Accurate time zero determination in an ultrafast transmission electron microscope without energy filter

Ultrafast transmission electron microscopy^1,2 has become a powerful technique for directly visualizing nanoscale dynamics with ultrafast time resolution. Its popularity is underlined by a wide variety of phenomena that have been studied, including mechanical oscillations of nanostructures,^3,4 the propagation of acoustic phonons,^5,6 ultrafast phase transitions,^7,8 and chemical transformations.^9–11 It has also given rise to new approaches for coherently manipulating free electron wavepackets.^12–16 In all of these experiments, it is crucial to accurately determine time zero, i.e., the time origin of the experiment at which the laser pump and electron probe pulse overlap in time at the sample. The methods for finding time zero typically employed in laser pump–electron probe experiments broadly fall into two categories. The first monitors a fast sample response at time zero, which ideally should resemble a step function. Suitable processes include the ultrafast lattice dynamics in 2D materials,^17,18 the expansion of a solid under impulsive laser heating,^19,20 the excitation of acoustic waves in an ordered solid,^6,21,22 and the magnetization dynamics observed in Lorentz microscopy.^23,24 The so-called transient electric field effect²⁵ is frequently employed as well, which arises when the sample is ionized at high pump laser intensities. The cloud of emitted electrons deflects or distorts the ultrafast electron pulse, which can be observed either in real space or in diffraction.^26–30 A common drawback of these methods is that the onset of the sample response may be slow or even delayed. For example, the deflection of the electron beam due to the transient electric field effect sets in gradually as the cloud of emitted electrons expands and begins to undergo complex motions.²⁶ Clearly, the sample response is not a step function in this instance, which renders an accurate determination of time zero more difficult.

Such complications are avoided in experiments that instead monitor electron–photon interactions, which are only present when the electron and laser pulse overlap in time. The ponderomotive interaction, which is quadratic in the electric field, accelerates electrons in the direction of lower fields. A tightly focused laser beam will, therefore, burn a hole into the electron cloud at time zero.^31–34 However, since the ponderomotive interaction is weak, this method requires high pulse energies (typically hundreds of microjoules), which may be impractical to use in an ultrafast transmission electron microscope. In contrast, the linear interaction of the electrons with the light field that gives rise to the photon-induced near-field electron microscopy (PINEM) effect can be observed at much lower laser intensities. It causes electrons traveling through the optical near fields of a nanostructure to gain or lose energy in multiples of the photon energy.^12,35,36 In some geometries, the energy spread can even reach up to hundreds of electron volts.^16,36,37 By monitoring the appearance of sidebands in the electron energy loss spectrum (EELS) or the depletion of the zero-loss peak, time zero as well as the instrument response can be accurately determined.^14,38–42 This measurement, however, requires an energy analyzer, a costly piece of equipment that is not available on all instruments.

Here, we demonstrate that time zero can be accurately determined without an energy filter by monitoring the width of the boundary of a nanoparticle, which increases when the electron and laser pulse overlap in time. Our experiments provide evidence that this phenomenon results from the PINEM effect, which causes electrons interacting with the near fields of the nanoparticle to acquire a large energy spread, which couples to the chromatic aberration of the imaging system and, thus, blurs the particle outline. This conclusion is supported by the pronounced angular dependence of the blurring, which closely resembles that of the PINEM effect. Moreover, we demonstrate that the magnitude of the blurring is consistent with the energy spread that the electrons have acquired.

Experiments are performed with a modified JEOL 2200FS transmission electron microscope.⁴³ Femtosecond time-resolved experiments are performed at 100 kHz repetition rate. The sample is illuminated at close to normal incidence with femtosecond laser pulses (515 nm, 200 fs) with the laser beam focused to a spot size of 24 ± 1 μm FWHM in the sample plane, as measured in situ with a knife edge scan. Time-delayed femtosecond electron pulses are generated by illuminating the emitter of the Schottky electron gun (∼850 nm tip diameter, 160 kV accelerating voltage) with UV laser pulses (257 nm). For the experiments in Fig. 3, the energy spread of the electron pulses is adjusted by varying the UV laser fluence. With increasing fluence, a larger number of electrons are extracted, which increase the energy spread due to space charge interactions occurring in the electron pulse.⁴³ Electron energy loss spectra are recorded with an in-column Omega-type energy filter. The width $σ$ of the boundary of a nanoparticle is measured by calculating a 26.5 nm wide intensity profile $Ix$ across the particle boundary and fitting it with the function $Ix=a1+erfx−x0σ2$⁠, where $erf()$ is the error function and $a$⁠, $x0$⁠, and $σ$ are fit parameters.

Figure 1 illustrates that the outline of a nanoparticle blurs when the electron and laser pulse overlap in time, which provides a straightforward approach for determining time zero. Figure 1(a) displays micrographs of a gold nanoparticle (340 nm diameter) on multilayer graphene under illumination with femtosecond laser pulses (22 mJ/cm²). The particle outline appears crisp when the electron pulse precedes the pump laser pulse by more than 1 ps (t < t₀, top panel) but blurs at time zero (t = t₀, bottom panel). Figure 1(b) reveals that this effect is only present when the electron and laser pulse overlap in time. The width of the particle boundary (blue dots) reaches a maximum at time zero, where it almost doubles. The temporal evolution of the width is reasonably well described by a Gaussian fit (red curve, 460 fs FWHM). It is centered at −7 ± 9 fs with respect to the time zero obtained from a measurement using the PINEM effect that is illustrated in Figs. 1(c) and 1(d). Figure 1(c) displays the evolution of the electron energy loss spectrum of the same particle as a function of the pump–probe delay. At the high laser intensity used in our experiment, the energy spread increases from 3.2 eV to over 20 eV when the electron and laser pulse overlap in time, while the intensity of the zero-loss peak drops to less than 40%, as shown in Fig. 1(d). A Gaussian fit is then used to find time zero (red curve, 620 fs FWHM, centered at 0 ± 1 fs). We conclude that the time zero obtained from monitoring the width of the particle outline agrees closely with that from a PINEM experiment.

