1. Introduction

To measure the ultrafast dynamics of photoinduced processes, light pulses with a duration shorter than the timescale of the process under investigation are necessary. Sub-10 fs pulses are routinely generated in part of the visible and near-IR spectral range, where Ti:Sapphire (Ti:Sa) lasers and visible non-collinear optical amplifiers (NOPAs) operate. However, a plethora of organic molecules do not possess absorption bands in these regions, but absorb light at shorter wavelengths, namely the blue VIS region between 400 and 500 nm and the near-UV range down to 250 nm. Notorious examples for blue-green absorbers are carotenoids, a class of pigments responsible for light harvesting and photoprotection in photosynthetic organisms [1–3], with ultrafast relaxation timescales as fast as 50 fs [4]. DNA and nucleobases [5–7] or metal tris(biphenyl) complexes [8] are known for ultrafast, i.e. sub 50 fs, relaxation dynamics after mid- or near UV absorption. For these spectral ranges, the generation of ultrashort pulses is more challenging.

The main difficulty in generating pulses suited for spectroscopic applications in the blue spectral region lies in obtaining broadband spectra, as no direct broadband laser sources exist. An additional complication is given by the fact that spectral broadening effects in the near UV are large. When considering excitation-frequency resolved ultrafast methods such as 2D electronic spectroscopy, ultrabroadband pulses are however a requirement in this spectral range [9]. Several groups so far have achieved sub-30 fs or even sub-20 fs pulses, but for relatively narrowband spectra [10–12]. So far, only a limited number of schemes to generate broadband pulses in the blue VIS and UV region have been implemented experimentally. One approach is to generate pulses with a large spectral bandwidth in the UV from commercial photonic crystal fibers using four-wave mixing [13–15]. In particular, pumping a gas-filled Kagomé fiber with ultrashort, few-$\mu$J, 400 nm pulses has been shown to efficiently generate deep-UV pulses with a Fourier limit of 15 fs [16]. Higher pulse energies are achieved by supercontinuum generation in gas-filled hollow core fibers. This approach entails well-established concepts for pulse compression by chirped mirrors [17]. Kobayashi and coworkers and Krausz and coworkers reported 7.5 - 8 fs pulses at 400 nm obtained by spectral broadening in a hollow core fiber (HCF) [18,19]. This method is however not tuneable in wavelength, which means that spectra below 350 nm are unattainable. High requirements are posed to the pump laser if the spectral bandwidth is to extend below 300 nm. Most importantly, ultrashort (sub-10 fs) pump pulses are necessary to obtain the desired broadening in this case [20].

Already in 1976, Volosov et al. suggested that mapping each color to its phase-matching angle in a nonlinear crystal (achromatic phase-matching) results in broadband second-harmonic generation with acceptable efficiencies [21]. In the nineties, Szabó and Bor [22] and Martínez [23] independently proposed frequency-doubling setups utilizing the angular dispersion of gratings to achieve achromatic phase matching in the nonlinear crystal and then eliminate the angular dispersion from the generated second-harmonic radiation. Later, Trebino and coworkers built the first ASHG setup based solely on prisms, successfully frequency-doubling the 5 ns output of an optical parametric oscillator to a blue beam with a bandwidth of 80 nm [24]. More recently, achromatic phase-matching was applied by Baum et al. to the frequency doubling of a 100 nm broad NOPA output spectrum, leading to UV pulses centered at 325 nm [25]. While this technique yields sub-10 fs pulses by the use of deformable mirrors, the achieved throughput energy is below 0.25 $\mu$J after the BBO crystal and 0.12 $\mu$J after compression. Additionally, the pulses are not tunable to central wavelengths longer than 375 nm. Manzoni and coworkers managed to generate high-energy (10 $\mu$J), sub-20 fs pulses in the UV by frequency doubling the output of a two-stage NOPA, but did so at the expense of spectral bandwidth [26]. Recently, Zigmantas et al. demonstrated the generation of broadband, (sub-)10 fs DUV pulses at high laser repetition rates by achromatic frequency doubling of the output of a visible NOPA [27]. While sub-10 fs pulses could only be achieved by using deformable optics in the compression routine, they were also able to obtain 10.5 fs pulses using only standard optics. Considering the blue part of the visible spectrum, Johnson et al. showed sub-10 fs pulses with 500 nm central wavelength range directly from a single-stage NOPA by making use of anamorphic focusing of the pump into the amplifying BBO crystal [28]. However, the available range of amplification in this method is restricted to a lower bound of 450 nm due to limitations set by phase matching.

