Hollow-core photonic crystal fibers (HC-PCFs) are a specific type of optical fiber with a micro-structured cladding and hollow core at the center [1]. A particular type of HC-PCF, known as anti-resonant guiding fiber, has been used extensively to study the nonlinear interaction between laser light and gases. The long interaction length and broadband transmission windows of such fibers, in combination with the ability to tune the nonlinearity and dispersion through the gas type and pressure, have allowed their use in ultrafast nonlinear studies such as supercontinuum generation, ultraviolet resonant dispersive wave generation, light-plasma interaction, and optical soliton propagation [1–3].

Pulse compression using soliton dynamics in HC-PCF has been studied extensively since the pioneering work by Ouzounov et al. [4] in photonic-bandgap fibers, and Mak et al. [5] in anti-resonant fibers [6–8]. It is also a key component of many ultrafast dynamics in such fibers [9,10]. Recently, pulse compression from 340 fs down to $\sim$3.8 fs was demonstrated in a two-stage experiment [8]. Subsequently, the same group also demonstrated pulse compression from 250 fs down to $\sim$5.4 fs at a repetition rate of up to 10 MHz in a single stage assisted by chirped mirrors [11].

The tight fundamental-mode confinement of the light in combination with soliton self-compression creates pulses with a peak intensity well beyond the photoionization threshold of light gases, leading to the formation of plasma which can strongly act back on the light field. The polarizability of the free electrons is much larger and opposite in sign to the bound electrons [3,12], leading to a complex intensity-dependent (and hence time-dependent) drop of the refractive index which in turn causes an asymmetric phase modulation and a continuous shift of the spectrum towards higher frequency [13]. When combined with soliton dynamics, this effect leads to the soliton self-frequency blueshift [14], which has also been observed in semiconductors [15]. Recently, Huang et al. studied the soliton–plasma interaction in HC-PCF in detail [16–18]. In Ref. [18], they demonstrated ionization-induced adiabatic soliton compression in HC-PCF starting with 20 fs pump pulses at 800 nm; similar dynamics were predicted in Ref. [19]. Initial experimental time-frequency characterization of plasma-induced blueshift during self-compression was carried out in the context of driving high-harmonic generation [20]. However, combined time-frequency measurements of clearly identifiable plasma-induced blueshifting solitons in gas-filled hollow fibers have not previously been reported.

Here, we experimentally demonstrate the compression of µJ-level pulses directly from a commercial 220 fs full-width half-maximum (FWHM) Yb:KGW laser to $\sim$13 fs in a single stage without the need for chirped mirrors. In addition, we demonstrate wavelength tunable sub-15-fs pulses through soliton compression and soliton–plasma interactions in the same system. We fully characterize the temporally compressed pulses using sum-frequency generation (SFG) time-domain ptychography (TDP) [21], and show that 13 fs pulses can be obtained at 1030 nm using soliton compression. Moreover, we characterize the blueshifting solitons in the time-frequency domain and show that they can compress down to 15 fs, and be tuned from the pump wavelength down to 700 nm simply by adjusting the input pulse energy. These measurements also provide the first unambiguous time-frequency measurement of the soliton self-frequency blueshift.

Figure 1(a) shows the experimental setup. A 1030 nm, 220 fs FWHM pump laser (Light Conversion PHAROS) was used at 1 kHz. Results at higher repetition rate, up to 200 kHz, are shown later. The pulse energy was controlled using a half-wave plate and a thin-film polarizer. The beam was then coupled into a 29 µm core diameter single-ring anti-resonant photonic crystal fiber (PCF) through a 5 cm focal length AR-coated plano–convex lens. The fiber is designed to suppress higher-order modes and maintain a clean fundamental Gaussian-like mode profile. The fiber, which had a core-wall thickness of $\sim$270 nm, was measured to have a broadband transmission window ranging from around 630 nm up to 1370 nm with less than 1 dB/m loss across that whole range, and around 0.1 dB/m at 1030 nm. The 2 m-long fiber was sealed into the gas cells, so that a pressure gradient along the fiber could be created. Optical access was provided by an AR-coated 2 mm-thick MgF₂ input window and uncoated 1 mm-thick MgF₂ output window. We achieved up to 76 % transmission at 1030 nm with both windows installed while the fiber was evacuated, this is less than expected from the linear loss, and we attribute the discrepancy to an imperfect focused spot size and launch conditions. The output spectrum was collected using an integrating sphere connected to a fiber-coupled CCD spectrometer. The spectrometer covers the spectral range 200 nm to 1160 nm (Avantes ULS2048XL). The whole system is calibrated on an absolute scale with NIST traceable lamps.

