Transient absorption (TA) spectroscopy has been used to study short-lived transient species via their dynamic absorption spectra and absorbance temporal changes, and has been an indispensable experimental method in fields such as photochemistry, photophysics, photobiology, and others since the pioneering work of flash photolysis by Porter and coworkers in 1949 [1]. Over the years, many techniques have been proposed to obtain TA signals under various experimental conditions. These techniques are classified into two main categories based on the form of the probe light: the pulse-probe method, which is often called the “pump-probe” (PP) method, and the continuous-wave (CW) probe method. In the PP method, a short probe pulse probes the instantaneous absorption of a sample at a given time delay after the sample is excited by a pump pulse. Since time resolution is determined by the width of those pulses, ultrafast spectroscopy employing PP techniques has advanced [2]. However, the measurable time range is restricted to several nanoseconds, because the time delay is generated by varying the optical path length between pump and probe pulses. In the CW method, the temporal profile imprinted on a CW probe beam by the TA of the sample is recorded in real time. Because the time resolution is determined by the response time of the detectors, the CW method has mainly been used to study relatively slow kinetics, typically longer than 10 ns. In addition, when using the CW method, nanosecond TA signals are often buried under strong fluorescence signals. Therefore, TA measurements in this time range require effective methods to correct for fluorescence.

Because of these difficulties, the important phenomena that traverse the nanosecond range, such as photoconversion in solar cells [3] or artificial photosynthesis [4] often lack experimental evidence of those dynamics around the nanosecond range. Several approaches have been proposed to seamlessly acquire TA spectra around the nanosecond range; some are based on the CW method [5,6], although most are extensions of the PP method [7–13]. Each proposal fulfills the goal to some extent, while each has weak points such as narrow time range [5–7], bulkiness [7,10,12], expensiveness [5,8–12], requirement of researchers’ craftsmanship [7–9,12], or limited probe wavelengths [5,11–13] for wider applications. Therefore, a need exists for a new technique of TA spectroscopy that seamlessly covers the wide time domain bridging the nanosecond range and the broad wavelength domain.

To address this need, we propose a TA spectrophotometer in which pulse train is used for probing (hereinafter, PT method). This approach has characteristics of both the PP and the CW methods and seamlessly covers a wide time domain. In the PT method, consecutive probe pulses are impinged upon the sample with a single pump pulse (i.e., the repetition frequency of the probe beam is much greater than that of the pump beam).

Figure 1 shows how a temporal TA profile is reconstructed in the PT method. By repeatedly pumping and probing with a variable time delay between the pump pulse and the probe pulse train, the requisite sampling density and wide dynamic range are obtained simultaneously. Note that the TA signal is derived by plotting the transmitted amplitude of each probe pulse, which is similar to the PP method. Accordingly, the time resolution is determined not by the speed of the detectors and/or the electronics (provided they are fast enough to discriminate between adjacent probe pulses) but by the largest of the pump-pulse width, the probe-pulse width, or the error in the delay time. An extremely broad temporal TA profile can thus be obtained by plotting of the peak intensities versus delay time. The PT method thus combines the advantages of the PP method (i.e., high time resolution) and the CW method (i.e., broad temporal range).

 

Fig. 1. Principle of PT and RIPT method. The TA temporal profile is reconstructed by plotting the probe amplitude from multiple probe-pulse trains versus corresponding PP delay time, $t_d$. When the first pump pulse excites the sample, the output signal of a probe-pulse train transmitted through the sample with a delay of $\mathrm{\Delta }t$[1] is recorded and the amplitude of each pulse is detected. This cycle is repeated many times while varying $\mathrm{\Delta }t[k]$ until the required time-delay density is obtained. If the pump pulse and the probe-pulse train are asynchronous, the method becomes the RIPT method.