This work is dedicated to the generation of ultrashort and tunable blue pulses by achromatic second-harmonic generation (ASHG) of a hollow-core fiber output. With this approach, we avoid the generation of UV-wavelengths in the hollow core fiber directly, alleviating the need for amplified 800 nm pulses of 10 fs pulse duration or less. Instead, we employ the 25 fs output pulses of a commercial Ti:Sa laser as an initial laser source. The generated supercontinuum is then frequency-doubled by ASHG as would generally be done for a NOPA output, yet with the advantage of higher pulse energies of several $\mu$J after the doubling crystal. Such high pulse energies are not necessary for the experiments presented here, but are advantageous for potential application in two-dimensional electronic spectroscopy in the UV [9,29], where low throughput efficiencies of pulse shaping devices are a reported problem [30]. The available tuning range is 250-550 nm. We show pulse durations of 12 fs at 475 nm central wavelength after compression, achieved by employing only conventional optics. Further, we demonstrate that the pulses obtained by ASHG are suitable for spectroscopic applications by integrating them into a broadband transient absorption (TA) experiment as pump pulses. As a proof-of-principle study, we measure the ultrafast relaxation dynamics of $\beta$-carotene in mesitylene after fully resonant excitation.

2. Experiment

The front-end of the system is a 5 kHz, 2.5 mJ, 25 fs amplified laser system (Coherent Legend Elite Duo HE+ Ti:Sapphire MOPA). To broaden the spectra of the Ti:Sa ouput pulses, a 180-$\mu$J fraction is sent into a 1 m long, 250 $\mu$m wide commercial hollow-core fiber (HCF) filled with 1 bar of Argon gas. This particular pump energy was chosen because it is the highest energy that can be used without observing ionization of the gas. This results in the generation of 110 $\mu$J pulses spanning the spectral range between 400 and 1000 nm. At the exit of the HCF, the pulses are collimated with a f =1000 mm spherical focusing mirror and split by a pair of fused silica wedges. The transmitted beam is frequency-doubled in the ASHG unit.

 

Fig. 1. Achromatic second harmonic generation unit as described in the text. Part of the output of a commercial HCF is laterally dispersed by a pair of fused silica prisms (FSP) and focused into the achromatic second harmonic generation crystal (ASHG) by a cylindrical mirror (CM). The frequency-doubled light is then collected by another CM and recollimated in a sequence of three CaF$_2$ prisms. The prisms, together with a grating stretcher, compress the pulse up to the third order of dispersion. The inset shows how different colors are mapped to their corresponding phase-matching angles in the crystal.

Download Full Size | PDF

2.1 Achromatic second harmonic generation

The ASHG unit (Fig. 1) in which part of the HCF output is frequency-doubled is similar to ones already described by Baum and Riedle [25] and Volosov and Goryachkina [21]. In order to achieve broadband conversion, each color of the incoming pulse is mapped to its corresponding phase-matching angle (inset in Fig. 1) in the nonlinear crystal. The necessary separation of the colors is obtained by laterally dispersing the beam with a fused silica prism pair (apex-to-apex distance: 1.3 m) before frequency doubling. As we employ a type I nonlinear crystal for SHG, the p-polarized HCF output is rotated by 90$^{\circ }$ with a periscope placed before the fused silica prism pair, which is AR-coated for 60$^{\circ }$ incident s-polarized light to minimize reflection losses.

