1. Introduction

A large number of scientific and industrial applications benefit from the use of laser pulses with moderate energy (~µJ) and ultrashort sub-100fs duration at the highest possible repetition rate. In scientific applications such as pump-probe multi-dimensional spectroscopy [1] for studying chemical/biological dynamics, or photoelectron spectroscopy, highly tunable laser pulses and extensive signal averaging are necessary to extract small signals. Ultrafast Ti:sapphire laser systems with multi-µJ pulse energy are ideal for pumping tunable optical parametric amplifiers which are tunable over a very large wavelength range. Other applications that benefit from moderate energy laser pulses at high repetition rates include ultrafast electron microscopy and diffraction imaging [2, 3], which are powerful techniques for investigating structural dynamics in condensed matter under unprecedented “atomic”-scale spatial and femtosecond-scale temporal resolutions. In these techniques, ultrafast laser pulses are used to generate < 100 fs electron pulses, which are severely limited in fluence to only a few electrons per pulse due to the space charge effects. Thus, a laser system with a MHz repetition-rate is essential for making this technique viable.

In the case of industrial applications, an excellent example is ultrafast laser micro/nano machining/surgery [4–7]. A tightly focused laser pulse requires only modest ~µJ energy with sub-100-fs pulse duration to ablate even hard-to-machine materials such as fused silica and diamond. The result is very precise ablation with low debris and minimal collateral damage, but also with a very small amount of material removed. The highest possible repetition rate is essential for maximizing machining throughput. Considerable evidence suggests that the shortest possible pulses with very “clean” pulse structure are necessary for machining the most-fragile materials, making a Ti:sapphire laser a viable contender in this area—especially if the repetition-rate of the laser can be increased to maximize throughput. Another example is the use of high harmonic generation to achieve a table-top “laser” like EUV/Soft X-ray source for applications such as extreme-UV (EUV) metrology, high-resolution EUV imaging, and microscopy, that benefit from high average power and high repetition rate. All of these applications can benefit from ultrashort pulses, high repetition rate, and high average power.

In the past, these needs have been addressed using a number of technologies. Ti:sapphire regenerative amplifiers were developed in the early 1990’s [8], with 100-300 kHz repetition rates, multiple microjoule pulse energies, and <100 fs pulse durations. Very high repetition-rate multipass amplifiers also have been implemented using Ti:sapphire [9]. However, these systems have not to-date succeeded in operating in the few MHz repetition-rate range and a few microjoule energy level. Cavity-dumped [10, 11] and long-cavity [12] Ti:sapphire oscillators can generate single-pulse energies significantly higher than a typical mode locked oscillator, but the pulse energies are still in the sub-µJ range and marginal for many of the above-mentioned applications. Although it has been shown previously that pulses from a cavity-dumped oscillator can be amplified in a CW-pumped regenerative amplifier, the maximum amplified pulse energies were less than a micro-joule [13]. Still other approaches have employed Yb based gain media, in either cavity-dumped [14]. free-space regenerative amplifier [15–17], fiber [18], or fiber-rod [19] amplifier configurations. However, these systems are at fixed wavelength(~1-1.03 µm) and have fairly long pulse durations (>200 fs). A satisfactory solution has not existed until now for applications that require ultrashort pulses at MHz repetition rates.

In this work, we demonstrate, to our knowledge, the first Ti:sapphire regenerative amplifier producing laser pulses with energies up to 3.7uJ, durations <60 fs and repetition rates up to 1.7 MHz. We studied the amplifier at various seeding pulse energies up to 150 nJ. Seed pulses below 6 nJ are produced by an ultrafast Ti:sapphire laser of standard design [20], while seed pulses of >6 nJ were produced using a cavity-dumped Ti:sapphire laser operating in the positive dispersion regime [11]. One challenge in implementing this system is the requirement for higher seed pulse energies to avoid dynamic instabilities. This challenge and the theoretical model developed to address this are discussed further.

