Nonlinear Optical Fibers and Effects

A single-mode fiber does something no bulk sample of glass can: it takes the linear waveguide picture built in the preceding articles — one confined transverse mode with a propagation constant $\beta(\omega)$ , a group-velocity dispersion $\beta_2$ , and a small residual loss $\alpha$ — and holds the light inside an area a few microns across for kilometers on end. Silica's third-order susceptibility $\chi^{(3)}$ , borrowed here from nonlinear optics, is feeble: the index it adds, $\Delta n = n_2 I$ , is one part in $10^{10}$ at a tenth of a watt. Yet confinement multiplies the intensity and length lets the phase accumulate, and a nonlinearity invisible in a lens becomes the dominant physics of the fiber. From that one fact a whole zoo of effects unfolds, and they all live inside a single master equation — the nonlinear Schrodinger equation for the envelope $A(z,T)$ , which pits loss, dispersion, and the Kerr term against one another. When dispersion is anomalous and the Kerr self-chirp exactly cancels dispersive spreading, the pulse becomes a soliton and propagates unchanged. When the delayed nuclear part of the response is provoked, energy scatters to lower frequencies as stimulated Raman scattering; on a far narrower band and only backward, sound waves give stimulated Brillouin scattering. And when photons trade energy coherently under a phase-matching condition, four-wave mixing appears — the parametric process behind amplifiers, wavelength converters, and octave-spanning supercontinua. These are at once the enabling mechanisms of modern photonics and the ceiling on how much light a fiber link can carry.

Key takeaways

The optical Kerr effect makes the index rise with intensity, $n=n_0+n_2 I$ , so a pulse phase-modulates itself — the origin of self-phase modulation (SPM), cross-phase modulation (XPM), and four-wave mixing (FWM).
A tiny bulk $n_2\approx2.6\times10^{-20}\ {\rm m^2/W}$ becomes decisive because confinement (small $A_{\rm eff}\sim50\!-\!80\ {\rm \mu m^2}$ ) and long effective length can accumulate the nonlinear phase shift to order unity over practical fibre lengths.
The nonlinear Schrodinger equation collects loss, dispersion, and the Kerr term; in anomalous dispersion its balance of SPM chirp against spreading is a soliton, $A=\sqrt{P_0}\,\mathrm{sech}(T/T_0)$ .
Stimulated Raman scattering shifts a broad ≈13.2 THz (both directions); stimulated Brillouin scattering shifts a narrow ≈11 GHz (backward only), with the lowest threshold of all.
FWM is parametric and coherent — it needs phase matching $\Delta\kappa\approx 0$ , tying its efficiency to the dispersion near the pump.
Key idea: a weak bulk nonlinearity becomes dominant in fibers because confinement and length multiply it — the accumulated nonlinear phase is $\phi_{\mathrm{NL}}=\gamma P_0 L_{\mathrm{eff}}$ with $\gamma=n_2\omega_0/(cA_{\mathrm{eff}})$ .

The core insight

A weak bulk nonlinearity ( $n_2 I \ll n_0$ ) becomes dominant in fibers because confinement (small $A_{\mathrm{eff}}$ ) and length (long $L_{\mathrm{eff}}$ ) multiply it: the accumulated nonlinear phase is $\phi_{\mathrm{NL}}=\gamma P_0 L_{\mathrm{eff}}$ with $\gamma=n_2\omega_0/(cA_{\mathrm{eff}})$ .

Where this lives today

The same nonlinearities can be either a limitation or the essential mechanism of the system, depending on the application. In high-capacity coherent transmission they are the dominant impairment: SPM and XPM among wavelength-division-multiplexed channels distort constellations and inter-channel FWM leaks crosstalk, so designers fight back with dispersion management (alternating fibers of opposite $\beta_2$ so local dispersion kills FWM while the path-averaged value stays low) and digital nonlinearity compensation. This penalty sets the practical, Shannon-like capacity ceiling of a link — you cannot simply keep raising the launched power. Turned the other way, the effects are the source. Launch femtosecond pulses into a photonic-crystal fiber near its zero-dispersion wavelength and SPM, FWM, soliton fission, dispersive-wave emission, and the Raman self-frequency shift cascade into an octave-spanning supercontinuum — the light that, self-referenced by $f$ – $2f$ interferometry, underlies the optical frequency comb and the most accurate clocks ever built. Distributed Raman amplifiers use the transmission fiber itself as gain medium; fiber optical parametric amplifiers exploit FWM for wide, wavelength-agile gain; and because the Brillouin shift tracks strain and temperature, distributed Brillouin sensing reads out kilometers of pipeline or bridge from the same ≈11 GHz line a laser link cannot ignore.

