Comparison between two methods

The experimental setup, consisting of the NOPA, step-scan stage, and fast-scan stage is shown in Fig. 2(a). We performed pump-probe measurements using the step-scan stage and the fast-scan stage.

First, a 20-μm-thick BBO crystal on a fused silica base plate was used to generate an autocorrelation signal, owing to the broadband spectral width of the laser pulse. While scanning the fast-scan stage, the autocorrelation signal was monitored by a photomultiplier tube (H9656-04 from Hamamatsu) connected to the last channel of the MLA system. When the step-scan stage moves, the autocorrelation signal on the fast-scan trace shifts with the same amount of delay as that of the step-scan stage. From the shift, the delay position of the fast-scan trace can be calibrated. The autocorrelation traces were obtained by changing the delay position of the step-scan stage in 50-fs steps. The peak positions of the autocorrelation signal are plotted in Fig. 2(b), giving a calibration curve showing a linear relationship

between the delay time and data points with 3.58  0.01 fs/point.

Pump-probe measurements of the MEH-PPV film were first performed by the step-scan method. The absorbance change ΔA in the wavelength region from 515 to 753 nm in 2.5-nm steps was obtained by scanning the delay time from -318 to 1482 fs in 0.6-fs steps (forward scan). For each delay point, ΔA was obtained by accumulating for 0.6 s. The mean ΔA spectra in 3-fs steps are shown in Fig. 3(a) in two-dimensional (2D) form. The data in Fig. 3(b) are for the corresponding backward scan. The poor reproducibility between Figs. 3(a) and (b) is indicative of sample damage and laser intensity instabilities.

The 2D traces give time-resolved ΔA spectra along their cross sections. The ΔA spectra at three delays (147, 326, and 1221 fs) are plotted in Figs. 3(c) and (d) for forward and backward scans, respectively. Damage accumulated in the sample causes scattering, which decreases ΔA in Fig. 3(d) compared to in Fig. 3(c).

Time-resolved ΔA traces at three probe wavelengths (568, 575, and 616 nm) were also extracted from the 2D traces in Figs. 3(a) and (b) and plotted in Figs. 3(e) and (f), respectively.

In the forward scans, the time-resolved traces start around ΔA=0 for negative delays. In contrast, there exists a large offset at negative delays for the backward scans. The offset is due to damage accumulated on the sample during the step-scan measurements.

Fourier power spectra of the ΔA traces reveal molecular vibrational modes of the sample.

Figures 4(a) and (b) are the 2D results for forward and backward scans, respectively. They show strong peaks around 1588 cm^-1, corresponding to the most intense Raman mode (see Fig.

2 of Ref. 39). However, they have low reproducibility between the forward and backward scans and show considerable difference from a Raman spectrum, because of the instability of the laser source.

Fourier power spectra probed at 616 nm were extracted from Figs. 4(a) and (b) and plotted in Figs. 4(c) and (d). Differences compared to a Raman spectrum [39] and the

irreproducibility of the step-scan data can be seen. Figs. 4(e) and (f) are Fourier power spectra of the most prominent mode at 1588 cm^-1, corresponding to the C=C stretching mode. They lack reproducibility at wavelengths below 590 nm.

To study the dynamics of the molecular vibrational modes, a spectrographic analysis was performed. Spectrograms were obtained by a sliding-window Fourier transform using the Blackman window function,

, ^∞ exp ,

0.42 0.5 cos 0.08 cos ,

where T is the gate width which we set to 600 fs, corresponding to a FWHM of 240 fs.

The calculation of the spectrogram was performed after applying a high-pass filter to the signal, because slow modulations of the signal (slower than tens of femtoseconds) are mainly caused by fluctuations in the laser power. For forward and backward scans, the results are shown in Figs. 5(a) and (b), respectively, using a time-resolved ΔA trace probed at 616 nm. It again fails to show good reproducibility.

After the preceding step-scan measurements, we moved the sample position to avoid the damaged area and began fast-scan measurements so that the experimental conditions remained identical. Twenty-four forward scans were obtained from -390 to 1396 fs. The mean of the 24 traces is plotted in Fig. 6(a), and backward scanned data from 1396 to -390 fs are shown in Fig. 6(b). In contrast to the step-scan results, these data show high reproducibility.

Figures 6(c) and (d) are time-resolved ΔA spectra for the forward and backward scans, respectively, at delays of 147, 326, and 1221 fs. Time-resolved ΔA traces probed at three wavelengths (568, 575, and 616 nm) are plotted for the forward and backward scans in Figs.

6(e) and (f), respectively. The improvement in the signal reproducibility was estimated by calculating a correlation coefficient between the time traces for the forward and backward (1) (2)

step-scan and fast-scan measurements. One can see that the reproducibility is much higher in the fast-scan measurements than in the step-scan measurements.

Fourier power spectra of the ΔA traces for the forward and backward scans were calculated and are plotted in Figs. 7(a) and (b), respectively, and show high reproducibility also. Similar to Figs. 4(c) and (d), ΔA traces probed at 616 nm are plotted in Figs. 7(c) and (d) for forward and backward scans, respectively. They confirm the reproducibility of the Fourier power spectra and show good agreement with Raman spectra from previous work [39].

Figures 7(e) and (f) are Fourier power spectra of the most prominent vibrational mode at 1588 cm^-1. Some small differences are evident at wavelengths longer than 640 nm. The small amplitude of ΔA in that spectral region [see Figs. 6(a) and (b)] is probably the origin of the error.

Between 600 and 640 nm, where the Fourier power is high, the signal reproducibility was quantified by calculating the error in the determination of the power for molecular vibrational modes of 963, 1112, 1261, 1316, and 1587 cm^-1. The relative difference in the Fourier power of a forward scan (If) and a backward scan (Ib) is |(If-Ib)/(If+Ib)|. Figs. 7(g) and (h) shows the errors calculated for step-scan and fast-scan measurements, respectively.

Evidently the Fourier power can be determined more precisely for the latter.

Using a Blackman gate with a FWHM of 240 fs, spectrograms were calculated for the time-resolved traces obtained in the fast-scan measurement. The results are plotted in Figs.

8(a) and (b) for the forward and backward scan, respectively, and show good reproducibility.

Therefore, the fast-scan method is preferable not only for the study of electronic dynamics but also for the study of vibrational dynamics.