Second-Harmonic Generation and χ⁽²⁾ Processes

Shine an intense infrared beam at 1064 nm into the right crystal and green light at 532 nm comes out the far side — light at exactly twice the frequency, a colour the laser never emitted. That is second-harmonic generation, the first and most iconic of the second-order nonlinear processes, seen within a year of the laser's invention and now sitting in every green laser pointer. Its origin is the squared term of the material polarization set up in the companion article Introduction to Nonlinear Optical Polarization: driving a $\chi^{(2)}$ medium with $E\propto\cos\omega t$ gives $P^{(2)}\propto E^2\propto\cos^2\omega t=\tfrac12(1+\cos2\omega t)$ , so the polarization carries a component oscillating at $2\omega$ that radiates the harmonic. Because $\chi^{(2)}$ is a third-rank tensor that vanishes under inversion, the effect lives only in non-centrosymmetric crystals — quartz, KDP, BBO, LiNbO₃, GaAs. Generating harmonic dipoles is the easy half. The hard half is making the billions of them radiate in step. Each atomic plane launches a $2\omega$ wavelet driven by the local fundamental, but the fundamental and the harmonic travel at different phase velocities — normal dispersion makes $n(2\omega)>n(\omega)$ — so freshly launched wavelets gradually fall out of phase with those ahead. After a coherence length $L_c$ they are half a cycle behind and begin to cancel what came before, and the output oscillates rather than grows. Defeating this dephasing — phase matching, by exploiting birefringence to make the two indices coincide or by periodically flipping $\chi^{(2)}$ (quasi-phase matching) — is the central engineering problem of the field. This article derives the master result $I_{2\omega}\propto I_\omega^2 L^2\,\mathrm{sinc}^2(\Delta k L/2)$ from the slowly-varying-envelope equation, pins down $L_c$ for a real crystal, and shows why a longer crystal is not always a brighter one. It also opens onto the wider $\chi^{(2)}$ family — sum- and difference-frequency mixing, parametric amplification — and forward to the third-order optical Kerr effect.

Key takeaways

Second-harmonic generation (SHG) is the prototypical $\chi^{(2)}$ process: two photons at $\omega$ combine into one at $2\omega$ , driven by the squared term $P^{(2)}=\varepsilon_0\chi^{(2)}E^2$ (recalled from Introduction to Nonlinear Optical Polarization).
It requires a non-centrosymmetric medium; in centrosymmetric materials $\chi^{(2)}=0$ identically and bulk SHG is forbidden.
Efficient conversion needs phase matching, $\Delta k=k_{2\omega}-2k_\omega=0$ , to defeat the dispersion that otherwise dephases the driving and radiated waves.
In the undepleted-pump limit the harmonic grows as $I_{2\omega}\propto I_\omega^2 L^2\,\mathrm{sinc}^2(\Delta k L/2)$ , so the coherence length $L_c=\pi/|\Delta k|=\lambda/(4\Delta n)$ caps the useful length.
For KDP doubling 1064→532 nm, $\Delta n=0.0207$ gives $L_c=12.85\,\mu\mathrm{m}$ ; an unmatched centimetre crystal is capped near $7\times10^{-7}$ of the phase-matched output.
Key idea: SHG is coherent scattering off the quadratic polarization; its brightness is set jointly by $\chi^{(2)}$ (a symmetry-permitted crystal property) and by phase matching $\Delta k\to0$ , which lets the $2\omega$ wavelets from every plane of the crystal add in step rather than cancel.

The core insight

SHG is coherent scattering off the quadratic term of the material polarization. Its efficiency is set jointly by the magnitude of $\chi^{(2)}$ — a symmetry-permitted, non-centrosymmetric property — and by phase matching, $\Delta k=k_{2\omega}-2k_\omega\to0$ , which lets the $2\omega$ dipoles radiated at every plane of the crystal interfere constructively. Without it the output is capped by the coherence length $L_c=\pi/|\Delta k|$ and oscillates as $\mathrm{sinc}^2(\Delta k L/2)$ with crystal length.

