Elsevier

Physics Reports

Volume 429, Issue 2, June 2006, Pages 67-120
Physics Reports

Passively modelocked surface-emitting semiconductor lasers

https://doi.org/10.1016/j.physrep.2006.03.004Get rights and content

Abstract

This paper will review and discuss pico- and femtosecond pulse generation from passively modelocked vertical–external-cavity surface-emitting semiconductor lasers (VECSELs). We shall discuss the physical principles of ultrashort pulse generation in these lasers, considering in turn the role played by the semiconductor quantum well gain structure, and the saturable absorber. The paper will analyze the fundamental performance limits of these devices, and review the results that have been demonstrated to date. Different types of semiconductor saturable absorber mirror (SESAM) design, and their characteristic dynamics, are described in detail; exploring the ultimate goal of moving to a wafer integration approach, in which the SESAM is integrated into the VECSEL structure with tremendous gain in capability. In particular, the contrast between VECSELs and diode-pumped solid-state lasers and edge-emitting diode lasers will be discussed. Optically pumped VECSELs have led to an improvement by more than two orders of magnitude to date in the average output power achievable from a passively modelocked ultrafast semiconductor laser.

Introduction

In the past few years, a novel type of laser has bridged the gap between semiconductor lasers and solid-state lasers. The vertical–external-cavity surface-emitting laser (VECSEL) [1], [2] combines the best of both worlds: the semiconductor gain medium allows for flexible choice of emission wavelength via bandgap engineering and offers a wealth of possibilities from the semiconductor processing world. Almost arbitrary optical layer structures can be integrated vertically with the gain section; in particular multi-purpose mirrors, combining functions such as dispersion control, dual wavelength reflection for pump and laser wavelength, antireflection layers etc. One important advantage of optically pumped VECSELs is that they can convert fairly low-cost, low-beam-quality optical pump power from high-power diode laser bars into a near-diffraction-limited output beam with good efficiency in wavelength regions which are not covered by established solid-state laser gain materials. Lateral integration promises high-performance photonic integrated circuits in the near future. The combination of the mature optical pumping technology extensively used for diode-pumped solid-state lasers with efficient heat removal of solid-state thin-disk lasers resulted in performance of VECSELs that surpasses anything possible to date with conventional semiconductor lasers. Continuous-wave output powers of up to 30 W with an M2 of 3 have been reported from such optically pumped VECSELs [3], and electrically pumped devices have reached 0.5 W single-mode output power [4].

Concerning high-performance passive modelocking, a domain where diode-pumped solid-state lasers using semiconductor saturable absorber mirrors (SESAMs) have been dominant for years [5], [6], [7], [8], the VECSEL possesses the advantage of a large gain cross-section which suppresses Q-switching instabilities [9]. VECSELs are therefore ideally suited for high-repetition-rate modelocking in combination with high average output powers. After the first demonstration of a passively modelocked VECSEL in 2000 [10], pulse width and output power have improved continuously to 486-fs pulses at 10 GHz with 30 mW [11] and 4.7-ps pulses at 4 GHz with 2.1 W average output power [12]. The comparison of various high-repetition-rate sources in Fig. 1 shows that optically pumped VECSELs have already pulled even with solid-state lasers in the regime between 1 and 10 GHz.

Novel SESAMs based on quantum dot saturable absorbers (QD-SESAMs) were developed to move towards an even more ambitious goal: the integration of the absorber into the VECSEL gain structure [13]. In a first step passive modelocking with the same mode area in the gain and the absorber had to be demonstrated for the full wafer-scale integration. We refer to this as “1:1 modelocking” which was successfully demonstrated using these new QD-SESAMs and therefore the viability of the integrated-absorber VECSEL concept has been demonstrated [13]. This could pave the way for the development of compact and rugged high-repetition-rate pulsed laser sources in the >100-mW power class which can be cheaply fabricated by wafer-scale mass production and therefore fill a gap in the performance spectrum of current laser technology. These novel QD-SESAMs also supported the scaling of pulse repetition rates to 50 GHz with 100 mW of average power [14].

