critical power, this can lead to catastrophic self-focusing and material damage.
• Self-phase modulation (SPM): The intensity-dependent index causes a time-varying phase shift across a pulse, broadening its spectrum. SPM is essential for supercontinuum generation and pulse compression.
• Kerr lens mode-locking: The self-focusing effect inside a laser cavity preferentially transmits the high-intensity peaks of a mode-locked pulse train, providing the mechanism for generating femtosecond pulses in Ti:sapphire lasers.

Typical values of $n_2$: ~$3 \times 10^{-16}$ cm²/W for fused silica, ~$10^{-14}$ cm²/W for CS₂.

Two-photon absorption

Two-photon absorption (TPA) occurs when a material simultaneously absorbs two photons whose combined energy matches an electronic transition, even though neither photon alone has enough energy.

• The absorption rate scales as $I^2$, making TPA a strongly intensity-dependent process. This quadratic dependence means TPA is confined to the focal volume of a tightly focused beam.
• Two-photon microscopy exploits this spatial confinement for deep-tissue fluorescence imaging with inherent optical sectioning, avoiding out-of-focus background.
• TPA also enables 3D microfabrication (two-photon polymerization) with sub-diffraction-limit resolution.
• TPA is characterized by the two-photon absorption coefficient $\beta$, related to $\text{Im}[\chi^{(3)}]$.

Phase matching conditions

Efficient nonlinear conversion requires that the generated wave stays in phase with the nonlinear polarization that drives it throughout the interaction length. Without phase matching, the generated field oscillates between constructive and destructive interference over a distance called the coherence length $l_c = \pi / \Delta k$, severely limiting efficiency.

For SHG as an example, the phase matching condition is:

$\Delta k = k_{2\omega} - 2k_{\omega} = 0$

which translates to $n(2\omega) = n(\omega)$. Because of normal dispersion, this is never satisfied automatically, so special techniques are needed.

Birefringent phase matching

Birefringent crystals have different refractive indices for different polarizations (ordinary and extraordinary rays). By choosing the right crystal orientation, you can make the ordinary index at one frequency equal the extraordinary index at another, satisfying $\Delta k = 0$.

• Type I: Both input photons have the same polarization (e.g., both ordinary), and the output has the orthogonal polarization (extraordinary).
• Type II: The two input photons have orthogonal polarizations.
• The phase matching angle and temperature can be tuned to select the operating wavelength.
• A limitation is the walk-off between ordinary and extraordinary beams, which reduces the effective interaction length in critically phase-matched geometries.

Quasi-phase matching techniques

Quasi-phase matching (QPM) takes a different approach: instead of making $\Delta k = 0$, you periodically reverse the sign of $\chi^{(2)}$ with a period $\Lambda$ such that the added grating vector compensates the mismatch:

$\Delta k - \frac{2\pi}{\Lambda} = 0$

• Periodic poling of ferroelectric crystals (PPLN, PPKTP) is the standard fabrication method. An applied electric field flips the ferroelectric domains at the desired period.
• QPM allows you to use the largest nonlinear coefficient of the crystal (often $d_{33}$), which may not be accessible via birefringent phase matching.
• Domain engineering enables more exotic designs: chirped gratings for broadband conversion, aperiodic structures for multiple simultaneous processes.

Phase matching bandwidth

The bandwidth over which efficient conversion occurs is set by how quickly $\Delta k$ deviates from zero as wavelength changes.

• Bandwidth is inversely proportional to the interaction length: longer crystals give higher efficiency but narrower bandwidth.
• Group velocity mismatch (GVM) between the interacting pulses is the dominant bandwidth-limiting factor for ultrashort pulses. If the fundamental and harmonic pulses walk off temporally, they stop interacting.
• Designing a nonlinear device always involves a trade-off between conversion efficiency (favoring long crystals) and bandwidth (favoring short crystals or engineered QPM structures).

Nonlinear optical materials

Choosing the right material is often the most consequential decision in a nonlinear optics experiment. The key figures of merit include the nonlinear coefficient, transparency range, damage threshold, and phase-matchability.

Inorganic crystals

Inorganic crystals remain the most widely used nonlinear materials due to their high optical quality and well-characterized properties.

