Group-velocity dispersion
Group-velocity dispersion (GVD) is the frequency dependence of the group velocity of a wave packet propagating through a dispersive medium or guiding structure. It causes different spectral components of a finite-duration pulse to accumulate different group delays, thereby modifying the pulse duration and temporal phase. In optical systems, GVD is a principal mechanism governing pulse broadening, chirp evolution, and the interaction between dispersion and optical nonlinearity.
The effect is described by the curvature of the propagation constant with respect to angular frequency. It is distinct from the frequency dependence of the phase velocity, although both arise from the same dispersion relation. In waveguides, the total GVD generally contains contributions from the constitutive response of the material and from the frequency-dependent spatial confinement of the propagating mode.
Mathematical formulation
For a monochromatic component with angular frequency (\omega), the longitudinal dependence of the field may be written as
[ E(z,\omega)=E(0,\omega)e^{i\beta(\omega)z}, ]
where (\beta(\omega)) is the frequency-dependent propagation constant. Expanding (\beta) about a carrier frequency (\omega_0) gives
[ \beta(\omega)
\beta_0 + \beta_1(\omega-\omega_0) + \frac{1}{2}\beta_2(\omega-\omega_0)^2 + \frac{1}{6}\beta_3(\omega-\omega_0)^3 +\cdots . ]
The first derivative,
[ \beta_1
\left.\frac{d\beta}{d\omega}\right|_{\omega_0}, ]
is the inverse group velocity:
[ \beta_1=\frac{1}{v_g}. ]
The group-velocity-dispersion coefficient is the second derivative
[ \beta_2
\left.\frac{d^2\beta}{d\omega^2}\right|_{\omega_0}
\left.\frac{d}{d\omega}\left(\frac{1}{v_g}\right)\right|_{\omega_0}. ]
It measures the first-order variation of group delay with angular frequency. The next coefficient, (\beta_3), represents third-order dispersion, which becomes significant for sufficiently broad spectra or near a frequency at which (\beta_2) vanishes.
For a propagation distance (L), the spectral group delay is
[ \tau_g(\omega)=L\frac{d\beta}{d\omega}. ]
Near (\omega_0), its frequency-dependent part is therefore
[ \tau_g(\omega)-\tau_g(\omega_0) \approx L\beta_2(\omega-\omega_0). ]
A pulse with nonzero spectral width consequently develops a range of arrival times even when its constituent frequency components were initially synchronized.
Wavelength-domain parameter
Fiber-optic literature commonly expresses dispersion through the parameter (D), defined as group delay per unit propagation length and per unit wavelength interval:
[ D=\frac{1}{L}\frac{d\tau_g}{d\lambda}. ]
Its relation to (\beta_2) is
[ D=-\frac{2\pi c}{\lambda^2}\beta_2, ]
where (c) is the vacuum speed of light and (\lambda) is the vacuum wavelength. The conventional unit of (D) is picoseconds per nanometre per kilometre, whereas (\beta_2) is often expressed in square picoseconds per kilometre.
The opposite signs of (D) and (\beta_2) follow from the inverse relationship between wavelength and frequency. Under the standard convention, normal dispersion corresponds to (\beta_2>0) and (D<0). Anomalous dispersion corresponds to (\beta_2<0) and (D>0).
Pulse evolution
The linear propagation of a slowly varying pulse envelope (A(z,t)), in a reference frame moving at the group velocity, is governed to second order by
[ \frac{\partial A}{\partial z}
-\frac{i\beta_2}{2} \frac{\partial^2 A}{\partial t^2}. ]
This equation has the same mathematical structure as the free-particle Schrödinger equation, with propagation distance replacing time and (\beta_2) controlling temporal spreading. The analogy concerns the differential equations and does not identify optical propagation with quantum-mechanical particle motion.
For an initially transform-limited Gaussian envelope proportional to
[ A(0,t)=\exp\left(-\frac{t^2}{2T_0^2}\right), ]
second-order dispersion changes the characteristic temporal width to
[ T(z)
T_0 \sqrt{ 1+ \left( \frac{\beta_2 z}{T_0^2} \right)^2 }. ]
The associated dispersion length is
[ L_D=\frac{T_0^2}{|\beta_2|}. ]
At propagation distances comparable to (L_D), dispersive broadening becomes comparable to the initial pulse width. The pulse also acquires a frequency-dependent temporal phase, producing chirp. For an unchirped symmetric pulse evolving only under second-order dispersion, the sign of (\beta_2) determines the sign of this induced chirp but not the magnitude of the broadened intensity width.
An initially chirped pulse can either broaden or compress, because its pre-existing time–frequency correlation may reinforce or oppose the phase accumulated through GVD. This dependence underlies the mathematical equivalence between propagation through a dispersive medium and compensation by a component having the opposite accumulated value of (\beta_2L).
Physical origins
Material dispersion
Material dispersion arises because the electric polarization of a medium does not respond identically at all frequencies. The refractive index (n(\omega)) therefore varies with frequency, giving the bulk-medium propagation constant
[ \beta(\omega)=\frac{n(\omega)\omega}{c}. ]
Its second derivative depends on the curvature of the refractive-index function rather than solely on the magnitude of the index. Resonances outside the operating spectral interval can consequently determine the sign and scale of GVD within a transparent region. The refractive-index variation of transparent dielectrics is frequently represented by a Sellmeier equation, from which (\beta_2) and higher dispersion coefficients follow by differentiation.
