1 Introduction

In recent years, solid state laser systems like the Ti:sapphire laser have more and more replaced mode-locked dye lasers as sources for ultrashort light pulses in time-resolved spectroscopy. Since for a number of applications pulse repetition rates of 80–100 MHz are disadvantageous (e. g. sample damage problems, heating effects, recovery time artefacts), attempts have been made in the past to transfer the concept of cavity-dumping with a Bragg cell which is well-established in fs-dye laser technology [1, 2] and more recently with the help of a Pockels cell [3] to the construction of cavity-dumped Ti:sapphire [4, 5, 6, 7] and chromium doped forsterite lasers [8]. It was shown that stable mode-locking of the lasers persisted despite the strong perturbations caused by the dumping process and that the pulse properties were not affected adversely. With a Ti:sapphire oscillator Ramaswamy et al. [4] succeeded in the generation of 50 fs/100 nJ pulses (at 950 kHz rep. rate, I_pump=8.8 W), Pshenichnikov et al. [5] reported 13 fs/60 nJ pulses (rep. rate up to 200 kHz, I_pump=5.5 W) and Scherer et al. [7] generated 15 fs/40 nJ (rep. rate up to 300 kHz). Repetition rates of up to 4 MHz at single pulse energies of some 10 nJ could be achieved without a dumping unit by increasing the optical length of the laser cavity to about 37,5 cm [9].

Because nonlinear effects are too small in the cw-operation regime, Kerr-lens mode-locked lasers are typically not self-starting and must be operated at the stability boundary to enhance self-amplitude modulation sufficiently to achieve a large saturable absorber action. However, if an additional saturable absorber element [10] is introduced, then saturable absorption and self-phase modulation can be optimized independently and reliable self-starting of the Kerr-lens mode-locking is achieved [11].

In this Letter we report on the novel design of a self-starting, cavity-dumped Ti:sapphire laser in which a semiconductor saturable absorber mirror (SESAM) is incorporated. The band gap of the specific semiconductor dertermines the wavelength of the produced pulses. Upon cavity-dumping, the single pulse energy is increased by more than a factor of 12 at 800 kHz rep. rate and 3.5 W pump power in comparison to the cw outcoupled pulse train (82 MHz).

2 Cavity design and Experiments

Our cavity design (Fig. 1) which has some similarity with that of Scherer et al. [7] differs from the designs described in early reports cited above by the location of the Bragg cell. It is placed in that arm of the cavity which also contains the prisms necessary for compensation of the group velocity dispersion. This relocation is made necessary because the second arm of the cavity which is chosen to be shorter is terminated by the SESAM (a detailed description of this device and its action is given in references [11, 12]). The chosen SESAM is a high-finesse antiresonant Fabry-Perot saturable absorber (AFPSA) which allows to produce pulses at λ≃800 nm. The gain medium is a 10 mm, Brewster-angled Ti:sapphire crystal (doping intensity 0.25 wt%) placed in an astigmatically compensated X-fold formed by two 10 cm radius-of-curvature mirrors. The tip-to-tip distance of the two SF10 glass prisms is increased to 27.5 cm in order to compensate also the additonal dispersion of the Bragg cell.

 

