AVS2015 Session SS+AS+EM+EN-ThM: Semiconductor Surfaces and Interfaces - I

Functionalization of semiconductor surfaces can provide tunable control of interfacial properties in organic-inorganic hybrid devices. In particular, multifunctional molecules have the potential to change the surface chemistry by leaving unreacted functional groups available after adsorption. Understanding the adsorption of these complex molecules could lead to various applications as sensors, selective film deposition, and molecular electronics.

In this work, the reaction of 1,2,3-benzenetriol on Ge(100)-2x1 surface was investigated. While the reaction of hydroxyl groups has been previously studied, differences in selectivity can be expected due to the position of the functional groups along the ring. The purpose of this study is to determine the extent of these differences and the effect on product distribution.

An analysis of the adsorption energetics was carried out by density functional theory. As expected, a proton transfer reaction was shown to be the most stable adsorbate configuration. However, after the adsorbate reacts with the surface through its first OH group, the energetics of the second OH dissociation showed differences based on two factors: (i) surface configuration (cross or diagonal trench and end or cross bridge) and more interestingly (ii) which two of the OH groups (1 and 2 or 1 and 3) are reacting with Ge. The latter constraint affects the adsorption energy of the second dissociation, where adsorption regardless of the surface configuration is less stable when the OH groups are next to each other. Finally, transition states for dissociation of the third OH were found to be limited by the configuration of the second dissociation, and in some cases were not possible to find without unrealistic distortions of the molecule.

Chemisorbed and physisorbed O(1s) and C(1s) spectra were obtained by X-ray photoelectron spectra. Differences between these spectra can be used to identify the reaction products. No change in the C(1s) spectra was observed, suggesting that no carbon forms a bond directly with the Ge surface. On the other hand, clear differences between the chemisorbed and physisorbed O(1s) spectra are observed. The presence of a second peak with a lower binding energy only in the chemisorbed spectra, assigned to oxygen bonded to Ge, confirms that 1,2,3-benzenetriol reacts with the Ge surface through OH dissociation. Quantitative analysis of the chemisorbed O(1s) spectra provides information on the fraction of OH groups reacting with the surface. Interestingly, about 66% of the total hydroxyl groups in 1,2,3-benzenetriol are involved in reaction with Ge, indicating that there is a significant fraction of unreacted OH groups.

In this work, the attachment of liquid and vapor-phase ethylenediamine on three types of oxide-free (H-, F- and Cl-terminated) Si(111) surfaces is examined by infrared absorption spectroscopy and X-ray photoelectron spectroscopy in conjunction with first-principles calculations. We find that chemisorption is only possible on F- and Cl-terminated Si surfaces, with H-terminated Si surfaces yielding only physisorbed diamine molecules. On Cl-terminated Si surfaces, diamines adsorb in a mixture of monodentate and bridging configurations (chemical reaction of both amine endgroups), while on partially F-terminated Si surfaces the adsorption occurs primarily at one end of the molecule. The reaction of ethylenediamine with Cl-terminated Si surfaces is also characterized by complete removal of Cl and partial Si-H (~25% ML) formation on the surface. This unexpected result suggests that a proton-chlorine exchange may take place, with the endothermic barrier possibly reduced via a silicon lattice assisted process after an initial attachment of ethylenediamine to the surface.

The monolayer coatings with aromatic functional groups can be used to tune mechanical, electronic, and chemical properties of semiconductor surfaces. This work focuses on obtaining well-defined surface of silicon functionalized with phenylhydrazine to produce an oxygen-free platform for further functionalization. Single crystalline Si(111) surface has been prepared using modified RCA procedure to produce well-ordered H-Si(111) surface. Next, Cl-terminated Si(111) surface is prepared from H-terminated Si(111) surface using PCl₅ in chlorobenzene solvent with trace amount of benzoyl peroxide as a reaction initiator under nitrogen atmosphere following previously established procedures. Phenylhydrazine-functionalized Si(111) sample is obtained from Cl-Si (111) surface with phenylhydrazine at 38^oC under N₂ atmosphere. To confirm the presence of Si-N bonds following this procedure, establish the structures of surface species produced and to investigate the oxidation mechanism, we followed the reaction by Fourier-transform infrared spectroscopy, X-ray photoelectron spectroscopy, and time-of-flight secondary ion mass spectrometry. To study the formation of Si-NH_x groups, this result was compared with the results of phenylhydrazine reactions on clean silicon surface under ultra-high vacuum (UHV) conditions. Density functional theory (DFT) calculations were performed to infer the mechanisms of surface reactions and further oxidation steps, and to compare the predicted vibrational spectra and core-level energies with the results of experimental studies.

Recently, oxide-free and partially methoxy-terminated Si surfaces¹ have been developed as a novel platform for surface reactions because of their superior reactivity compared to hydrogen termination². As a result, strong polar bonds such as Si-F could be stabilized on these surfaces. Since the electrical quality is critical for many applications (i.e. surface defects can degrade the device performance), we performed contactless surface recombination velocity measurements to examine the electronic quality of partially covered surfaces. Interestingly, we found that the carrier lifetime is significantly increased after fluorine termination, with the carrier lifetime 10 times higher than that of hydrogen terminated Si surfaces, approaching 1.5 ms. This anomalously long carrier lifetime can be explained either by a better surface passivation or by surface band bending effects. We therefore performed UPS and kelvin probe measurements to investigate the band structure of these surfaces after fluorine termination and found evidence for band bending. A potential model of a surface dipole layer induced band bending is supported by DFT calculations. Regardless of the mechanism controlling the recombination time, this method is well suited to explore the fluorination mechanism of H-terminated surfaces.