 

Fig. 1. (a) Optical layout of the experiment: $\lambda$/2, half-wave-plate; TFP, thin-film polarizer. Panels (b) and (c) show the cross section of the PCF used in the experiment and the near-field of the beam in a micrograph image for the end face of the fiber. (d),(e) Experimental spectrum evolution for increasing coupled input energy, for 220 fs pulses launched into a 29 µm core diameter, Ar-filled fiber with an increasing pressure gradient from vacuum to (d) 800 mbar and (e) 1.2 bar. The red dashed lines indicate the energy used for the time-domain characterization measurements.

Download Full Size | PDF

Figure 1(d) shows the evolution of the experimental spectra for argon-filled HC-PCF with an increasing pressure gradient from vacuum at the input cell to 0.8 bar at the output cell. These parameters correspond to a soliton order of five at the output end of the fiber. The pulse spectrum broadens due to self-phase modulation which is accompanied by pulse compression in the time domain due to the effect of the anomalous group velocity dispersion (GVD). The pump wavelength lies in the anomalous GVD region throughout the fiber length. In addition, above 5 µJ, a strong spectral expansion toward the blue can be seen. This is due to the formation of plasma by the compressed pulse, leading to the soliton self-frequency blueshift, and is examined in more detail later.

We characterized the output pulses in the time domain using a home-built TDP setup shown in Fig. 2. We used the original laser pulse (the residual transmitted beam from the TFP) as the gate pulse. Both the PCF output (test) pulse and the gate pulse were focused into a BBO crystal in a non-collinear configuration. The BBO crystal was 10 µm thick and cut for type-I (o-o-e) phase-matching ($\theta = {29.2}^{\circ}$), resulting in a phase-matching window that extends from around 400 nm to beyond 1200 nm as shown in inset (i) of Fig. 2. Spectral measurements for the test and gate pulses were made at the crystal position to obtain the frequency marginal for the TDP trace retrieval. We also measured the background noise of the TDP signal to help to get a cleaner spectrum. After the BBO, the signal was isolated from the test and gate beams by using a combination of a Rochon prism and an aperture and then re-focused onto the spectrometer fiber tip. We used the extended ptychographic engine (ePIE) for pulse retrieval [22].

 

Fig. 2. (a) TDP characterization setup. The inset (i) shows the phase matching efficiency for SFG between the broadband test pulse and the narrowband gate pulse at 1030 nm using a 10 µm barium borate (BBO) crystal for $\theta ={29.2}^{\circ}$.

Download Full Size | PDF

Figures 3(a) and 3(b) show the measured and the retrieved TDP traces for 4.75 µJ input pulses with a retrieval error of $0.18{\% }$. The overall retrieved pulse spectrum matches the measured spectrum well, as shown in Fig. 3(c). Note the asymmetry in the spectrum due to the self-steepening effect. The pulse FWHM at the fiber output is 11 fs—obtained by back-propagating the retrieved pulse numerically through the optics and air-path from the fiber output to the characterization crystal. The retrieved pulse duration at the crystal position, without back-propagation, is still short, with a duration of around 13 fs, as shown in Fig. 3(d). The main pulse is accompanied by a broad pedestal component, an inherent feature of soliton-effect pulse compression [23]. The energy contained within the $1/e^2$ width of the compressed pulse is around 43${\% }$ of the total output pulse energy.

 

Fig. 3. (a) Measured and (b) retrieved SFG TDP traces on a logarithmic color scale for 4.75 µJ input pulses, using an increasing argon pressure gradient from vacuum to 0.8 bar. The root mean square (RMS) error between the measured and retrieved traces is $0.18{\% }$. (c) The measured and retrieved spectrum of the test pulse. (d) Temporal profile of the retrieved pulse from (b) at the crystal position. The FWHM duration of the pulse is 13 fs.

Download Full Size | PDF

Figure 1(e) shows the experimentally measured output spectrum as a function of pump energy when the output pressure is increased to 1.2 bar, increasing the nonlinearity. As the pulse propagates inside the fiber, it compresses from 220 fs to $\sim$10 fs and the spectrum broadens as shown. With further increase in the pulse energy, the intensity becomes sufficient to ionize the gas and form free electrons. The reduction in the refractive index due to liberated electrons is 1.8 times stronger than the refractive index change induced by the Kerr nonlinearity at 4.25 µJ [13]. Hence, plasma dynamics are more pronounced and a blueshifting pulse can be observed. Due to the significant anomalous dispersion experienced by the pulses, combined soliton–plasma dynamics occur, and the pump breaks up into self-frequency blueshifting solitons [24]. Above 5 µJ, the rate of the soliton blueshift becomes slower, because ionization-induced loss decreases the pulse energy. By controlling the gas pressure and pulse energy, the central wavelength of the blueshifting soliton can be easily tuned to the desired wavelength as shown in the two presented cases in Figs. 1(d) and 1(e). We checked the stability of the blueshifted soliton by measuring the relative intensity noise (RIN). We found that the RIN at the central wavelength of the blueshifted soliton was 0.25${\% }$, similar to the RIN of the pump laser system.