2. Laser system

Figure 1 shows a schematic of the laser amplifier system. The amplifier crystal is pumped by two 11-W continuous wave green lasers at 532 nm and is cryogenically cooled to avoid thermal lensing [21, 22]. The seed pulse is generated from a femtosecond oscillator and stretched to ~10 ps using a grism pair [23, 24] that introduces negative group velocity dispersion (GVD) as well as negative third order dispersion (TOD). The stretched pulse is then sent through a Faraday isolator and injected from a thin film polarizer (TFP) into the regenerative amplifier. The amplifier cavity consists of two flat mirrors, two 20cm radius-of-curvature mirrors, a cryogenically-cooled Ti:sapphire rod, and an electrical optical (EO) switch. The EO switch consists of a TFP, a quarter waveplate and a 2 cm long BBO Pockels cell. The Pockels cell high voltage driver operates at quarter wave rotation voltage and repetition rate up to 1.7 MHz. Once injected into the laser cavity, the pulse is trapped inside the cavity by turning on the EO switch. After passing through the amplifier crystal for 30 to 40 times (15 to 20 round trips), the amplified laser pulse is switched from the amplifier at the polarizer by turning off the EO switch. The pulse is then dumped by the Faraday isolator into a SF6 glass compressor, which adds positive GVD and TOD. After passing through the compressor glass for 1 m path length (5x20 cm), as shown in Fig. 2 , the amplified laser pulse is measured to be ~51 fs using a SHG FROG device. The pulse duration is currently bandwidth limited by the optics inside the laser cavity. The compressed output of the amplifier is 3.7 W; i.e. 3.7 uJ pulse energy at 1MHz repetition rate. In this case, the seed pulse energy required was 200 nJ. We also measured an amplified spatial beam quality of 1.3 M², a long-term power stability of 0.71% RMS over 8 hours (Fig. 3b ), and a pulse-to-pulse energy stability of 0.5% at 1.5 MHz.

 

Fig. 1 Schematic of cryogenically-cooled MHz regenerative amplifier system.

Download Full Size | PDF

 

Fig. 2 Laser pulse measurement. a) Laser spectral intensity and phase; b) Laser temporal intensity and phase; Inset: SHG FROG trace. The 51fs duration of the measured pulses is close to the transform limit of the spectrum centered at 793 nm.

Download Full Size | PDF

 

Fig. 3 a) Amplifier output energy versus repetition rate for different seed pulse energies; b) laser power long-term stability for 8 hours at 1.5 MHz using 2 nJ seed energy.

Download Full Size | PDF

Cryogenic cooling of the Ti:sapphire crystal results in both higher overall efficiency of the regenerative amplifier (for example, the Coherent RegA employs 12W pump energy to generate 1.6W output [25]), as well as increasing the versatility of the laser system by allowing for tunable repetition-rate. At lower repetition rate the pulse duration is typically slightly shorter that 60 fs since the higher gain requires fewer passes in the amplifier introducing less higher order dispersion from the Pockels cell and amplifier crystal. Figure 3 shows the pulse energy and average power of the laser as a function of repetition rate, and also as a function of seed pulse energy. We found that higher seeding energy yields higher amplifier output for fewer round trips in the cavity. Furthermore, we also found that the higher seeding energy can saturate the gain of the amplification in fewer round trips, which yields a more stable pulse train, as discussed later in this paper.

3. Challenges

There are two main challenges to achieve this high repetition rate while using CW pump lasers in the amplifier system. The first challenge is that the gain of this amplifier system is limited by CW pumping, which requires a tightly focused laser beam (<100 μm diameter) in the amplifier laser crystal to reach a reasonable gain. The pump lasers can be focused well with M² values less than 1.1 and are combined by positioning the two parallel pump beams close together on the focusing lens. Given a 22 W pump and an 80µm focal spot on the crystal, we estimated the focal length of the thermal lens in the crystal to be 3 mm if the crystal were kept at room temperature. This strong lensing would make any type of laser amplifier virtually impossible. This thermal lens focal length increases to 5.4 m with the crystal at cryogenic temperature, which is very manageable and in fact does not require that the regenerative amplifier cavity be specifically engineered for a particular focal length. This laser uses a closed-loop, helium, cryogenic cooler with 90W capacity for cooling the crystal below 50K. As a result, we did not observe any significant thermal lens in our amplifier crystal and achieved 0.8% RMS power stability in long term (12 hr) tests. Likely the pump power can be increased by several times as more powerful green pump lasers become available. This is likely to further increase the extraction efficiency of the amplifier.

