Sub-10 fs Time-Resolved Vibronic Optical Microscopy.

We introduce femtosecond wide-field transient absorption microscopy combining sub-10 fs pump and probe pulses covering the complete visible (500-650 nm) and near-infrared (650-950 nm) spectrum with diffraction-limited optical resolution. We demonstrate the capabilities of our system by reporting the spatially- and spectrally-resolved transient electronic response of MAPbI3-xClx perovskite films and reveal significant quenching of the transient bleach signal at grain boundaries. The unprecedented temporal resolution enables us to directly observe the formation of band-gap renormalization, completed in 25 fs after photoexcitation. In addition, we acquire hyperspectral Raman maps of TIPS pentacene films with sub-400 nm spatial and sub-15 cm-1 spectral resolution covering the 100-2000 cm-1 window. Our approach opens up the possibility of studying ultrafast dynamics on nanometer length and femtosecond time scales in a variety of two-dimensional and nanoscopic systems.

Nanostructured electronic materials such as organic semiconductors[1−3] and organic–inorganic metal halide perovskites[4−6] have attracted considerable attention both from the point of view of fundamental electronic properties and for applications in next-generation photovoltaics (PVs), light-emitting diodes (LEDs), and other optoelectronic devices. In contrast to conventional semiconductors such as silicon, one of the defining features of these materials is the spatial inhomogeneity of thin films on μm and sub-μm length scales. This has been widely studied by electron, X-ray and photoluminescence (PL) microscopy,[7−14] yet an understanding of how these structural and chemical properties control and influence exciton and charge dynamics remains unknown. This lack of understanding can be largely traced back to the fact that transient electronic and vibrational spectroscopies are traditionally and predominantly performed at the ensemble level due to the necessity of combining the response from a significant amount of material to generate an observable signal in the presence of measurement noise. Critically, no currently available technique is able to correlate femtosecond time scale transient microscopic data with local molecular structure and composition, thereby considerably limiting our ability to rationalize the origins of spatial variations in ultrafast electronic dynamics. Given the growing importance of these nanostructured materials and numerous indications of the existence and potential importance of spatial inhomogeneity,[7,15−18] there is a clear need for moving ultrafast electronic and vibrational spectroscopy beyond the ensemble average and toward high spatial resolution.

As a result, considerable efforts have been made at combining femtosecond temporal with sub-μm spatial resolution. An ideal tool to address these aspects is (far-field) transient absorption microscopy (TAM), where a sample is photoexcited by a pump pulse with a duration typically of a few hundred femtoseconds followed, at a variable time delay, by a probe pulse that records the pump-induced signal response. Using a microscope objective-based imaging setup directly after the sample, this response can be directly mapped onto a camera, revealing light-induced dynamics with sub-μm spatial resolution.[17−19] Several experimental studies have demonstrated the enormous potential of TAM for a wide range of systems encompassing among others, solar cell materials,[15,16,20−23] graphene sheets,[24−27] and carbon nanotubes[28−30] as well as nanoplasmonic structures[31−34] and the monitoring of drug delivery.[35] Apart from a few exceptions,[36] all experiments to date employ a traditional transmission microscope setup, in which both pulses pass through an illumination objective prior to interaction with the sample, limiting the achievable pulse durations to ∼100 fs due to temporal dispersions introduced by the objective.[17,18] Application of sophisticated pulse-shaping techniques have recently shown that it is possible to combine ultrashort near-IR pulses (∼800 nm, sub-10 fs), which have become ubiquitous in ultrafast spectroscopy, with high spatial resolution provided by high numerical aperture (NA) objectives.[37] The extension toward the visible spectral region, where many of the most studied materials are strongly active, however, has been lacking.

