Intense, few-cycle laser sources in the short-wave infrared (SWIR) region (1.4 $\mathrm{\mu} \text {m}$–3.0 $\mathrm{\mu} \text {m}$) have drawn considerable attention as driving sources for nonlinear frequency conversion processes. Compared with well-established near-infrared (near IR) laser sources (0.75 $\mathrm{\mu} \text {m}$–1.4 $\mathrm{\mu} \text {m}$), the longer driving wavelengths enable an efficiency boost in terahertz generation using two-color plasma [1,2], and they offer the ability to employ nonoxide crystals with high nonlinear coefficients for mid-infrared generation beyond 5 $\mathrm{\mu} \text {m}$ [3]. Moreover, for high-harmonic generation, the SWIR driving wavelength is considered a promising trade-off in achieving a high photon energy cutoff toward the important water window spectral region (280 eV–540 eV) while maintaining a reasonable single-atom response [4]. Table-top laser sources in these spectral regions with a high conversion efficiency and high photon flux are attractive tools for biological, medical, and industrial applications, such as imaging, spectroscopy, and sensing [5–7]. This implies driving sources providing simultaneous high peak and average power.

A traditional method to generate SWIR, few-cycle pulses is optical parametric chirped pulse amplification (OPCPA) driven by near-IR laser sources, which can offer a broad phase-matching bandwidth and stable carrier-envelope phase [8]. To date, 100 kHz repetition rate SWIR OPCPA systems have been reported, providing 16.5 fs pulses at 2.2 $\mathrm{\mu} \text {m}$ with 25 W average power [9], and 166 fs pulses at 1.75 $\mathrm{\mu} \text {m}$ with 106 W average power [10]. Meanwhile, 2.3 TW (100 mJ, 31 fs) at 1.7 $\mathrm{\mu} \text {m}$ [11] and 2.5 TW (30 mJ, 11.6 fs) at 1.8 $\mathrm{\mu} \text {m}$ [12] (both with 10 Hz repetition rate) represent, so far, the highest peak powers reported in the SWIR spectral region. Another viable approach to achieve few-cycle pulses in the SWIR region is via direct laser emission with a suitable laser-host material. In recent years, bulk lasers based on chromium (Cr)-doped chalcogenide crystals (such as ZnSe, ZnS) have undergone rapid development. Based on the broad emission bandwidth of these gain materials, there have been a number of reports on mode-locked oscillators or amplifiers directly delivering few-cycle pulses with multi-watt average power at high repetition rates (70 MHz–1 GHz) [13–15]. However, owing to their low thermal conductivity, large thermo-optic coefficient, and relatively low optical damage threshold, the maximum output power of Cr:ZnS and Cr:ZnSe lasers is currently limited to the 10 W-level, without spinning the gain element or the laser beam [13]. Another promising SWIR laser type is a thulium (Tm)-doped fiber chirped pulse amplification (Tm:FCPA) system followed by a nonlinear pulse compression stage. Tm-doped fused silica has a broad emission bandwidth around 1.9 $\mathrm{\mu} \text {m}$, which supports sub-100 fs pulses. In terms of power scaling, Tm:FCPA systems have demonstrated 265 fs pulses with 1 kW average power at 80 MHz repetition rate [16], and 85 fs pulses with 1.65 mJ pulse energy and 167 W average power [17]. To further reduce the pulse duration to a few optical cycles, nonlinear pulse compression based on self-phase-modulation-induced spectral broadening can be employed. The most notable recent achievements of this technique are 13 fs pulses at 1.82 $\mathrm{\mu} \text {m}$ with 34.4 $\mathrm{\mu}$J pulse energy and 43 W average power, out of a gas-filled anti-resonant hollow-core fiber (HCF) [18], and 35 fs pulses at 1.94 $\mathrm{\mu} \text {m}$ with 122 $\mathrm{\mu}$J pulse energy and 51 W average power, based on a gas-filled multi-pass cell [19].

In this Letter, we report on the nonlinear pulse compression of a high-power Tm:FCPA system using a gas-filled HCF. This scheme reduces the input pulse duration from 90 fs to 10.2 fs, with 1.3 mJ output pulse energy, 80 GW peak power, and 132 W average power. This sub-two cycle system can be readily used for subsequent nonlinear frequency conversion experiments.

