Fiber lasers offer a reliable, economical, and easy-to-use alternative to solid-state lasers and optical parametric frequency conversion.
JAN POSTHUMUS AND FLORIAN TAUSER
Of all lasers, femtosecond lasers are probably the most versatile, but their application is nevertheless still largely confined to physics and chemistry departments in academia. So what makes this technology so interesting and why is it not used more often?
Femtosecond pulses are required whenever there is a need for the shortest possible pulses-in ultrafast pump-probe spectroscopy, or in time-domain terahertz spectroscopy, for instance. The application to multiphoton processes is also quite obvious because the concentration of energy in the shortest possible pulse width results in the highest peak power. Probably less known is the wavelength tunability based on various forms of nonlinear frequency conversion schemes. Parametric amplifiers, for instance, allow the generation of laser radiation at virtually any desired wavelength in the visible and infrared. Such a widely tunable laser is certainly not only interesting for physics and chemistry, but also for the life sciences and industrial applications.
Femtosecond technology is not used more often, however, because of practical factors, notably effort and cost. Traditionally, ultrafast technology belongs to the domain of expert laser users working in dedicated optics laboratories. Over and above the initial investment, the maintenance costs are often substantial. Applications in the life sciences and in industry, however, require cost-effective turnkey lasers with few constraints as to the place of installation (see Fig. 1). Such criteria have proven difficult to realize for solid-state lasers, but fiber lasers have now evolved to the point that they can provide a competitive alternative. Novel frequency conversion techniques based on nonlinear fibers and periodically poled crystals are playing an important role in this technological evolution.
Established telecom technology
Fiber-laser development has accelerated of late in large part by taking advantage of the many high-quality components originally developed for the telecom industry. These components, which operate at 1.55 µm, the wavelength of lowest loss in silica, are already optimized for reliability and cost. With appropriate care, it is possible to construct a femtosecond laser from such components without losing their favorable properties.
A typical femtosecond fiber laser consists of a passively modelocked oscillator operating at a repetition frequency in the range 80 to 110 MHz, and a power amplifier. The laser gain medium is erbium-doped single-mode fiber with a gain bandwidth of several tens of nanometers such that it can support ultrafast pulses. A nonlinear process ensures that all longitudinal waves propagate in phase (modelocking) and short pulses emerge. Since fibers guide the beam, the alignment is very stable, and the pulsed operation is self-starting. The dispersion in the oscillator is compensated by combining fibers with normal and anomalous dispersion at 1.55 µm. Femtosecond lasers operating at shorter wavelengths generally require bulk optics for dispersion compensation and are therefore less stable.
The erbium-doped fibers are pumped by fiber-coupled diode lasers. In telecom applications these diodes operate continuously for many years. Even in the unlikely event of failure, diode-laser replacement is still no big deal compared to the maintenance of pump lasers for solid-state femtosecond lasers. Typical modelocked fiber oscillators produce many milliwatts of power, which are then boosted to a few hundreds of milliwatts by a fiber amplifier. Pulse durations shorter than 100 fs are possible.
Industry-standard lasers for terahertz
Terahertz (THz) spectroscopy and imaging appears to be a bourgeoning technology. Terahertz radiation, lying between the infrared and microwave wavelengths, can be generated and detected in various ways, such as difference frequency generation in nonlinear crystals, optical rectification at surfaces, and photoconductive switches. These methods exploit the large bandwidth of femtosecond light pulses to generate frequencies for which many organic materials are transparent. Nevertheless, large molecules still have unique absorption spectra in the terahertz range so it is possible to monitor and analyze objects and goods through opaque packaging or clothing using their terahertz “fingerprint.” Hence, terahertz imaging and spectroscopy is of particular interest for quality-control and security applications. Obviously, such applications require instruments with reliable turnkey operation at affordable prices.