After passing through the first prism pair, the pulses are focused into the nonlinear crystal by an Ag-coated, 50 mm cylindrical mirror. The frequency-doubled output is then collected by another 50 mm cylindrical mirror with Al coating. Spherical and parabolic mirrors were also tested, but limit the energy of the input pulses to 5 $\mu$J, as filamentation in air is observed at higher energies, leading to distortions in the spectral phase of the pulse after ASHG. While this effect can be avoided by a longer focal length of the focusing mirror, this leads to an increase of the necessary prism separation distance, making the system too long and impractical in its operation. The advantage of the employed cylindrical mirrors lies in the fact that they focus the beam in a line instead of a single point, thus keeping the energy far below the critical power ($P_{critical}$) for filamentation in air, even at input energies as high as 100 $\mu$J. As $P_{critical}$ is orders of magnitude lower in solids than in gases, supercontinuum generation can still be observed in the second harmonic crystal under these conditions. However, it can be avoided by shifting the crystal slightly out of the focus. We define the optimum operation condition as the position for which we observe the highest intensity output without generating whitelight in the BBO. To exclude the presence of aberrations introduced by the focusing cylindrical mirror, the crystal was translated by 1.2 $\mu$m in 0.3 $\mu$m steps in both directions relative to the optimum operation position. This resulted in a change in the pulse intensity, but no difference was observed in the pulse spectrum. This means that the focal spot in the nonlinear crystal is homogeneous and free of aberrations for the purpose of the experiment. Additionally, the beam profile was measured up to 0.3 $\mu$m before and after the operation position (0.1 $\mu$m steps) and no distortion of the (near-Gaussian) spatial mode was observed.

The second harmonic generation takes place in a 300-$\mu$m type-I BBO crystal cut at 28.4$^{\circ }$. This particular crystal thickness was chosen as a compromise between the obtained conversion efficiency and the quality of collimation of the output beam. Under these conditions, the fundamental HCF spectrum shown in Fig. 2(a) typically leads to a second harmonic spectrum spanning from 380 nm to 530 nm (tail-to-tail, blue line in Fig. 2(b)). The conversion efficiency for the frequency-doubling process is 11%, resulting in 6 $\mu$J pulses for an input energy of 52 $\mu$J. The dip at 425 nm results from a correspondent dip in the HCF supercontinuum. To facilitate pulse compression, the region of SHG of the fundamental beam around 800 nm is blocked mechanically, which means that only wavelengths longer than 430 nm enter the pulse compression unit of the setup (blue shaded area in Fig. 2).

 

Fig. 2. Comparison between HCF spectra in (a) and ASHG spectra (b). In b, the blue line outlines the broadest spectrum achievable on the red side. The shaded blue area represents the spectrum used for the experiments. Finally, the shaded purple area represents the broadest spectrum achievable on the blue side.

Download Full Size | PDF

The spatial dispersion of the pulses after frequency doubling is compensated in a 3-prism configuration (prism material: CaF$_2$), which also acts as a prism compressor to add negative group delay dispersion. Compensation of third order dispersion is achieved in a grating-based compressor in $4f$ geometry (f = 200 mm). The diffraction efficiency of the grating in use (1200 grooves/mm, centered at 350 nm, Oriel instruments)is 58%. Additional reflection losses from the protected aluminum mirrors result in a throughput of the $4f$-stretcher of about 8%, which could be significantly improved by the use of a grating with higher diffraction efficiency. The compressed pulses show an energy of $\sim$0.2 $\mu$J/pulse at the sample position of the transient absorption setup (cf. section 2.3).

In terms of pulse energy stability, the presented ASHG setup performs only marginally worse than the HCF fundamental. Simultaneous measurements of the ASHG and HCF pulse powers over 100 000 shots by means of a two-channel photo-diode digitizer (Glaz-PD photo-diode digitizer, Synertronic Designs) result in a root-mean-square error (RMSE) of 1.1 % for the ASHG and a RMSE of 0.93 % for the HCF, respectively.