The physical picture

Start with the Kerr effect and forget dispersion for a moment. The index at any instant is $n_0+n_2 P(T)/A_{\mathrm{eff}}$ , so the crest of a pulse travels through the slightly largest index and lays on itself the most phase, $\phi_{\mathrm{NL}}(T)=\gamma P(T)L_{\mathrm{eff}}$ . A time-varying phase is a frequency shift, $\delta\omega(T)=-\partial\phi_{\mathrm{NL}}/\partial T$ : the rising leading edge is pushed red, the falling trailing edge blue. The temporal intensity envelope remains approximately unchanged in the ideal SPM-only picture — the pulse still propagates through the fibre, while its time-dependent phase broadens the spectrum. This is self-phase modulation, and the first figure shows the chirp it stamps on a pulse.

Figure

Self-phase modulation stamps a frequency chirp on a pulse. Top the power envelope P(T) is unchanged; bottom the instantaneous frequency shift δω = −∂φ/∂T runs red on the rising leading edge and blue on the falling trailing edge. The peak nonlinear phase γP₀L_eff ≈ 2.9 rad for P₀ = 100 mW over L_eff = 21.5 km of standard fiber

Now let dispersion back in. In anomalous group-velocity dispersion ( $\beta_2<0$ ) blue outruns red, and SPM has parked red at the front and blue at the back — precisely the order dispersion needs to pull the pulse back together. Tune the peak power so that self-phase modulation and anomalous group-velocity dispersion balance each other: the pulse becomes a fundamental soliton, a self-trapped sech-shaped pulse that can propagate without changing its ideal temporal profile. Push higher and integer soliton orders breathe; lower, and dispersion wins. The delayed part of the response opens a second front. A pump photon can scatter off a molecular vibration and emit a lower-energy Stokes photon plus a phonon (Raman), or off a travelling acoustic wave (Brillouin). Both are positive-feedback loops — once a Stokes wave exists it stimulates more of itself — so above a threshold the Stokes field grows exponentially and drains the pump. The two differ enormously in scale, and the second figure maps them onto one frequency-offset axis alongside the instantaneous Kerr effects.

Figure

Where each fiber nonlinearity lives on the frequency-offset axis. The Kerr effects (SPM, XPM, FWM) are instantaneous and coherent at zero shift; stimulated Brillouin scattering is a narrow backward line at ≈11 GHz with gain g_B ≈ 5×10⁻¹¹ m/W; stimulated Raman scattering is a broad ≈40 THz band peaking at ≈13.2 THz with g_R ≈ 1×10⁻¹³ m/W

Four-wave mixing is a parametric process: in the degenerate case, two pump photons generate a signal and idler satisfying $\omega_s+\omega_i=2\omega_p$ . Efficient conversion requires phase matching, $\Delta k\approx0$ , which is often favoured near a zero-dispersion wavelength but can also be engineered through the fibre dispersion profile.

The core equations

Everything begins with the intensity-dependent index. The Kerr coefficient $n_2$ is the ratio of the index change to intensity, fixed by the real part of $\chi^{(3)}$ averaged over the optical carrier.

n(I) = n_0 + n_2 I,\qquad n_2 = \dfrac{3\,\mathrm{Re}\,\chi^{(3)}}{4\varepsilon_0 c\, n_0^2},\qquad n_2^{\mathrm{silica}} \approx 2.6\times10^{-20}\ \mathrm{m^2/W} \qquad (1)

where: $n_0$ — linear index (≈1.45 for silica); $n_2$ — nonlinear index coefficient; $I$ — optical intensity in W/m². The index tracks the instantaneous intensity, so a pulse carries a moving index bump that phase-modulates itself. To turn intensity into guided power we fold the transverse mode into a single number, the nonlinear parameter $\gamma$ . It merges the material response $n_2$ with the geometry $A_{\mathrm{eff}}$ , and its size decides how quickly nonlinear phase accumulates.

\gamma = \dfrac{n_2\,\omega_0}{c\,A_{\mathrm{eff}}} = \dfrac{2\pi n_2}{\lambda\, A_{\mathrm{eff}}},\qquad A_{\mathrm{eff}}=\dfrac{\left(\int|F|^2\,\mathrm{d}A\right)^2}{\int|F|^4\,\mathrm{d}A} \qquad (2)

where: $\gamma$ — nonlinear coefficient in W⁻¹km⁻¹, typically 1–10 in standard fiber; $\omega_0$ — carrier frequency; $A_{\mathrm{eff}}$ — effective mode area; $F(x,y)$ — transverse mode profile; $\lambda$ — vacuum wavelength. A smaller mode means higher intensity for a given power, hence larger $\gamma$ . The master equation collects the three propagation effects — loss, dispersion, and Kerr nonlinearity — acting on the slowly varying envelope $A(z,T)$ . It is written in a frame co-moving with the pulse, so only reshaping, not the uniform drift, appears.