The experimental setup is depicted schematically in Fig. 1. The input for the nonlinear pulse compression stage is based on a custom-built Tm:FCPA system with a coherent combination of four rod-type amplifiers, similar to the one described by Heuermann et al. [17]. In this experiment, the Tm:FCPA system delivers 90 fs pulses at 1.92 $\mathrm{\mu} \text {m}$ central wavelength and about 180 W average power with 101 kHz repetition rate. The nonlinear pulse compression stage comprises two vacuum chambers for the input and output coupling, and a high-pressure chamber filled with argon for the nonlinear broadening. The bottom side of the high-pressure chamber is water-cooled, and a rod-type HCF is placed on a V-groove in the middle of the chamber. To mitigate gas ionization effects, we choose a HCF inner diameter of 500 $\mathrm{\mu} \text {m}$, which is well above the minimum calculated radius for the onset of ionization effects described in Vozzi et al. [20]. The 1 m-long HCF can provide sufficient nonlinearity while offering good power handling with a theoretical transmission of 89.5% [21]. Anti-reflection coated, fused-silica lenses ($f =750\,\text {mm}$) are used for the input and output coupling of the HCF. The collimated output beam is then sampled by an uncoated wedge reflection for characterization purpose, and the output power is measured from the transmission of the wedge. The input and output sides of the high-power section are placed in vacuum chambers with <1 mbar pressure to mitigate detrimental effects caused by water-vapor absorption [22,23]. The low-power characterization path is in a laboratory atmosphere.

 

Fig. 1. Experimental setup of nonlinear pulse compression stage. Input- and output-coupling lenses: anti-reflection coated, fused-silica lenses ($f =750\,\text {mm}$). Windows: 1 mm thick, anti-reflection coated, fused-silica windows.

Download Full Size | PDF

To determine the optimal compression conditions, we have performed an argon gas pressure scan with an input average power of 177 W. The HCF output powers at different argon gas pressures are plotted in Fig. 2(a). The output power is nearly constant around 139 W for argon pressures ranging from 1 to 3 bar, which corresponds to an overall transmission of 78.5%. When the applied argon pressures rises above 3 bar, the output power decreases noticeably with increasing gas pressure, reaching a minimum output power of 125 W at 4.25 bar. Additionally, the near-field image of the HCF output beam [Fig. 2(b)] is unchanged at applied argon pressures from 1 bar to 3 bar. The imaged beam becomes unstable starting from 3.5 bar, and becomes clearly distorted at 4 bar. Moreover, it is noteworthy that neither the transmission drop nor the output beam degradation is observed at low input average power levels (at the same repetition rate) with the same applied gas pressures. Therefore, we believe that these two effects arise from an increasing Kerr lensing associated with high pulse energies and high applied argon pressures. The estimated peak power of the input pulses (at 177 W) is about $\hat {P}=13\,\text {GW}$. Compared with the calculated critical peak power ($\hat {P}_{\text {crit}}$) for self-focusing [24], the ratio $\hat {P}/\hat {P}_{\text {crit}}$ [Fig. 2(a)] reaches 70% at 3.0 bar argon and is almost 100% at 4.25 bar. A detailed investigation of the underlying physical processes will be conducted in the future. Apart from that, the spectral broadening [Fig. 2(c)] is measured with a grating spectrometer (measurement range: 1.2–2.4 $\mathrm{\mu} \text {m}$) and a multimode delivery fiber, to avoid nonlinear propagation within the measurement setup. The Fourier-limited pulse durations [full width at half maximum (FWHM)] corresponding to the measured spectra are shown in Fig. 2(d). Significant spectral broadening is observed with an increase of the applied pressure up to 3 bar, where the Fourier-limited pulse duration is 9.8 fs. Increasing the gas pressure further leads to a less efficient spectral broadening. The Fourier-limited pulse duration drops to 8.1 fs at 4 bar, and increases slightly to 8.3 fs at 4.25 bar. This also agrees with the power drop, which leads to a reduction of the accumulated $B$ integral in the nonlinear broadening stage.

 

Fig. 2. Results of argon gas pressure scan. (a) Blue: output average power ($P_{\text {avg}}$) as a function of argon pressure. Red: ratio between input pulse peak power and calculated self-focusing critical peak power ($\hat {P}/\hat {P}_{\text {crit}}$) as a function of argon pressure. (b) Near-field images of HCF output beam at argon pressures of 1 bar, 3 bar, and 4.25 bar. (c) Measured normalized broadened spectra (in logarithmic scale) at different argon pressures. (d) Calculated Fourier-limited pulse durations (FWHM) ($\tau _{\text {FT}}$) based on measured output spectra in (c).