The ultimate goal is to extend the excellent results with optically pumped VECSELs to electrical pumping. However, this is not a simple extension, even though very promising results have been achieved in the cw regime, with 500 mW average output power demonstrated in a near-diffraction-limited beam [4]. Initial modelocking results reported however only very moderate average output power well below 100 mW [15].

From an application point of view, telecom transmission systems at 10 Gb/s and higher mostly use return-to-zero (RZ) formats [16] and soliton dispersion management techniques [17], both of which rely on clean optical pulses. The much higher contrast ratio of directly pulsed lasers (compared to externally modulated continuous-wave sources) improves overall system signal-to-noise and allows further scaling to higher repetition rates through optical time-division multiplexing (OTDM). Apart from the transmitter side, there are also other important applications of pulsed lasers in the receivers of transmission systems, e.g. optical switching for demultiplexing and clock recovery [18]. Research in optical clocking and interconnects [19], [20] quantum cryptography, high-speed electro-optic sampling [21], [22], frequency metrology [23], [24], or generation of polarized electron beams for particle accelerators [25] has proved the need and the applicability of clean and stable high repetition-rate optical pulses in a variety of different fields. Although the span of possible applications is very broad, the requirements on an ideal pulse source are similar in each field and can be summarized as follows: The emitted pulse train has to consist of femto- or picosecond optical pulses with high contrast ratio, high pulse energy and low timing jitter. Additionally, wavelength tunability or setability in the regime of interest should be feasible. The instrument itself needs to be affordable, reliable, compact and robust.

Multi-gigahertz pulse sources to date have almost always involved either an edge-emitting semiconductor laser [26], which is usually actively or hybrid modelocked, or a harmonically modelocked fiber ring laser [27]. Edge-emitting semiconductor lasers can appear to be attractive due to their very compact optical setup, but expensive electronics are required for active modelocking, the structures required for a semiconductor laser with this performance are complicated and difficult to produce, and their average power levels are relatively low. Additionally, there is little of the expected cost saving from semiconductor manufacturing, due to low yield/relatively low production volumes, and the dominant packaging/testing costs. Multi-gigahertz fiber lasers can also generate high-quality pulses, but they have very long and complex laser cavities, requiring sophisticated means to obtain stable “harmonic” modelocking—which consists of a large number of precisely equidistant pulses in the cavity. Additionally, individual pulses generated by harmonic modelocking do not necessarily exhibit a fixed phase relation. This excludes promising and important coding formats such as return-to-zero differential phase shift keying [28], [29] which carry the data in the phase of the pulses, rather than in the amplitude.

Up to a few years ago, the repetition rate of passively modelocked solid-state lasers was limited to a few gigahertz. Q-switching instabilities impaired performance at the highest pulse repetition rates [9]. In recent years, the consequent exploitation of the flexibility of semiconductor saturable absorber mirrors (SESAMs) [5], [7], [30] allowed the Keller group to develop passively modelocked lasers with multi-GHz pulse repetition rates, very good pulse quality, comparatively high output powers, and wavelength tunability in the areas of interest (for example the ITU-specified C-band from approximately 1525 to 1565 nm) (Table 1). Passive modelocking means that the pulses are achieved without using any expensive multi-gigahertz electronics. In addition, the pulses originate from fundamental modelocking. Thus, every output pulse is a copy of the same single pulse, which travels back and forth in the cavity. Therefore, pulse-to-pulse variations are minimized and the phase of the pulses is constant. For the first time, pulse repetition rates above 10 GHz from passively modelocked ion-doped solid-state lasers have been generated with Nd:YVO4 lasers at a center wavelength around 1μm [31]. This laser has a large gain cross section and therefore Q-switching instabilities are more strongly suppressed. Shortly afterwards the frontier was pushed up to 77 GHz [32] and 160 GHz [33] (Fig. 1). The average power has been optimized at a 10 GHz pulse repetition rate to as high as 2.1 W [34]. The peak power was sufficient for efficient nonlinear frequency conversion. For example, a synchronously pumped optical parametric oscillator (OPO) was demonstrated producing picosecond pulses broadly tunable around 1.55μm with up to 350 mW average output power [34], [35]. Such all-solid-state synch-pumped OPOs can reach the S-, C- and L-bands in telecommunications. With an additional Yb-doped fiber amplifier, the repetition rate was pushed up to 80 GHz [36]. In the telecom wavelength ranges (around 1.3 and 1.55μm), where only few solid-state gain media are available, it was not initially possible to demonstrate multi-GHz pulse repetition rates [37], [38]. However with improved SESAM designs [30], and a deeper understanding of the Q-switching instabilities [39], [40], full C-band tuning [41] and pulse repetition rates up to 50 GHz [42] have been demonstrated with a diode-pumped Er:Yb:glass laser (Fig. 1). At 1.3μm both Nd:YLF [43] and Nd:YVO4 [44] have been passively modelocked at GHz repetition rates. In addition, the timing jitter of diode-pumped solid-state lasers is very close to the quantum noise limit [45]. Compared to the ultrafast solid-state lasers, VECSELs offer the prospect of even more compact high-power sources, in which the SESAM is integrated into the VECSEL gain chip, and the structure is pumped electrically. Both of these innovations, however, are yet to be demonstrated.