\dfrac{\partial A}{\partial z} + \dfrac{\alpha}{2}A + \dfrac{i\beta_2}{2}\dfrac{\partial^2 A}{\partial T^2} = i\gamma |A|^2 A \qquad (3)

where: $A(z,T)$ — envelope with $|A|^2$ the instantaneous power; $T=t-z/v_g$ — retarded time; $\alpha$ — power attenuation; $\beta_2$ — group-velocity dispersion in s²/m, negative for anomalous. The left side spreads and damps the pulse; the right side is the Kerr self-action. The competition among those terms is governed by three lengths — how far the pulse travels before nonlinearity, dispersion, or loss matters. Their ratios, not their absolute values, decide the regime.

L_{\mathrm{NL}} = \dfrac{1}{\gamma P_0},\qquad L_D = \dfrac{T_0^2}{|\beta_2|},\qquad L_{\mathrm{eff}} = \dfrac{1-e^{-\alpha L}}{\alpha} \qquad (4)

where: $P_0$ — peak power; $T_0$ — pulse half-width (1/e); $L$ — physical length; $L_{\mathrm{eff}}\to 1/\alpha$ for long fibers. Comparing $L$ to these lengths fixes the regime. Self-phase modulation is read directly off the Kerr term in the dispersionless limit. The pulse acquires a power-shaped phase, and the time derivative of that phase is a frequency chirp.

\phi_{\mathrm{NL}}(T) = \gamma P(T) L_{\mathrm{eff}},\qquad \delta\omega(T) = -\dfrac{\partial \phi_{\mathrm{NL}}}{\partial T} = -\gamma L_{\mathrm{eff}}\dfrac{\partial P}{\partial T} \qquad (5)

where: $\phi_{\mathrm{NL}}$ — nonlinear phase; $\delta\omega$ — shift from carrier; $P(T)$ — instantaneous power. The rising edge $(\partial P/\partial T>0)$ is red-shifted, while the falling edge is blue-shifted — the characteristic SPM chirp. A fundamental soliton arises when self-phase modulation and anomalous group-velocity dispersion exactly balance, preserving the pulse shape. The dimensionless soliton order $N$ is the square root of the ratio of the two lengths.

N^2 = \dfrac{L_D}{L_{\mathrm{NL}}} = \dfrac{\gamma P_0 T_0^2}{|\beta_2|},\qquad A(z,T)=\sqrt{P_0}\,\mathrm{sech}\!\left(\dfrac{T}{T_0}\right)e^{i\gamma P_0 z/2}\quad (N=1) \qquad (6)

where: $N$ — soliton order; $N=1$ is the fundamental soliton, integer $N>1$ the breathing higher-order solitons; $\mathrm{sech}$ — the shape-invariant amplitude profile. $N=1$ means $L_D=L_{\mathrm{NL}}$ : dispersion and nonlinearity cancel. The delayed response gives the stimulated scatterings. Raman has a broad gain band peaking far out in the terahertz; its threshold is set by how much gain accumulates over the effective length.

\Omega_R/2\pi \approx 13.2\ \mathrm{THz},\qquad P_{\mathrm{th}}^{\mathrm{SRS}} \approx \dfrac{16\,A_{\mathrm{eff}}}{g_R\,L_{\mathrm{eff}}} \qquad (7)

where: $\Omega_R$ — Raman peak shift; $g_R$ — Raman gain, ≈ $10^{-13}$ m/W peak in silica, over a ≈40 THz band; $P_{\mathrm{th}}^{\mathrm{SRS}}$ — critical pump power for forward SRS. The factor 16 comes from requiring the Stokes wave to grow from noise to the pump level. Brillouin scattering shifts by a tiny amount set by the acoustic velocity, and it travels only backward. Its threshold is by far the lowest of any fiber nonlinearity because its gain coefficient is orders larger.

\nu_B = \dfrac{2 n_0 v_A}{\lambda} \approx 11\ \mathrm{GHz}\ (1.55\,\mu\mathrm{m}),\qquad P_{\mathrm{th}}^{\mathrm{SBS}} \approx \dfrac{21\,A_{\mathrm{eff}}}{g_B\,L_{\mathrm{eff}}} \qquad (8)

where: $\nu_B$ — backward Brillouin shift; $v_A\approx5960$ m/s — acoustic velocity in silica; $g_B\approx5\times10^{-11}$ m/W — Brillouin gain, roughly 500× larger than $g_R$ . The narrow ≈tens-of-MHz linewidth means SBS is fully available only for narrow-line sources. Four-wave mixing builds up only when the total wavevector mismatch — linear dispersion plus the Kerr correction — is near zero. That condition ties its efficiency directly to the fiber's dispersion at the pump.

