Chirped mirrors push pulses to their limits

17 Jun 2002

Scientists at Austrian and Swiss laboratories are carrying out cutting-edge research to produce the world's shortest laser pulses. The work is opening up new vistas to chemists and biologists, who can now observe the world on a sub-atomic level. Rob van den Berg talks to the main players.

From Opto & Laser Europe September 2001

Field patterns

Using laser pulses of a few tens of femtoseconds, chemists can study the motion of atoms during chemical reactions and biologists can witness the earliest steps in natural processes, such as photosynthesis. However, the only way to catch a glimpse of even faster natural processes - the motion of electrons, for instance - is to use shorter pulses. Short pulses also have a broader spectrum, which is important in medical imaging techniques, and can concentrate the peak energy for materials processing. The generation of light pulses of less than 10 fs is no easy task. In the mid-1980s the the barrier was broken for the first time with a dye laser. It was 10 years before this result was improved on using a solid-state Ti:sapphire laser. Today's ultrafast all-solid-state sources continue to demonstrate an unsurpassed performance in terms of pulse duration, pulse repetition rates, average power and wavelength range.

Optical pulses of around 5 fs - corresponding to just two cycles of visible light - are produced using various methods, all of which are reliant on the subtle interplay between two effects: the spectral broadening of the pulses and the subsequent realignment of the spectral components. The latter of these effects has been the cause of some headaches. Only since the development of a special kind of chirped mirror in 1994 has it become possible to generate so few-cycle light pulses reliably.

One of the inventors of this optical element is Ferenc Krausz, who is still very much involved in the science of ultrashort laser pulses at the Vienna University of Technology in Austria.

Krausz explained what led him and his Hungarian colleague Robert Szipöcs to the idea: "To shorten a pulse in time, you need to increase its spectral bandwidth. To do this we use a nonlinear optical phenomenon - the Kerr effect - in which the refractive index of a material is dependent on the light intensity. The time-dependent intensity of a laser pulse leads to the modulation of the refractive index and thus to a time-dependent phase shift. The end result is a red shift in the leading part of the pulse and a blue shift in its trailing part."

Such so-called "self phase modulation" (SPM) occurs either in an optical fibre or inside the lasing material itself, and it leads not only to a greater bandwidth, but also to a pulse that is initially broadened in time, because the different frequency components travel at different speeds. This positive "group delay dispersion" can be a tremendous nuisance. In a typical femtosecond Ti:sapphire laser, one round trip through the laser crystal leads to a 100 fs delay between the leading and trailing edge of the pulse. So, to exploit the extra bandwidth produced for short-pulse generation, the red and blue components in the wings of the pulse have to be temporarily delayed and advanced, respectively.

However, dispersion comes in two flavours and, fortunately, certain optical components have a negative dispersion. In an optical fibre, where SPM and dispersion occur simultaneously, this could even give rise to solitons - optical pulses that can propagate over long distances in a nonlinear medium with a constant or periodically changing shape.

In many Ti:sapphire lasers the necessary dispersion compensation is supplied in the form of a prism pair. According to Krausz: "This introduces a dispersion that varies linearly as a function of frequency, which fits well with the positive variation that is induced by SPM. However, as the pulses become shorter, higher order dispersion effects start to play a role."

Cubic dispersion effects can still be suppressed by a suitable prism material, such as fused silica, but at around 800 nm - the centre wavelength of the Ti:sapphire laser - a strong fourth-order dispersion sets in that can no longer be corrected for.

Another limiting factor is the finite high-reflectivity bandwidth of the dielectric mirrors used in this scheme. Standard dielectric mirrors are composed of identical quarter-wave layer pairs. They provide high reflectance over about 200 nm around the central wavelength. Outside this region these relatively simple mirror structures have considerable transmission. Both effects lead to rapid pulse degradation for pulse widths approaching 10 fs, and a strong pulse-width sensitivity to cavity and prism alignment.

In 1994 Krausz and Szipöcs found a solution to these problems. They developed dispersion-compensated or chirped mirrors, in which the multilayer periodic structure is modulated across the mirror.

Szipöcs developed a computer program that calculates the correct mirror structure for the materials used and the bandwidth and reflectivity required. In this way the optimal dispersion to compensate for material effects is obtained, in addition to a much broader reflection bandwidth than is found in standard mirrors. With chirped mirrors the Ti:sapphire oscillator is capable of producing bandwidth-limited pulses with a duration of as little as 8 fs. Moreover, the bandwidth is centred around the peak of the pulse. This is important if further amplification is required: in this case a Ti:sapphire amplifier could be used. During the last few years, the work of other groups has resulted in improvements being made to the original design of the chirped mirror. One significant limitation that arose was owing to interference between light reflected at different depths inside the mirror structure. This results in residual oscillations in the group delay dispersion, which ultimately limits pulse shortening. Ursula Keller's group at the ETH in Zurich therefore developed double-chirped mirrors (DCMs) to combat the problem.