For comparison, it is also instructive to consider briefly the performance of pulsed edge-emitting semiconductor diode lasers, which can exhibit the highest pulse repetition rates of any optical source. The obvious advantages of compactness, efficiency of pumping, and ease of manufacture and integration make these sources primary candidates for applications such as optical time-domain multiplexing, microwave carrier generation and optical clock recovery. The efficiency of direct modulation of the diode current falls off exponentially with increasing frequency above the diode relaxation resonance, which lies typically in the range 1–10 GHz: thus the highest repetition frequencies are achieved using modelocking of monolithic diode lasers, with gain, saturable absorption and/or external modulation all built into a single chip. The various schemes developed to realise lasers of this type have been reviewed by Avrutin et al. [58]. Passive modelocking, with a reverse-biased saturable absorber section included in the monolithic cavity, is particularly well-adapted to the generation of ultrashort pulses at high repetition rate because it does not require electrical modulation, which imposes a bandwidth limitation on repetition rate, and also impresses phase structure on the pulses. The first demonstration of such a monolithic device was reported by Vasil’ev et al. [59], who reported a 100-GHz train of 2.5-ps pulses from an AlGaAs/GaAs injection laser, corresponding to fundamental modelocking of the 380-μm long cavity. Repetition frequencies higher than 250GHz have not been reported using fundamental modelocking; the short gain section imposes severe power limitations on the device, which also becomes challenging to fabricate. Higher frequencies are achieved by harmonic modelocking of either the colliding pulse [60] or the compound-cavity [61] type. Yanson et al. [62] have reported modelocking of 860-nm AlGaAs/GaAs double quantum well ridge waveguide laser diodes at pulse repetition frequencies up to 2.1 THz, corresponding to the 33rd harmonic of the round-trip frequency. The modelocking performance of these devices relied critically on accurate control of the sub-cavity length ratios, which were lithographically defined. The pulses were near sinusoidal, and the devices emitted up to 2.2 mW per facet in cw operation. For applications that require the shortest possible pulses it is generally more practical to compress chirped picosecond pulses externally than to generate femtosecond pulses directly from a modelocked diode source. Tamura et al. [63] were able to generate a 50-GHz train of 280-fs pulses at a wavelength of 1557 nm with average power more than 100 mW by combining a modelocked diode with an external all-fiber amplifier and pulse compressor; their modelocked edge-emitter had an average power of 17 mW. Scollo et al. [64] have reported the generationof  600 fs pulses directly from a 42-GHz modelocked diode, albeit accompanied by numerous satellite pulses; these authors make use of a novel saturable absorber design [65] with estimated sub-picosecond recovery time. A two-section quantum dot diode laser produced strongly chirped 2-ps pulses with 45 mW average power and 400-fs pulses with 25 mW at 21 GHz pulse repetition rates [66]. Modelocked edge-emitting diodes are thus immensely versatile in repetition frequency, from individual gain-switched pulses, through the microwave region of the spectrum and up to THz. Their power scalability, however, is limited, with about 10 mW appearing currently to be about the practical limit; and they typically emit self-phase-modulated pulses of picosecond duration that can be externally compressed to the femtosecond regime if the pulse has a suitable phase structure.

