Femtosecond light pulses in the ultraviolet (UV) spectral range are required to study fundamental chemical and biological processes in biomolecules, such as DNA [1] and proteins [2], which exhibit strong absorption bands in this region. High time resolution and two-dimensional (2D) electronic spectroscopy experiments [3,4], in particular, call for ultrabroadband, few-optical-cycle pulses. The generation of sub-10 fs UV pulses is, however, not straightforward, since this spectral region poses serious challenges in terms of: (i) broad pulse bandwidth; (ii) pulse energy; (iii) spectral phase handling; and (iv) pulse characterization. There are no gain media that produce femtosecond pulses in the UV region, and broadband optical parametric amplifiers (OPAs) are hampered by the occurrence of two-photon absorption of the short-wavelength pump pulse [5]. Therefore, most approaches rely on nonlinear frequency conversion to shift ultrashort visible and infrared pulses to the UV range [6–17].

Using gases as nonlinear media, few-optical-cycle UV pulses have been obtained by third [6–8] or higher order [9] harmonic generation or by four-wave mixing between the second harmonic (SH) and the fundamental frequency (FF) of the driving laser, taking place either in a hollow fiber [10–12] or in a filament [13]. Frequency upconversion of broadband visible pulses in nonlinear crystals needs to compromise between the competing requirements of phase-matching bandwidth and efficiency [14]: broadband phase matching in fact requires thin crystals, which limit the efficiency of the process. This limitation can be overcome by achromatic phase matching [15–17], at the expense, however, of a significant complication of the experimental setup. In addition, linear dispersion due to propagation in transparent media is particularly severe in the UV spectral range, and careful control of the spectral phase to achieve transform-limited (TL) pulse duration is difficult, due to higher order dispersion introduced by pulse compressors. Finally, pulse characterization in the UV is complicated by the short-wavelength absorption of nonlinear materials and the unfavorable phase-matching conditions.

In this Letter, we introduce a simple scheme for the generation of tunable ultrabroadband UV pulses which effectively overcomes all these challenges, including spectral phase handling and pulse characterization. Our system is based on collinear sum-frequency generation (SFG) between a broadband visible pulse and a narrowband pulse, tunable from the visible to the infrared, and uses a suitable phase-matching configuration to balance the trade-off between bandwidth and efficiency and upconvert the whole visible bandwidth to the UV. This process also allows us to efficiently manipulate the spectral phase of the UV light through parametric transfer [18,19]. Finally, we successfully apply two-dimensional spectral shearing interferometry (2DSI) [20] to the UV pulse characterization.

We start by analyzing the phase-matching bandwidth of the SFG process between a narrowband pulse centred at frequency ${\overline{\omega }}_1$ and a broadband pulse centred at ${\overline{\omega }}_2$, to generate a broadband pulse at ${\overline{\omega }}_3={\overline{\omega }}_1+{\overline{\omega }}_2$. Since the efficiency of the process is limited by the wave vector mismatch $\mathrm{\Delta }k$, we can estimate the SFG bandwidth by evaluating $\mathrm{\Delta }k$ for the detuned process ${\overline{\omega }}_1+({\overline{\omega }}_2+\mathrm{\Delta }\omega )=({\overline{\omega }}_3+\mathrm{\Delta }\omega )$. To the first order

Equation (1) indicates that the SFG process has broader acceptance bandwidth for smaller values of the GVM ${\delta }_{32}$. Figure 1(a) shows the calculated GVM values for upconversion of a broadband visible pulse (centered at 600 nm) by a narrowband pulse with wavelength ${\lambda }_1$ ranging from 0.6 to 1.1 μm; data refer to SFG occurring in a $\beta$-barium borate (BBO) crystal for the three possible polarization schemes. The configuration that provides the smallest GVM is Type II (${\mathrm{e}}_1+{\mathrm{o}}_2\to {\mathrm{e}}_3$), which has a broad acceptance bandwidth for the ordinary axis and considerably narrower one for the extraordinary axis, thus allowing to efficiently transfer the broad visible bandwidth to the UV.

 

Fig. 1. (a) GVM ${\delta }_{32}$ for the SFG process in BBO between a narrowband pulse at ${\lambda }_1$ and a broadband pulse at ${\lambda }_2=600\mathrm{nm}$. (b) The dashed–dotted line shows expected TL UV pulse durations for the ${\mathrm{e}}_1+{\mathrm{o}}_2\to {\mathrm{e}}_3$ SFG process in a 50 μm crystal, starting from 6.2 fs visible pulses and monochromatic light at ${\lambda }_1$. The circles show TL pulse durations of the measured UV spectra of Fig. 3.