\Delta\kappa = \Delta\beta + 2\gamma P_p,\qquad \Delta\beta = \beta(\omega_s)+\beta(\omega_i)-2\beta(\omega_p) \approx \beta_2\,\Omega^2 \qquad (9)

where: $\Delta\kappa$ — total mismatch; $\Delta\beta$ — linear mismatch with $\Omega=\omega_s-\omega_p$ ; $P_p$ — pump power, giving Kerr contribution $2\gamma P_p$ . Efficient FWM needs $\Delta\kappa\approx 0$ , which near the pump requires slightly anomalous $\beta_2$ .

Derivation ADeriving the nonlinear Schrodinger equation from Maxwell's equations

The goal is to descend from the full wave equation with a nonlinear polarization to the compact envelope equation (3), and to see exactly where each of its three terms is born. The strategy is to factor the field into a fixed transverse mode times a slow envelope, expand the propagation constant so dispersion appears order by order, and keep the cubic polarization only to lowest order — the approximation chain every result here rests on.

Step 1. Split the polarization into linear and nonlinear parts, $P=\varepsilon_0(\chi^{(1)}E+\chi^{(3)}E^3)$ , and retain the component of the cubic response oscillating at the carrier frequency. For a monochromatic field, $E^3$ contains both a fundamental-frequency term and a third-harmonic term; the former produces the intensity-dependent index $\Delta n=n_2I$ , with $n_2=3\operatorname{Re}\chi^{(3)}/(4\varepsilon_0n_0^2c)$ under the stated convention. The third-harmonic component belongs to third-harmonic generation and is omitted from the envelope equation.

Step 2. Separate the field as $E(x,y,z,t)=F(x,y)A(z,t)e^{i(\beta_0z-\omega_0t)}$ , where $F$ solves the waveguide eigenproblem from the previous equations. For a single guided mode whose transverse profile is approximately unchanged along the fibre, the overlap integrals of $|F|^2$ and $|F|^4$ reduce to the effective area $A_{\rm eff}$ . The remaining pulse dynamics therefore depend primarily on $z$ and retarded time.

Step 3. Expand $\beta(\omega)=\beta_0+\beta_1(\omega-\omega_0)+\frac12\beta_2(\omega-\omega_0)^2+\cdots$ , and move to the retarded frame $T=t-\beta_1z$ , which removes the uniform group-delay term. Retaining only the second-order dispersion leaves $\beta_2$ as the leading dispersive contribution and hence the group-velocity dispersion (GVD). Why: the pulse's own frame strips away the uniform drift, exposing dispersion as the leading reshaping effect.

Step 4. Invoke the slowly varying envelope approximation, neglecting $\partial^2A/\partial z^2$ relative to $\beta_0\partial A/\partial z$ , and use $(\omega-\omega_0)\rightarrow i\partial_T$ to convert the frequency expansion into time derivatives. Why: the envelope varies slowly compared with the optical carrier and over one wavelength, so its rapid second-order longitudinal variation can be neglected.

Step 5. Collect the pieces: linear loss enters as $\alpha A/2$ , second-order dispersion as $i\beta_2\partial_T^2A/2$ , and the instantaneous Kerr response as $i\gamma|A|^2A$ , with $\gamma=n_2\omega_0/(cA_{\mathrm{eff}})$ . Assembling these terms gives equation (3). Why: adding the delayed Raman response and other higher-order effects upgrades this to the generalized NLSE used, for example, in supercontinuum modelling.

Go deeper · Derivation B · 1 minSPM chirp and the fundamental soliton condition

Derivation BSPM chirp and the fundamental soliton condition

The goal is to show, without solving the full equation, why a specific power turns a spreading pulse into a shape-invariant soliton. The plan is to isolate pure self-phase modulation in the lossless dispersionless limit, read off its chirp, then switch on anomalous dispersion and demand that its reshaping exactly undo that chirp — a balance that reduces to equating two lengths and fixes the peak power for a given width.

Step 1. With $\beta_2=0$ the NLSE integrates trivially: $A(z,T)=A(0,T)\exp[i\gamma P(T)z]$ , so the amplitude is untouched and the phase is $\phi_{\mathrm{NL}}=\gamma P(T)z$ , peaking at $\phi_{\max}=\gamma P_0 L$ , and the broadened spectrum shows roughly $M\approx\phi_{\max}/\pi+\tfrac12$ peaks. Why: pure SPM is phase-only — it cannot move energy in time, only paint new frequencies onto the envelope.

Step 2. Differentiate to get the chirp, $\delta\omega(T)=-\gamma L\,\partial P/\partial T$ : the leading edge red-shifts, the trailing edge blue-shifts, peak chirp growing with $\phi_{\max}/T_0$ . Why: this red-then-blue ordering is exactly what anomalous dispersion can reverse, which is why a balance is possible.

Step 3. Turn on anomalous dispersion, $\beta_2<0$ , in which blue travels faster than red. The blue trailing edge chases the red leading edge, compressing the pulse against dispersive spreading. Why: the sign of $\beta_2$ is decisive — in normal dispersion the two effects add and the pulse only broadens, so solitons live solely on the anomalous side.