• LiNbO₃ (lithium niobate): Large nonlinear coefficients ($d_{33} \approx 25$ pm/V), wide transparency (0.35–5 μm), and excellent for periodic poling. Susceptible to photorefractive damage at visible wavelengths, which can be mitigated by MgO doping.
• KTP (KTiOPO₄): High damage threshold, good for high-power SHG of Nd:YAG lasers (1064 → 532 nm). Also available in periodically poled form (PPKTP).
• BBO (β-BaB₂O₄): Broad phase matching range extending into the UV, high damage threshold, but relatively small nonlinear coefficient and significant walk-off.
• AgGaS₂, AgGaSe₂: Used for mid-infrared generation (transparency extending beyond 10 μm).

Organic materials

Organic nonlinear materials can exhibit very large nonlinear coefficients because their $\chi^{(2)}$ response originates from highly polarizable π-electron systems in conjugated molecules.

• Conjugated polymers and molecular crystals (e.g., DAST) can have electro-optic coefficients several times larger than LiNbO₃.
• Advantages include fast electronic response times (femtosecond scale), ease of thin-film processing, and structural tunability through molecular design.
• The main challenges are lower laser damage thresholds, potential photodegradation, and difficulty growing large, high-quality single crystals.
• Most promising for electro-optic modulators and THz generation rather than high-power frequency conversion.

Semiconductor nanostructures

Quantum confinement in low-dimensional semiconductors enhances nonlinear optical responses by concentrating oscillator strength into discrete transitions.

• Quantum wells (e.g., GaAs/AlGaAs) are used in saturable absorbers for mode-locking and in intersubband nonlinear devices in the mid-infrared.
• Quantum dots exhibit large $\chi^{(3)}$ per unit volume due to their atom-like density of states, and their nonlinear response is tunable via dot size.
• Silicon photonics leverages $\chi^{(3)}$ effects (FWM, Raman amplification) in CMOS-compatible waveguides. Silicon's centrosymmetry means $\chi^{(2)}$ is nominally zero, though strain and electric-field-induced symmetry breaking can introduce an effective $\chi^{(2)}$.
• 2D materials like graphene and transition metal dichalcogenides (MoS₂, WSe₂) show strong nonlinear responses per unit thickness, with broken inversion symmetry in monolayer TMDs enabling SHG.

Applications of nonlinear optics

Nonlinear optical processes underpin a wide range of technologies spanning telecommunications, precision measurement, biomedical imaging, and quantum information science.

Frequency conversion devices

• Optical parametric oscillators (OPOs): Place a nonlinear crystal inside a resonant cavity to build up parametric gain from DFG. OPOs provide continuously tunable coherent output across wavelength ranges inaccessible to conventional lasers.
• Optical frequency combs: Mode-locked lasers combined with nonlinear broadening (SPM in fibers) produce evenly spaced spectral lines spanning an octave or more. These are the basis of optical clocks and precision spectroscopy (2005 Nobel Prize to Hänsch and Hall).
• Terahertz generation: Optical rectification in ZnTe or LiNbO₃, or DFG between two near-infrared beams, produces coherent THz radiation for imaging and spectroscopy.
• Supercontinuum generation: Pumping a photonic crystal fiber with a short pulse produces a white-light continuum spanning hundreds of nanometers, driven by SPM, FWM, and soliton dynamics.

Optical parametric oscillators

An OPO works by placing a $\chi^{(2)}$ crystal inside an optical cavity. A pump photon at $\omega_p$ splits into a signal ($\omega_s$) and idler ($\omega_i$), with $\omega_p = \omega_s + \omega_i$.

• Tuning is achieved by adjusting the phase matching condition (crystal angle, temperature, or QPM period).
• OPOs can operate continuous-wave or pulsed, with output spanning from the UV to the mid-infrared depending on the crystal.
• They are widely used in molecular spectroscopy, LIDAR, and as pump sources for further nonlinear processes.

Ultrafast pulse generation

Nonlinear optics is inseparable from ultrafast science. Several key techniques rely directly on nonlinear effects:

1. Kerr lens mode-locking in Ti:sapphire lasers produces pulses as short as ~5 fs by exploiting the intensity-dependent self-focusing inside the gain medium.
2. Optical parametric amplification (OPA) can amplify broadband seed pulses to high energy while preserving (or even compressing) pulse duration, reaching the few-cycle regime.
3. Nonlinear pulse compression uses SPM in a fiber or gas-filled hollow capillary to broaden the spectrum, followed by dispersive optics to compress the pulse.
4. High harmonic generation (HHG) in noble gases driven by intense femtosecond pulses produces coherent extreme-UV and soft X-ray radiation in the form of attosecond pulse trains.