Because Ti:sapphire sources have dominated the femtosecond-laser market for the last decade, nearly all terahertz equipment has been optimized for wavelengths around 800 nm. This wavelength range is easily accessed with erbium-fiber lasers by frequency doubling the output with periodically poled lithium niobate (PPLN) crystals. These very efficient crystals generate approximately 100 mW of average power at 775 nm, with pulse durations close to 100 fs. Because of the search for reliable and cost-effective systems, however, interest is shifting toward emitters and detectors for operation at 1.55 µm, and first results are promising.1
Universal lasers for microscopy
Various forms of modern microscopy rely on lasers, whether pulsed or continuous wave. The short and intense pulses of femtosecond laser energy allow time-resolved studies as well as multiphoton excitation. The most interesting property, though, is the tunability based on nonlinear frequency conversion. Thus far, confocal microscopy has relied on diode and gas lasers, because they are compact, economical, and easy to operate. However, because these lasers have fixed wavelengths microscopes are often equipped with more than one laser. Nevertheless, the user must always look for a fluorescent marker that fits one of the available wavelengths. With a tunable laser, the user can choose a marker that is preferred from a scientific point of view and then adjust the laser accordingly.
The femtosecond fiber laser can be turned into an easy-to-operate tunable laser in the visible to meet this need. Nonlinear processes inside a specialty fiber generate a beam that is continuously tunable from 1050 to 1400 nm, with an average power of 20 to 30 mW and pulse durations shorter than 30 fs. The beam is then frequency doubled to the visible, where it reaches an average power of several milliwatts with a spectral width of 1 to 2 nm (see Fig. 2). For continuous tuning, proprietary PPLN crystals with fan-out poling designs have been fabricated. The narrow bandwidth ensures that the Stokes-shifted fluorescence signal is easily distinguishable from the scattered excitation light. The entire tunable laser system can be packaged in a compact laser head that is relatively insensitive to its environment.
Because the visible laser beams are pulsed, they also enable novel microscopy methods such as fluorescence lifetime imaging (FLIM). Here the temporal decay of the fluorescence is used as an additional source of information, in addition to the intensity of the fluorescence. Two-color fiber lasers, with automatically synchronized pulses, facilitate coherent anti-Stokes Raman-scattering (CARS) microscopy and pump-and-probe spectroscopy (see Fig. 3). Finally, multiphoton excitation is possible with the powerful beams at 1.55 µm and 775 nm. In addition to adding new colors to confocal microscopy, these lasers therefore also allow a user to investigate opportunities that arise from these new developments.
Reliable lasers for metrology
Half of the most recent Nobel Prize in physics was assigned to work based on precision spectroscopy with femtosecond lasers. Passively modelocked lasers emit a regularly spaced comb of frequencies, described by fn = n × ƒrep + υCEO, where υCEO is an offset frequency, ƒrep is the laser’s repetition rate, and n is an integer. Falling into the microwave domain, υCEO and ƒrep can be measured and controlled electronically. In this way, both quantities can be referenced to current frequency standards, such as commercial microwave atomic clocks, GPS receivers, or hydrogen masers. In effect, the output of the modelocked laser forms an absolute frequency ruler in the optical domain, which finds application in high-precision frequency metrology.2
Recent tests have shown that the instability of a frequency measurement carried out with a modelocked fiber laser is solely limited by currently available frequency standards, and can be lower than 2 × 10-14 after 1 second of integration time.3 In contrast to previous technology, fiber-based systems are readily transportable and can support measurements over days and weeks without interruption. Nonlinear frequency-conversion techniques extend the useful range from 530 to 2000 nm. An uninterrupted measurement over more than 80 hours at the wavelength of the calcium optical frequency standard (657 nm) has been reported.4
Fiber vs. solid-state
Wherever highest peak power and shortest pulse durations are indispensable, a Ti:sapphire-based solution will remain the first choice. Despite the successes in fundamental research, however, most attempts to establish these sources elsewhere have failed.
Fiber-based laser technology, on the other hand, offers major benefits in terms of cost-effectiveness, reliability, and environmental stability. Although fiber lasers may not hold femtosecond world records, they nevertheless fulfill the requirements in many fields, such as those described earlier. With respect to confocal microscopy, the concept of a continuously tunable and narrowband laser offers entirely new opportunities. Using a single laser source, a user is now practically free to choose the optimal fluorescent dye for the experiment.❏
REFERENCES
1. J. Lloyd-Hughes, E. Castro-Camus, M.B. Johnston, Solid State Communications136, 595 (2005) and references therein.
2. Th. Udem, R. Holzwarth, T.W. Hänsch, Nature416, 233 (2002).
3. E. Benkler, H.R. Telle, A. Zach, F. Tauser, Optics Exp.13, 5662 (2005).
4. F. Adler et al., Optics Exp.12, 5872 (2004).