Step 4. Impose the balance quantitatively: dispersive spreading acts over $L_D=T_0^2/|\beta_2|$ , nonlinear self-action over $L_{\mathrm{NL}}=1/(\gamma P_0)$ ; setting $L_D=L_{\mathrm{NL}}$ gives $N^2=\gamma P_0 T_0^2/|\beta_2|=1$ . Why: when the two characteristic lengths coincide the two effects proceed at the same rate and cancel step for step, leaving the shape stationary.

Step 5. The exact fundamental solution is $A=\sqrt{P_0}\,\mathrm{sech}(T/T_0)\,e^{i\gamma P_0 z/2}$ , invariant in shape; integer $N>1$ give higher-order solitons that breathe with period $z_0=\tfrac{\pi}{2}L_D$ . Why: the $\mathrm{sech}$ is the eigenshape for which SPM and GVD are pointwise opposite, so only an overall phase accumulates and the intensity profile never changes.

Go deeper · Derivation C · 1 minParametric gain and phase matching in four-wave mixing

Derivation CParametric gain and phase matching in four-wave mixing

The goal is to explain why FWM is exquisitely sensitive to dispersion while the Kerr effects are not — deriving the exponential parametric gain and the condition that switches it on. The approach is to seed a strong undepleted pump with weak signal and idler waves, keep the resonant terms of the coupling, and solve for a growth rate, which turns out real only inside a narrow window of phase mismatch.

Step 1. From the NLSE with three frequency components the signal obeys $\partial_z A_s=i\gamma[2P_pA_s+A_p^2A_i^*e^{i\Delta\beta z}]$ , with the conjugate equation for the idler. Why: the $2P_p$ term is XPM from the pump, while $A_p^2A_i^*$ is the parametric coupling that actually transfers energy between signal and idler.

Step 2. Absorb the pump-induced phase by writing $A_s=a_s e^{2i\gamma P_p z}$ and likewise for the idler; the coupling then carries the total mismatch $\Delta\kappa=\Delta\beta+2\gamma P_p$ . Why: the pump accumulates the self-phase shift $\gamma P_p z$ , while the signal and idler acquire cross-phase shifts $2\gamma P_p z$ , so the Kerr contribution enters the phase-matching condition as $2\gamma P_p$ .

Step 3. Seek solutions growing as $e^{gz}$ ; the two coupled equations give the eigenvalue $g^2=(\gamma P_p)^2-(\Delta\kappa/2)^2$ , so real (exponential) gain requires $|\Delta\kappa|<2\gamma P_p$ . Why: outside that window the eigenvalue is imaginary and the sidebands merely oscillate — energy sloshes back to the pump instead of growing.

Step 4. Gain is maximized at $\Delta\kappa=0$ , giving $g_{\max}=\gamma P_p$ , which since $\Delta\beta\approx\beta_2\Omega^2$ means $\beta_2\Omega^2=-2\gamma P_p$ — slightly anomalous dispersion near the pump. Why: this is why FWM and parametric amplifiers are engineered right at the zero-dispersion wavelength, where the phase-matching window is widest and the gain bandwidth largest.

Step 5. The same solution names the products: the idler appears at $2\omega_p-\omega_s$ , the basis of wavelength conversion, and extended over an octave the cascade underlies supercontinuum generation. Why: because photon number is conserved, FWM is coherent and reversible — unlike Raman and Brillouin, it deposits no energy in the medium.

Limiting cases

Which effect rules is decided by comparing the fiber length to $L_D$ and $L_{\mathrm{NL}}$ , and the launched power to the scattering thresholds.

Case	Result	Meaning
Dispersion-dominated	$L_{\mathrm{NL}} \gg L_D$	Low power or short pulses: GVD reshapes the pulse and nonlinearity is negligible — ordinary linear transmission analysis holds.
Nonlinearity-dominated	$L_{\mathrm{NL}} \ll L_D$	High power, broad pulse: pure SPM broadens the spectrum with a nearly frozen temporal shape — used for spectral broadening and compression.
Soliton regime	$L_D \sim L_{\mathrm{NL}},\ \beta_2<0$	Fundamental soliton at $N^2=L_D/L_{\mathrm{NL}}=1$ : SPM chirp exactly cancels dispersive spreading.
Below scattering thresholds	$P_0 \ll P_{\mathrm{th}}^{\mathrm{SBS}},\,P_{\mathrm{th}}^{\mathrm{SRS}}$	Kerr effects (SPM, XPM, FWM) act coherently with no exponential power transfer to Stokes waves.
Narrow-line continuous wave	$\Delta\nu \ll \Delta\nu_B \sim 20\text{-}100\ \mathrm{MHz}$	Full Brillouin gain is available; SBS has the lowest threshold and caps the launchable power in coherent links.

