Femtosecond time-resolved two-photon photoemission studies of ultrafast carrier relaxation in Cu2O photoelectrodes

Cuprous oxide (Cu2O) is a promising material for solar-driven water splitting to produce hydrogen. However, the relatively small accessible photovoltage limits the development of efficient Cu2O based photocathodes. Here, femtosecond time-resolved two-photon photoemission spectroscopy has been used to probe the electronic structure and dynamics of photoexcited charge carriers at the Cu2O surface as well as the interface between Cu2O and a platinum (Pt) adlayer. By referencing ultrafast energy-resolved surface sensitive spectroscopy to bulk data we identify the full bulk to surface transport dynamics for excited electrons rapidly localized within an intrinsic deep continuous defect band ranging from the whole crystal volume to the surface. No evidence of bulk electrons reaching the surface at the conduction band level is found resulting into a substantial loss of their energy through ultrafast trapping. Our results uncover main factors limiting the energy conversion processes in Cu2O and provide guidance for future material development.

shows the photoemission process involving multiple photons. In a single-color experiment (probe only, two blue arrows, right) a two-photon transition lifts an electron from the VBM 0.5 eV below the Fermi level to above the vacuum level reaching the VBM (1) as indicated level on the kinetic energy scale. For the two-color experiment (probe and pump, blue and green arrow, center) the VBM is projected onto a lower level indicated as VBM (2) due to the lower combined photon energy. Both cases are mediated via virtual states and the energetic band alignment of the occupied states (valence band) is measured. In contrast, two photon transitions via resonant intermediates probe the energetic structure of the unoccupied states. This process is illustrated by where EKin is the directly measured electron kinetic energy and F the work function of the spectrometer. As an example, the binding energy scale that is valid for this specific transition the Fermi level was found, similar to previously reported values 5,6 . The spectral shape as well as the position of single spectral features was not changed when the probe intensity was varied by more than two orders of magnitude.

Surface band bending
The surface band bending (BB) was evaluated in the Cu2O semiconductor due to the either adsorption of the terminal molecular oxygen described above as well as due to the formation of a  Figure 7a). An exponential decay of the potential within the semiconductor was assumed for the generation of the Cu 2p and O 1s core levels 11 . The simulations were performed considering the BB potential and the SSC width as free parameters, to obtain the same ΔBE for the simulated core levels as that observed experimentally. The inelastic mean free paths (IMFP) were obtained using the Tanuma For reconstructed samples, XPS did not provide evidence for formation of surface CuO or solid Cu, which are prone to recombination and Schottky barrier formation, respectively, [13][14][15][16] both of which may reduce the photovoltage and/or photocurrent.

Photoluminescence (PL)
Photoluminescence (PL) measurements were conducted to investigate the origin of the observed defect levels (see Supplementary Figure 12). PL provides bulk information and therefore complements the surface-sensitive 2PPE technique. The experiments were conducted at different stages of sample preparation (reconstructed, Pt-covered) while maintaining UHV conditions throughout. The pump laser pulse conditions were identical to those in the 2PPE measurements.
Both PL spectra exhibited a sharp peak at ~ 620 nm that has been assigned to exciton luminescence in Cu2O 17,18 (Supplementary Figure 12a). This peak was used to normalize the spectra, as it is not expected to be influenced by the surface treatment. The second, much broader contribution, in the range between 800 and 1050 nm, can consistently be attributed to defect- with expectations for final state effects that account for the changes in the response of the valence electrons to the core-hole as a function of the cluster size 26 . The decrease in the line shape asymmetry (i) for low Pt coverage is consistent with expectations for a loss of screening of the core-hole potential 26 , whose reduction causes also a decrease of the cluster work function 9 and the consequent positive shift of the 4f binding energy (ii). The linewidth broadening (iii) is instead ascribable to the plasmon excitation at the boundary surface of the cluster 26 . The smaller the Pt cluster, the higher the surface-to-bulk ratio, and the higher the plasmon excitation effect on the acquired spectral signal. The observation of the three aforementioned size effects in the case of the lower Pt coverage is consistent with expectations for a non-conformal Pt overlayer, which presumably consists of separate nanometer-sized islands. A thin, continuous Pt layer exposing under-coordinated Pt atoms may also produce an increase in the 4f binding energy 27 . In contrast, compared to a bulk Pt phase, the spectral change observed in this latter case typically is characterized by only a shift in binding energy, without modification of the spectral linewidth and asymmetry. The fitting procedure also did not yield evidence for oxidized Pt species at the semiconductor/metal interface.
To summarize, X-ray photoelectron spectroscopy (XPS) data and its theoretical modeling indicated a non-conformal growth for Pt deposition via UHV evaporation of films with thicknesses < 1 nm. For films with theoretical thickness equivalent to 1 nm and larger -as determined by monitoring the rate of deposition using a quartz crystal microbalance (QCM) in UHV -the fitting procedure of the XPS data revealed a decrease in the size effects (line shape asymmetry, shift of the Pt 4f binding energy, linewidth broadening) associated with nonconformal growth (Supplementary Figure 5b). Based on these results, Pt films with QCM thickness equivalent to 1 nm were used for all of the experiments discussed herein.

AFM+SEM/EDX
Atomic force microscope (AFM, NT-MDT Ntegra) images showed a non-conformal structure after Pt deposition on the reconstructed Cu2O (100) surface (see Supplementary Figure   8). Non-laminar island growth is often observed for vacuum metal deposition on metal oxides because of weak adatom-substrate interactions 28 . Three-dimensional clusters with diameters of ~ growth was centered around pinholes that extended tens of nm into the Cu2O (100) surface. This type of growth behavior has been previously observed and ascribed to a trapping of adatoms at surface defects, thereby forming nuclei for a subsequent heterogeneous nucleation growth process 29 . The AFM images were obtained in semicontact mode and the phase signal showed a clear contrast that resembled the domain boundaries observed in the height map. This correlation is consistent with the presence of different mechanical properties, and hence different materials, inside and outside the domains.
The presence of Pt islands on the Cu2O surface was verified by means of scanning-electron microscope (SEM) imaging and EDX elemental distribution mapping (Supplementary Figure 9), performed at a beam energy of 2 kV to minimize the excitation volume. Data were obtained using a Zeiss UltraPlus scanning electron microscope and an Oxford Instruments Ultim Extreme EDX system with Aztec software suite, at a beam current of ~ 10 nA for optimal signal-to-noise ratio). The Pt-N and C-K lines are very close in energy to each other, thus, the EDX data alone cannot unambiguously confirm the presence of Pt in the islands. However, additional SEM imaging at much lower beam current (about 10 pA) than for the SEM image in Supplementary   Figure 9, indicated that the islands were composed of small agglomerates (about 100 nm in diameter), which is typical for Pt but not for C (Supplementary Figure 10).