Where this lives today

The most familiar $\chi^{(2)}$ device is the green laser pointer: a diode pumps a Nd:YVO₄ or Nd:YAG chip to 1064 nm, and an intracavity KTP or periodically poled crystal doubles it to 532 nm, exploiting the high circulating intensity and the $I^2$ law. The same physics, phase-matched in BBO or LBO and sometimes cascaded, reaches deep into the ultraviolet for laser cooling of ions, precision spectroscopy, and lithography sources where no direct laser exists. Periodically poled lithium niobate has made quasi-phase matching a commodity, powering the widely tunable optical parametric oscillators used for trace-gas sensing and mid-infrared generation. Second-harmonic generation is also a measurement tool. Non-collinear SHG in a thin crystal is the heart of the autocorrelators and FROG devices that measure femtosecond pulse shapes, since the $\chi^{(2)}$ signal tracks the product of two pulse replicas' intensities. And because bulk SHG is symmetry-forbidden, it is intrinsically surface-sensitive: SHG microscopy images non-centrosymmetric structures such as collagen fibrils with no staining and no background from the surrounding tissue.

The physical picture

Think of the crystal as a stack of atomic planes, each one a tiny antenna. The intense fundamental sweeps through and drives every plane's electrons into the asymmetric part of their potential, so each plane radiates a small wavelet at $2\omega$ . The question is not whether the wavelets exist but whether they add up. If every wavelet arrives at the exit face in phase with all the others, their amplitudes sum — the harmonic amplitude grows linearly with crystal length and its power as $L^2$ . If the wavelets slip progressively out of phase, later ones subtract from earlier ones and the harmonic never accumulates.

Figure

SHG as a phased array of 2ω antennas. Top: phase-matched, wavelets in step, phasors add to a long resultant (power ∝ L²). Bottom: mismatched, phasors rotate and after L_c curl back to a near-zero resultant.

Because normal dispersion always makes the harmonic index a little larger than the fundamental's, the wavelets do slip, and the natural fate of a plain crystal is the oscillating, capped output on the bottom. Quantum-mechanically the same picture reads as photon bookkeeping: SHG annihilates two pump photons and makes one at $2\omega$ , so energy conservation $\hbar\omega+\hbar\omega\to2\hbar\omega$ is automatic, while the phase-matching condition $\mathbf k_{2\omega}=2\mathbf k_\omega$ is nothing but momentum conservation for the photons inside the medium. The whole art of practical SHG is arranging that momentum condition to hold. The next figure shows exactly what it buys.

Figure

The punchline: harmonic output vs crystal length. Phase-matched (dashed) grows as L²; mismatched (solid) is capped and oscillates as sinc²(ΔkL/2), peaking at L_c = 12.85 µm and nulling at 2L_c = 25.7 µm for KDP 1064→532 nm.

The core equations

The starting point is recalled by name from Introduction to Nonlinear Optical Polarization: the quadratic term of the polarization, driven by a monochromatic field, splits into a static piece and one oscillating at twice the frequency. It is that $2\omega$ term that acts as the antenna current for the harmonic.

P^{(2)}(t)=\varepsilon_0\chi^{(2)}E_0^2\cos^2\omega t=\tfrac12\varepsilon_0\chi^{(2)}E_0^2\left(1+\cos 2\omega t\right) \qquad (1)

where: $E_0$ — fundamental field amplitude; $\omega$ — fundamental angular frequency; $\chi^{(2)}$ — second-order susceptibility (tables quote the contracted coefficient $d=\tfrac12\chi^{(2)}$ ). The constant term is optical rectification; the $\cos2\omega t$ term radiates the second harmonic. Two conservation laws govern whether that radiation builds. Energy conservation fixes the output colour automatically; momentum conservation is the delicate one, because the photon momenta $\hbar k$ depend on the refractive index, which differs at $\omega$ and $2\omega$ .

\omega+\omega=2\omega,\qquad \mathbf k_{2\omega}=2\mathbf k_\omega \qquad (2)

where: $\mathbf k_\omega,\ \mathbf k_{2\omega}$ — wavevectors of fundamental and harmonic in the medium. Energy conservation is guaranteed; momentum conservation (phase matching) is not, and its failure is measured by the mismatch below. Because the two waves see different indices, the harmonic wave and the polarization wave that feeds it travel at different phase velocities. The rate at which they dephase is the wavevector mismatch, and normal dispersion ( $n(2\omega)>n(\omega)$ ) makes it nonzero for any plain crystal.