Worked examples

1 · How a 10⁻²⁰ nonlinearity becomes radians of phase (confidence builder)

Standard single-mode fiber at 1550 nm: $n_2=2.6\times10^{-20}\ \mathrm{m^2/W}$ , $A_{\mathrm{eff}}=80\ \mu\mathrm{m}^2$ , loss $0.2$ dB/km, length $L=100$ km, launched peak power $P_0=100$ mW. Find $\gamma$ , the effective length, and the accumulated nonlinear phase — and check how tiny the raw index change is.

Step 1. Nonlinear parameter from equation (2): $\gamma=\dfrac{2\pi n_2}{\lambda A_{\mathrm{eff}}}=\dfrac{2\pi(2.6\times10^{-20})}{(1.55\times10^{-6})(80\times10^{-12})}=1.32\times10^{-3}\ \mathrm{W^{-1}m^{-1}}=\mathbf{1.32\ W^{-1}km^{-1}}$ .

Step 2. Convert loss to base- $e$ : $\alpha=0.2/4.343=0.0461\ {\rm km}^{-1}$ , so $1/\alpha=21.7\ {\rm km}$ . The effective length is $L_{\rm eff}=(1-e^{-\alpha L})/\alpha=(1-e^{-4.61})/0.0461=21.5\ {\rm km}$ . Thus the nonlinear phase accumulated over the 100 km fibre is equivalent to that produced by about 21.5 km of lossless fibre at the input power, because the optical power decays exponentially along the fibre.

Step 3. Accumulated nonlinear phase, equation (5): $\phi_{\mathrm{NL}}=\gamma P_0 L_{\mathrm{eff}}=(1.32)(0.100)(21.5)=\mathbf{2.84\ rad}\approx 0.9\pi$ . Yet the raw index bump is minuscule: $I=P_0/A_{\mathrm{eff}}=0.1/80\times10^{-12}=1.25\times10^{9}\ \mathrm{W/m^2}$ , giving $\Delta n=n_2 I=3.3\times10^{-11}$ . A one-part-in- $10^{10}$ index change accumulates to nearly a radian of phase — because confinement and length multiply it.

2 · Peak power and energy of a fundamental soliton

Same fiber, now in the anomalous window: $\beta_2=-20\ \mathrm{ps^2/km}$ , $\gamma=1.32\ \mathrm{W^{-1}km^{-1}}$ . For a soliton of width $T_0=10$ ps (full-width $\approx1.76\,T_0=17.6$ ps), find $L_D$ , the required peak power, and the pulse energy.

Step 1. Dispersion length, equation (4): $L_D=T_0^2/|\beta_2|=(10\ \mathrm{ps})^2/(20\ \mathrm{ps^2/km})=\mathbf{5.0\ km}$ .

Step 2. Set $N=1$ in equation (6): $P_0=|\beta_2|/(\gamma T_0^2)=20/[1.32(100)]=0.152\ {\rm W}=152\ {\rm mW}$ . This is the peak power of the fundamental soliton. Consistency check: $L_{\rm NL}=1/(\gamma P_0)=4.99\ {\rm km}\approx L_D$ , as the soliton condition demands.

Step 3. A $\operatorname{sech}^2$ pulse has energy $E=\int P_0\operatorname{sech}^2(T/T_0)\,dT=2P_0T_0=2(0.152)(10\times10^{-12})=3.04\ {\rm pJ}$ . Thus a $17.6$ ps FWHM fundamental soliton in standard fibre requires a peak power of about $152$ mW and a pulse energy of about $3.0$ pJ to maintain its ideal soliton shape.

3 · Why Brillouin, not Raman, caps a narrow-line laser

Same fiber, $A_{\mathrm{eff}}=80\ \mu\mathrm{m}^2$ , $L_{\mathrm{eff}}=21.5$ km $=2.15\times10^4$ m. Take $g_R=1\times10^{-13}$ m/W and $g_B=5\times10^{-11}$ m/W. Compare the Brillouin shift and the two stimulated-scattering thresholds for a narrow-linewidth CW source.

Step 1. Brillouin shift, equation (8): $\nu_B=\dfrac{2n_0v_A}{\lambda}=\dfrac{2(1.45)(5960)}{1.55\times10^{-6}}=\mathbf{11.2\ GHz}$ , in the backward direction.

Step 2. SBS threshold: $P_{\mathrm{th}}^{\mathrm{SBS}}=\dfrac{21A_{\mathrm{eff}}}{g_B L_{\mathrm{eff}}}=\dfrac{21(80\times10^{-12})}{(5\times10^{-11})(2.15\times10^4)}=\mathbf{1.6\ mW}$ .