Optical switching and logic

Nonlinear optical effects offer a path to all-optical signal processing, potentially surpassing the speed limits of electronics.

• Kerr-based switches: The intensity-dependent index change can route signals between output ports in a Mach-Zehnder interferometer configuration, with switching times limited by the electronic response (~fs for bound-electron Kerr effect).
• Nonlinear optical loop mirrors (NOLMs): Sagnac interferometers exploiting SPM for ultrafast demultiplexing in telecom systems.
• Optical bistability: A nonlinear medium inside a Fabry-Pérot cavity can exhibit two stable output states for a given input, functioning as an optical memory element.
• Practical challenges include the high optical powers typically required and the difficulty of cascading nonlinear logic gates without signal degradation.

Measurement techniques

Characterizing nonlinear optical properties requires specialized methods that isolate the (often weak) nonlinear signal from the dominant linear response.

Z-scan method

The Z-scan technique is one of the simplest and most widely used methods for measuring both the nonlinear refractive index $n_2$ and nonlinear absorption coefficient $\beta$.

1. A sample is translated along the propagation axis (z-axis) of a focused Gaussian beam.
2. As the sample moves through the focus, the local intensity changes, modifying the nonlinear phase shift and absorption.
3. Closed-aperture Z-scan: A small aperture before the detector converts the nonlinear phase shift (self-focusing or defocusing) into a measurable transmission change. The characteristic peak-valley (or valley-peak) signature directly gives the sign and magnitude of $n_2$.
4. Open-aperture Z-scan: All transmitted light is collected, so only nonlinear absorption (TPA, saturable absorption) affects the signal. A dip at the focus indicates TPA; a peak indicates saturable absorption.
5. Dividing the closed-aperture data by the open-aperture data isolates the purely refractive contribution.

Pump-probe spectroscopy

Pump-probe experiments resolve nonlinear dynamics on femtosecond to nanosecond timescales.

1. A strong pump pulse excites the sample, creating a transient change in its optical properties.
2. A weak probe pulse, delayed by a variable time $\Delta t$, measures the resulting change in transmission or reflection.
3. By scanning $\Delta t$, you map out the temporal evolution of carrier relaxation, phonon dynamics, exciton formation, or energy transfer.
4. Variants include transient absorption spectroscopy (broadband probe for spectral information), degenerate pump-probe (same wavelength for both), and four-wave mixing geometries (e.g., photon echo) for studying dephasing processes.

Nonlinear interferometry

Interferometric techniques provide high-sensitivity measurements of nonlinear phase shifts.

• A Mach-Zehnder or Michelson interferometer is configured so that the nonlinear medium sits in one arm. The intensity-dependent phase shift appears as a fringe shift at the output.
• This approach can distinguish between electronic (instantaneous) and thermal (slow) contributions to the nonlinear index by varying the pulse repetition rate or using time-gated detection.
• Nonlinear interferometry is particularly useful for characterizing waveguide and fiber nonlinearities, where long interaction lengths make even small $n_2$ values measurable.

Nonlinear optics in waveguides

Confining light to a small cross-sectional area over a long interaction length dramatically enhances nonlinear effects. The effective nonlinear interaction scales as $L_{\text{eff}} / A_{\text{eff}}$, where $L_{\text{eff}}$ is the effective length (limited by loss) and $A_{\text{eff}}$ is the effective mode area. Waveguides can achieve ratios orders of magnitude larger than bulk crystals.

Guided wave nonlinear optics

• Integrated waveguides in materials like thin-film LiNbO₃ (TFLN) achieve mode areas of ~1 μm², compared to ~100 μm² in bulk crystals, boosting nonlinear efficiency by orders of magnitude.
• Dispersion engineering through waveguide geometry allows precise control of phase matching, enabling broadband frequency conversion and comb generation on chip.
• Silicon and silicon nitride (Si₃N₄) waveguides exploit $\chi^{(3)}$ for FWM-based comb generation (microresonator Kerr combs) and parametric amplification.
• These platforms are central to the development of integrated quantum photonic circuits, where on-chip photon pair sources and squeezed light generators are needed.

Nonlinear fiber optics

Optical fibers, despite having a relatively small $n_2$ (fused silica), achieve strong nonlinear effects because of their small mode area (~80 μm² in standard single-mode fiber) and long interaction lengths (km scale).

