New fiber lasers deliver pulses at tens of gigahertz

Time and frequency are the most accurately measured physical quantities, largely thanks to the precision available from ultrafast lasers. A train of femtosecond laser pulses can generate a coherent broadband spectrum that with suitable optics is resolvable into a comb of equally spaced reference frequencies. Indeed, optical clocks based on such frequency combs can be more precise than the best atomic clocks.

The titanium:sapphire laser is the standard among femtosecond lasers, able to produce a spectrum spanning more than an octave of frequency. But the fiber laser, whose cavity is a length of fiber-optic glass, has its own natural appeal: It’s durable and robust, far less expensive, and exceedingly compact—about the size of a paperback novel—and it produces superb-quality subpicosecond pulses. What’s more, when composed of glass doped with erbium ions, fiber lasers emit at telecommunication frequencies. Those advantages make them ideal for portable metrology, optical signal processing, and communications applications, particularly if the pulse rate can be driven into the tens of gigahertz, the data-transmission rate that fiber optics currently handle.

A fiber laser is usually pumped by a laser diode that excites a superposition of many cavity modes. Because the relative phases of those modes rapidly change, any steady-state emission is more or less a continuous wave. The presence of an intensity-dependent component known as a saturable absorber in the cavity, however, can lock the phases of those modes together. The absorber, typically composed of semiconducting quantum wells, enforces that locking by being opaque to light below some threshold intensity but increasingly transparent above it. The result is a cavity door, in essence, that transmits a stream of high-intensity pulses at a repetition rate f _rep set by the length L of the cavity: f _rep = c/2nL, where c is the speed of light in vacuum and n the fiber’s index of refraction.

Typically, the lasers operate at tens of megahertz due to the long lengths—on the scale of meters—needed for sufficient gain in Er-doped fiber. To raise that into the gigahertz regime, one could tap into a higher harmonic of the fundamental repetition rate. But that approach relies on having more than one pulse in the cavity at a time and leads to pulse jitter and noise. Instead, several engineers have opted to steadily push the fundamental rate higher by simply shortening the cavity while keeping its losses below the accumulated gain.

Amos Martinez and Shinji Yamashita, both at the University of Tokyo, report the latest milestone in that effort: a pulse rate of 20 GHz from a fiber laser just 5 mm long. ¹ Key to that achievement is the incorporation of semiconducting single-wall carbon nanotubes as the saturable absorber, which abuts one of two highly reflective Fabry–Perot mirrors used to form the laser cavity, as shown in the figure on page 15. Yamashita’s group introduced nanotubes into a fiber-laser cavity six years ago and have been refining the approach ever since by optimizing the system’s parameters and efficiency.

Pulse shaping

Researchers have long known that the gain from silicate fiber rises dramatically when the fiber is doped not just with Er, which has a low absorption cross section and tends to cluster at high concentration, but also with ytterbium, which doesn’t cluster and whose absorption cross section is two orders of magnitude higher. Martinez and Yamashita exploit those properties by loading their fiber with Yb³⁺ ions, which allows them to shorten the fiber’s length.

But although the gain can be made great enough to beat losses, the removal of silicate fiber reduces the material’s nonlinear interactions with light, interactions that keep the pulses from stretching out in the cavity. Moreover, as the cavity shortens, the energy per pulse drops proportionately; on average, the net fiber nonlinearity decreases as the square of the drop.

That loss of intensity and of the fiber’s pulse-shaping nonlinearity places a greater burden on the saturable absorber, explains MIT’s Erich Ippen. “You want this door to be stiff enough to quickly close behind the pulse as the intensity drops, but as it’s made stiffer, it can be harder to open.” Carbon nanotubes seem ideal in that respect; so fast are their charge-carrier dynamics that once the nanotubes absorb light, electrons relax from excited states in less than a picosecond. And yet, as Martinez and Yamashita’s demonstration confirms, the absorption can begin to saturate even at the lower light intensities of a short fiber.

The nanotubes also exhibit low losses and consume essentially no space. The researchers just spray them directly onto one of the mirrors after ultrasonically dispersing the nanotubes to break up bundles formed by van der Waals forces. The film is kept thin (about 100 nm) to reduce light scattering and minimize the output pulse width. In addition, the nanotubes’ high thermal conductivity, Martinez speculates, may help sink the heat that builds when the pump laser drives the fiber at higher powers.

Tuning the absorption

The researchers’ nanotubes were produced with diameters such that the fiber-laser frequency roughly matches that of the tubes’ absorption. The quest to precisely tune those diameters and control whether the nanotubes are semiconducting or metallic remains a significant challenge, though. Having nanotubes out of resonance adds to cavity losses.

Nevertheless, even a wide distribution of tube diameters can prove advantageous. Three years ago Cambridge University’s Andrea Ferrari and his colleagues designed a nanotube mode-locked laser tunable over a much broader range of wavelengths than other systems. ² Although the nonsaturable losses are greater and degrade the mode locking, Ferrari argues, they’re tolerable in fiber lasers with relatively large gain coefficients. And the mixture of semiconducting and metallic nanotubes in the Tokyo team’s system lowers the absorption recovery time, as excited-state electrons can relax more quickly by tunneling to a nearby metallic nanotube.

More recently three independent groups—Ferrari’s, ³ Yamashita’s, ⁴ and Dingyuan Tang’s ⁵ at Nanyang Technological University in Singapore—each developed ultrafast lasers mode locked using graphene. Thanks to the linear dispersion of graphene’s Dirac electrons (see PHYSICS TODAY, January 2006, page 21 ), there always exist electron–hole pairs in resonance with light of any frequency, a property that makes the material ideal for a tunable system. Unlike semiconductor quantum wells and nanotubes, graphene requires none of the bandgap engineering or diameter control to optimize performance. ⁶

Even so, nonsaturable losses remain on par with those from nanotubes. To overcome those losses, the challenge will be to improve graphene’s synthesis.