\Delta k=k_{2\omega}-2k_\omega=\frac{2\omega}{c}\left[n(2\omega)-n(\omega)\right] \qquad (3)

where: $n(\omega),\ n(2\omega)$ — indices at the two frequencies; $c$ — vacuum speed of light. $\Delta k=0$ is exactly the phase-matched condition; the larger $\Delta k$ , the shorter the crystal length over which the harmonic can grow. The dynamics live in one first-order equation. Starting from Maxwell's wave equation with the nonlinear polarization as a source, and keeping only the slowly-varying envelope (Derivation A), the harmonic amplitude obeys a driven equation whose only subtlety is the phase factor $e^{i\Delta k z}$ .

\frac{\mathrm dA_{2\omega}}{\mathrm dz}=\frac{i\,\omega\,d_{\mathrm{eff}}}{n_{2\omega}c}\,A_\omega^2\,e^{i\Delta k z} \qquad (4)

where: $A_\omega,\ A_{2\omega}$ — slowly-varying complex amplitudes; $d_{\mathrm{eff}}$ — effective nonlinear coefficient; $n_{2\omega}$ — harmonic index; $z$ — propagation coordinate. In the undepleted-pump limit $A_\omega$ is constant and the equation integrates elementarily. Integrating and squaring gives the master result of the subject: the harmonic power is quadratic in the fundamental intensity and in the length, modulated by a $\mathrm{sinc}^2$ factor that is unity only when phase-matched.

I_{2\omega}(L)=\frac{2\omega^2 d_{\mathrm{eff}}^2 L^2}{n_\omega^2 n_{2\omega}\varepsilon_0 c^3}\,I_\omega^2\,\mathrm{sinc}^2\!\left(\frac{\Delta k L}{2}\right),\qquad \mathrm{sinc}\,x=\frac{\sin x}{x} \qquad (5)

where: $L$ — interaction length; $I_\omega$ — fundamental intensity. The intensity enters squared — a central practical fact: doubling the input intensity quadruples the harmonic. At $\Delta k=0$ the sinc $^2$ is one and $I_{2\omega}\propto L^2$ ; otherwise the output oscillates. The last equation names the length scale that sets everything. The coherence length is where the harmonic and polarization dephase by $\pi$ ; a real crystal is useful only up to about this length unless it is phase-matched.

L_c=\frac{\pi}{|\Delta k|}=\frac{\lambda}{4\,|n(2\omega)-n(\omega)|} \qquad (6)

where: $L_c$ — coherence length; $\lambda$ — fundamental vacuum wavelength; $\Delta n=|n(2\omega)-n(\omega)|$ . For KDP doubling 1064→532 nm, $\Delta n=0.0207$ gives $L_c=12.85\,\mu\mathrm{m}$ . Quasi-phase matching restores growth by flipping $\chi^{(2)}$ every $L_c$ , poling period $\Lambda=2L_c$ .

Derivation AThe sinc² law from the coupled-wave integral

This is the owned result of the article — the single formula that explains why bright second harmonic demands phase matching. The goal is to integrate the slowly-varying-envelope equation for the harmonic amplitude across a crystal of length $L$ , assuming the pump is strong enough not to be noticeably depleted, and to show that the harmonic intensity comes out proportional to $L^2\,\mathrm{sinc}^2(\Delta k L/2)$ — a factor equal to one when the waves are matched and oscillating down to zero when they are not. The whole calculation reduces to one elementary integral of a complex exponential.

Step 1. Take the envelope equation $\mathrm dA_{2\omega}/\mathrm dz=(i\omega d_{\mathrm{eff}}/n_{2\omega}c)\,A_\omega^2\,e^{i\Delta k z}$ , in which the phase factor $e^{i\Delta k z}$ records how far the locally driven polarization has slipped from the harmonic wave. Why: dropping the second $z$ -derivative is legitimate because the envelope changes little over one wavelength, and it turns the wave equation into a first-order ODE we can integrate directly.

Step 2. Treat the pump as undepleted, so $A_\omega$ (and hence $A_\omega^2$ ) is constant along $z$ . The harmonic at the exit is then $A_{2\omega}(L)=(i\omega d_{\mathrm{eff}}/n_{2\omega}c)\,A_\omega^2\int_0^L e^{i\Delta k z}\,\mathrm dz$ . Why: early in the interaction almost all the energy is still in the fundamental, so holding $A_\omega$ fixed is an excellent approximation and isolates the phase physics.