It is important to underline that, if optimized accordingly, the HCF output spectrum could in principle reach wavelengths down to 420 nm in the blue VIS region without the need for frequency doubling. However, the HCF spectrum offers limited options when it comes to spectral shaping of the pulses. In the case of the ASHG, selected frequencies can be readily filtered out by mechanically blocking them after the light is dispersed in the first prism pair (see Fig. 2(b)). The ASHG spectrum is thus easily tuned by choosing the spectral range of the HCF spectrum to be doubled and by adjusting both the distance between the dispersing prisms and the angle of the BBO crystal accordingly. The spectrum of the HCF can be shaped by bandpass filters. However it has been shown that, for the most common cases of interference and notch filters, a substantial phase shift is introduced to the pulses, distorting their temporal shape [31]. This renders pulse compression problematic, whereas the ASHG pulses have no such issues even after the desired spectral range was selected as described above. Additionally, the ASHG scheme has the advantage of being able to generate pulses below 400 nm (see Fig. 2(b)).

It is important to note that the frequency-doubled spectrum in the region of the the Ti:Sa fundamental (760-860 nm, ASHG spectrum between 380-430 nm) is, in the general case, too modulated to be compressable. This limits the tunability of the ASHG setup to the regions between 280-380 nm and 430-530 nm (tail-to-tail) or 300-360 nm and 450-510 nm central wavelength. The bluest spectrum we could generate is given in Fig. 2 (filled purple). The spectrum supports 7 fs pulses, however in this work we have not attempted to compress it.

2.2 ASHG pulse measurement

The duration of the ASHG pulses was measured by means of transient grating FROG (TG-FROG) [32]. The transient grating was created in a 100 $\mu$m thick fused silica glass slide with two non-collinear pulses obtained by spatially filtering the 25 fs Ti:Sa fundamental with a mask [33]. To avoid the introduction of additional dispersion in air, the glass slide was kept at the sample position of the TA setup (cf. section 2.3). The angle between the ASHG pulses and the near-IR pulses was set to 4$^{\circ }$. The pulses were focused into the glass sheet by a 300 mm lens. As the Fourier limit of the ASHG pulses (8 fs) is shorter than the duration of the gate pulses, the pulse duration was obtained with a TG X-FROG retrieval software (FROG 3.1.2, Femtosoft Technologies). The error of a retrieval with a 512x512 matrix is 0.0054. The retrieved pulse duration is 12 fs (FWHM), which compares well to the 8 fs transform limit (Fig. 3).

 

Fig. 3. TG-FROG measurements of the ASHG pulses. (a) Measured TG-FROG trace, (b) retrieved TG-FROG trace, retrieved (c) spectral intensity and phase and (d) temporal intensity and phase.

Download Full Size | PDF

2.3 Transient absorption

To test the suitability of the generated pulses for spectroscopic applications, the compressed ASHG pulses were integrated into a broadband TA-experiment as pump pulses. They are focused into the sample via a 250 mm focusing mirror to a spot size of 60 $\mu$m as measured by a CinCam CMOS beam profiler (Cinogy Technologies). The energy of the pump can be adjusted with a variable neutral density filter.

Details of the pump-probe setup are given in Fig. 4. For creation of the probe pulse, a 2-$\mu$J fraction of the Ti:Sapphire fundamental is used to generate white light in a 5 mm thick CaF$_2$ crystal, which is continuously linearly translated back and forth by a piezo stepper. The fundamental light passes through an adjustable aperture and is focused into the crystal with a 100 mm focal length AR-coated lens. The resulting supercontinuum spans the spectral range between 400 nm and 1050 nm. An OD4 700 nm short-pass filter (#84-701, Edmund Optics) suppresses the residual 800-nm light, leaving a probe spectrum covering the range between 400 nm and 700 nm. Alternatively, an OD2 900 nm long-pass filter (DMLP900T, Thorlabs) can be used to transmit the long-wavelength components of the CaF$_2$ white light. The probe is focused into the sample with a 100 mm lens to a spot size of 40 $\mu$m and its intensity can be adjusted by a variable neutral density filter. We note that in principle, the output of the HCF can also serve as both pump and probe in this TA setup. It limits the accessible spectral probing range as compared to CaF$_2$ white light, but provides compressed pulses for both pump and probe [34]. This configuration gives transient absorption data for which post-processing steps such as chirp correction are not necessary.