Download Full Size | PDF

The experimental setup for ultrabroadband UV pulse generation is shown in Fig. 2(a). The system starts with a regeneratively amplified Ti:sapphire laser (Libra, Coherent), which delivers 100 fs pulses at 1 kHz repetition rate and at 800 nm wavelength. A fraction of the laser light drives a single-pass visible noncollinear OPA (NOPA), pumped by the SH and seeded by a white-light continuum. This NOPA produces broadband pulses with a spectrum extending from 525 to 740 nm, corresponding to a TL pulse duration of 6.2 fs, and with energy higher than 10 μJ. The spectral phase of the NOPA pulses is then manipulated by a series of reflections onto a pair of dielectric double-chirped mirrors (DCMs) [21].

 

Fig. 2. (a) Experimental setup for the generation of the UV pulses. DL, delay line; DM, dichroic mirror. (b) 2DSI setup based on DFG between UV and stretched replicas of the FF. BS, beam splitter; FS, fused silica plate.

Download Full Size | PDF

The visible pulses are mixed with narrowband pulses, either derived from the FF beam or generated by a second narrowband OPA, tunable from 0.65 to 1.05 μm. SFG occurs in a 50 μm thick BBO crystal cut for Type II (${\mathrm{e}}_1+{\mathrm{o}}_2\to {\mathrm{e}}_3$) interaction. This crystal thickness is a good trade-off, balancing efficiency and bandwidth of the SFG process and allowing us to upconvert the whole NOPA spectrum. A sequence of SFG spectra, for different narrowband upconverting pulses, is shown in Fig. 3. The TL durations of the UV pulses are reported as circles in Fig. 1(b) and closely match the expected values. The spectral phase of the UV light is controlled by the mechanism of indirect phase transfer [18,19]: the SFG process transfers the nearly flat spectral phase of the NOPA pulse to the upconverted light. To compensate for the positive chirp acquired by propagation to the measurement point, we impart a slight negative chirp to the visible pulses, resulting in negatively chirped UV pulses. Fine tuning of the spectral phase of the UV is then achieved by adding a small amount of material dispersion.

 

Fig. 3. Tunability of the UV pulses from SFG between the 600 nm broadband NOPA pulse and a narrowband pulse with wavelength indicated on top.

Download Full Size | PDF

Particular attention has been devoted to the optimization and characterization of the SFG process between the visible NOPA and the FF light. To this aim, a 30 μJ fraction of the FF laser light was used. To focus the two collinear beams in the BBO crystal we used spherical mirrors with 250 mm focal lengths to prevent self-phase modulation effects induced by the high energy of the interacting pulses. Three UV dielectric mirrors were employed in order to remove the nonconverted light. The SFG stage provides UV pulses with energy up to 1.5 μJ. A typical spectrum is shown as a dashed line in Fig. 3: it extends from 315 to 380 nm and corresponds to 8 fs TL pulse duration.

The measurement of the duration of the UV pulses is challenging since, due to the lack of transparency and of a suitable phase-matching configuration, no traditional techniques based on SH or SFG can be applied. Methods based on frequency-resolved optical gating (FROG) or spectral interferometry (SI) in combination with a well-characterized reference pulse have been exploited for the measurement of UV pulses, including XFROG [22], MFROG [23], XTURTLE [24], and SI with an auxiliary pulse [25]. Self-referenced techniques such as self-diffraction FROG [6,14] and autocorrelation based on two-photon absorption [26,27] have also been successfully applied. Finally, Baum et al. introduced the ZAP-SPIDER [28] and demonstrated the measurement of 7 fs pulses in the UV with this method [15].

