 Etalon coating group delay

An etalon coating group delay is the frequency derivative of the reflection phase contributed by the two coatings during one internal round trip. At fixed incidence and polarization, write the internal reflection coefficients as $r_j(\omega )=|r_j(\omega )|e^{i{\theta }_j(\omega )}$. With the phase convention of etalon round-trip phase,The phases must refer to consistently chosen surfaces: moving a reference plane transfers a propagation contribution between the terms. Only the consistently defined total is directly comparable with the resonance spacing.

A constant coating phase changes the positions of resonances but contributes zero group delay. A linear phase produces constant group delay. Curvature produces group delay dispersion:A highly reflecting coating therefore cannot be characterized by reflectance alone. The manufacturer's ultrafast-mirror documentation separately specifies reflection dispersion and shows multiple-bounce arrangements for compensation.^[ref]

For an original constant-slope example, suppose propagation contributes $10\mathrm{ps}$ and the coatings together contribute $0.20\mathrm{ps}$. The local free spectral range is approximately $1/(10.20\mathrm{ps})=98.04\mathrm{GHz}$, rather than $100\mathrm{GHz}$. Ignoring coating delay would overassign that $0.20\mathrm{ps}$ to propagation when inferring a spacer thickness or group index. The values are stipulated, not a coating specification.

For an independent curvature example, let the summed coating phase contain $D(\omega -{\omega }_0{)}^2/2$, with $D=100\mathrm{f}{\mathrm{s}}^2$. Across a frequency interval $10\mathrm{THz}$, its group delay changes by $D2\pi \Delta f\approx 6.28\mathrm{fs}$. The derivative is with respect to angular frequency; omitting $2\pi$ changes the result.

Experimentally, one may measure adjacent resonance frequencies and compare their spacing with an independently characterized spacer and group index. A discrepancy alone cannot uniquely identify coating delay: spacer dispersion, incidence angle and geometry also affect the result. Etalon delay inferred from free spectral range gives an interval-averaged phase slope; identifying it with a single-frequency derivative requires sufficiently slow variation. The simple manufacturer's spacer relation explicitly neglects spacer dispersion.^[ref]