To alleviate these shortcomings, we build on our recent work[38,39] and present a wide-field femtosecond transient absorption microscope (fs-TAM) capable of delivering sub-10 fs pump and probe pulses to the sample while maintaining diffraction-limited spatial resolution (Figure a). The corresponding experimental setup is highly reminiscent of a standard transient absorption spectrometer where pump and probe pulses are focused onto a sample by a dispersion-free concave mirror (Figure b).[40,41] Rather than collecting the probe light using point- or spectrally-dispersed detection, we now collect the illumination beams with a high-NA microscope objective and subsequently image them onto a digital camera. Alternatively, inserting a slit in an image plane combined with dispersion by a prism for maximum detection efficiency enables simultaneous 1D spatial and spectral imaging.[38] Placement of the imaging optics after the pulses have interacted with the sample ensures that the strong dispersion induced by the objective does not affect the pulse duration at the sample.[42] Consequently, the microscope simply images the transmission of the sample with a spatial resolution defined by λ/2NA, where λ is the carrier wavelength of the probe pulse and NA the numerical aperture of the objective. Modulation of the pump pulse by a chopper synchronized with the camera exposure thus returns differential transmission images of the sample. Delaying the probe relative to the pump pulse with a motorized translation stage retrieves transient kinetics completely analogous to that of standard pump–probe spectroscopy.[40,41]

Figure 1

Concept and schematic experimental setup of wide-field fs-TAM. (a) A short focal length concave mirror focuses sub-10 fs pump and probe pulses onto a sample deposited on a cover glass slide. A high-NA objective collects both pulses to form a differential absorption image. The transient response is recorded with high spatial resolution to reveal the transient morphology of the sample, as illustrated here for a triple grain boundary in a MAPbI3–Cl perovskite film. (b) Schematic of the setup based on white light (WL)-derived pump (500–640 nm) and probe (650–950 nm) pulses. TS, translation stage; C, chopper; FM, flip mirror; P, prism; LP, 650 nm long-pass filter.

To highlight the capabilities of fs-TAM, we begin by discussing the dispersed detection modality, applied to investigate the transient response of the mixed iodide–chloride methylammonium lead perovskite, MAPbI3–Cl.[5,43] The normalized transmission image of a grain boundary region (Figure a) recorded using a probe pulse covering 650–950 nm exhibits almost no absorption at the boundary but a strong reduction in the signal directly adjacent to it, most likely arising from strong scattering (Figure S1). Our perovskite grains appear to be regularly structured with brighter and darker absorption regions distributed throughout the grain (Figure a). We emphasize that this structure may originate from much smaller subdiffraction inhomogeneity that is not resolved by our optical microscope.

Figure 2

fs-TAM on MAPbI3–Clx perovskite films. (a) Normalized transmission image showing a horizontally aligned grain boundary. (b) Dispersed fs-TAM image at 1 ps pump–probe delay and (c) steady-state PL image of a diffraction-limited slice selected by the slit, as indicated in (a). (d–f) Transient absorption maps retrieved from the lower grain, the grain boundary, and the upper grain, as indicated in (a). (g) Transient absorption map averaged over a 20 × 20 μm2 region equivalent to an ensemble measurement. The scale bar in (a–c) is 2 μm. The original data were binned to 146.5 nm/px in space and 3.8 nm/px in wavelength. All experiments were carried out with the full probe bandwidth (650–950 nm) and fluences of 16 and 120 μJ/cm2 for pump and probe pulses, respectively.

For dispersed detection, we close the slit positioned in the intermediate image plane to select a diffraction-limited slice of the image and direct the probe onto a prism for spectral dispersion along an axis perpendicular to the slit (Figure b). The corresponding transient absorption spectrum of our perovskite sample after photoexcitation with a 10 fs pump pulse at 560 nm at a time delay of 1 ps exhibits a prominent bleach signal at 760 nm, in agreement with previous reports (Figure b).[44,45] Interestingly, the TA spectra exhibit large spectral variations as a function of space, indicating a significant underlying structural inhomogeneity even within the grain. To deduce whether this behavior is intrinsic to our sample, we blocked the probe pulse and measured the steady-state PL intensity generated by the pump pulse (Figure c). While some spectral changes remain evident, the spectrally resolved TAM image is significantly more inhomogeneous, suggesting a highly nonlocal environment of charge carriers directly after photoexcitation compared to the relaxed excited state measured with PL.