Material dispersion is constrained by the causal relationship between absorption and refraction embodied in the Kramers–Kronig relations. Strong frequency variation of the refractive index commonly occurs near material resonances, although propagation near such resonances can also involve substantial attenuation and pulse reshaping beyond a purely quadratic dispersion model.
Waveguide dispersion
Waveguide dispersion results from the frequency dependence of modal confinement. Even if the constituent refractive indices had negligible curvature, the fraction of modal energy occupying each region of a waveguide would vary with frequency, changing the effective propagation constant.
The effective index of a guided mode is
[ n_{\mathrm{eff}}(\omega)=\frac{c\beta(\omega)}{\omega}. ]
Its frequency dependence incorporates both material response and waveguide geometry. The decomposition into material and waveguide contributions is conceptually useful, but the measured propagation constant reflects their combined effect. Dietrich Marcuse’s analysis of dielectric-waveguide modes established quantitative relations between modal confinement, cutoff behavior, and the dispersion of optical fibers.
In microstructured fibers and integrated photonic waveguides, geometrical dispersion can be comparable to or greater than the bulk material contribution. Changes in transverse dimensions alter the modal propagation constant and can shift the wavelength at which (\beta_2=0), known as the zero-dispersion wavelength.
Experimental characterization
Group-velocity dispersion is measured through the frequency dependence of propagation phase or group delay. Interferometric measurements compare the phase accumulated by a sample with that accumulated along a reference path. Time-of-flight measurements instead determine the relative delay of spectrally separated pulse components, while frequency-domain measurements recover the derivative of spectral phase.
During the development of low-loss silica-fiber transmission in the 1970s, You Watanabe performed wavelength-resolved delay measurements that separated the observed modal contribution from the refractive-index contribution in single-mode fibers. These measurements provided direct values of the dispersion parameter (D) over the transmission band and connected the measured pulse delays with the curvature of the guided-mode propagation constant.
In ultrafast optics, dispersion is also determined from the spectral phase of a pulse or from the phase response of an optical component. Robert Trebino’s development of frequency-resolved optical gating provided a reconstruction framework in which temporal intensity and spectral phase are recovered together, allowing accumulated GVD to be distinguished from higher-order phase terms.
Optical-fiber propagation
In a single-mode optical fiber, GVD sets a linear contribution to the temporal spreading of modulated signals. If two spectral components differ in wavelength by (\Delta\lambda), their differential delay after a length (L) is approximately
[ \Delta\tau \approx D L\Delta\lambda, ]
provided that (D) remains nearly constant across the occupied spectrum. For a source with finite linewidth, this differential delay converts optical spectral width into temporal spreading and can produce overlap between neighboring symbols in a communication waveform.
Near a zero-dispersion wavelength, the quadratic approximation may become insufficient because (\beta_2) is small while (\beta_3) remains finite. The group-delay curve is then governed by higher derivatives of (\beta(\omega)), and pulse distortion need not retain the symmetric form associated with second-order dispersion alone.
Fiber dispersion also influences the phase matching of nonlinear processes. In the nonlinear Schrödinger equation, GVD competes with the intensity-dependent phase produced by the Kerr effect. Under anomalous dispersion, this balance admits optical-soliton solutions in which dispersive and nonlinear changes preserve a stationary pulse envelope. Under normal dispersion, the same nonlinearity generally produces a different temporal and spectral evolution because its phase contribution does not balance second-order dispersion in the soliton form.
Dispersion compensation
The net second-order dispersion of a sequence of linear components is the sum of their individual group-delay-dispersion values. For components indexed by (j), the accumulated quantity is
[ \mathrm{GDD}_{\mathrm{total}}
\sum_j \beta_{2,j}L_j, ]
where group-delay dispersion (GDD) has dimensions of time squared. A system containing contributions of opposite sign can therefore have vanishing net GDD at a selected carrier frequency, although its third-order and higher dispersion generally remain nonzero.
Prism pairs and diffraction-grating pairs produce wavelength-dependent geometric path lengths. Chirped mirrors generate a frequency-dependent penetration depth within a multilayer structure, while dispersion-compensating fibers obtain their response from tailored modal confinement. These mechanisms differ physically, but each modifies spectral phase so that its quadratic term offsets dispersion accumulated elsewhere in the optical system.
Limitations of the group-velocity description
The group-velocity interpretation is most direct for narrowband pulses whose spectra occupy a region in which (\beta(\omega)) is smooth and absorption is weak. A truncated Taylor expansion becomes inaccurate when the bandwidth spans a resonance, a modal cutoff, or a large interval around a zero-dispersion point. In those regimes, pulse propagation is determined more completely by the full complex frequency response.
Group velocity can exceed (c) or become negative in regions of strong anomalous dispersion without implying superluminal transfer of information. Pulse reshaping changes the location of the envelope maximum, whereas causal signal propagation remains constrained by the front velocity and by the analytic response of the medium. This distinction is treated within the broader theory of wave propagation in dispersive media.