This paper will review and discuss pico- and femtosecond pulse generation from passively modelocked VECSELs both optically and electronically pumped. After surveying the different semiconductor material systems in Section 2, we continue with a brief description of the VECSEL gain medium in Section 3, followed by a review of the different semiconductor saturable absorber nonlinearities that can be integrated into a SESAM structure in Section 4. The basic physical principles of passive modelocking of VECSELs will be discussed in Section 5, and the results achieved to date in Section 6. We shall conclude with an outlook towards wafer scale integration in Section 7.

Section snippets

Semiconductor materials

Semiconductor materials offer a wide flexibility in choosing the laser emission wavelength, which can range from 400nm in the UV using GaN-based material, to 2.5μm in the mid-infrared using GaInAsSb-based materials. More standard high-performance semiconductor material systems which can be grown today cover the infrared wavelength range from 800 nm up to 1.5μm. Semiconductor compounds used for these wavelengths are AlGaAs (800 to 870 nm), InGaAs (870 to about 1150 nm), GaInNAs (1.1 to 1.5μm), or

Optically pumped VECSELs

The layer structure of a generic VECSEL gain chip is shown schematically in Fig. 5, which depicts two commonly used geometries for an optically-pumped active mirror. In Fig 5(a) a Bragg mirror, of typically 25–30 periods, is grown next to the substrate. The active region consists of a few half-wavelengths thickness of a material which combines the functions of pump absorber, optical spacer, and quantum well barrier. Quantum wells are embedded in the active region, singly or in pairs, at λ/2

Semiconductor dynamics

Semiconductors are well suited absorber materials for ultrashort pulse generation. In contrast to saturable absorber mechanisms based on the Kerr effect, ultrafast semiconductor nonlinearities can be optimized independently from the laser cavity design [6], [136]. In addition, ultrafast semiconductor spectroscopy techniques [137] provide the basis for many improvements of ultrashort pulse generation with semiconductor saturable absorbers.

In ultrafast semiconductor dynamics, it is often

Dynamic gain saturation

Passive modelocking mechanisms are well-explained by three fundamental models: fast saturable absorber modelocking [213], [214] (Fig. 19(a)), slow saturable absorber modelocking without dynamic gain saturation in the picosecond [148] and femtosecond regime which is described by soliton modelocking [215], [216] (Fig. 19(b)) and slow saturable absorber modelocking with dynamic gain saturation [217], [218] (Fig. 19(c)). Dynamic saturation of the gain is only assumed in Fig. 19(c) where the gain

Overview

A quick overview of the results achieved to date in passive modelocking of optically-pumped VECSELs is provided by the data of Table 5. It is notable that lasers with, for the most part, rather similar gain structures have been investigated in such widely-varying dynamical regimes; with pulse durations from tens of ps to less than 500 fs; repetition rates from a few hundred MHz to 50 GHz, average powers from the few-mW to the few-W regime, and pulse energies ranging from few-pJ to hundreds of pJ.

Final remarks

In recent years, optically-pumped semiconductor laser research has emerged as a hot current field, with the demonstration of quantum well lasers that emit many Watts, or even tens of Watts, in near-diffraction-limited beams. Much interest therefore currently focusses on the design, fabrication and thermal management of wafers for high power cw devices, which have enabled the development of a new class of visible laser; the intracavity-doubled VECSEL. With the extension of VECSELs to the red

Acknowledgment

U. Keller acknowledges financial support from an Intel-sponsored research agreement and from ETH Zurich. A. Tropper acknowledges financial support from the Engineering and Physical Sciences Research Council, UK, and from the project Active Terahertz Imaging for Security in the EU Preparatory Action on the enhancement of the European industrial potential in the field of security research. Both of them also acknowledge support through the European Network of Excellence on Photonic Integrated

References (237)

  • U. Keller

    Progress in Optics

    (2004)
  • K. Sato

    IEEE J. Lightwave Technol.

    (2002)
  • R. Mellish et al.

    Electron. Lett.

    (1998)
  • R. Grange et al.

    Electron. Lett.

    (2004)
  • F. Kuznetsov et al.

    IEEE Photon Technol. Lett.

    (1997)
  • F. Kuznetsov et al.