We now shift our focus toward the grain boundary, which shows almost no intensity in the fs-TAM or steady-state PL map (Figure b,c), in agreement with previous findings.[46] To understand the transient kinetics, we compare spectrally-resolved TAM kinetics derived from the upper grain, the lower grain, and the grain boundary (Figure d–f). All maps exhibit an early dispersive line shape with a photoinduced absorption at 780 nm and a bleach at 740 nm, which evolves into a single bleach signal at 760 nm on the picosecond time scale. This behavior was previously assigned to band-gap renormalization occurring after above-band-gap excitation.[44] Interestingly, the transient absorption response of the grain boundary (Figure f) shows a significantly reduced bleach contribution albeit with a similar intensity in the photoinduced absorption feature compared to the grain (Figure d,e).

In addition, we can directly compare our fs-TAM results to ensemble TA spectroscopy experiments by averaging the transient absorption response over the full beam diameter (Figure g). The resulting TA map matches previous studies but lacks the characteristic signature of the grain boundary, emphasizing the importance of an imaging approach for samples with microscopic or nanoscopic inhomogeneity. Our results furthermore illustrate that the photoinduced absorption feature, corresponding to band-gap renormalization, grows in within 25 fs, irrespective of the spatial position within the sample, refining previous reports of a sub-100 fs time constant.[44] These initial results illustrate that the high spatiotemporal resolution promises to be an important asset in determining the optical responses of perovskite films to help uncover the reasons for their exceptional optoelectronic behavior. We remark that the unusual combination of a higher probe compared to pump power is a consequence of the interplay between maintaining a sufficiently low differential absorbance caused by the pump while maintaining sufficient photon counts in the probe to reveal the resulting signatures limited by shot noise. Although this is unconventional, we emphasize that the resulting transient absorbance spectra and kinetics are indistinguishable from those obtained at the ensemble level with a weaker probe compared to the pump, suggesting that our approach does not alter the observed dynamics.

To illustrate the chemical and structural sensitivity of our approach, we utilize the nondispersed imaging modality of fs-TAM and apply it to investigate the transient morphology of TIPS pentacene films,[47−49] which we have previously studied using vibronic spectroscopy with sub-10 fs resolution at the ensemble level.[50] Here, we investigate films aged for 1 month under ambient light, which introduces considerable microscopic inhomogeneity to a previously smooth and homogeneous film (Figure S2), accompanied by local shifts of the absorption spectrum (Figure a).[51] In addition, we introduced a 740 nm band-pass filter in the detection channel, to maximize our sensitivity to such local spectral shifts (see SI) because the macroscopic transient absorption spectrum exhibits a sign-inversion from a ground-state bleach signal to a photoinduced absorption signal at this wavelength with a steep gradient.[50]

Figure 3

fs-TAM on aged TIPS pentacene films. (a) Normalized transmission image of a 15 × 15 μm2 area of an aged TIPS pentacene film. (b) Corresponding transient absorption image recorded at a pump–probe time delay of 1.25 ps. (c) Extracted Fourier transform power map for a diffraction-limited slice of the image as indicated by the vertically dashed line shown in (a) and (b). (d–f) Normalized Fourier transform power map of TIPS pentacene ground-state Raman bands at 267, 1167, and 1375 cm–1, respectively. Regions 1 and 2 illustrate large spatial differences in Fourier intensity. All fs-TAM images and corresponding spectra were recorded with a 740 nm band-pass filter (fwhm = 10 nm) in detection. The scale bar corresponds to 5 μm, and the data were binned to 175.9 nm/px. Pump and probe fluences were adjusted to 240 and 120 μJ/cm2, respectively.

