Cross-phase modulation (XPM)

Cross-phase modulation is the change in phase that one optical wave experiences because another wave, at a different wavelength or polarization, raises the refractive index of the shared medium through the Kerr effect. In a co-polarized pair the effect is twice as strong per unit intensity as self-phase modulation. In standard single-mode fiber at 1550 nm, with $\gamma \approx 1.3$ W⁻¹km⁻¹, a 10 mW neighboring channel over an 80 km span would impose up to about 0.55 rad on a probe if the two stayed aligned in time; dispersion-induced walk-off reduces this considerably in practice.

For two waves at frequencies ${\omega }_A$ and ${\omega }_B$, the index seen by wave A is

$$n_A=n_0+n_2{I}_A+2n_2{I}_B$$

and in a waveguide the corresponding nonlinear phase on A is

$${\varphi }_A=\gamma L_{\text{eff}}(P_A+2P_B)$$

The factor of 2 on the cross term arises from the symmetry of the ${\chi }^{(3)}$ tensor: for distinct frequencies there are twice as many permutations contributing to the cross term as to the self term. For linearly polarized waves with orthogonal polarizations in an isotropic medium the factor falls to 2/3. For $N$ co-propagating waves, each wave acquires a nonlinear phase proportional to its own power plus twice the sum of all other co-polarized channel powers.

Walk-off

Waves at different wavelengths travel at different group velocities in a dispersive medium. After a walk-off length they no longer overlap temporally, and XPM from any given symbol averages out. The group-delay difference per unit length is $d=D\mathrm{\Delta }\lambda$, and the walk-off length for a pulse or symbol of duration $T$ is

$$L_W\sim \frac{T}{D\mathrm{\Delta }\lambda }$$

Worked numbers. For SMF-28 with $D=17$ ps/(nm·km) and channels 100 GHz apart (0.8 nm at 1550 nm), $d=13.6$ ps/km. For 100 ps pulses (10 Gb/s), $L_W\approx 7.4$ km. For a 64 GBd signal (15.6 ps symbols) with a neighbor 75 GHz away (0.6 nm), $L_W\approx 1.5$ km. Both are much shorter than the effective length of an 80 km span (about 21 km), so XPM accumulates over the walk-off length and then averages out; channels farther away contribute progressively less.

A complete collision between two pulses in lossless fiber produces no net frequency shift, because the chirp imposed on the leading and trailing halves of the collision cancels. Fiber loss, amplification in the middle of a collision, or a collision that is only partly complete at the fiber ends leave a residual shift, which dispersion then converts into timing jitter.

Effects in WDM systems

In WDM systems, XPM produces:

Scaling with power and spacing. Higher per-channel power produces stronger XPM, in proportion to the neighbor's power. Closer channel spacing increases the number of nearby channels contributing and lengthens the walk-off for each.

Dispersion management. In links without per-span dispersion compensation, accumulated chromatic dispersion smears pulses, reducing peak intensities and consequently reducing XPM. In dispersion-compensated links (legacy 10G systems), peak power remains high and XPM is more severe. Partial dispersion mismatch was deliberately maintained in some designs to suppress XPM. Links carrying intensity-modulated channels next to phase-modulated coherent channels are particularly exposed, because on-off keying produces large power fluctuations that XPM transfers directly into phase noise on the neighbors.

Coherent receiver compensation. Digital signal processing in coherent receivers can partially compensate XPM through multi-channel nonlinear backpropagation, but at substantial computational cost, and only for channels the receiver actually detects. The remaining uncompensated XPM contributes to the effective noise floor that limits per-channel reach.

XPM is the principal nonlinear inter-channel penalty in DWDM transmission systems. Self-phase modulation acts within a channel, whereas XPM couples channels. In single-channel systems only SPM applies; in DWDM with $>$ 10 channels at high per-channel power, XPM typically exceeds SPM as the limiting nonlinearity.

Measurement and uses

XPM is measured with a pump-probe arrangement: an intensity-modulated pump and a weak CW probe at another wavelength are launched together, and the phase modulation transferred to the probe is detected with an interferometer or converted to intensity by a dispersive element or optical filter. The ratio of probe phase to pump power gives $2\gamma L_{\text{eff}}$ and so an independent measurement of $\gamma$.

The same process is useful in other settings:

• Optical switching and Kerr gating: a control pulse shifts the phase of a signal in one arm of an interferometer, as in the nonlinear optical loop mirror
• Wavelength conversion: XPM imprints the data of a pump onto a CW probe, and a filter on one probe sideband selects the converted signal
• Supercontinuum: XPM between solitons and dispersive waves extends the blue edge of fiber supercontinua

Pitfalls

The factor of 2 applies only to co-polarized waves. In real fibers the polarization states of channels evolve randomly, and averaged models (the Manakov equation) use a reduced coefficient. Estimates that ignore walk-off, like the 0.55 rad upper bound above, overstate XPM in dispersive fiber by a large factor. In low-dispersion fiber, such as dispersion-shifted fiber near 1550 nm, walk-off is slow and the upper bound is a closer approximation.

Common questions

Does XPM transfer energy between channels?

No. XPM alters only phase; energy transfer between channels comes from four-wave mixing and stimulated Raman scattering. Combined with dispersion, however, XPM phase modulation turns into intensity distortion.

References: Agrawal, Nonlinear Fiber Optics 6th ed. 2019; Boyd, Nonlinear Optics 4th ed. 2020; Saleh & Teich, Fundamentals of Photonics 3rd ed. 2019.