Every ultrafast laser needs a way to undo the dispersion its own gain medium and optics impose. Prism pairs and gratings do it with path length, at the cost of alignment drift, throughput loss and a footprint that grows with the pulse energy. Chirped mirrors do it inside a coating a few micrometres thick, with no alignment to drift and reflectivity that can exceed 99.9 percent. The trade-off is that the dispersion is designed into the layer structure, and whatever the design gets wrong cannot be adjusted afterwards with a knob.

How a mirror becomes dispersive
A conventional high reflector is periodic: pairs of high and low index layers, each a quarter wave thick at the design wavelength. Light at that wavelength reflects from the first few pairs and never penetrates far. Move away from the design wavelength and the reflection depth changes, which is precisely the mechanism chirped mirrors exploit.
Instead of a periodic stack, a chirped mirror uses an aperiodic sequence in which the layer pairs get progressively thicker toward the substrate. Short wavelengths reflect from the top pairs, long wavelengths penetrate deeper before they find a matching period. The extra penetration depth translates directly into group delay, and because the depth increases with wavelength, the group delay decreases with wavelength, which is negative group delay dispersion.
The ripple problem
The penalty for aperiodicity is impedance mismatch. Every transition between the front structure and the substrate, and every internal resonance in the stack, produces an oscillation in group delay as a function of wavelength. That oscillation, usually called GDD ripple, is what turns a beautifully compressed 30 fs pulse into a 45 fs pulse with satellite structure.
Designers fight it in two ways. The first is numerical: optimise the layer sequence so that the residual ripple is small and, critically, so that its period is short compared with the pulse bandwidth. The second is physical: design matched mirror pairs whose ripple curves are equal and opposite, and use them in sequence so the oscillations cancel. Paired designs routinely reach plus or minus 20 fs squared across an octave, which a single mirror cannot do.
Bandwidth, angle and the design triangle
Three requirements pull against each other. Broad bandwidth needs deep penetration, which increases ripple. Flat dispersion needs a long, carefully tapered structure, which increases the total layer count and the deposition time. High reflectance needs enough layer pairs at every wavelength, which again adds thickness.
The fourth constraint is angle. A chirped mirror specified at 5 degrees behaves differently at 15, and the difference is not a simple shift: the penetration profile changes, and the GDD curve distorts asymmetrically. Compressors that fold a beam through several bounces at slightly different angles accumulate that distortion. Designs should therefore be specified at the actual incidence angle, and the optomechanics should hold it to within a degree.
Manufacturing to a fraction of a nanometre
A chirped mirror for 800 nm might carry 60 to 120 layers, with individual thicknesses between 40 and 400 nm. A one percent thickness error on a 100 nm layer moves the phase by several degrees, and the accumulated error across a hundred layers destroys the dispersion curve even though the reflectance still looks acceptable.
This is why chirped mirrors are ion-beam-sputtered with in-situ broadband optical monitoring. The deposition is controlled by watching the actual spectral response and terminating each layer when the fit says so, rather than by timing a rate. Even so, the calibration of the monitor's own dispersion is a source of systematic error, and the good coating houses characterise it with reference samples rather than trusting the software.
Measuring what was made
Reflectance is easy and nearly useless on its own: a chirped mirror can hit 99.9 percent reflectance with the wrong dispersion entirely. The measurement that matters is group delay versus wavelength, usually by white-light spectral interferometry against a reference surface. From that, GDD is obtained by differentiation, and the ripple becomes visible as structure on the curve.
Two practical notes for buyers. Ask for the measured GDD curve over the full specified band, not a single number at the centre wavelength. And ask for the measurement at the specified angle, on the finished part, with the reference mirror identified, because a GDD curve without those three pieces of context cannot be compared between suppliers.
Where chirped mirrors earn their cost
Femtosecond oscillators, chirped-pulse amplifiers, optical parametric oscillators, high-harmonic generation and two-photon microscopy all rely on them. In each case the value proposition is the same: dispersion that does not drift, alignment that never needs touching, and throughput that stays above 99 percent for years. Against a grating compressor that must be re-optimised whenever the room temperature moves, a well-made chirped mirror pair is not an expensive component. It is the cheapest stability in the system.
Product page: precision front-surface mirrors — specifications, coating options and lead times.