Download Full Size | PDF

To acquire TA data with the PT method, the delay time between the pump pulse and the probe pulse must be determined precisely. If the two beams are electrically synchronized, electronic jitter determines the time resolution. To improve on this approach, we expand the idea of the PT method into the randomly-interleaved-pulse-train (RIPT) method, which does not require synchronization. The large difference in repetition frequency between the two independent light sources naturally results in an essentially random timing between the pump pulse and probe-pulse trains; therefore, the required delay times can be obtained simply by repeated pumping cycles. In this method, the delay times are passively evaluated shot by shot; therefore, the time resolution is determined by the accuracy of the delay-time detection and/or the intrinsic pulse widths. Delay-time measurement, which is also termed delay-interval measurement (DIM), has been well established and the accuracy reaches subnanosecond by several methods [14] aimed for various applications such as time-correlated-single-photon counting (TCSPC) in fluorescence lifetime measurements.

Figure 2 shows the basic scheme of the RIPT apparatus. The configuration is rather similar to that of a monochromatic CW system as opposed to a PP system; it requires no optical delay line or synchronization unit between the pump source and the probe source. The pump-pulse source is a picosecond laser, PL2210A (EKSPLA, 1 kHz, 25 ps, 355 nm, 0.3 mJ), and is split into two beams: One beam is detected with a Si PIN photodiode (PD) (S5972, Hamamatsu, PD1) to measure the time delay, and the other beam irradiates the sample to induce photoreactions. The probe source is a supercontinuum (SC) radiation source (SC-450, Fianium, 20 MHz, 50–100 ps pulse width depending on the wavelength, 450–2000 nm) and is also split into two beams: one is detected with another Si PIN PD (PD2) to measure the time delay and the other is monochromatized with a predispersive monochromator (MD200, Unisoku). Note that the pump and probe repetition frequencies are 1 kHz and 20 MHz, respectively, which represents a ratio of 1:20,000. After the monochromator, the probe beam is further split into two beams: one beam irradiates the sample at the same position as the pump beam. The probe energy just in front of the sample was less than 50 pJ/pulse. The beam transmitted through the sample is detected by an InGaAs PD (G10899-01K, 400–1600 nm, Hamamatsu) with a 26 MHz preamplifier (PD3), which allows a more precise readout of the amplitude of each peak. An aperture and optical filters are positioned before the detector to minimize fluorescence signals from the sample. To correct shot-to-shot intensity fluctuations of the SC radiation source, the amplitude of the split light after the monochromator is also measured by another amplified InGaAs PD (PD4). All PD outputs are recorded with an oscilloscope (HDO8038, Teledyne Lecroy) with 2.5 GSamples/s triggered by the pump pulse, and then transferred to a PC to construct the TA temporal profile.

 

Fig. 2. Prototype of TA spectrometer for RIPT method. BS1 and BS2 are beam splitters; PD1 and PD2 are fast photodiodes; PD3 and PD4 are detectors with preamplifiers; VND is a continuously variable reflective neutral-density filter. No timing controller is needed to synchronize the probe light source with the pump laser.

Download Full Size | PDF

Note that $\mathrm{\Delta }t[k]$ in Fig. 1 is the delay time between the pump pulse of the given number and the probe pulse immediately following the pump pulse. We calculated the delay time by measuring the difference between the intersections of the rising edge of two fast-detectors’ output (PD1 and PD2 in Fig. 2) with an arbitrary threshold voltage. We evaluated the accuracy of the delay times by measuring identical light pulses with the two detectors. The FWHM of the Gaussian distribution of the calculated time difference between two detectors was found to reach 8.2 ps using an oscilloscope with a sampling interval of 400 ps. Thus, all required PP delay times in a pulse train are calculated with better than 10 ps accuracy. Denoting this accuracy as $T_{\sigma }$, the time resolution $T_{\text{reso}}$ of the RIPT method is determined by the following error-propagation equation:

The delay times also give the precise positions of the peak of the output signals from the amplified photodiodes (PD3 and PD4 in Fig. 2). Thus, the peak intensities of multiple probe pulses in a pulse train are acquired. The results from PD3, $I_{\mathrm{PD}3}(t_d)$, are divided by the corresponding results from PD4, $I_{\mathrm{PD}4}(t_d)$, to correct for fluctuations in pulse intensity as