• Self-phase modulation (SPM): Intensity-dependent phase accumulation broadens the pulse spectrum symmetrically. This is the first step in many supercontinuum and pulse compression schemes.
• Cross-phase modulation (XPM): The intensity of one wavelength channel modifies the phase of a co-propagating channel. This is a source of crosstalk in wavelength-division multiplexed (WDM) telecom systems.
• Stimulated Raman scattering (SRS): Transfers energy from a pump to a frequency-downshifted Stokes wave, providing distributed amplification in fibers. The Raman gain bandwidth in silica is ~13 THz.
• Stimulated Brillouin scattering (SBS): Scatters light backward via interaction with acoustic phonons. SBS has a very narrow bandwidth (~20 MHz in silica) and is the lowest-threshold nonlinear effect in long fibers.

Soliton formation and propagation

A soliton forms when the nonlinear phase shift from SPM exactly compensates the dispersive broadening of a pulse, producing a shape-preserving waveform.

• In the anomalous dispersion regime of a fiber ($\beta_2 < 0$), the fundamental soliton propagates without changing its temporal or spectral shape. Its peak power is $P_0 = |\beta_2| / (\gamma T_0^2)$, where $\gamma = n_2 \omega / (c A_{\text{eff}})$ is the nonlinear parameter and $T_0$ is the pulse width.
• Higher-order solitons ($N > 1$) undergo periodic breathing but return to their original shape after each soliton period.
• Dissipative solitons arise in systems with gain and loss (e.g., mode-locked fiber lasers) and are stabilized by the interplay of nonlinearity, dispersion, gain, and spectral filtering.
• Soliton concepts underpin long-haul optical communication system design and are central to understanding Kerr comb generation in microresonators.

Quantum aspects of nonlinear optics

Nonlinear optical processes are the primary tools for generating and manipulating non-classical states of light. At the quantum level, the same $\chi^{(2)}$ and $\chi^{(3)}$ interactions that produce classical frequency conversion also create quantum correlations between photons.

Squeezed states of light

A squeezed state has reduced quantum noise in one field quadrature (e.g., amplitude) at the cost of increased noise in the conjugate quadrature (phase), consistent with the Heisenberg uncertainty principle.

• Squeezed vacuum is generated by parametric down-conversion ($\chi^{(2)}$) or degenerate FWM ($\chi^{(3)}$) below the oscillation threshold.
• The degree of squeezing is quantified in dB relative to the shot noise level. State-of-the-art experiments achieve >15 dB of squeezing.
• LIGO uses squeezed light injection to improve gravitational wave detection sensitivity beyond the standard quantum limit.
• Other applications include quantum-enhanced sensing, continuous-variable quantum information, and quantum imaging.

Entangled photon generation

• Spontaneous parametric down-conversion (SPDC): A pump photon in a $\chi^{(2)}$ crystal spontaneously splits into a signal-idler pair that are entangled in energy-time, polarization, and/or spatial mode. This is the most common source of entangled photons.
• Spontaneous four-wave mixing (SFWM): The $\chi^{(3)}$ analog in fibers or waveguides. Two pump photons annihilate to create a correlated signal-idler pair. SFWM sources are attractive for integration into fiber networks and silicon photonic chips.
• Entangled photon pairs are the key resource for quantum key distribution (QKD), Bell inequality tests, quantum teleportation, and linear optical quantum computing.
• Source quality is characterized by pair generation rate, heralding efficiency, spectral purity, and entanglement fidelity (measured via quantum state tomography).

Quantum nonlinear optics

Classical nonlinear optics requires intense fields because $\chi^{(2)}$ and $\chi^{(3)}$ effects are weak at the single-photon level. Quantum nonlinear optics aims to make single photons interact with each other, which would enable deterministic quantum logic gates.

• Cavity QED: Placing an atom (or artificial atom) inside a high-finesse optical cavity enhances the light-matter coupling to the point where a single photon can saturate the atomic transition, producing photon blockade and effective photon-photon interactions.
• Artificial atoms: Semiconductor quantum dots coupled to photonic crystal cavities, and superconducting qubits coupled to microwave resonators, are leading platforms for achieving strong single-photon nonlinearities.
• Rydberg atoms in cold atomic ensembles provide giant nonlinearities via dipole blockade, enabling single-photon switches and transistors.
• These capabilities are essential building blocks for quantum repeaters, deterministic photonic quantum gates, and scalable quantum networks.