Step 3. SRS threshold: $P_{\mathrm{th}}^{\mathrm{SRS}}=\dfrac{16A_{\mathrm{eff}}}{g_R L_{\mathrm{eff}}}=\dfrac{16(80\times10^{-12})}{(1\times10^{-13})(2.15\times10^4)}=\mathbf{0.60\ W}$ . SBS trips at ≈1.6 mW — nearly 400× below the ≈0.6 W Raman threshold — so for a narrow-line CW laser Brillouin scattering is the first and lowest ceiling on launched power.

Physics of the portrait

Nonlinear fiber optics is a young field with a clear founder: almost every effect in this article was first seen in a fiber by one experimentalist at Bell Labs within a single remarkable decade.

Origins · 1 minRoger Stolen and how this was discovered

Roger Stolen

1938– · experimental founder of nonlinear fiber optics at Bell Labs

Roger H. Stolen is, more than anyone, the person who turned the abstract $\chi^{(3)}$ of nonlinear optics into the working physics of the optical fiber. When Corning demonstrated low-loss silica fiber in 1970, it created for the first time a medium in which weak nonlinearities could accumulate over kilometers, and Stolen — with collaborators including E. P. Ippen, Chinlon Lin, and James Gordon — went on to observe essentially the entire catalogue. In 1972 he and Ippen reported stimulated Brillouin scattering in single-mode fiber, and with Ippen and Tynes stimulated Raman oscillation the same year; the mid-1970s brought the first fiber four-wave-mixing experiments; and in 1978, with Lin, he measured self-phase modulation directly, mapping the broadened spectrum against the $M\approx\phi_{\max}/\pi$ count. His crowning collaboration came in 1980, when Mollenauer, Stolen, and Gordon observed the fundamental optical soliton that Hasegawa and Tappert had predicted in 1973 — the experimental birth of soliton communications. Stolen also made the careful measurements of $n_2$ , $g_R$ , and the silica Raman gain spectrum that every fiber engineer still uses. His work is why the abstractions of this article are also a laboratory reality.

History

1970

— Corning's Kapron, Keck and Maurer reach 20 dB/km in silica fiber, giving kilometre interaction lengths

1972

— Ippen and Stolen observe stimulated Brillouin and Raman scattering in single-mode fiber

1973

— Hasegawa and Tappert predict optical solitons from the nonlinear Schrodinger equation

1978

— Stolen and Lin measure self-phase modulation directly in silica fiber

1980

— Mollenauer, Stolen and Gordon observe the fundamental fiber soliton

Going further

Texts. G. P. Agrawal, Nonlinear Fiber Optics, 5th ed. (Academic Press) — the definitive treatment, Ch. 2 (NLSE), 4–5 (SPM, solitons), 8–9 (Raman, Brillouin), 10 (FWM) · Saleh & Teich, Fundamentals of Photonics, 3rd ed., Ch. 21 and 24 · R. W. Boyd, Nonlinear Optics, 4th ed., Ch. 4 (Kerr), 9–10 (stimulated scattering, parametric) · R. Loudon, The Quantum Theory of Light, 3rd ed. (scattering) · Demtröder, Laser Spectroscopy (Raman and Brillouin). Key numerical benchmarks: $n_2^{\mathrm{silica}}\approx2.6\times10^{-20}\ \mathrm{m^2/W}$ with $n_2=3\,\mathrm{Re}\,\chi^{(3)}/(4\varepsilon_0 c n_0^2)$ ; $\gamma=1.32\ \mathrm{W^{-1}km^{-1}}$ at $A_{\mathrm{eff}}=80\ \mu\mathrm{m}^2$ , $\lambda=1.55\ \mu\mathrm{m}$ ; $L_{\mathrm{eff}}=21.5$ km and $\phi_{\mathrm{NL}}=2.84$ rad for $\alpha=0.2$ dB/km, 100 km, 100 mW; the sech soliton $N^2=\gamma P_0 T_0^2/|\beta_2|$ gives $P_0=152$ mW, $E=3.0$ pJ at $T_0=10$ ps, $\beta_2=-20\ \mathrm{ps^2/km}$ ; Raman peak ≈13.2 THz (≈440 cm⁻¹), $g_R\sim10^{-13}$ m/W, $P_{\mathrm{th}}^{\mathrm{SRS}}\approx0.60$ W; Brillouin $\nu_B=2n_0v_A/\lambda=11.2$ GHz at $v_A=5960$ m/s, $g_B\sim5\times10^{-11}$ m/W, $P_{\mathrm{th}}^{\mathrm{SBS}}\approx1.6$ mW; FWM parametric gain $g=\sqrt{(\gamma P_p)^2-(\Delta\kappa/2)^2}$ .