In practice, each delay time is rounded off to an arbitrary precision (10 ps to 50 ns, termed bin width), and the results for $\mathrm{\Delta }\mathrm{OD}$ are averaged for the same rounded delay times. The bin width of 10 ps needs at least several times of $5000(=50\mathrm{ns}/10\mathrm{ps})$ pumping to drop at all the bins. Average numbers are not equal in all bins and missing bins are not acceptable. Time windows can be set from 100 ns to 500 μs. One measurement cycle with multiple pumping is executed in sequence mode of the oscilloscope up to maximum memory size (50 MByte/ch). For example, when the time window is 200 ns, 50 MByte/(200 ns*2.5 GSamples/s*2 bytes/Sample) = 50,000 pumpings can be stored maximally, though only 100 pumpings can be with the time window of 100 μs in one cycle. Recorded data in all channels are transferred to the PC with the transfer rate of $>100\mathrm{Mbytes}$. Then delay-time detection and $\mathrm{\Delta }\mathrm{OD}$ calculation, which are the most time consuming and cause extra pumping of samples during processing at present, are performed by home-built software. This cycle is repeated until acceptable signal-to-noise ratio is obtained. Then, by scanning the probe wavelength, we can obtain the TA spectra. Actual measurement times required in the current stage are specified in the caption of Fig. 3, with the measurement conditions, though they could be improved by various technical approaches.

 

Fig. 3. TA results obtained by RIPT method from $C_{60}$ solution. (a) Overall TA decay curves at 755 nm (red; $T_1$ absorption peak), 805 nm (green; isosbestic point), and 985 nm (blue; $S_1$ absorption peak); (b) subnanosecond results with bin-width of 10 ps ($\sim 200$ pumping/bin); (c) time delay from $-1$ to 5 ns. with bin width of 50 ps ($\sim 100$ pumping/bin); (d) time delay up to 50,000 ns; (e) 2D false-color contour plot of TA; (f) TA spectra in the sub-5-ns domain reconstructed from (e). The temporal gap not covered by the conventional TA methods is highlighted in pink in (a) and (c). Bin width was lengthened gradually along the time axis in (a), (d), and (e). For each figure [collected time window, collection time per one decay] were [100 μs, 6 min] for (a) and (d), [200 ns, 37 min] for (b), [200 ns, 4 min] for (c) and [100 μs, 2 min] for (e), respectively. To obtain the whole feature of (e), 177 min were needed.

Download Full Size | PDF

Figure 3 shows the TA of $C_{60}$ in toluene. The TA of $C_{60}$ and its derivatives have been performed [15] because they are promising materials for solar energy conversion devices. To figure out all photoinduced processes of $C_{60}$, observing singlet excited state ($S_1$) and triplet excited states ($T_1$) is essential. However, the intersystem crossing time in $C_{60}$ ($\sim 1\mathrm{ns}$) is an “irritating” value for the PP and CW methods. Consequently, fluorescence-lifetime measurements have often been used as an alternative approach to study $S_1$ of $C_{60}$ [16], though no evidence of intersystem crossing to $T_1$ can be obtained. Figure 3(a) shows the TA results obtained by the RIPT method. At the isosbestic point [Fig. 3(f)], the time resolution of the system is evaluated to be 80 ps [Fig. 3(b)], which is 2 orders of magnitude higher than typical nanosecond CW systems. Therefore, the decay of $S_1$ and the rise of $T_1$ with a time-constant of $\sim 1\mathrm{ns}$ are clearly detected without deconvolution [Fig. 3(c)]. Because the TA at longer time delays can be acquired simultaneously, the lifetime of $T_1$, $\sim 5000\mathrm{ns}$ is also obtained [Figs. 3(a) and 3(d)]. In addition, broadband TA spectra spanning from 500 to 1250 nm are acquired by scanning the probe wavelength [Figs. 3(e) and 3(f)]. Thus, the RIPT method seamlessly covers a dynamic range spanning more than 6 orders of magnitude (0.1–100 000 ns) and a spectral range going from the visible to the near-infrared.