Fig. 1. Schematic of the cavity-dumped Ti:sapphire laser: pump laser (Millenia Spectra Physics); lens, f=50 mm; TS, 10 mm Ti:sapphire crystal (Roditi); M1 -M5, mirror high reflectance, R=10 cm (Laser Optik); OC, output-coupling mirror R≃96% (Laser Optik); P1,P2, SF10 glass prism; RF, cavity-dumper driver unit. In further contrast to previously described experimental arrangements [4, 5, 6, 8] (which employ an additional flat end mirror) the acousto-optic cavity-dumper comprises only the two spherical mirrors M4, M5 (R=10 cm) and the 3.2 mm thick fused-silica cell placed under Brewster’s angle in the focus of M4. The distance between the Bragg cell and M5 is twice the focal length of M5. The cavity-dumper is operated in a double-pass configuration with the deflected beam being displaced vertically by about 3 mm at the position of the outcoupling mirror OC. This setup has the advantage that the deflected pulses transverse the prism pair along an analogous path. The RF input to the Bragg cell is provided by the standard Cavity Dumper Driver unit (Spectra Physics model 454) with a reference signal generated by a photodiode in combination with constant fraction discrimantor (CFD), amplifier and frequency divider (D/2). The advantage of using the CFD is that it produces stable trigger pulses, even though the intracavity pulse energy is low and consequently the photodiode signal small. The intensity ratio between the optimally outcoupled pulse and the preceeding and trailing pulses is on the order of 350 : 1 as determined by the single photon counting technique (Fig. 2). Its value depends sensitively on the careful choice of the timing and phase setting of the driver electronics, as well as on the proper focussing into the Bragg cell. For this purpose, a special mounting frame was buildt allowing for adjustment along 3 rotational and 3 translational degrees of freedom [13].

Download Full Size | PDF

Introduction of the SESAM should change the modelocking mechanism from Kerr-lens modelocking (KLM) to soliton modelocking in which the pulse is shaped by soliton formation, but stabilized by a slow saturable absorber [14]. To minimize the losses, the absorber should be saturable with the expected pulse fluence, e. g. the pulse energy in the laser should be several times more than the saturation energy, but not too high because then the laser tends to exhibit multiple pulsing. In the chosen design, the incident pulse intensity onto the SESAM can be adjusted by varying the focus condition of Mirror M1.

If the intracavity pulse powers are low, e. g. because of low pump power, then tighter focussing helps to achieve the necessary saturation fluence of some ten µJ/cm². Since our main goal was to achieve stable operation with a minimum of pump power, we limited our experiments to pump powers of about 3.5 W. With higher pump energies (>4 W) the laser led indeed to multiple pulse formation in the cavity [15].

 

Fig. 2. Determination of the suppression ratio for the preceeding and consecutive pulses applying the single photon timing technique (dumping rate 800 kHz).

Download Full Size | PDF

The fluence in the cavity is, on the other hand, also determined by the reflectivity of the outcoupling mirror OC. Multiple pulses at somewhat higher pump powers can be suppressed if the reflectivity is reduced thus yielding higher cw output at OC. The intensity of the cavity-dumped pulses however does not decrease significantly. Therefore we restricted ourselves to optimization of the cavity at pump levels of about 3.5 W, R ≃96%.

 

Fig. 3. Intracavity laser pulse dynamics during the dumping process at bottom 80 kHz, middle 800 kHz and top 4.1 MHz showing the overshoot during the recovery at 80 kHz (I_pump=3.5 W).

Download Full Size | PDF

The traces in Fig. 3 show the intracavity pulse dynamics in dependence on dumping frequency. With dumping rates of 80 and 800 kHz a large fraction of the single pulse energy can actually be deflected out of the resonator (≃70%). Records with higher time resolution prove that mode-locking is not lost despite the large drain in single pulse energy. Under our experimental conditions it takes about 1 µs until the intracavity pulse energy reaches the equilibrium value. This implies that a dumping rate of 800 kHz is close to the optimum if the maximum energy in the whole pulse train is requested. At lower dumping rate the well-known overshooting of the pulse energy is observed. Under the chosen experimental conditions, it takes about 2 µs before the steady state is reached again (Pshenichnikov et al. [5] report a relaxation time of about 4 µs at 80 kHz repetition rate). If a higher dumping rate is applied then the next pulse is deflected before the build-up in pulse energy is completed. In accordance with this intracavity pulse dynamics, we find the following pulse energies: 1.3 nJ at 82 MHz, 20 nJ at 4.1 MHz and 34 nJ at 800 kHz (pump power is always 3.5 W). These values correspond to average powers of the pulse trains of 107 mW (82 MHz), 82 mW and 27 mW, respectively. Pshenichnikov et al. [5] found also a reduction in single pulse energy from 62 nJ (80 kHz) via 50 nJ (400 kHz) to 20 nJ (4 MHz) applying a pump power of 5.5 W. The conversion efficiency of our system compares therefore favorably with the cited one. Our conversion efficiency is also in accord with the data presented by Ramaswamy et al. [4]. They produce about 100 nJ pulses with 8.8 W of pump power for dumping rates up to ~950 kHz. Under these operation conditions they claim a time of ~500 ns for the intracavity pulse energy to recover to the unperturbed level. In viewof our lower pump power, a recovery time of 1 µs is reasonable (s. Fig. 3).

