Access to Ultrafast Surface and Interface Carrier 1 Dynamics Simultaneously in Space and Time 2

We thank Prof. Dr. Xue-Wen Fu and Dr. Shan-Shan Zhang for their insightful suggestions and discussion on this work. The research reported in this publication was supported by King Abdullah University of Science & Technology (KAUST).

dominant factor that significantly suppresses the photoconversion efficiency (PCE) of 1 optoelectronic devices. [16][17][18] In this regard, minimizing surface defects is an extremely crucial task 2 for a large variety of light conversion applications. [19][20][21][22] Thus, to pursue a higher efficiency of light 3 conversion, one needs to profoundly understand both theoretically and experimentally how surface 4 treatment and engineering affect the ultrafast surface carrier dynamics at the device's interface. To 5 precisely investigate the ultrafast surface carrier dynamics of photoactive materials, excellent 6 surface sensitivity methods are required.  pump-optical probe configuration with excellent spatiotemporal resolutions. [32][33][34] Finally, time-18 resolved near-field scanning optical microscopy (SNOM) with a sharp optical-fiber tip can achieve 19 ~ 100 of nm and ~ 100 fs spatiotemporal resolutions, respectively. 35 However, the main 20 shortcoming of these powerful techniques is their poor surface sensitivity caused by the intrinsic 21 deep penetration characteristic of the optical pump and probe pulses, which usually range in depth 22 from a single digit nm for metals, 36 and 10s of nm to a few μm in semiconductor materials and 23 insulators; 37 thus, the carrier dynamics information is largely obtained from the bulk, not surface 1 exclusively. Ultrafast thermomodulation microscopy (UTM) has recently addressed this 2 shortcoming by adopting a novel probe strategy, scanning the ultrafast changes in reflectivity due 3 to temperature differences at various distances from the excitation region of the specimen. It also 4 achieved an excellent 20 nm spatial resolution and a time resolution of 250 fs; 38 however, due to 5 the requirements of rapid heat dissipation, its application scope is limited to materials with a high 6 thermal conductivity, such as metals. These emerging techniques can provide abundant 7 information on charge carrier dynamics, but none of them are capable of selectively tracking 8 surface carrier dynamics with the required resolutions in real space and real time, simultaneously. better spatial and temporal resolutions. 5, 59-61 Since then, S-UEM has demonstrated its superior 23 2. Instrumentation, working principles and data interpretation 1 In conventional electron imaging techniques, a continuous beam of thermoelectrons is 2 employed as an imaging source to either scan the surface of a specimen (scanning electron 3 microscopy, SEM) or to transmit through a thin specimen (transmission electron microscopy, 4 TEM) to form an image. With a well-designed electron microscope, an excellent spatial resolution 5 down to ~ 1 nm and atomic scales can be obtained by SEM and TEM, respectively, providing 6 abundant morphological and structural information for a variety of materials, chemicals and 7 biological systems. [69][70][71] Unfortunately, due to the continuous characteristic of the electron beam, 8 these techniques have very limited temporal resolution and thus lack the capability of imaging 9 ultrafast phenomena in real time and real space simultaneously, which is highly demanded in 10 modern ultrafast sciences. By using a computer-controlled beam chopper or similar electronic 11 components, the continuous electron beam was separated into pulsed electron packets in 12 stroboscopic electron microscopes, which were invented in the 1960s to enable the temporal 13 resolution of electron microscopes. 72 However, since each electron packet still contains a large 14 number of electrons and due to the space-charge effect (Coulomb repulsion among electrons within 15 the packet), the temporal resolution of this type of stroboscopic electron microscope is limited to 16 sub-ns. 73 This shortcomings was addressed successfully by introducing a laser source to generate  Fig. 1a) or a field emission 5 gun (Fig. 1b) to generate ultrashort packets of photoelectrons. Note that the temperature of the 6 field emission gun of conventional SEM is about 1800 K, and contains enough energy for the gun 7 tip to emit electrons thermally for imaging, while in S-UEM, we significantly reduce the tip 8 temperature by reducing the filament current before laser illumination to generate pulsed electrons 9 for ultrafast imaging. 59 In S-UEM, an optical delay line controlled by a computer is used to provide 10 delay times in the range of -1 ns to ~ + 6 ns. It should be noted that a negative time means that the 11 electron-probe pulse arrives at the sample surface prior to the optical excitation pulse, while a capable of escaping into the vacuum as SEs, which are eventually captured by SEs detectors 20 (Everhart-Thornley detector) for imaging. The current of these pulsed photoelectrons is in the 10s 21 of picoamperes (pA) range 59 or even down to almost a single electron per packet, 40 suggesting that 22 the pulse-broaden effect due to the spatial repulsion of electrons is minimized compared to the 23 ones generated by beam blanker in other time-resolved SEM. 74 To enhance the signal-to-noise 1 ratio, the time-resolved images were recorded as an integration of multiple (usually 64 or 128) 2 frames with a dwell time of 300 ns at each pixel.