Step 3. Do the integral: $\int_0^L e^{i\Delta k z}\,\mathrm dz=\dfrac{e^{i\Delta k L}-1}{i\Delta k}=L\,e^{i\Delta k L/2}\,\mathrm{sinc}\!\left(\dfrac{\Delta k L}{2}\right)$ , using $e^{i\theta}-1=2i\,e^{i\theta/2}\sin(\theta/2)$ . Why: the factored form separates a pure phase $e^{i\Delta k L/2}$ (irrelevant to intensity) from the real amplitude envelope $L\,\mathrm{sinc}(\Delta k L/2)$ that carries all the physics.

Step 4. Convert to intensity, $I_{2\omega}=\tfrac12 n_{2\omega}\varepsilon_0 c\,|A_{2\omega}|^2$ . The phase drops out and the modulus-squared gives $I_{2\omega}\propto I_\omega^2\,L^2\,\mathrm{sinc}^2(\Delta k L/2)$ . It peaks at $\Delta k=0$ (where $\mathrm{sinc}^2\to1$ and $I_{2\omega}\propto L^2$ ) and first vanishes at $\Delta k L=2\pi$ , i.e. $L=2L_c$ . Why: this is the entire story — quadratic growth when matched, and an oscillation that returns to zero every $2L_c$ when not. *Verified in verify_aq-optics-second-harmonic-generation-chi2.py: the integral $\int_0^L e^{i\Delta k z}\mathrm dz$ equals $L\,e^{i\Delta k L/2}\mathrm{sinc}(\Delta k L/2)$ (sympy), $|{\cdot}|^2=L^2\mathrm{sinc}^2$ , the $\Delta k\to0$ limit gives $L^2$ , and the first null sits at $L=2L_c$ .*

Go deeper · Derivation B · 1 minCoherence length and the phase-matching penalty

Derivation BCoherence length and the phase-matching penalty

Equation (5) says nothing is bright without $\Delta k\approx0$ ; this card turns that into a length and a number. The goal is to express the coherence length $L_c$ — the distance over which the harmonic dephases by $\pi$ — in terms of the index difference, evaluate it for KDP doubling 1064 nm to 532 nm, and quantify how badly a long unmatched crystal underperforms a phase-matched one of the same length. The strategy is to rewrite $\Delta k$ through the dispersion and then compare the capped mismatched output with the ideal $L^2$ growth.

Step 1. From $\Delta k=(2\omega/c)\,\Delta n$ with $\omega=2\pi c/\lambda$ , the coherence length is $L_c=\pi/|\Delta k|=\pi c/(2\omega\,\Delta n)=\lambda/(4\Delta n)$ . Why: writing $L_c$ in terms of $\lambda$ and $\Delta n$ alone makes it a property of the material dispersion, independent of how one expresses the frequency.

Step 2. Put in KDP numbers: $\lambda=1064\,\mathrm{nm}$ , $n(\omega)=1.4942$ , $n(2\omega)=1.4735$ , so $\Delta n=0.0207$ and $L_c=1064\,\mathrm{nm}/(4\times0.0207)=12.85\,\mu\mathrm{m}$ . Why: a coherence length of only about thirteen microns is why unmatched SHG is hopeless in a millimetre crystal — the useful length is a hundredth of a millimetre.

Step 3. Compare capped to ideal. The mismatched output $L^2\mathrm{sinc}^2(\Delta k L/2)=(4/\Delta k^2)\sin^2(\Delta k L/2)$ is bounded by $(2/\Delta k)^2=(2L_c/\pi)^2$ , whereas the matched output is $L^2$ . Their ratio is $(2L_c/\pi L)^2$ . Why: the mismatched harmonic can never exceed what it built in the first coherence length, so as $L$ grows the fractional shortfall worsens as $1/L^2$ .

Step 4. Evaluate for a centimetre crystal: $(2L_c/\pi L)^2=(2\times12.85\,\mu\mathrm m/(\pi\times10^4\,\mu\mathrm m))^2=6.7\times10^{-7}$ . The unmatched crystal delivers under a millionth of the phase-matched output. The fix is to make $\Delta n=0$ — birefringent phase matching, tuning the crystal angle so the harmonic's extraordinary index equals the fundamental's ordinary index (Type I, o+o→e) — or quasi-phase matching. Why: this single factor is why birefringent (or quasi-) phase matching is not a refinement but the entire enabling technology of frequency conversion. *Verified in verify_aq-optics-second-harmonic-generation-chi2.py: $L_c=\lambda/(4\Delta n)$ symbolically, the KDP value $12.85\,\mu\mathrm m$ , and the centimetre penalty $(2L_c/\pi L)^2=6.7\times10^{-7}$ .*