The normalized transmission image of the aged film at 740 nm is spatially highly inhomogeneous (Figure a). The corresponding transient absorption image at a time delay of 1.25 ps following excitation by a 10 fs pump pulse centered at 560 nm generally reproduces this inhomogeneity, which is dominated by photoinduced absorption signals (red, Figure b) with occasional ground-state bleach regions (blue, Figure b). A possible origin for this behavior is degradation of the TIPS pentacene film upon aging (Figures S2–S4), resulting in spatial regions where only tightly coupled pentacene molecules still exhibit dynamics close to the original ultrafast ground-state bleach response. This causes a significantly altered transient response more similar to uncoupled TIPS pentacene molecules, which is expected to blue-shift the transient response by ∼100 nm (Figure S5).[51]

An intrinsic advantage of our ability to use extremely short pump and probe pulses in fs-TAM is the impulsive action of the pump pulse on the sample, which triggers synchronized molecular motion modulating the recorded time-domain signal via impulsive stimulated Raman scattering.[52−54] Applying a Fourier transformation for recorded transients subsequently enables us to retrieve spatially-resolved Raman spectra.[40,41] To illustrate this aspect of fs-TAM, we have selected a cut along the center of the image (dashed line in Figure a,b) and carried out a Fourier transform analysis for each recorded transient signal (Figure c). For homogenous samples, we would expect a homogeneous intensity distribution along the system’s Raman frequencies (Figure S4). For the aged TIPS pentacene sample, however, we find large variations in Fourier intensity as a function of space. Comparing the relevant regions with the TAM image at 1.25 ps (Figure b) shows that ground-state bleaching signals result in measurable Fourier intensity, while photoinduced absorption signals lead to considerably reduced intensity. A comparison of the obtained Fourier spectra with a previously published time-domain ground-state spectrum further strengthens this correlation (see the SI for a more detailed analysis).

Beyond changes in the overall Fourier intensity, close inspection of the Fourier map reveals significant differences in the relative intensities of the individual bands (Figure c). To further emphasize this observation, we computed the Fourier transform power spectrum at each imaged pixel and mapped the spatial dependence of the three dominant ground-state bands located at 267, 1163, and 1375 cm–1 (Figure d–f). Despite overall similarity in the spatial distribution of the recorded Fourier intensity, we detect regions in which the relative Fourier intensities vary substantially as a function of space and wavenumber (compare regions 1 and 2 in Figure d–f), highlighting the potential of fs-TAM to detect spatial inhomogeneity simultaneously with electronic and structural sensitivity (compare Figure b and d–f and see the SI). We remark that the detection of vibrational coherences in the time domain is closely linked to the noise-floor of the experiment. Typical vibrational coherence amplitudes for the employed pulse parameters range from ΔT/T = 10–3 to 10–4,[40,41] while the current setup is routinely capable of achieving sensitivities below ΔT/T = 10–4, making it widely applicable to a variety of systems (see the SI).

We have presented a fs-TAM capable of acquiring transient vibronic responses of nano-structured materials simultaneous sub-10 fs temporal and diffraction-limited spatial resolution. Initial experimental results for MAPbI3–Cl perovskite and TIPS pentacene films demonstrate the potential of fs-TAM to investigate ultrafast dynamics following photoexcitation on microscopic and potentially nanoscopic length scales. The intrinsic ability of fs-TAM to record impulsively generated Raman spectra will prove extremely useful in determining not only the electronic but also the structural origins and outcomes of optical responses of materials. With the current experimental parameters, fs-TAM can achieve a shot-noise-limited localization precision of sub-30 nm,[55,56] which, combined with the structural information, makes it ideally suited to study energy flow and transfer through nanoscale systems and across boundaries. We note that the current setup can be readily extended to include a super-resolution modality upon addition of a third saturation pulse.[25] Furthermore, combination of focused pump with wide-field probe pulses will enable studies of spatially evolving vibronic dynamics with unprecedented simultaneous spatial and spectral resolution.