Furthermore, the RIPT method makes it easy to correct for fluorescence that pollutes the TA spectrum in the form of “negative” absorption. As illustrated in Figs. 4(a)–4(h), the baseline, which is distorted by fluorescence, can be reconstructed in the “randomly-interleaved” manner. We used the highly fluorescent molecule tetraphenylporphyrin (TPP) as a model compound to verify the proper functioning of this procedure. Figure 5(a) shows the TA signal obtained by the CW method and with no correction for fluorescence. Sharp negative signals [highlighted in pink in Fig. 5(a)] are attributed to the strong fluorescence of TPP, so the true TA dynamics are severely masked. However, when fluorescence correction is applied (i.e., baseline subtraction) in the RIPT method, fluorescence is no longer observed in this same time range as in Fig. 5(b). Accordingly, the fast-decay component due to $S_1$ can be clearly observed in contrast to the CW method. Thus, the RIPT technique can be applied to a wide variety of samples with strong fluorescence.

 

Fig. 4. Baseline subtraction procedure. Panels (a)–(c) show the simulated output of the detector for the probe-pulse train transmitted through the sample (PD3 in Fig. 2). The bump around time zero seen in each figure is due to the fluorescence of the sample induced by the pump light. Picking up the value just before the rise of each pulse [solid green circles in panels (a)–(c)] which should be zero if no fluorescence enters the detector [open circles in panels (a)–(c)] results in reconstructing a nonzero baseline curve in a “randomly-interleaved” manner when the procedure is repeated many times. If the fluorescence signal is superposed over the probe-pulse trains, then the reconstructed baseline curve reproduces the temporal profile of the fluorescence. By subtracting this profile from each data set from the probe-pulse train [panel (e), which shows an enlarged view of the region highlighted in pink in panels (a) and (f)], the signal due to the fluorescence completely disappears, as shown in panels (f)–(h). After subtracting fluorescence, the peak intensity of each pulse [solid blue circles in panels (f)–(h)] allows the reconstruction of the TA curve, as illustrated in Fig. 1.

Download Full Size | PDF

 

Fig. 5. Typical results of fluorescence pollution by the CW method and correction by the RIPT method with highly fluorescent sample, TPP; (a) TA temporal profile at several wavelengths from 460 to 580 nm by the CW method with no fluorescence correction; (b) corresponding TA temporal profile by the RIPT method with fluorescence correction. The temporal gap not covered by the conventional TA methods is highlighted in pink.

Download Full Size | PDF

In summary, we introduced ideas of exploiting the asynchronicity between two pulsed light sources with large different repetition frequency and passively detecting the PP delay times, which leads us to name this method RIPT. This approach not only bridges the temporal gap between the PP and CW methods, but also, with a single apparatus, covers a wide time dynamic range from subnanosecond to millisecond range. As for the concept of asynchronicity between a pump and (a) probe(s), asynchronous optical sampling (ASOPS) technique is known and some advanced forms of ASOPS provide time delays from picoseconds to milliseconds [11,12]. The critical difference is that two femtosecond oscillators are required in ASOPS, while availability of various type of light sources such as a SC light source in the RIPT method results in wide spectral range measurements. Also, shot-by-shot passive detection of delay times allows us to use pulsed light sources with large jitter such as passive Q-switched microchip lasers. Furthermore, the TA profiles obtained can easily be corrected for fluorescence contamination. The time domain covered by the RIPT method is of great interest because many photoinduced reactions have not been well studied because of a lack of appropriate instrumentation. Therefore, the RIPT method can contribute significantly to advancing research related to these processes. We believe that the RIPT methods will become a mainstream TA measurement technique and expect that the concept of PT and RIPT will be applied to various other techniques in a number of fields.