 

Fig. 4. Two-color pump-probe setup used for the transient absorption measurements of $\beta$-carotene. BS - beam splitter; HCF - hollow core fiber; ASHG - achromatic second harmonic generation unit; TG-FROG - transient grating FROG unit.

Download Full Size | PDF

The time delay between the pulses is controlled with a mechanical translation stage (M-VP-25XA, Newport Corporation). The spectra are collected by a spectrometer consisting of a Kymera 328i polychromator and a Zyla-4.2 camera (Oxford Instruments, Andor), synchronized with the laser. The camera is operated in the shot-to-shot regime at 5kHz, with a region of interest of 8 x 2048 pixels on the 2048 x 2048 CMOS-chip. To suppress scattered light, the spectra were measured in a double-chopping scheme with a 1:2 duty cycle.

The pump-probe spectra were measured at room temperature in the time range between -2 ps and 28 ps in 0.025 ps steps so as to capture the entire range of dynamic processes. The kinetics of the data was analyzed by performing a multiexponential fit (global analysis) with the open-source software Glotaran [35,36]. To avoid a disproportional weighting of the longer lifetimes due to a larger number of points at longer delays, not all the data points at longer delays were included in the global fit. For delays above 300 fs, the data was sampled in logarithmic step sizes.

2.4 Sample preparation and experimental conditions

$\beta$-carotene dissolved in mesitylene was used as a test sample. Apart from absorbing in the spectral range of the ASHG pulse (Fig. 5(a)), this particular carotenoid has the advantage of being well-studied, making it an ideal reference sample to test the functionality of the TA setup. Both $\beta$-carotene and mesitylene were purchased from Sigma-Aldrich and used without further purification. For the TA measurements, a cuvette with 400 $\mu$m path length and 200 $\mu$m wall thickness was used to introduce a minimum amount of additional dispersion. The 200 $\mu$m wall thickness was accounted for during pulse pre-compression by placing a window of the same material and equal thickness in the beam path. The concentration of $\beta$-carotene was set so that the maximal optical density of the sample at 450 nm was below 0.3 OD (see Fig. 5(a)). To avoid photodegradation, the sample was flowed continuously by a microgear pump. The pump pulses were attenuated to 0.006 $\mu$J/pulse. (3x10$^{14}$ photons/cm$^2$/pulse when taking into account the size of the beam spot).

The chirp correction of the TA traces was achieved by assuming that the zero-delay between pump and probe coincides with the maximum of the coherent artefact [37] during pump- and probe pulse overlap. The dependence of the zero-delay value on the detection wavelength was then fitted by a polynomial. The obtained polynomial was used to correct for the chirp-dependent zero-delay for each detection wavelength by means of a two-dimensional interpolation over the data grid.

The instrument response function of the TA experiment is given by a measurement in pure solvent (hexane, data not shown), and shows a near-Gaussian coherence spike with a FWHM of 40 fs.

3. Discussion

The kinetics of $\beta$-carotene have been analyzed in detail by several groups [38–49]. The general energy transfer scheme for carotenoids is shown in Figure 5(b). Upon excitation, population transfer occurs from the $S_{0}$ to the $S_{2}$ state, as the direct one-photon transition to $S_{1}$ is forbidden. The $S_{0}\rightarrow S_{2}$ transition is responsible for the absorption band in the 400–520 nm (25-19.2$\cdot 10^{3}$ cm$^{-1}$) region with its characteristically strong vibronic modulation [1]. From $S_{2}$, optical transitions back to the ground state or to a higher lying state $S_{m}$ may occur. Additionally, a transition to the optically dark (with respect to $S_{0}$) state $S_{1}$ via ultrafast internal conversion on a sub-200 fs timescale takes place. Subsequently, $S_{1}$ deactivates back to the ground state on a few-picosecond timescale. This process can be probed via a strong excited state absorption (ESA) in the visible from $S_{1}$ to $S_{n}$.