3
A difference image approach has been applied in S-UEM. More specifically, we first take a 4 reference image at a very negative time (when the PEs beam arrives much earlier than the optical 5 pump beam), then, we take images at different time delays and subtract the reference image to 6 obtain the time-resolved SEs difference images. By this way, the difference between the irradiated 7 and unirradiated areas by optical pulse can be clearly seen, and the charge carrier dynamics can be In addition to the bright contrast illustrated in Fig. 1b, dark contrasts 76 or dark-bright coexisting 8 contrasts 77-79 were also observed in diverse photoactive materials. The mechanism of bright 9 contrast is straightforward; electrons are promoted from the valence band (VB) to conduction 10 bands (CB) via optical excitation pulses. As a result, these excited electrons with a higher kinetic 11 energy compared to unexcited electrons will have a higher probability of SEs emission, resulting 12 in a bright contrast in the difference images (the so-called "energy-gain" mechanism). 76 In contrast, 13 there are a few proposed mechanisms for dark contrast formation (the so-called "energy-loss" 14 mechanism). For instance, in p-n junction semiconductors, after photoexcitation the 15 photogenerated electron-hole pairs will separate and transport toward the junction. The generated 16 holes will then give rise to a lower yield of SEs compared to the unexcited area (shown as a dark 17 contrast in the S-UEM snapshots). 78 For other semiconductors, the understanding of dark contrast 18 has developed into several stages. The first possible mechanism is the energy loss, which was 19 attributed to the diverse scattering events of photon-carrier and carrier-carrier interactions, as 20 observed in GaAs single crystals 76 and InGaN nanowires. 5 In other words, scattering processes 21 with photogenerated electron-hole pairs are most likely responsible for the dark contrast or energy 22 loss mechanism. As the effective cross-section for the scattering of SEs with conduction electrons 23 is much higher than that with valence electrons, 76 a decrease in the SEs emission is observed, 24 resulting in a dark contrast. 5, 60 Recently, our research group found out that with the removal of 1 ultrathin surface oxides, a contrast change from dark to bright was clearly observed in several 2 semiconductor material surfaces, including Si, CdTe, CdZnTe and GaAs. 62 In that work, the results 3 of density function theory (DFT) calculations revealed that the work function will increase by 4 approximately 0.2 eV in the oxygen-passivated Si (100) single crystals (Fig. 2), leading to a 5 stronger resistance for the escape of SEs from the sample surface and thus a lower yield of SEs is 6 generated as compared to the unilluminated areas, which eventually causes dark contrast formation 7 in oxygen-passivated Si single crystals. Revealing that the surface defects, such as oxygen 8 vacancies in the band gap aroused by the surface oxide formation, should contribute to the dark 9 contrasts observed in these materials. 62 Although the DFT calculations suggest that the oxide layer 10 leads to the defect formation and work function changes, the band bending and nature of the 11 surface with and without oxide layer could also contribute to the image-contrast change. To 12 conclude, we still believe that more efforts should be made to fully uncover the mechanism(s) that 13 result in dark contrast, and theoretical simulations, such as ab initio and Monte Carlo simulations, 14 could be promising tools in this regard.