[1] D. Michalak, S. Amy, D. Aureau, M. Dai, A. Esteve, and Y. J. Chabal, Nature Materials, 9, (2010)

[2] P. Thissen, T. Peixoto, R. Longo, W. Peng, W. Schmidt, K. Cho, and Y.J. Chabal, JACS, 134 (2012)

It is known that native oxides of III-V semiconductors are consumed during atomic layer deposition using certain subsets of precursors. It was believed these surface oxides were completely removed during the first few deposition cycles because once the surface was covered by a coalesced film the native oxides would be protected. It has been observed, however, that native oxide consumption in systems such as ALD TiO₂ on GaAs(100) and InAs(100) proceeds continuously well after the surface is completely covered. Therefore there must be a transport mechanism that continuously moves these oxides through the developing film in order to interact with the precursor at the surface and be removed.

The aim of this work was to find unequivocal evidence of the transport mechanism needed for continuous oxide removal during ALD at typical processing conditions. ALD processes using metal organics and H₂O were used to deposit TiO₂, Al₂O₃ and HfO₂ films on GaAs(100). The experiments were designed so as to decouple the native oxide consumption from the native oxide transport and provide convincing evidence for the existence of this unacknowledged thus far mechanism. We will provide results that solidify the hypothesis that native oxide diffusion is a critical component in the complete and continuous removal of the interfacial layer.

Time resolution, surface sensitivity and element specificity are technical ingredients required to investigate ultrafast photoinduced processes and charge localization at semiconductor surfaces. All these requirements are fulfilled by a new experimental apparatus that consists of a tunable femtosecond high harmonics XUV source, a pump-probe setup, and an ultra-high vacuum surface science chamber for surface preparation and investigation.

The present contribution focuses on the charge carrier dynamics at the surface of a bare p-type GaAs(100) as well as Ti overlayers on p-type GaAs(100). The charge transfer between the bulk and the surface of the bare GaAs(100) is produced by the pump laser pulse at the central wavelength of 800 nm and is investigated by monitoring the surface photovoltage through the shift of the Ga 3d photoemission peak with the XUV probe laser pulse as a function of the pump-probe time delay. A transient shift of the Ga 3d photoemission peak to lower binding energy at early pump-probe time delay, with a magnitude of 0.3 eV, is observed and is attributed to transport of the electrons from the bulk to the surface. Upon increasing the pump-probe time delay, a restoration of the Ga 3d peak is observed, which corresponds to the recombination of the positive and negative carriers.

When a Ti overlayer is deposited on the p-type GaAs(100) surface, a Schottky diode is formed. If the 800 nm pump laser pulse has sufficient intensity to produce a photoemission process via multi-photon excitation, non-equilibrium effects occur at the Ti-GaAs interface independently from the presence of the surface photovoltage. In this case, positive charges accumulate at the surface and are not effectively screened by the electrons coming from the bulk, and the Schottky diode is transiently driven into a reversed bias mode. The formation of the reverse bias Schottky diode, which is studied in real time with the XUV probe laser pulse by monitoring the Ti Fermi level photoemission shift as a function of the pump-probe time delay will be presented and discussed.

AlGaN layers prepared by metal-organic chemical vapor deposition, with varying composition of Al (6 – 17%), were studied using the surface photovoltage (SPV) technique. Previous SPV studies on both n and p-type GaN allowed us to calculate the value of the surface band bending, by applying a thermionic model to explain the transfer of charges over the near surface barrier in various conditions (air, vacuum, and for a wide range of temperatures, T = 80 – 600 K). [1,2] The band bending was estimated to be 1.0 eV and – 2.0 eV, for n-type GaN and p-type GaN, respectively. SPV measurements on p-type AlGaN layers were expected to have similar behaviors to their p-type GaN counterparts. However, numerous measurements showed that this was not the case. The SPV transients (upon turning on or off the excitation source) showed significantly slower transients and smaller values than expected from the thermionic model. Moreover, the restoration of the band bending, as indicated by the restoration of the SPV signal to its dark value, did not occur within a reasonable amount of time. The data could not be fit by the thermionic model, and thus we were unable to calculate the band bending. We attribute the slow transients and lack of restoration to a defective surface region which interferes with thermionic processes. To verify this assumption, the top 40 nm of the AlGaN layer was etched using a reactive-ion etch (RIE). After etching, the SPV behavior exhibited substantially different behavior. Fast transients and close-to-thermionic behavior was recovered. Additionally, the effect of annealing the samples after etching provided even closer values to what is predicted by the thermionic model. From this study, it can be concluded that a defective, near surface region is inhibiting the transfer of holes over the near surface barrier under illumination, and hole trapping may be occurring during restoration. In both cases, this behavior cannot be modeled by theory. Etching removes the defective layer, and reveals a region of presumably higher quality as evidenced by the subsequent thermionic behavior.

[1] M. A. Reshchikov, M. Foussekis, and A. A. Baski. J. Appl. Phys. 107, 113535 (2010).

[2] M. Foussekis, J. D. McNamara, A. A. Baski, and M. A. Reshchikov, Appl. Phys. Let.101, 082104 (2012).