 

Fig. 5. (a) Absorption spectrum of $\beta$-carotene in mesitylene (orange) overlaid with the ASHG pump spectrum (blue shaded area); (b) General energy level scheme of $\beta$-carotene; (c) Chirp-corrected TA trace of $\beta$-carotene for probing in the visible spectral range; (d) Chirp-corrected TA trace of $\beta$-carotene with NIR probing.

Download Full Size | PDF

In terms of transient absorption, these dynamics translate into a ground-state bleach (GSB) signal in the S$_{0}$ to the S$_{2}$ absorption region (400-520 nm/25-19.2$\cdot 10^{3}$ cm$^{-1}$), a stimulated emission (SE) signal slightly red-shifted with respect to the GSB and ESA both in the NIR (ca. 850-1200 nm/11.7-8.3$\cdot 10^{3}$ cm$^{-1}$ for $S_{2}$ to $S_{m}$ absorption) and visible spectral range (ca. 520-650 nm/19.2-15.4$\cdot 10^{3}$ cm$^{-1}$ for the transition from $S_{1}$ to $S_{n}$). Time-dependent spectral narrowing and blue-shifting of this feature are associated with vibrational cooling in $S_{1}$ [1,50]. The vast majority of transient absorption experiments on carotenoids are performed by exciting the very red edge of the absorption spectrum due to the lack of suitable pump sources [39,41,51,52]. Such near-resonant excitation generally leads to the presence of a pronounced "coherent artefact" during the temporal overlap of pump and probe-pulses. [37,53] as well as to pronounced excitation of ground-state wavepackets [34,54]. Both phenomena can obscure the excited state dynamics of the system. There are a few reported studies employing 400 nm pump pulses, derived from SHG of the 800 nm fundamental [42,45]. Such excitation conditions are not physiologically relevant and lead to the involvement of additional excited states contributing to the dynamics [45]. Using the ASHG pulse described here as the pump pulse allows for a fully resonant excitation of the lowest vibronic band in the absorption spectrum (cf. Fig. 5(a)), leading to more efficient excitation than for near-resonant pulses and to overall ’cleaner’ dynamics.

Fig. 5(c) shows the chirp-corrected TA trace for $\beta$-carotene in mesitylene taken with the ASHG pump and the CaF$_{2}$ whitelight probe. In the low wavenumber/long-wavelength part of the TA signal, the broad and positive ESA corresponding to the $S_{1}\rightarrow S_{n}$ transition dominates the signal. The observed spectral narrowing in the first hundreds of fs is attributed to vibrational cooling on $S_{1}$. The strong negative contribution at the high energy side of the spectrum in Fig. 5(c) stems from both GSB and SE-contributions. The latter explain the marginal broadening of the negative signals at early times with respect to the absorption spectrum in Fig. 5(a).

The dataset was fitted using a global analysis procedure employing a multiexponential function of the form $F(\lambda,t) = \sum _{i}^{} A_{i}(\lambda )exp(-t/\tau _{i})$ [35,36]. Here, A$_{i}$ is an amplitude factor associated with decay component i having a time component $\tau _{i}$. The sum of the amplitude factors gives the so-called decay-associated spectra (DAS), which provide information regarding the growth and decay of individual components. In a sequential decay model as was assumed for $\beta$-carotene, approximate first guesses concerning the spectral profiles of the transient species can also be made in the form of evolution-associated spectra (EAS) by taking a weighted product of the amplitude factors. Although EAS in complicated systems do not necessarily correspond to the spectra of particular excited states, they provide information about the time evolution of the whole system and are particularly useful in the assignment of time constants to specific species. A global multiexponential fit of the data reveals three lifetime components of $\tau _{1}$ = 0.12 ps, $\tau _{2}$ = 0.3 ps and $\tau _{3}$ = 9.8 ps. The corresponding DAS and EAS are shown in Fig. 6.