Figures 4(a) and 4(b) show the measured and retrieved traces for 7 µJ pulse energy [indicated by the dashed line in Fig. 1(e)]. We apply 40 iterations of the ePIE retrieval algorithm to retrieve the pulse profile from our measured traces with the spectral projection of the gate pulse applied for the first 20. The retrieved trace shows excellent agreement with the measured trace with an error of $0.48{\% }$. The retrieved pulse spectrum is shown in Fig. 4(c) along with the separately measured spectrum. Again, an excellent agreement between the measured and retrieved spectrum is obtained. The retrieved pulse in the time domain is shown in Fig. 4(d). The blueshifting soliton is located in the shaded area at the front of the pulse due to the blueshift induced pulse acceleration (higher group velocity for higher frequencies in the anomalous dispersion region). The FWHM duration of the blueshifting soliton is around 15 fs at the measured position. When back-propagated to the fiber end, the soliton duration is shorter, with a FWHM of 13.5 fs. The blueshifting soliton contains 0.5 µJ of energy when the input pulse energy is 7 µJ. This gives a conversion efficiency of around 7${\% }$ (calculated as the ratio of blueshifting soliton energy to input energy). The evolution of the spectrograms shown in Figs. 4(e)–4(g) clearly confirms the solitonic nature of the blueshifting pulse, as it remains parallel to the wavelength (frequency) axis, despite propagating over many dispersion lengths inside the fiber. This is the first unambiguous time-frequency measurement of the soliton self-frequency blueshift in hollow-core fibers.

 

Fig. 4. (a) Measured and (b) retrieved SFG TDP traces on a logarithmic color scale for 4.75 µJ input pulses using an increasing argon pressure gradient from 0 bar to 1.2 bar. The RMS error between the measured and retrieved traces is $0.48{\% }$. (c) The measured and retrieved spectrum of the test pulse. (d) Temporal profile of the retrieved pulse from (b). Panels (e), (f), and (g) show the spectrogram evolution for the retrieved and back-propagated pulses at three different energies 3 µJ, 4 µJ, and 7 µJ, respectively. The spectrograms are reconstructed from the retrieved and back-propagated pulses.

Download Full Size | PDF

All results shown so far were obtained using 1 kHz repetition rate. We also investigated the behavior of the plasma–soliton shaped spectrum as the repetition rate was increased. Figure 5 shows the output spectrum at different repetition rates for two energies and an increasing pressure gradient from vacuum to 0.8 bar. For 4.5 µJ, the output spectrum is similar for the three repetition rates presented. At this energy, the peak intensity of the compressed pulse is not sufficient to ionize the gas and no inter-pulse effects can be seen in the spectra as the repetition rate is increased. On the other hand, when the pulse energy is increased to 5.3 µJ, an obvious difference in the output spectra can be seen between the 200 kHz case and lower repetition rates. We investigated this for different parameters at different repetition rates and found that when the pulse energy is high enough to start ionizing the gas inside the fiber, the spectrum at high repetition rates starts to deviate from lower repetition rates and its bandwidth decreases. This suggests that at high repetition rates, the plasma effects created inside the fiber by one pulse do not fully dissipate in between pulses and affect the following pulses, reducing or even inhibiting the blueshift. Such an effect has been observed and studied in detail in Refs. [25–27]. In these articles, the authors suggested that plasma post-recombination heating and refractive index changes can strongly affect soliton compression starting from as low as 50 kHz repetition rate, as confirmed by our observations. Hence, care has to be taken to avoid plasma formation when developing a high-repetition-rate source, as described in Ref. [11].

 

Fig. 5. Experimental output spectrum 4.5 µJ and 5.3 µJ at three different repetition rates.

Download Full Size | PDF

In summary, we experimentally demonstrate the flexible generation of short ultrafast laser pulses at different wavelengths from a commercial Yb laser in a single system based on gas-filled HC-PCF. Through soliton self-compression, we obtain pulses at 1030 nm with 13 fs duration on target without dispersion compensation after the fiber. By increasing the gas pressure, we enter the regime of soliton–plasma interactions and generate frequency-tunable pulses down to 700 nm with 15 fs on-target duration through the soliton self-frequency blueshift. We characterize the time-frequency structure of the frequency-tunable compressed pulses using SFG TDP, and confirm the full soliton plasma dynamics in hollow-core fiber for the first time. Finally, we show that these dynamics can be scaled to 50 kHz in the presence of the plasma, but that at higher repetition rates inter-pulse plasma buildup inhibits the blueshift. Our results enable simple sources of sub-15-fs pulses using commercial and industrialized Yb-based ultrafast laser systems.