Limiting cases

Case	Result	Meaning
Phase-matched, undepleted pump	$I_{2\omega}\propto L^2 I_\omega^2$	Ideal regime: harmonic grows quadratically with length and with fundamental intensity.
Mismatched, undepleted pump	$I_{2\omega}\propto L^2\,\mathrm{sinc}^2(\Delta k L/2)$	Output oscillates with period $2L_c$ , capped near $(2L_c/\pi)^2$ ; useful length $\approx L_c$ .
Phase-matched, depleted pump	$I_{2\omega}\propto\tanh^2(\Gamma L)$	Strong conversion: the fundamental transfers energy monotonically, efficiency $\to100\%$ .
Centrosymmetric medium	$\chi^{(2)}=0$	Inversion symmetry forbids bulk SHG; a signal appears only at surfaces or via weaker $\chi^{(3)}$ .

Common trap“A bigger χ⁽²⁾ or a longer crystal always means brighter second harmonic”

Both halves of that instinct fail. A large $\chi^{(2)}$ buys nothing without phase matching: an unmatched crystal caps its output at the coherence length $L_c=\pi/|\Delta k|$ and then oscillates, so lengthening a mismatched crystal from $L_c$ to $2L_c$ sends the harmonic all the way back to zero — the second half radiates exactly out of phase with the first and cancels it. And do not confuse phase matching ( $\Delta k=0$ ) with energy conservation ( $2\omega$ is fixed automatically); they are independent conditions. The related power trap: because $I_{2\omega}\propto I_\omega^2$ , doubling the fundamental power quadruples the harmonic, and halving it cuts the green to a quarter — the conversion is quadratic, not linear, which is why intracavity doubling (where the circulating fundamental is huge) is so effective.

Worked examples

1 · KDP coherence length and the cost of not phase matching (confidence builder)

A KDP crystal doubles 1064 nm light to 532 nm, with $n(\omega)=1.4942$ and $n(2\omega)=1.4735$ . Find the coherence length, and the fraction of the phase-matched output that a 1 cm unmatched crystal produces.

Step 1. Index difference: $\Delta n=|1.4735-1.4942|=0.0207$ . This normal-dispersion gap is what dephases the waves.

Step 2. Coherence length: $L_c=\lambda/(4\Delta n)=1064\,\mathrm{nm}/(4\times0.0207)=\mathbf{12.85\,\mu m}$ . Equivalently $\Delta k=2\pi/(2L_c)=2.44\times10^{5}\,\mathrm{m^{-1}}$ and $L_c=\pi/|\Delta k|$ .

Step 3. Penalty for $L=1\,\mathrm{cm}$ : the mismatched output is capped at $(2L_c/\pi L)^2=(2\times12.85\,\mu\mathrm m/(\pi\times10^4\,\mu\mathrm m))^2=\mathbf{6.7\times10^{-7}}$ of the phase-matched value. A centimetre of unmatched KDP does worse than thirteen microns of matched KDP.

Step 4. The lesson: the crystal length that matters is $L_c$ , not the crystal you can grow. Birefringent phase matching removes $\Delta n$ so the whole centimetre contributes coherently. $L_c=12.85\,\mu\mathrm m$ ; an unmatched 1 cm crystal is capped near $7\times10^{-7}$ of the matched output.

2 · Maker fringes and the power-squared law (exam level)

The same mismatched crystal is polished into a thin wedge so its thickness varies. Describe the second-harmonic pattern versus thickness, and find how the green output changes when the 1064 nm power is doubled.

Step 1. The output follows $\mathrm{sinc}^2(\Delta k L/2)=(2/\Delta k L)^2\sin^2(\Delta k L/2)$ , oscillating in $L$ — the Maker fringes. Sweeping thickness sweeps $\Delta k L$ , tracing the interference between harmonic and polarization waves.

Step 2. Peaks fall at odd multiples of $L_c$ ( $\Delta k L/2=\pi/2,3\pi/2,\dots$ ) and nulls at even multiples, so the fringe period is $2L_c=\mathbf{25.7\,\mu m}$ . At each peak $\mathrm{sinc}^2(\pi/2)=(2/\pi)^2=\mathbf{0.405}$ , and because $(2/\Delta k)^2$ is fixed the fringe envelope is flat — every peak the same height.