Download Full Size | PDF

Taking the spectral broadening and output beam stability into consideration, we chose 3 bar of argon gas for the nonlinear pulse compression scheme. In this case, with 180 W input average power, the output average power is 132 W, corresponding to 1.3 mJ output pulse energy at a repetition rate of 101 kHz. The near-field image of the output beam is depicted in Fig. 3, together with the spectral and temporal characterization of the output pulses. The light-blue areas in Fig. 3 are the measured spectrum and the corresponding calculated Fourier-limited pulse. The sharp edge of the measured spectrum in the short-wavelength side is due to the limited measurement range of the spectrometer (1.2–2.4 $\mathrm{\mu} \text {m}$). The broadened spectrum is centered at a wavelength of 1.87 $\mathrm{\mu} \text {m}$, with a $-40$ dB spectral width, spanning over one optical octave. The output pulses are compressed by propagating through fused-silica optical elements (one window and the output-coupling lens), which amount to a total thickness of about 3 mm and a group delay dispersion (GDD) of about $-225$ fs$^{2}$ [25]. A pair of chirped mirrors is used to compensate the dispersion of the second fused-silica window (1 mm thick, placed before the characterization path). The compressed pulses are characterized by a commercial (SourceLab) TIPTOE device (tunneling ionization with a perturbation for the time-domain observation of an electric field) [26]. The temporal measurement window is 1.6 ps, with a step size of 0.07 fs, and the rms error of the reconstruction is $3\times 10^{-5}$. The retrieved spectral and temporal data are shown in Fig. 3 with solid lines. The retrieved output pulse duration is 10.2 fs (FWHM), corresponding to 1.64 cycles at the central wavelength of 1.87 $\mathrm{\mu} \text {m}$. The retrieved pulse shows a good temporal pulse contrast, with about 66% of the pulse energy confined in the main temporal feature. Normalizing the retrieved pulse on the measured pulse energy, we retrieved a pulse peak power of about 80 GW.

 

Fig. 3. (a) Measured spectrum (light-blue area), simulated spectrum (green) and spectral phase (orange), and retrieved spectrum (blue) and spectral phase (red). Inset: Near-field image of output beam. (b) Fourier-limited pulse (light-blue area) calculated from measured spectrum in (a) and simulated (red) and retrieved (blue) temporal pulse profiles. All temporal pulse profiles are normalized to the output pulse energy.

Download Full Size | PDF

Furthermore, we have also performed a numerical simulation of the nonlinear pulse compression stage. The input pulse profile of the simulation is calculated from the retrieved spectrum and spectral phase of the Tm:FCPA measured via a second-order, frequency-resolved optical gating (FROG) setup [17]. The pulse propagation along the gas-filled HCF is simulated based on the unidirectional pulse propagation equation [27], with an assumption of fundamental mode propagation. The simulation tool includes both gas and waveguide dispersion, third-order nonlinearity, and plasma formation (by calculating the Ammosov–Delone–Krainov ionization rates [28]). Afterward, the pulse compression in the simulation is achieved by adding the dispersion equivalent to 2.8-mm-thick fused silica based on the Sellmeier formula [25], which is very similar to the experimental condition. The difference of the simulated and measured spectra can be explained by the residual third-order phase present on the input pulses. Apart from this, the simulated results match the measured and retrieved ones well (Fig. 3); this indicates that the impact of the Kerr lensing in our experiment is not significant under this condition.

Figure 4 depicts the result of the short-term stability measurement. The measurement and evaluation method is the same as that of Heuermann et al. [17]. The relative intensity noise (RIN) of the Tm:FCPA output is 0.75%, integrated in a frequency range from 20 Hz to 50 kHz (half of the repetition rate). In the nonlinear pulse compression stage, the major noise contributions are mostly in the low frequency range up to 2 kHz, which, we believe, originate from mechanical vibrations coming from water cooling and vacuum pumps. Integrating the trace, the RIN of the sub-two cycle pulses is 1.0%.

 

Fig. 4. Results of short-term stability measurement. (a) Power spectral density of laser noise over frequency. (b) Dark-corrected, integrated noise from 20 Hz to 50 kHz. Red: measurement result of Tm:FCPA output. Blue: measurement result of nonlinear pulse compression output. Gray: Dark trace.

Download Full Size | PDF

In conclusion, we have demonstrated a sub-two cycle laser source at a central wavelength of 1.87 $\mathrm{\mu} \text {m}$, with 1.3 mJ of pulse energy, 80 GW of peak power, and a record average power of 132 W. This performance is enabled by a Tm:FCPA system operated at 101 kHz repetition rate and a subsequent HCF nonlinear pulse compression based on a gas-filled HCF. An argon pressure scan allows us to observe limitations, potentially resulting from Kerr lensing, which result in a transmission drop, output beam degradation, and inefficient spectral broadening. According to this, we carried out nonlinear pulse compression with 3 bar of argon, which compresses the input pulses from 90 fs pulse duration to 10.2 fs, with stable output pulses ($\approx 1\%$ RIN). This source thereby marks the highest average-power few-cycle source in the SWIR to date, to the best of our knowledge. Apart from that, based on this result, further performance scaling toward multi-mJ, several 100 W, sub-two cycle sources is feasible in the near future, by using lighter noble gases and larger HCFs. Furthermore, featuring sub-two cycle pulses with mJ-level pulse energy and 80 GW pulse peak power at 101 kHz repetition rate, this laser source is well suited to drive nonlinear frequency conversion into the terahertz, mid-infrared, and soft X-ray spectral regions.