The Hidden Cost of a Wrong Mirror:
Why Dispersion Control Defines Your Pulse

In femtosecond optics, a mirror is never just a mirror. The choice between a standard dielectric reflector and a dispersion-compensating mirror can stretch a 30 fs pulse into 200 fs — or preserve it intact through a complex beam path. This tutorial explains why.

A femtosecond pulse is a fragile object. Its temporal compactness — sometimes only ten optical cycles — depends on the precise phase relationship between every spectral component it contains. Disturb that phase, and the pulse falls apart in time, even though no photon has been lost.

Every optical element a pulse encounters imprints additional phase. Substrates, coatings, even air itself behave as frequency-dependent delay lines: blue light arrives at a slightly different time than red. For a continuous-wave laser this is invisible. For a 30 fs pulse, it is catastrophic. After bouncing off a handful of "ordinary" mirrors, your sub-50 fs source can emerge stretched to several hundred femtoseconds, with peak power collapsed by a factor of ten.

This article walks through the physics of dispersion in reflection, quantifies the damage caused by uncontrolled mirrors, and shows how Group Delay Dispersion (GDD) compensation — implemented via chirped or Gires-Tournois Interferometer (GTI) mirrors — restores the pulse you actually paid for.

§ 01The Physics: Phase, Not Power

Consider a transform-limited Gaussian pulse with electric field

Its spectrum, obtained by Fourier transform, has flat spectral phase φ(ω) = 0. All frequencies are phase-locked at t = 0, which is precisely why the pulse is short.

When the pulse reflects off a real mirror, the coating imprints a frequency-dependent reflection phase φ(ω). Expanding around the carrier frequency ω₀:

The first term is a constant (an unobservable carrier-envelope offset). The second is a group delay (the whole pulse arrives a bit later — also harmless). The damage starts at the second-order term φ″, called the Group Delay Dispersion (GDD), with units of fs². This is where the trouble begins.

A non-zero GDD means the instantaneous frequency drifts linearly across the pulse — the pulse becomes chirped. The temporal envelope broadens according to:

Note the inverse-square dependence on input duration: shorter pulses are exponentially more sensitive to dispersion. A 100 fs pulse barely notices 500 fs² of GDD. A 20 fs pulse, hit with the same GDD, more than doubles in width.

§ 02Where the GDD Comes From

Three sources dominate the dispersion budget of a typical femtosecond beamline:

1. Substrate-traversed coatings

Even "high-reflector" coatings have penetration depth into the dielectric stack. The longer wavelengths typically penetrate more deeply than the shorter ones, generating a small but non-zero GDD that is positive in sign and varies with wavelength. A standard quarter-wave-stack mirror at 800 nm typically contributes +30 to +80 fs² per bounce — not catastrophic alone, but it accumulates.

2. Beam-steering through transmissive optics

Lenses, windows, beam splitters, and waveplates all add positive GDD via material dispersion. A 5 mm fused silica window adds approximately +180 fs² at 800 nm. A pair of 25 mm thick laser line filters? Over +1800 fs² per pass.

3. Air itself

Often forgotten. One meter of air at standard conditions adds about +22 fs² at 800 nm. In a long delay-line setup, this becomes consequential.

A pulse is not destroyed by absorption. It is destroyed by phase.

§ 03Two Mirrors, Two Fates

Below we run a numerical experiment. A transform-limited 30 fs Gaussian pulse at 800 nm is propagated through 8 reflections. In Case A, each mirror is a standard high-reflector (+50 fs²/bounce). In Case B, each mirror is a dispersion-compensating chirped mirror (−50 fs²/bounce). The remaining beamline contributes a residual +400 fs² of material dispersion in both cases.

§ 04The Chirped Mirror Solution

A chirped mirror is a multilayer dielectric stack engineered such that longer wavelengths penetrate deeper into the structure before being reflected. Because they travel a longer optical path inside the coating, red components emerge with an additional negative group delay relative to the blue. The result: negative GDD, opposite in sign to the positive material dispersion of glass.

Modern designs achieve nearly any target GDD curve over bandwidths exceeding an octave. Two-mirror complementary pairs cancel residual GDD oscillations to deliver flat, broadband negative dispersion across the full Ti:Sapphire emission band (650–1100 nm) and the 1030 nm Yb amplifier band.

How a chirped mirror differs from a GTI mirror

§ 05Practical Decision Framework

How should you choose? Three quick rules of thumb, calibrated to real beamline practice:

Rule 1 — The 100 fs threshold

If your pulse duration is longer than ~100 fs and your beamline has fewer than 6 reflections, standard HR mirrors are usually fine. The accumulated GDD won't exceed ~500 fs², which broadens a 100 fs pulse by less than 5%.

Rule 2 — The 50 fs caution zone

For pulses between 30 fs and 100 fs, audit your dispersion budget element-by-element. If the total exceeds 0.3 × τ², plan compensation. A common mistake: assuming a "low-dispersion" coating is enough. It rarely is.

Rule 3 — Sub-30 fs is a different regime

For pulses below 30 fs, every transmissive element matters and chirped mirrors become non-negotiable. Sub-10 fs work additionally requires third-order dispersion (TOD) management — typically using complementary chirped mirror pairs or prism-pair compressors.

§ 06Live Demonstration: A Real HR Mirror

The plots above used idealized GDD values. Below we run the same calculation with real measured phase data from a commercial high-reflector coating designed for 750–850 nm. We compare two operating points on the same physical mirror:

• Operating in the design band (780–830 nm) — phase is well-controlled, GDD stays within ±8 fs²
• Operating outside the design band (690–740 nm) — phase is wildly chaotic, GDD spikes exceed +770,000 fs² at coating resonances

The pulse below is a transform-limited 30 fs Gaussian, propagated through 4 reflections by applying the measured spectral phase φ(ω) = ½·GDD·(ω−ω₀)² (locally) and inverse-Fourier-transforming back to time domain. This is the same calculation a real beamline performs — except here it runs live in your browser on actual coating data.

The asymmetric, multi-peak structure in the chaotic case is not a calculation artifact — it is the inevitable consequence of a phase function with sharp local features. Each spike in φ″(ω) acts as a small chirped delay line for the spectral component sitting on top of it, scattering portions of the pulse energy into time-shifted satellites. This kind of damage cannot be undone by adding more compensation downstream. It can only be avoided by operating the mirror within its design band.

Practical takeaway: When sourcing mirrors, the reflectance curve is not enough. Always request the GDD and TOD curves over the full wavelength range your pulse spectrum covers, not just the center wavelength. A 30 fs pulse at 800 nm has measurable spectral content from 750 nm to 870 nm — the mirror must behave well across that entire window, not just at 800 nm.

§ 07Closing: Phase Is the Hidden Asset

The investment a femtosecond user makes in a laser oscillator or amplifier is fundamentally an investment in spectral phase coherence. Every additional fs² of uncompensated GDD silently erodes that asset. A "saved" $400 on standard mirrors can transform a $200,000 system into one that performs like a $30,000 source.

The good news: dispersion is deterministic. It is calculable, measurable, and — with the right reflectors — fully cancellable. Treat the GDD budget the way an electronics designer treats an impedance budget: account for every component, sign each contribution, and engineer the sum to zero.

Your pulse will thank you. So will your application — whether it is two-photon imaging, attosecond science, micromachining, or quantum gate operations. In every one of those cases, peak intensity is the figure of merit, and peak intensity follows from short pulses, and short pulses follow from controlled phase.

For tailored advice on dispersion compensation in your specific setup, contact the WaveQuanta applications team. We routinely produce custom GDD specifications for OEM and research integrators.