Step 3. Now the power law. Since $I_{2\omega}\propto I_\omega^2$ , doubling the fundamental intensity multiplies the harmonic by $2^2=\mathbf{4}$ . Doubling the input power (same beam area) likewise quadruples the green. This quadratic scaling is why femtosecond pulses — enormous peak intensity — double so efficiently, and why halving the pump cuts the green fourfold, not twofold.

Step 4. Together the two facts define practical SHG: get $\Delta k$ to zero so the fringes give way to $L^2$ growth, then push $I_\omega$ as high as damage allows to exploit the square. Fringe period $2L_c=25.7\,\mu\mathrm m$ , peak height $(2/\pi)^2=0.405$ ; doubling the fundamental quadruples the harmonic.

Physics of the portrait

The equations above were written down in the early 1960s, but only because one experiment first showed that a laser could change the colour of its own light — a result so faint it nearly disappeared into legend.

Origins · 1 minPeter Franken and how this was discovered

Peter Franken

1928–1999 · University of Michigan (nonlinear optics)

Peter Franken was an American physicist at the University of Michigan when, in 1961, he and his collaborators A. E. Hill, C. W. Peters and G. Weinreich focused a pulsed ruby laser (694.3 nm) into a quartz crystal and detected a trickle of ultraviolet at 347.2 nm — the second harmonic, the first demonstration that intense light could generate a new optical frequency. The conversion efficiency was about one part in $10^{8}$ , so faint only a spectrograph could catch it. A celebrated legend attaches to the paper: the tiny harmonic spot on the published photographic plate was reportedly so weak that in reproduction it was mistaken for a speck of dirt and removed, leaving only the arrow that had pointed to it — a story often retold, and worth flagging as the kind of anecdote that grows in the telling. Franken went on to help found and direct the Optical Sciences Center at the University of Arizona and served in science-advisory roles including work with the Institute for Defense Analyses. Admired as a vivid teacher and experimentalist, he died in 1999.

History

1960

— Maiman operates the first laser (ruby, 694.3 nm), the source Franken's group would use

1961

— Franken, Hill, Peters and Weinreich observe second-harmonic generation in quartz (694.3 → 347.2 nm)

1962

— Armstrong, Bloembergen, Ducuing and Pershan give the coupled-wave theory, including the quasi-phase-matching idea

1962

— Maker, Terhune, Nisenoff and Savage report the Maker fringes of SHG versus crystal angle

1962

— Giordmaine, and independently Maker and coworkers, demonstrate birefringent phase matching

1990

s — Fejer and collaborators develop periodically poled lithium niobate, making quasi-phase matching practical

Going further

Texts. Saleh & Teich, Fundamentals of Photonics, ch. 21 (nonlinear optics) · Boyd, Nonlinear Optics, 3rd ed., chs. 1–2 · Hecht, Optics, 5th ed., §13.4 · Loudon, The Quantum Theory of Light, 3rd ed., ch. 9 · Demtröder, Laser Spectroscopy, Vol. 1, ch. 5 · Franken, Hill, Peters & Weinreich, Phys. Rev. Lett. 7, 118 (1961).

Provenance. Equations are cross-checked against Saleh & Teich ch. 21, Boyd chs. 1–2, and Loudon ch. 9. A verifier should: confirm the $\cos^2$ identity giving DC + $2\omega$ ; verify the inversion argument forcing $\chi^{(2)}=0$ in centrosymmetric media; reproduce the $\mathrm{sinc}^2$ SHG law by integrating the coupled-wave ODE (undepleted pump); check the $\tanh^2$ depleted-pump solution satisfies $u_1^2+u_2^2=$ const; and verify the quasi-phase-matching period $\Lambda=2L_c$ and the $2/\pi$ effective-coefficient factor. The 532/1064 nm doubling and the 1961 Franken discovery are standard textbook facts; the KDP coherence length $L_c=12.85\,\mu\mathrm m$ follows from $\Delta n=0.0207$ .

Prerequisites. Electric Polarization and the Displacement Field · Dielectric Susceptibility and Permittivity · Maxwell's Equations in Dielectric Media · The Refractive Index · Dispersion: Normal and Anomalous · The Lorentz Oscillator Model. Leads to. Sellmeier and Cauchy Dispersion Equations · Optical Parametric Oscillation · Group Velocity and Group Index.

Why does doubling need phase matching when energy is already conserved?