    IEEE J. Sel. Top. Quantum Electron.

    (1999)
  • J. Chilla, S. Butterworth, A. Zeitschel, J. Charles, A. Caprara, M. Reed, L. Spinelli, in Photonics West 2004, Solid...
  • J.G. McInerney et al.

    Electron. Lett.

    (2003)
  • U. Keller et al.

    Opt. Lett.

    (1992)
  • U. Keller et al.

    IEEE J. Sel. Top. Quantum Electron.

    (1996)
  • U. Keller

    Nature

    (2003)
  • C. Hönninger et al.

    J. Opt. Soc. Am. B

    (1999)
  • S. Hoogland et al.

    IEEE Photon Technol. Lett.

    (2000)
  • S. Hoogland et al.

    IEEE Photon Technol. Lett.

    (2005)
  • A. Aschwanden et al.

    Opt. Lett.

    (2005)
  • D. Lorenser et al.

    Appl. Phys. B

    (2004)
  • D. Lorenser, D.J.H.C. Maas, H.J. Unold, A.-R. Bellancourt, B. Rudin, E. Gini, D. Ebling, U. Keller, IEEE J. Quant....
  • Q.A. Zhang et al.

    IEEE Photon Technol. Lett.

    (2004)
  • H. Taga et al.

    Electron. Lett.

    (1998)
  • L.F. Mollenauer et al.

    IEEE J. Quantum Electron.

    (1998)
  • S. Bigo et al.

    IEEE J. Sel. Topics in Quantum Electron.

    (1997)
  • D.A.B. Miller

    Opt. Lett.

    (1989)
  • M.J. Kobrinsky et al.

    Intel Technol. J.

    (2004)
    (Online) Available: http://www.intel.com/technology/itj/2004/volume08issue02/art05_on-chip/p01_abstract.htm.)...
  • K.J. Weingarten et al.

    IEEE J. Quantum Electron.

    (1988)
  • J.A. Valdmanis et al.

    Appl. Phys. Lett.

    (1982)
  • T. Udem et al.

    Nature

    (2002)
  • H.S. Margolis, C.S. Edwards, G.P. Barwood, P. Gill, in: F. X. Kärtner (Ed.), Conference Proceedings, 15th Annual...
  • A. Hatziefremidis et al.

    Nucl. Instrum. Methods A

    (1999)
  • E. Yoshida et al.

    IEEE Photon Technol. Lett.

    (1999)
  • J. Leibrich et al.

    IEEE Photon Technol. Lett.

    (2002)
  • C. Xu et al.

    IEEE Photon Technol. Lett.

    (2003)
  • G.J. Spühler et al.

    Appl. Phys. B

    (2005)
  • L. Krainer et al.

    Appl. Phys. B

    (1999)
  • L. Krainer et al.

    Electron. Lett.

    (2000)
  • L. Krainer et al.

    IEEE J. Quantum Electron.

    (2002)
  • S. Lecomte et al.

    J. Opt. Soc. Am. B

    (2004)
  • S. Lecomte et al.

    Opt. Lett.

    (2002)
  • S. Lecomte et al.

    IEEE Photon Technol. Lett.

    (2005)
  • B.C. Collings et al.

    Opt. Lett.

    (1997)
  • P. Laporta et al.

    IEEE Photon Technol. Lett.

    (1995)
  • A. Schlatter et al.

    J. Opt. Soc. Am. B

    (2004)
  • R. Grange et al.

    Appl. Phys. B

    (2005)
  • G.J. Spühler et al.

    Electron. Lett.

    (2003)
  • S.C. Zeller et al.

    Electron. Lett.

    (2004)
  • S.C. Zeller, R. Grange, V. Liverini, A. Rutz, S. Schön, M. Haiml, U. Keller, S. Pawlik, B. Schmidt, Advanced...
  • G.J. Spuhler et al.

    IEEE Photon Technol. Lett.

    (2005)
  • A. Schlatter et al.

    Opt. Lett.

    (2005)
  • A.J. Kemp et al.

    Electron. Lett.

    (2001)
  • T. Tomaru

    Opt. Lett.

    (2001)
  • T. Tomaru

    Opt. Commun.

    (2003)
  • Cited by (0)

    View full text