12
It is well known that many factors, such as the environmental atmosphere, surface morphology, 13 absorber-layer thickness, surface passivation/coating, crystal orientations, doping types, and 14 doping levels of photoactive materials will affect the charge carrier dynamics and the overall PCE 1 of corresponding optoelectronic devices. S-UEM, with its unique surface sensitivity, can provide 2 valuable and controllable information from all of these aspects, which will be discussed in several 3 selected applications in the following section.

4
The different operation modes in S-UEM for visualizing the surface and interface dynamics of 5 specimens under different conditions are presented in Fig. 3. Some experiments can be conducted 6 under extreme environments, such as in humid and acidic/basic environments and in the presence 7 of organic solvents. Since evaluating the impacts of the environmental medium on the surface 8 dynamic processes of specimens is crucial, S-UEM in the environmental mode ( Fig. 3a) can 9 simulate the effects of these factors by introducing certain amounts of gaseous solvents into the 10 SEM sample chamber (we will discuss a case of this scenario in detail later). 62 S-UEM in high 11 vacuum mode (Fig. 3b) can be facilely employed to study the intrinsic properties of the specimen.

12
Moreover, there are many electron-beam-sensitive photoactive materials, such as nanocrystals and 13 perovskite solar cells, have important and promising applications. Characterizing electron-beam-14 sensitive materials using electron imaging techniques is a challenging task. 80 To overcome this 15 obstacle, S-UEM at a low AV (1 kV) has been recently developed by our research group. We will 16 discuss its superior surface selectivity and promising capability in characterizing electron-beam-17 sensitive materials in a later section. Also, the combination of S-UEM in high-and low-voltage 18 modes ( Fig. 3c) to comprehend the layer/thickness-dependent photophysics of materials is also an 19 interesting direction that will be discussed.   The most significant application of S-UEM is the visualization of the charge carrier dynamics of 3 photoactive material surfaces in real space and real time simultaneously. In the following section 4 we will discuss such applications in a variety of semiconductor systems.  In the case of uniform n-and p-type Si single crystals and Si p-n junctions, significant 11 differences in the SEs images are observed (Fig. 4). 78 For both lightly doped n-type Si and heavily respectively. This rapid cross junction transportation arrives to its maximum at approximately +80 20 ps after photoexcitation. Later, excess electrons/holes on both sides recombine gradually. It is 21 worth noting that the transport distance expanded to more than 80 μm during the first 80 ps, which 22 is much larger than the depletion layer of Si (varying from dozens to hundreds of nm 84 to a few 23 μm, largely depending on the doping concentrations 85 ) and violates the well-accepted drift-1 diffusion theory. 86 Alternatively, the authors proposed a model based on ballistic-type motion, 2 where the calculated localization of charge density around the p-n junction ( Fig. 4d-g) was in good 3 accordance with the experimental data.