Here, we propose a system based on the 2DSI technique [20], which allows full characterization of ultrabroadband pulses. The 2DSI method relies on spectral shearing interferometry with zero delay; with respect to the SPIDER technique, which encodes phase as a sensitively calibrated fringe in the spectral domain, 2DSI robustly encodes phase along a second dimension. In traditional 2DSI, two highly chirped replicas of a broadband gate pulse are mixed with the test pulse in a nonlinear crystal, to generate two spectrally sheared upconverted signals by SFG. The delay of one of the chirped auxiliary pulses is then scanned over a few optical cycles and the spectrum of the upconverted signal is recorded as a function of this phase delay, yielding a 2D map which encodes the group delay (GD) of the test pulse. In our approach, instead, the two replicas of the gate are mixed with the UV pulse to generate two spectrally down-converted signals by difference frequency generation (DFG). The gate is provided by a strongly chirped portion of the FF light [Fig. 2(b)]; the DFG shifts the UV light into the visible spectral range, where favorable phase-matching conditions exist. In order to chirp the 100 fs FF pulse, we employed a grating-based compressor, introducing a group delay dispersion (GDD) of the order of $-{10}^5{\mathrm{fs}}^2$. Two collinear replicas of this gate are then generated by means of a Michelson interferometer and noncollinearly mixed with the UV pulses in a 10 μm thick Type-I BBO crystal. The resulting difference frequency (DF) signal is detected by a spectrometer (HR2000, Ocean Optics). A delay stage in one of the arms of the Michelson interferometer [${\tau }_{\mathrm{\Omega }}$ in Fig. 2(b)] controls the spectral shear $\mathrm{\Omega }$ between the two gate pulses. The other arm is equipped with a high-precision translation stage (SmarAct, SLC-2460), which scans the delay $\tau$ between the two gate pulses by $\pm 10\mathrm{fs}$.

The results of 2DSI measurements are shown in Fig. 4. Figure 4(a) displays the 2DSI interferogram: by DFG with the gate pairs, the UV spectrum is shifted to the visible spectral range and acquired at various delays $\tau$, giving rise to fringes along the vertical delay axis. The wavelength-dependent delay of the fringes [white line in Fig. 4(a)] is extracted by Fourier analysis along the scanning delay. This delay is proportional to the GD of the UV pulse, which is detailed in Fig. 4(b), through the spectral shear $\mathrm{\Omega }$. The accurate determination of $\mathrm{\Omega }$ is crucial in order to properly deduce the GD from the measured fringe delay and is obtained either by comparing the DF spectra separately arising from the two arms of the interferometer or from the delay ${\tau }_{\mathrm{\Omega }}$ and the dispersion of the gate pulse. These methods are sensitive, respectively, to the exact characterization of the shift between the two DF spectra and to the precise determination of the GDD introduced by the compressor. We tested the reliability of the deduced spectral shear by measuring the GD introduced by various fused silica plates on the UV path. With both approaches, the retrieved GD was hardly matching the expected dispersion. For this reason, we calibrated the device by a different approach: we measured the dispersion introduced on the UV pulse by a 3 mm thick ${\mathrm{CaF}}_2$ plate. We first evaluated the fringe delay of the 2DSI map due to the plate. Then, comparison of this delay with the plate GD deduced from the Sellmeier equations allowed us to calculate the ratio between fringe delay and GD. Also in this case, we tested the retrieved conversion factor by measuring the dispersion of several fused silica plates inserted in the UV path. The excellent matching between the measured and the calculated GD reported in Fig. 4(c) validates the reliability of the approach.

 

Fig. 4. (a) Experimental 2DSI map of the compressed UV pulses. (b) Detail of the retrieved GD. (c) Dispersion of thin fused silica plates as deduced from 2DSI measurements (solid lines) and from Sellmeier equations (dashed lines).

Download Full Size | PDF

Preliminary 2DSI measurements showed that the UV pulse immediately after the SFG stage had a small negative chirp, confirming that the negative dispersion introduced on the visible pulse by the DCMs was nonlinearly transferred to the UV light. Compressed UV pulses were obtained by adding a 400 μm thick fused silica plate to the UV path. The 2DSI map shown in Fig. 4(a) refers to the optimally compressed UV pulse. The retrieved GD detailed in Fig. 4(b) is almost flat and reproducible from day to day; the little ripples are attributed to the DCMs and the dielectric mirrors we used to remove the unconverted light. By integrating the GD, we could estimate the spectral phase [Fig. 5(a), solid red line] and the temporal profile of the corresponding pulse [see Fig. 5(b)]. The resulting pulse has 8.4 fs width, very close to the 8 fs TL value.

 

Fig. 5. (a) Spectral intensity and phase for the compressed UV pulse. (b) Retrieved pulse in the temporal domain.

Download Full Size | PDF

In conclusion, we have introduced a simple and robust scheme for the generation of tunable sub-10 fs UV pulses with energy in excess of 1 μJ. The system allows easy control of the spectral phase in the UV range. In addition, we have demonstrated a technique for the characterization of the UV spectral phase and demonstrated the generation of 8.4 fs UV pulses, applying for the first time, to the best of our knowledge, the DFG process to the 2DSI technique. Due to the simplicity and reliability of their generation method, we expect that these pulses will become valuable tools for high-time-resolution and 2D UV spectroscopy.