 

Fig. 4. Autocorrelation trace of cavity-dumped pulse (collinear second harmonic generation) recorded with dumping rate 800 kHz.

Download Full Size | PDF

Figure 4 shows the autocorrelation trace of the cavity-dumped pulses recorded by the conventional collinear second harmonic generation in a 0.5 mm thick potassium dihydrogen phospate crystal. The pulse width is determined as 90 fs assuming a sech² pulse shape.

In Fig. 5, the sensitivity corrected spectrum of the laser output is shown for the same typical operating condition (I_pump=3.5 W, repetition rate=800 kHz). The calculated time-bandwith product of 0.33 shows that the pulses are nearly transform limited.

 

Fig. 5. Spectrum of a dumped output pulse at 800 kHz. The center wavelength of 799 nm is determined by the band gap of the saturable absorber.

Download Full Size | PDF

In this study we use an antiresonant fabry perot saturable absorber (AFPSA), where the center wavelength of the laser radiation is determined by the bandgap in the semiconductor saturable absorber. To change the central wavelength, the absorber device can be easily substituted by another one requiring only a small readjustment of the prism position and the folding optics of the Bragg cell. Experiments with another AFPSA yielded pulses centered around 830 nm with similar pulse energies. Thus there are no limits to use the new broadband SESAM to achieve stable operation in the tuning range of the Ti:sapphire crystal. The pulse energy fluctuations are typically less than 5% like in the case of the unperturbed oscillator. We found a superior operation of the system if pumped with the Millenia (Spectra Physics, frequency doubled Nd:YVO₄) rather than with an Ar⁺ laser (Coherent, INNOVA 90-6). This improvement is in part due to the higher pointing stability, in part to the lower intensity fluctuations of the pump laser.

It should be mentioned also that the pulses produced in our system are sufficiently intense to generate the second and third harmonic with efficiencies of 33% and 7%, respectively, using a commercial doubler/tripler unit (Spectra Physics, GWU- 23FS). The described oscillator is, therefore, very well suited to provide excitation pulses not only for pump-probe experiments with the fundamental output but also for single photon counting experiments using the higher harmonics. We get a similiar suppression ratio as Scherer [6] for preceeding and consecutive pulses. (Only Gibson et al. [3] report a higher contrast ratio, because of using a Pockels cell in combination with Brewster angle prisms.) This facilitates the evaluation of measurements by time-correlated single photon counting.

3 Conclusions

We have described a novel design for a self-starting, cavity-dumped mode-locked Ti:sapphire laser oscillator that delivers pulses of about 90 fs duration at λ=800 nm with energies up to about 34 nJ for pump powers of only 3.5 W. One advantage of the described system is (in our opinion) that in applications like time-correlated single photon counting, where a stable operation over many hours is required due to the self-starting mechanism a stable mode-locked operation is guaranteed. Since pulses with durations of less than 10 fs have been produced applying the technique of soliton modelocking with broad band saturable absorber [12], we are positive that in our design the pulse durations can be reduced by changing to a shorter Ti:sapphire crystal and another prism material which allows for a better cubic-phase dispersion compensation [16]. The newer generation of SESAMS exhibits a much wider transmission range. Their implementation should allow one to make use of the full width of the gain profile of the Ti:sapphire crystal, either by generating shorter pulses or tunable pulses with larger pulse width.