13
The carrier dynamics of heavily doped n-type and p-type Si single crystals 87 and amorphous Si 77 14 ( Fig. 5) are also unique and different from one another and from those of the p-n junction discussed 15 in Fig. 4.The contrast of n-type Si transforms from the bright contrast of lightly doped Si 78 ( Fig.   16 4a) into a dark contrast for heavily doped Si 87 (Fig. 5a). Note that the dark contrast was attributed 17 to the dynamics of photogenerated holes; 78 however, the heavily doped p-type Si (Fig. 5b) shows 18 a bright contrast, similar to its counterpart in Fig. 4b. Furthermore, the ultrafast superdiffusion in 1 the initial 200 ps ( Fig. 5a-b) was attributed to the fast diffusion of the hot photogenerated charge 2 carrier. 87 For amorphous materials, such as hydrogenated Si (a-Si:H), unexpected ultrafast (40 ps) 3 formation and diffusion of hot carriers followed by efficient electron-hole separation and trapping 4 of both electrons and holes were directly imaged (Fig. 5c). This phenomenon was ascribed to a 5 transient high temperature triggered by laser illumination, suggesting that S-UEM is capable of  As discussed in Figs. 4 and 5, being a single crystalline or amorphous form can dramatically 8 influence the charge carrier dynamics of Si. It is worth mentioning that our group found that 89 9 surface native oxides have significant impacts on the recombination of charge carriers on Si 10 samples. As shown in Fig. 6a, most of the photoexcited charge carriers recombined within 3 ns on 11 the Si single crystal with native oxide layers (top row). In contrast, most photoexcited charge 12 carriers survived much longer after 3 ns in the same type of Si sample, but were freshly etched by 13 hydrofluoric acid (HF), which removed the native oxide layers (bottom row). These results directly 14 proved the crucial impacts of native oxides on the charge carrier recombination dynamics. Because 15 the native oxide layer will be reformed on a clean Si surface when it is exposed to oxygen and 1 moisture, even for a short period of time, Kelvin probe force microscopy (KPFM) was conducted 2 to study the surface potential of freshly etched Si (Fig. 6b-c). It showed a large difference in the 3 surface potential between areas with and without laser illumination that was measured immediately 4 after HF etching; however, after exposure to air for 40 min, there was no difference between the 5 laser-on and laser-off conditions ( Fig. 6b and 6d), suggesting that the fast regrowth of native  The study of charge carrier dynamics at the surface of GaAs single crystals (Fig. 7) 76 showed 6 that the dark contrast is the only phenomenon observed at both negative time delays and positive 7 time delays (as noted before, positive delay times, means optical pump beam arrives prior to PEs 8 beam, and negative delay times means PEs beam arrives before the optical pump beam). The 9 reason for this phenomenon is as follows: both PEs pulses and optical excitation pulses can 10 increase the average energy of SEs via collisions (this will promote the SEs yield in the 11 photoexcited area compared to the unexcited area, "energy gain" mechanism). However, both the 12 electron and optical pulses can also scatter the SEs via collisions and restrain the SEs from reaching 13 the sample surface and escaping into the vacuum (this will suppress the SEs yield in the 14 photoexcited area, "energy loss" mechanism). The dark contrast indicates that the scattering 15 process plays the dominant role and completely counteracts the energy gain effect benefiting from 16 the collision of SEs and electron/optical pulses. Note that the effective cross-section for the 17 scattering of SEs with conduction electrons is much higher than that with valence electrons, 18 leading to a decrease in the SEs emission and subsequently a dark contrast formation.   It is worth noting that in our recently published work, 62 similar to Si, surface native oxides also 8 play a significant role in the carrier dynamics of GaAs (110) single crystals because after the 9 removal of surface native oxides, the contrast changes from dark to bright at positive time delays 10 (Fig. 8). The XPS spectra proved the removal of native oxides, and the SEM images, before and 1 after argon etching of the GaAs single crystal showed there were no observable changes in the 2 surface morphology. 62 These results unambiguously proved that the origin of the contrast 3 transformation was surface defects due to native oxides formation.   The surface morphology (crystalline or thin films) and the thickness of the absorber layers in 4 various optoelectronic applications is of great concern and interest. Taking CdSe as a model system 5 to address this concern, S-UEM provides direct evidence of faster charge carrier recombination in 6 powder film compared to the single crystal form (Fig. 10a), which could be attributed to the 7 presence of much higher concentrations of surface trapping states including nanostructure features 8 (acts as quenching centers) in the thin film sample. We also observed different contrasts and 9 different carrier dynamics from CdSe powder films with different thicknesses (Fig. 10b-c). For   The specimens investigated using the S-UEM discussed above are bulk materials with simple 8 structures, morphologies and compositions, whereas the devices used in practical applications may 9 have much more complex structures and architecture. In this regard, we chose indium gallium 10 nitride nanowires (InGaN NWs) as a model system for nanoscale photoactive semiconductor 11 materials ( Fig. 11a-b). The observed dark contrast can be ascribed to the energy loss mechanism 12 that originated from the scattering processes. 5, 60 The evaluation of synthetic strategies, such as the 13 influence of surface passivation and doping conditions on the charge carrier dynamics can also be 14 conveniently visualized via S-UEM. We passivated the InGaN nanowires using 1-octadecanethiol 15 (ODT), a popular surface assembly reagent that has been reported to eliminate defects and suppress 1 the dangling bonds at the surface of InGaN nanowires that lead to a significant enhancement of 2 the PCE of InGaN-based light-emitting devices. 95 Being in this regime, slower charge carrier 3 recombination rates (Fig. 11c-d) were unambiguously demonstrated for ODT-passivated InGaN 4 nanowires. Therefore, the enhanced PCE can be reasonably attributed to the lower density of the 5 surface states after treatment, which suppresses nonradiative recombination channels. 96 Moreover, 6 S-UEM is also capable of exploring how Si doping affects the photophysics of InGaN/GaN 7 nanowires ( Fig. 11e-f). 97 Interestingly, it was demonstrated that with Si doping, the specimen and delays the growth to a larger degree for the first 200 ps (Fig. 11f). Meanwhile, a higher 14 concentration of trapping states can also quicken the subsequent electron-hole recombination after 15 the initial photoexcited charge carrier separation, revealing the reason why the decay rates for 16 doped sample with Si at 1200 °C is larger than both doped sample at 1150 °C and undoped sample 17 in the longer time regime (Fig. 11e). Moreover, these results were corroborated by numerical