 

Fig. 6. (a) Smoothed evolution-associated spectra (EAS) and (b) decay associated spectra (DAS) of $\beta$-carotene in mesitylene solution.

Download Full Size | PDF

The EAS of the 120 fs component (black line in Fig. 6(a)) shows a broad negative peak in the spectral region of the GSB/SE. The corresponding DAS component (black line in Fig. 6(b)) features a broad negative (rise) profile spanning the entire range between 480 and 620 nm (20.8-16.1$\cdot 10^{3}$ cm$^{-1}$). These features are indicative of energy transfer from the initially excited $S_2$ state to the lower-lying $S_1$ state and the lifetime of 120 fs can therefore be assigned to the $S_{2}$ state [40,41,46].

The $S_1$ state is populated in a vibrationally excited state of $v^{'}$ = 4-5, as discussed in literature [43]. As a consequence, vibrational cooling in the $S_1$ state can be observed as a blue-shift and narrowing of the ESA feature corresponding to a lifetime component of 300 fs [41,50,55]. The DAS and EAS for this component are consistent with vibrational relaxation from higher vibronic states in $S_1$ to a vibrationally equilibrated state. The DAS shows a negative (rise) feature corresponding to the main ESA peak and a small, positive (decay) feature at the red edge of the ESA signal. At the same time, the EAS shows a broad, positive peak in the ESA region and another broad, negative feature in the GSB region. It is important to note that, although the assignment of this lifetime component to vibrational cooling can be made confidently, there is a large uncertainty in the determined lifetime of 300 fs. Spectral shifts due to vibrational cooling are inherently difficult to fit in a model with discrete energy states showing well separated spectra and decay times, as assumed by global target analysis [56]. However, the value we obtain agrees well with literature in magnitude, as lifetimes ranging between 250-600 fs have been reported for the vibrational cooling of the $S_1$ state of $\beta$-Carotene [41,51].

The spectra of the long-lived component (light blue lines in Fig. 6(a) and (b)) show a broad, positive-amplitude (decay) feature in the ESA spectral region and a negative (rise) for the GSB. The EAS of this component is qualitatively identical to the DAS and can be interpreted as the spectrum of the $S_1\to S_n$ transition from the vibrationally equilibrated $S_1$ state. Literature values match the measured lifetime of 9.8 ps very well, attributing it to the lifetime of the $S_{1}$ state [39,40,42,46,51,55].

Fig. 5(d) shows a TA trace taken with the long-wavelength portion of the CaF$_{2}$ whitelight as the probe. Only one broad, positive feature can be observed, extending from 885 nm (11.3$\cdot 10^{3}$ cm$^{-1}$) beyond 1050 nm (9.5$\cdot 10^{3}$ cm$^{-1}$). Based on multiexponential fitting at different wavelengths, this component decays with 140 fs (fits not shown). Based on literature [52,57], this feature is assigned to ESA from $S_2$ to a higher $S_m$ state manifold. Its decay time then corresponds to the lifetime of the $S_2$ state and is in good agreement with the 120 fs lifetime extracted from the dataset taken for visible probing discussed above.

4. Conclusions

The experiment presented here provides ultrashort pulses in the wavelength region between 400 and 500 nm. By applying achromatic frequency doubling of an Ar-filled hollow core fiber supercontinuum, we generate 12 fs pulses at 475 nm central wavelength and a pulse energy of 0.2 $\mu$J. We show that the approach is generally applicable to central wavelengths down to 300 nm. As a proof-of-principle experiment, we measured the ultrafast relaxation dynamics of $\beta$-carotene dissolved in mesitylene. The transient absorption spectrometer described here is ideally suited for excitation wavelength dependent studies in molecules and molecular complexes exhibiting absorption below 500 nm and ultrafast relaxation dynamics. Prototypical system with such properties are natural light harvesting complexes and their constituents [43] as well as organic photovoltaic systems [58].