Günter Steinmeyer, a colleague of Keller, explained: "The 'classic' chirped mirror still suffers from two problems. One is the air/mirror interface, where a broadband antireflection coating is applied to prevent interference effects. But there are also oscillations that arise inside the mirror stack."

The group applied an additional chirp to the thickness ratio of the low- and high-index layers to smooth the backward coupling of the light. For such carefully optimized broadband DCMs, the peak-to-peak amplitude of residual dispersion oscillations has been reduced to less than 1 fs per reflection, to give only a small measured phase oscillation of the laser pulses.

Chirped mirror development

Several companies have the capability to manufacture high-tech DCMs. Steinmeyer said: "We carefully adjust the mirror design to a specific application and calculate the required layer thicknesses based on the materials' indices of refraction. During manufacturing, this design is corrected for growth errors and small deviations in the index of the sputtered material. The thickness of some of the layers has to be accurate to within 2 Å. Our main supplier, Nanolayers in Rheinbreitbach, uses an ion-beam sputtering technique, but Layertec near Jena in Germany attains a similar precision using magnetron sputtering."

Even though there is a trade-off between dispersion oscillations and reflection bandwidth, the DCM makes bandwidths of up to 200 THz accessible for pulse generation, which could, in principle, support 4 fs pulses. Steinmeyer said: "It is simply impossible to make an antireflection coating over the full range of an optical octave. You can avoid the problem of the bandwidth of the antireflection coating in the DCM by choosing an ambient medium with an index identical to that of one of the coating materials, such as fused silica. Then double-chirping works ideally for matching from silica to the coating. This means reverting the structure and coating it on the back of a substrate." Interference from the front surface can be avoided if the two substrate faces are not parallel. When Steinmeyer and colleagues patented this idea for a back-side coated chirped mirror they proposed to use a thin-lens substrate, but it is also possible to use a thin wedge - the solution put forward by Gabriel Tempea from the Vienna group. Their tilted-front-interface (TFI) mirrors can introduce a smooth dispersion over a full octave (of more than 300 THz) from 500 to 1000 nm, offering potential for 3 fs, quasi-single-cycle optical-pulse generation.

There are two schools of thought on chirped mirrors. On the one hand, exclusive use of mirrors allows for compact dispersion-compensation schemes, which is of particular importance for oscillators. It is this philosophy that guides Krausz and his colleagues in their spin-off company, Femtolasers, which prides itself in offering the commercial laser system with the shortest pulses - of around 10 fs - available to date.

If, on the other hand, one wants to generate the shortest pulses possible, the continuous adjustment of dispersion is required, which only becomes feasible through the use of chirped mirrors in combination with adjustable prisms. Although this "adds a significant amount of complexity", it has the advantage of a more favourable higher-order dispersion, which results in a broader bandwidth of the chirped coatings. One of the great benefits of these ultrafast lasers is their huge bandwidth. Krausz said: "This is useful for a novel diagnostic technique in medicine, called optical coherence tomography. It allows the study of the interior of the human eye, by shining a laser onto it and recombining the reflected light with a reference beam. The depth of the layer at which the light was reflected can then be determined, allowing tomographic imaging of the eye. This enables the diagnosis of diseases, such as glaucoma, at the earliest possible stage."

The resolution of the technique - currently 15 to 20 µm - is inversely proportional to the bandwidth of the laser light used, which is around 10 nm. Using the 10 fs pulses with a bandwidth of 120 to 150 nm, resolution could be increased by a factor of 10. Krausz said: "Using our laser, Wolfgang Drexler from the University of Vienna has recently demonstrated that it is possible to image individual cells."

So chirped mirrors really are making a difference. But what are the fundamental limits to this scheme? Will we ever see attosecond pulses being generated using this method? Krausz is doubtful: "In this wavelength range one oscillatory period of the light takes 2 to 3 fs. Since light is an electromagnetic wave, it is only capable of propagating in an oscillatory manner. With our 4 to 5 fs pulses - 1.5 to 2 oscillations - we have now almost reached that limit, but we might actually reach it with the TFI mirrors. If you really wanted to go beyond that, however, you would have to go to shorter wavelengths."

And that is exactly what they have done. Recently, Krausz's group generated ultra-short X-ray pulses via higher harmonic frequency generation lasting just a little more than 1 fs (Science 291 1923-7).