Indium Gallium Selenide nanocrystals (CIGSe NCs)
3 Similar to the InGaN nanowires discussed above, the surface trapping states can act as 4 recombination centers for the photoexcited charge carriers that affect their performance in 5 optoelectronic devices, and efficient surface passivation techniques need to be facilely designed to 6 minimize these types of trapping states. S-UEM is also powerful in accurately mapping the surface 7 trapping states due to its superior surface selectivity. Taking the CIGSe NCs photodetector as an 8 example, the surface passivation of the CIGSe NCs by zinc sulfide (ZnS) 60 was directly visualized 9 via S-UEM (Fig. 12). After ZnS passivation, a slower charge carrier recombination was achieved 10 (Fig. 12a), and the photoresponsivity was enhanced three times as compared to the unshelled 11 CIGSe-based devices (Fig. 12b). It should be noted that the high-resolution TEM (HRTEM) 12 images verified that the size and the shape of CIGSe NCs before and after ZnS passivation 13 remained almost the same (Fig. 12c-d), indicating that observed changes in carrier recombination 14 before and after surface treatment is mainly related to the difference in the surface traps of the two 15 samples.    A thin film made from a semiconducting polymer, poly (3-hexylthiophene-2, 5-diyl) (P3HT, 11 molecular structure shown in the bottom right corner of Fig. 14a), was optically excited (515 nm) 12 at the center, and both the photoexcited and unexcited regions were scanned by PEs (Fig. 14a). A 13 rectangular region nearly 200 μm from the center was chosen to visualize the generation and 14 evolution of surface acoustic waves (SAWs) after photoexcitation (Fig. 14b). Two representative 15 time delays (0.1 ns and 3.4 ns) are presented to show the generation and propagation of SAWs 10 16 ( Fig. 14c-d). By mapping these photoresponse behaviors, the surface elastic properties of materials 17 can be mapped and evaluated accurately, which is of great importance for thin-film devices and 18 materials that require a specific stiffness. 105 Another important example is the periodic interference 19 patterns that can be observed on the surface of aluminum after optical excitation 57 (Fig. 14e). 20 Similar periodic patterns were only observed for some metal surfaces, but not all of them. This 21 astonishing phenomenon was proposed to be related to the plasmonic behaviors and magnonic 22 characteristics of rough metal surfaces, but the exact mechanism requires further investigation. 57

23
These cases proved that S-UEM is capable of probing not only the ultrafast charge carrier 1 dynamics of material surfaces and interfaces, but also the surface mechanical features.   16 Optoelectronic devices sometimes work in complex surroundings with abundant water vapor, 106 1 ammonia, acidic gases or organic solvents. In some extreme cases, the material interface with the 2 environment is even "wet". 107 S-UEM in high vacuum mode is well suited for exploring the 3 intrinsic carrier dynamics of materials, but is not suitable for evaluating the effects of ambient 4 conditions. In this regime, S-UEM in environmental mode (schematic illustrated in Fig. 3a) was 5 introduced by Zewail and co-workers to address this shortcoming. 81 Taking an CdSe single crystal 6 as the prototype (surface structures of the corresponding orientations shown in Fig. 15a), compared 7 with high vacuum conditions, the ambient conditions, especially a polar solvent environment, will 8 affect not only the contrasts (Fig. 15b-d) but also the dynamics (Fig. 15e-h   We further used LVS-UEM to evaluate the impact of ultrathin surface oxide layers on surface 7 charge carrier dynamics, which universally exist in most photoactive materials. For instance, the 8 natural surface oxide (SiO2) on Si (100) was reported to be ~1 nm thick and located at the 9 uppermost surface of as-received Si specimens. 122 The enhancement and fading of a dark contrast 10 were observed for Si (100) surfaces with surface oxides (Fig. 17a); the as-received Si samples 11 were kept in diluted (5%) HF (hydrofluoric acid) aqueous solution for half a minute, then washed 12 by deionized water and dried by ultrapure nitrogen to remove the native SiO2 layer and avoid any 13 significant surface morphology changes, after these surface cleaning operations, the samples were carriers on the clean surface of Si (Fig. 17b). The transformation from a dark contrast to a bright