Builds on: Waveguide Modes and the Wave Equation · Chromatic and Modal Dispersion in Fibers · Single-Mode and Multimode Operation · Fiber Attenuation and Loss Mechanisms. Leads to: Dispersion Management and Compensation · Microring Resonators and Optical Cavities on Chip · Integrated Photonic Waveguides and Circuits · Polarization-Mode Dispersion and Birefringent Fibers.

Ask the tutor: Ask me to walk through why a 152 mW soliton and a 1.6 mW Brillouin ceiling can coexist in the very same fiber, and what that tension means for a real transmission link.

If SPM does not change the pulse shape, how do fibers compress pulses?

SPM by itself is a phase operation: it paints a chirp on the pulse — red at the front, blue at the back — but leaves the temporal envelope untouched. Compression happens only when that chirp meets dispersion. In an anomalous-dispersion fiber the blue back travels faster than the red front and catches up, squeezing the pulse in time; this is the soliton mechanism, and pushed harder it is how fiber compressors reach femtosecond durations. The recipe is always two steps: SPM to create the chirp, dispersion to turn that chirp into a shorter (or longer) pulse.

What is the effective area, and why does a smaller core raise the nonlinearity?

The effective area $A_{\mathrm{eff}}$ is the mode's power divided by its peak intensity — roughly the cross-section over which the light is spread, formally $(\int|F|^2)^2/\int|F|^4$ . Because the nonlinear index depends on intensity, not power, packing the same power into a smaller area raises the accumulated phase: $\gamma=2\pi n_2/(\lambda A_{\mathrm{eff}})$ scales as $1/A_{\mathrm{eff}}$ . Standard telecom fiber has $A_{\mathrm{eff}}\approx80\ \mu\mathrm{m}^2$ and $\gamma\approx1.3\ \mathrm{W^{-1}km^{-1}}$ ; highly nonlinear and photonic-crystal fibers shrink $A_{\mathrm{eff}}$ to a few $\mu\mathrm{m}^2$ and raise $\gamma$ ten- to hundredfold, which is why they generate supercontinuum in centimeters.

Why is the effective length so much shorter than the fiber?

Because nonlinear effects need intensity, and intensity fades with loss. The power decays as $e^{-\alpha z}$ , so the nonlinear phase integrates to $\phi_{\mathrm{NL}}=\gamma P_0\int_0^L e^{-\alpha z}\,\mathrm{d}z=\gamma P_0 L_{\mathrm{eff}}$ with $L_{\mathrm{eff}}=(1-e^{-\alpha L})/\alpha$ . For any fiber much longer than $1/\alpha$ this saturates at $L_{\mathrm{eff}}\to1/\alpha$ , about 21.7 km at $0.2$ dB/km. So a 100 km span behaves nonlinearly like only its first ≈21 km, where the light is still bright. This is also why distributed amplification, by keeping the power up, actually increases the accumulated nonlinearity.

How do stimulated Raman and Brillouin scattering differ in practice?

Both are stimulated scattering off a material vibration, but on wildly different scales. Raman scatters off high-frequency molecular vibrations: a broad ≈40 THz gain band peaking ≈13.2 THz below the pump, in both directions, with a high threshold (≈0.6 W in a long standard fiber). Brillouin scatters off low-frequency acoustic waves: a narrow line only ≈11 GHz below the pump, backward only, but with a gain ≈500× larger, so its threshold is a few milliwatts for a narrow-line source. In practice Brillouin is the first ceiling with a narrow CW laser and is suppressed by dithering the source; Raman matters for broadband or high-power signals and is harnessed for distributed amplification.

Why does four-wave mixing care about dispersion when the Kerr effects do not?

SPM and XPM are phase effects on co-propagating light — they do not need different frequencies to stay in step, so dispersion barely gates them. FWM transfers energy between distinct frequencies ( $\omega_s+\omega_i=2\omega_p$ ), and that transfer only accumulates if the waves keep a fixed phase relationship. Dispersion makes them travel at different phase velocities, so unless the total mismatch $\Delta\kappa=\Delta\beta+2\gamma P_p$ is near zero, energy that flows into the idler on one stretch flows back out on the next. That is why FWM is efficient near the zero-dispersion wavelength — and why WDM systems keep a little local dispersion to starve inter-channel FWM.

Is the fiber Kerr effect the same physics as the bulk optical Kerr effect?

Yes — the microscopic origin is identical: the real part of $\chi^{(3)}$ in silica, which makes the index rise with intensity as $\Delta n=n_2 I$ . What changes is the geometry of exposure. In a bulk sample the beam diffracts and the interaction length is centimeters, so the nonlinear phase is negligible at ordinary powers. A fiber removes both limits: the mode is guided (no diffraction) and confined to tens of $\mu\mathrm{m}^2$ , and the interaction runs for kilometers. The product of small $A_{\mathrm{eff}}$ and long $L_{\mathrm{eff}}$ is what promotes the same weak $n_2$ to the governing physics of the fiber.