The second challenge with this amplifier system is that adjacent amplified pulses compete for gain when the repetition rate is higher than 312.5 kHz, where the inter-pulse spacing at this repetition rate is shorter than 3.2 μs which is the spontaneous lifetime of the Ti:sapphire laser level. The gain in the laser thus does not fully recover to a steady-state level between cycles of regenerative amplifier, resulting in the potential for a pulse bifurcating instability. If the amplifier gain is not sufficient to saturate the amplifier, a “runt pulse” will be generated and excited state population will not be fully depleted. Continued pumping means that on the next cycle of the regenerative amplifier, the stored energy will be higher. The result is saturated amplification on every other cycle of the amplifier. The solution to this issue is to ensure saturated amplification by using a high injection energy and operating at a gain that allows for saturated amplification within a limited number of passes of the amplifier. Figure 4 shows experimental oscilloscope traces of the laser output as a function of the number of passes through the amplification crystal between switching the pulse in and out of the cavity. With less than 20 passes, the laser pulse train is normal and stable at 1 MHz. However, Fig. 4b shows that one pulse gains more energy than the other for 35 passes through the laser crystal. Figure 4c shows that the previous pulse depletes the gain for the next pulse so there is no significant amplification of the later pulse for 50 passes through the laser crystal.

 

Fig. 4 Oscilloscope traces of amplified laser pulse train at 1 MHz with 2 nJ seed pulses for a) 20 passes, b) 35 passes, and c) 50 passes through the amplifier laser crystal.

Download Full Size | PDF

To understand how to optimize the amplifier system, we developed a simple model based on the recurrence relation developed in reference [26] for regenerative amplifiers:

After the laser pulse is amplified, the gain is reduced. We can simplify the buildup of the laser gain under the influence of a continuous pumping rate and spontaneous decay as a function of time by

Using the recurrence relation (1) and (2), we calculate the laser intensity and gain after each time a pulse passes through the amplifier crystal in the regenerative amplifier cavity. Then we can use Eq. (3) to estimate the gain recovery after the pulse is dumped out of the laser cavity. In order to compare with experimental results, we measured our gain in our laser as $g_0=1.2$. As demonstrated in Fig. 5 , we calculated the pulse energy ratio of the two adjacent pulses, defined as$(J_{previous}-J_{later})/J_{previous}$, as a function of number of passes through the amplifier crystal at different repetition rates—ideally this parameter is zero for a stable pulse train. The results in Fig. 5 are obtained using a pump power of 22W at 532 nm, an 80 µm pump mode at the crystal, and 2% loss in the amplifier cavity, which saturates the laser pulse amplification at 50 passes. At 10 kHz (the black curve in Fig. 5), the pulse energy ratio is constant at zero even with up to 200 passes. At 100 kHz, instability begins to appear as the number of passes exceeds 25. Interestingly, as the number of passes keeps increasing, the ratio eventually returns to zero at ~62 passes. This is due to the loss introduced in the laser cavity that reduces the energy of the previous pulse to the level that the later pulse can reach at lower gain. For repetition rates above 312.5 kHz, the ratio can be as high as 1 above 50 passes when the previous laser pulse amplification is completely saturated; i.e. as in the case of Fig. 4c. From this we conclude that in order to achieve good pulse-to-pulse energy stability, the pulse should be switched-out before its energy significantly saturates the laser gain. This also results in a (modest) reduction in amplifier efficiency. For example, we obtain an overall efficiency of 16.8% at 1 MHz, while the typical efficiency at a lower repetition rate using a pulsed pump laser is >20%.

 

Fig. 5 Laser pulse energy ratio $(J_{previous}-J_{later})/J_{previous}$ v.s. number of passes through the laser crystal at different repetition rates. When the ratio is zero, the two adjacent pulses have the same pulse energy. When the ratio is 1, the first pulse completely suppresses the gain of a later pulse.

Download Full Size | PDF

4. Conclusion and summary

We demonstrate the first ultrafast amplifier that can produce 52 fs, 3.7 uJ pulses at 1 MHz and 800 nm. The repetition rate is continuously tunable from 50 kHz up to 1.7 MHz. We discussed the technical challenges and how they were overcome. Although the output power is currently limited by the available pump laser power at 22 W, the maximum pump power this laser can accommodate is likely to be >100 W, limited by crystal cooling capacity.