Summary and Outlook 1
The characteristics of a superior surface selectivity, excellent temporal and spatial resolutions, 2 and easy sample preparation enable S-UEM to serve as a unique methodology for characterizing material, but also enable intricate optimization of synthetic and surface engineering strategies for 12 advancing high-performance optoelectronic, photovoltaic and photocatalytic devices. It is worth 13 pointing out that the influence of compositional defects, impurities and adsorbates species on the 14 surface dynamics can be directly explored and deciphered using S-UEM. 15 We would like to provide an outlook on future theoretical and experimental development 16 directions (Fig. 18). From a theoretical perspective, the balance between scattering processes (both 17 electron-electron and electron-photon scattering processes) and electron-beam-promoted SEs 18 emission from the photogenerated carriers in the CB of materials that lead to "energy gain" or 19 "energy loss" pattern remains unclear. We propose that with the assistance of scattering simulation 20 approaches (such as ab initio simulations) and mesoscopic carrier transport simulation models  The current optical excitation wavelength is a fixed wavelength (515 nm/343 nm/257 nm, etc.), 1 switchable excitation wavelengths (in other words, adding an optical parametric amplifiers (OPA) 2 module into the pump light section) will be useful for extending the application scope, especially 3 for visualizing the carrier dynamics of materials with large bandgap. It is also worth mentioning 4 that the performance of the emission gun in case of S-UEM measurements strongly depends on 5 the incident energy of the laser pulse. For instance, using shorter wavelength light will prolong the 6 period of steady photoemission, reducing the need for tip refreshment to avoid the formation of 7 nature oxides or other tip contaminations that reduce the efficiency of photoemission. 8 Implementing a temperature control module inside the chamber (for example, using liquid 9 nitrogen as a coolant and a heating module to heat up the specimen in order to enable measurement 10 under a temperature range of -196 to 100 ℃) to examine temperature-dependent surface dynamics 11 is also a very promising direction because many functional materials show temperature-dependent S-UEM experiments has to be compromised between 1500× to 3000× folds (for comparison, the 23 upper limit of magnification is 1,000,000× in FEI Quanta 650 FEG), so the detailed structural 1 features of the specimen are not as clearly imaged as one would expect. In addition, this hinders 2 the accurate evaluation of the charge carrier diffusion/transport lengths on the specimen surface, 3 because the diffusion lengths of most photoactive materials are in the range of 0 to 10s of μm; 131 4 only in very rare cases the length reaches the order of 100s of μm and even mm. 132 Therefore, the 5 current spot size is not suitable for precisely tracking the charge carrier dynamics of materials with 6 short diffusion lengths (0 to a few μm). To address this shortcoming, the size of the pump beam 7 should be reduced. Thus, we are currently developing a new optical conveyance system for the 8 pump laser to reduce the beam size by more than an order of magnitude. More specifically, we set 9 an objective lens inside the SEM specimen chamber to replace the focal lens outside of the sample 10 chamber (focal length is 175 mm, as shown as L2 in Fig. 16), introduce the optical pump pulses 11 into the objective lens and focus them on the sample surface with a much smaller size than that of

18
Where R is the radius of the final laser spot size, λ is the wavelength of the optical laser (0.515 19 μm), F is the focal length of lens and D is the diameter of the incident laser beam. 20 We believe that with these improvements S-UEM will be more powerful tool for a better