Ultrafast Surface Carrier Recombination in Semiconductors Evaluated by Two-Photon Photoemission Spectroscopy

Recently, ultra-small sized optical and electronic devices have actively been developing, and evaluation of ultrafast carrier recombination processes on semiconductor surfaces is important technique to control the device performances. Time-resolved photoluminescence spectroscopy and microwave photoconductivity decay are typically used to characterize carrier recombination lifetime in semiconductors. However, the detected signals include information from both surface and bulk states due to penetration depth of the excitation light. Time-resolved two-photon photoemission (Tr-2PPE) spectroscopy can be a novel technique to directly detect excess electrons as photoelectrons emitted from sample surfaces. In this approach, the first pulsed-light excites electrons from a valence band to a conduction band, and the second pulse ionizes the excited-electrons beyond a vacuum level after controlled time-delay. The detected signals are limited by surface carriers due to the short electron mean free paths (less than several nanometers). In this paper, we report on the analyses of surface recombination of semiconductors based on the Tr-2PPE spectroscopy where GaAs (110) is evaluated as an example, and show a strong impact of unintentional native oxides on the surface on carrier recombination processes.<br/>A mode-locked Ti:Sapp laser (pulse width: ~100 fs, repetition frequency: 80 MHz) was used as an excitation light source, and the fundamental wave (1.55 eV) with photon energy slightly higher than the bandgap energy of GaAs, 1.42 eV, was used as a pump light. The third-harmonic (4.65 eV) was generated as a probe light, and each pulsed light was input into a Mach-Zehnder interferometer. By controlling the time-delay, we constructed a Tr-2PPE system with high time resolution (the cross-correlation width of the pump and probe pulses was ~175 fs). The pump and probe pulses were irradiated on the samples in a vacuum chamber.<br/>Using the Tr-2PPE system, we evaluated carrier recombination lifetimes at un-doped GaAs (110) surfaces varying air-exposure time. After a clean GaAs (110) surface obtained by cleaving an un-doped (001) GaAs substrate in the vacuum chamber were characterized by the Tr-2PPE, the sample was exposed to the atmosphere for 0.1 to 2612 hours (temperature 20°C, humidity 40%). Then, it was re-introduced in the vacuum chamber for measurements. All the measurements were performed at room temperature.<br/>After excitation by the pump light, nonequilibrium excited electrons in the conduction band around <i>Γ</i>-valley were found to relax to the conduction band minimum with 0.6 ps. This relaxation process is attributed to intravalley relaxation associated with electron-phonon scattering due to Fröhlich interaction[1]. Surface carrier recombination processes of the excited electrons after being relaxed to the conduction band minimums were subsequently evaluated. From the long-ranged Tr-2PPE decay curves, the carrier recombination lifetimes were estimated by fittings of the decay curves with a single exponential component. Although the surface carrier lifetime at the as-cleaved GaAs (110) was 5100 ps, those after air-exposure drastically reduced as air-exposed time increased, and it ranged from 1670 to 73 ps. The carrier lifetimes are extremely short compared with the typical recombination lifetime based on band-to-band transitions in GaAs, and they clearly originate from nonradiative Shockley-Read-Hall processes due to surface recombination. Furthermore, we found that the formation of native oxides on the GaAs (110) surfaces significantly promotes nonradiative recombination. These results quantitatively explain that GaAs-based optical devices deteriorate significantly with air-exposure[2], and also demonstrate that the Tr-2PPE can be a powerful spectroscopic method directly evaluating the surface recombination lifetime of semiconductors.<br/>[1] H. Tanimura<i> et al</i>., <i>Phys. Rev. B</i> <b>93</b>, 161203(R) (2016).<br/>[2] G. Brammertz, <i>et al</i>., <i>Appl. Phys. Lett.</i> <b>90</b>, 134102 (2007).