MÓDULO FORMATIVO 4

The first step of the time-resolved experiment is the Pd 3d core level acquisition during the P.U.N.D. train of pulse. The palladium pad connected to the top electrode is probed. The aim is to verify if the response of the electrode follows the dynamic of the circuit used to apply the bias. Since the palladium is the topmost layer of the studied heterostructure, the Pd 3d core level spectra give very good signal/noise ratio. This allowed us to develop an automatic peak fitting procedure to access the displacement of the Pd 3d core level as a function of time. This procedure is also used on the less intense Ba core levels, where effects of the polarization switching are expected.

Figure 4.8 a) shows a Pd 3d photoemission core level reference spectrum taken with both electrodes grounded before the application of the P.U.N.D. train. The energy window of the acquisition was chosen so that the Pd 3d5/2 peak was in the center of the window. The total width of the peak was about 4 eV (with a KE between about 760 eV and 764 eV), and we expected an energy shift of ± 4 eV since the applied bias is ± 4 V. The width of the snapshot window is 6% of the Pass Energy. Therefore, we used a pass energy of 200 eV giving a window of 12.54 eV. Figure 4.8 b) is an image of the Pd 3d core level spectral intensity as a function of time; the color scale represents the intensity of the photoemission signal. One vertical line

of this image corresponds to a Pd 3d spectrum, similar to the one presented figure 4.8 a). The time abscissa between 0 and 9 µs corresponds to the duration of one P.U.N.D. train. As a single P.U.N.D. train pulse gives rise to spectra with low signal to noise ratio, the P.U.N.D. was repeated 7000 times. Therefore, each vertical line of the image is the average of seven thousands spectra. The P.U.N.D. signal can easily be identified on this image, as the Pd 3d shifts along with the applied bias. In the upper part of the figure 4.8 b), we schematically represented the P.U.N.D. train.

Figure 4.8 –a) Pd 3d core level photoemission spectrum b) Image of the Pd 3d core level photoemission spectra

as the function of time, following a PUND train. The color scale represents the intensity of the photoelectrons.

We checked that the general shape for every recorded spectrum did not change with time. An automated fitting procedure was then developed to subtract the secondary electron background and to fit each spectrum of the image with a Gaussian function. A correct peak shape would consist in using the asymmetric Doniach-Sunjic function [190], best represetative of a metal. However, since in this study we are only interested in the position of each peak, and not to its exact shape, Gaussian functions were used to simplify and speed up the automatic fitting procedure. The width of each peak is kept constant after a preliminary fit procedure using one of the snapshot spectra. A typical 7000 shot averaged snapshot spectrum along with the corresponding Gaussian fit are presented in Figure 4.9. The FWHM of the Gaussian peak was 0.80 eV and was fixed at this value for the rest of the procedure (i.e. for the fit of all 450 Pd 3d core level snapshot of the P.U.N.D train).

Figure 4.9 –Snapshot of the Pd 3d core level acquired during the time-resolved experiment (red crosses) and

Gaussian fit (black line).

from the automatic peak fitting procedure (replacing the raw data in image Figure 4.8 a)). The Gaussian peaks are displayed in binding energy as a function of time. The energy shift measured with respect to the position at zero bias for each Gaussian peak is presented as a function of time in figure 4.10 b). This allows following the shift of core level due to the electrical excitation of the capacitor while it is going through the P.U.N.D. train.

Figure 4.10 –a) Map of the Pd 3d5/2and 3d3/2peaks, resulting by a systematic peak fitting for each spectrum

of image Figure 4.8 b) with a Gaussian function. b) Energy shift of the Gaussian peaks from their energy at zero bias as a function of time.c) Zoom of the P pulse showing the energy shift spectra (red) and the applied bias (blue).

When observing the energy shift of the Pd 3d5/2 core level as a function of time, we can directly notice that the shift does not exactly follow the shape of the P.U.N.D. train. Indeed,

when looking closely at each one of the pulses, as presented on the zoom of the P pulse in figure 4.10 b), it gives a non-instantaneous response to the applied bias. This means that there is a time constant in the system slowing down the core level response to a bias pulse. We fitted the spectra 4.10 b) to access the time constants of the charges for the four pulses P, U, N and D. The resulting spectra along with these fits, and the values of the time constants are compiled in figure 4.11. In the case of the metallic palladium, the time constants for each pulses are expected to be identical, as they represent the electrical circuit. However, as we can see, they are different for the P and N pulses versus the U and D pulses. Furthermore, for both bias directions, the switching time constant is significantly different from the non-switching one. The circuit response is therefore quite complex. This behavior, different to the one expected for the palladium core level, will be discussed in section 4.2.5.

The ripples observed on the image obtained from the time resolved measurements are pro- bably due to a finite impedance in the circuit. Indeed, the samples are connected via gold micro wiring to a copper plate, but silver paste was also used to connect this plate to the sample holder, which might provide imperfect connection with a residual impedance. To ensure that those ripples are indeed a characteristic of the electrical circuit and not the polarization swit- ching, we checked that they disappear after subtracting the displacement measured during a non-switching pulse to the one measured during a switching pulse.

Figure 4.11 – Energy shift of each Gaussian peak from its energy at zero bias as a function of time (red

crosses), and fits of the charges and discharges of the P.U.N.D. circuit (black lines). The values of the charac- teristic times for each pulse, τP, τU, τN and τD, obtained from the fits are indicated on the figure.