Figure 2(a) provides a hint as to the underlying mechanism of the blurring, revealing that it is most pronounced in the direction of the laser polarization. Whereas the particle boundary shows a uniform width of about 2 nm when the electron and laser pulse do not overlap in time (t ≠ t₀, blue dots, the solid curve provides a guide to the eye), the blurring at time zero is highly anisotropic (t = t₀, red dots). The width of the particle boundary increases to 6 nm in the horizontal direction, while it barely changes in the vertical. The angular dependence of the blurring closely resembles that of the dipolar near fields of the particle, which are revealed in Fig. 2(b) by an energy gain-filtered micrograph recorded at time zero.^12,16,35,44 We note that when we rotate the laser polarization [double-headed arrow in Fig. 2(b)], the angular dependence of the blurring rotates accordingly. This suggests that the blurring results from electrons that interact with the near fields of the particle. Since such electrons acquire a large energy spread through the PINEM effect, we propose that the blurring is a consequence of the chromatic aberration of the imaging system.

We verify this hypothesis by demonstrating that the amount of image blurring observed is consistent with the energy spread that the electrons acquire through the PINEM interaction. To this end, we measure the width of the outline of a nanoparticle in the absence of laser irradiation as a function of the energy spread. Figures 3(a) and 3(b) show that the chromatic aberration blurs the image of a gold nanoparticle (without laser irradiation) when we increase the energy spread of the electron pulses from 6.6 eV [Fig. 3(a)] to 28.9 eV [Fig. 3(b)]. The corresponding energy loss spectra are shown below the micrographs. Figure 3(c) reveals that the average width of the particle outline (blue dots) increases by 0.88 nm per 10 eV energy spread, as determined from a linear fit (blue line). The data points highlighted with blue and green circles correspond to the micrographs in Figs. 3(a) and 3(b), respectively.

When the energy spread of the electrons is increased through the PINEM interaction, the width of the particle outline increases by a similar amount. The red diamonds in Fig. 3(c) indicate the average boundary width of the particle in the experiment of Fig. 1 as a function of the energy spread that is observed at a given time delay. The data point highlighted with an orange circle corresponds to the micrograph recorded at time zero [Fig. 1(a), t = t₀], with the corresponding energy loss spectrum in Fig. 3(d) exhibiting an energy spread of 20.2 eV FWHM. A linear fit of the data (red line) shows that the boundary width increases by 1.1 nm per 10 eV energy spread, in good agreement with the value obtained above. This supports our conclusion that the blurring of the particle boundary at time zero is caused by the large energy spread that the electrons acquire as they traverse the near fields of the particle.

We note that the two linear fits predict different widths of the particle boundary at zero energy spread, which is likely the case because both experiments were performed with different nanoparticles and under different imaging conditions. The difference in slope between both data sets likely arises from the fact that the energy spread acquired through the PINEM interaction strongly depends on the local field intensity. The reported energy spread represents an average over regions of different field strengths and includes electrons that pass at a large distance from the particle and, thus, do not gain or lose energy. Our measurement, therefore, underestimates the energy spread of the electrons close to the particle boundary, which are responsible for the observed blurring and which experience the highest fields. This causes the boundary width to increase more rapidly as a function of the measured energy spread. Finally, we note that at the high laser intensity in our experiment, the nanoparticles undergo ionization.⁴⁵ It is, however, unlikely that the transient electric fields of the emitted electrons can explain the blurring of the particle outline since these fields are known to persist for tens of picoseconds,^26,30 whereas in our experiment, the blurring can only be observed when the laser and electron pulse overlap in time. Another small contribution to the blurring may arise from the transverse deflection of the electrons in the time-varying near fields,⁴⁶ which couples to the spherical aberration.

We have presented a straightforward approach for determining time zero in an ultrafast transmission electron microscope that does not require an energy filter and can, therefore, find applications in most instruments. It relies on the blurring of the outline of a nanoparticle at time zero, an effect that is only present when the electron and laser pulse overlap in time and, therefore, allows for an accurate measurement. We have shown that this blurring arises from the PINEM effect, which causes electrons that interact with the near fields of a nanoparticle to acquire a large energy spread. Because of the chromatic aberration of the imaging system, these electrons blur the outline of the particle. It should be noted that this effect limits the spatial resolution that can be obtained at time zero, unless, of course, the chromatic aberration can be corrected. A similar issue arises when the interaction with a time-varying field is employed to compress the electron pulses.^47–52 The blurring of the particle outline at time zero can also be used to visualize the near fields on its surface without using an energy filter and can, thus, serve as a poor man's PINEM experiment (Fig. 2). Moreover, from the orientation of the dipolar near fields of a spherical nanoparticle [Fig. 2(a)], the laser polarization can easily be determined in situ. Finally, our method also provides an estimate of the instrument response (Fig. 1), which is otherwise more difficult to obtain without an energy filter. In the limit of small laser pulse energies, this estimate approaches the actual value.

This work was supported by the Swiss National Science Foundation through Grant No. PP00P2_163681.