Energy conservation only fixes the colour of the output — two $\omega$ photons make one $2\omega$ photon, automatically. It says nothing about whether the harmonic wavelets from different depths in the crystal arrive in step. That is a separate, momentum-conservation condition, $\mathbf k_{2\omega}=2\mathbf k_\omega$ , and it fails whenever $n(2\omega)\neq n(\omega)$ , which normal dispersion guarantees. When it fails, the harmonic launched deep in the crystal is out of phase with that launched near the entrance, and they interfere destructively beyond a coherence length. Phase matching forces $\Delta k=0$ so all depths contribute coherently — the difference between microwatts and watts of green.

What exactly is the coherence length, physically?

It is the propagation distance over which the second-harmonic wave and the polarization wave that drives it drift $\pi$ radians out of phase, $L_c=\pi/|\Delta k|=\lambda/(4\Delta n)$ . Up to $L_c$ the harmonic builds constructively; past it the newly generated harmonic starts to subtract from what is already there, and by $2L_c$ the crystal's second half has exactly cancelled its first. So in an unmatched crystal $L_c$ is the only length that ever contributes — making a longer crystal buys nothing. For KDP at 1064 nm this is a mere 12.85 µm, which is why unmatched frequency doubling is essentially useless and phase matching is mandatory.

How does birefringent phase matching actually achieve Δn = 0?

By using a crystal's two different indices to cancel dispersion. In a birefringent crystal the extraordinary index depends on the angle between the beam and the optic axis, while the ordinary index does not. Normal dispersion pushes $n(2\omega)$ above $n(\omega)$ , but if the fundamental travels as an ordinary wave and the harmonic as an extraordinary wave, one can tune the crystal angle until the harmonic's (angle-dependent) extraordinary index drops to equal the fundamental's ordinary index — Type I, o+o→e phase matching. At that angle $\Delta n=0$ , $\Delta k=0$ , and the sinc $^2$ factor becomes one. The price is walk-off: the extraordinary beam drifts sideways from the ordinary one, limiting the useful length in bulk crystals.

What is quasi-phase matching, and why accept a smaller effective coefficient?

Quasi-phase matching does not remove the mismatch — it periodically undoes its effect. By flipping the sign of $\chi^{(2)}$ every coherence length (in practice by periodically poling the ferroelectric domains of lithium niobate), the crystal reverses the harmonic's phase just as it would begin to cancel, so growth resumes each period. The poling grating supplies the missing momentum $2\pi/\Lambda=\Delta k$ with $\Lambda=2L_c$ . A square-wave sign flip only projects a factor $2/\pi\approx0.64$ of the full nonlinearity onto the useful Fourier component, so $d_{\mathrm{eff}}$ is reduced — but this is more than repaid by letting you use non-birefringent materials and the largest tensor element (e.g. $d_{33}$ of LiNbO₃), which birefringent matching often cannot access.

Can second-harmonic conversion really approach 100%, or is there a fundamental cap?

There is no fundamental cap. The undepleted-pump $\mathrm{sinc}^2$ result assumes the fundamental barely changes; once conversion is strong, energy conservation feeds back and the phase-matched solution is $I_{2\omega}\propto\tanh^2(\Gamma L)$ , which asymptotically transfers all the power to the harmonic. Practical devices routinely exceed 50% and approach unity. What limits real conversion is engineering, not the two-wave physics: finite phase-matching bandwidth, spatial walk-off, crystal absorption and thermal dephasing, and — for pulses — group-velocity mismatch between fundamental and harmonic. Push those away and the green nearly extinguishes the infrared.

Why is second-harmonic generation such a good surface and interface probe?

Precisely because bulk $\chi^{(2)}$ vanishes in any centrosymmetric material. Most liquids, glasses, and many crystals are centrosymmetric on average, so their interiors produce no second harmonic at all. But an interface — a surface, a membrane, a buried junction — necessarily breaks inversion symmetry, so a thin sheet there has a nonzero $\chi^{(2)}$ and radiates $2\omega$ . The signal is therefore background-free: everything you detect comes from the symmetry-broken region. This makes SHG a uniquely clean probe of adsorbates, molecular orientation at surfaces, and, in microscopy, of ordered non-centrosymmetric structures like collagen — imaged without any dye against a dark bulk.

Second-Harmonic Generation and χ⁽²⁾ Processes

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · KDP coherence length and the cost of not phase matching (confidence builder)

2 · Maker fringes and the power-squared law (exam level)

Physics of the portrait

Going further

Before this

Where this leads