AVS2009 Session SS1+EM-MoA: Semiconductor Surfaces and Interfaces I: Ge and III-V's

Field effect transistors (FETs) remain at the heart of integrated circuit technology, and are forecasted to do so for at least the next decade. Silicon has been the material of choice for this purpose, but appears to be reaching significant performance limitations with further device dimension shrinking. As a result, the use of alternative semiconductor materials has again become of interest for FE T s. However, the native oxides (As-O and Ga-O) of these materials have been shown for more than thirty years to be of poor quality for metal-oxide-semiconductor (MOS) device performance. Furthermore, deposition of any gate oxide onto a clean III-V surface results in the oxidation of the substrate to detrimental effects. Despite the extensive research of III-V materials, there is still much to be understood about these oxides. In particular, the individual oxidation states of As (5+ and 3+) and Ga (3+ and 1+) are rarely considered despite evidence that they are quite different in forming defect states.

Recent work [1,2,3] will be presented on the detection and control of each of these surface oxidations states through carefully managed interfacial reactions and depositions on GaAs and InGaAs. The fabrication of MOS capacitors and FETs with these studied interfaces has led to a correlation between the spectroscopy and electrical measurements. An emphasis on controlling or eliminating each oxidation state through a variety of techniques has allowed for a detailed understanding of these native oxides and how each one affects device performance. The presence of the Ga 1+ oxidation state is spectroscopically detected for the first time at these interfaces and a dramatic increase in device performance is demonstrated by controlling the Ga 3+ surface concentration. This work is supported by the FCRP Materials, Structures, and Devices (MSD) Center, SEMATECH, FUSION funded by System IC 2010 (COSAR), The Texas Enterprise Fund, and NIST, Semiconductor Electronics Division.

[1] Hinkle et al., APL 94, 162101 (2009).

[2] Sonnet et al., APL. 93 122109 (2008).

[3] Hinkle et al., IEEE EDL 30, 316 (2009).

The formation of a semi-ordered oxide passivation layer between hafnium oxide and In_0.53Ga_0.47As(001)-(4x2)/c(8x2) and InAs(001)-(4x2)/c(8x2) was studied using scanning tunneling microscopy/spectroscopy (STM/STS), and density functional theory (DFT) calculations. A passivation layer is needed to protect the surface from disruption during bulk amorphous oxide deposition for a high-κ gate insulator. Two methods of forming low coverage of HfO₂ were investigated: reactive oxidation of the e-beam deposited Hf metal and e-beam deposition from an HfO₂ target. STM results show that Hf atoms must cluster to be reactive to O₂. DFT suggests there is a high tendency for Hf to displace substrate atoms, which is undesirable. Direct deposition of the oxide is a better method. At submonolayer coverage, STM has identified individual bonding sites for the HfO₂ molecule; the HfO₂ forms small structures of mostly monolayer height with a high nucleation density. Density functional theory (DFT) calculations have been employed to assign the bonding structure. The DFT simulations show that for HfO₂/InAs(001)-(4x2), the most likely sites are stable by about -4.5 eV and the calculated density of states (DOS) shows no evidence of Fermi level pinning (no mid-gap states). At submonolayer coverage, the HfO₂ molecule bonds via group III-oxygen bonds and group V-hafnium bonds. STS measurements of clean InGaAs(001)-(4x2) reveal that the surface has significant band bending, showing p-type character for both n-type and p-type samples. Deposition of > 1 ML of HfO₂ is enough to move the Fermi level towards the conduction band for n-type InGaAs(001)-(4x2), as shown in results of STS vs. HfO₂ coverage. For p-type material, the Fermi level remains near the valence band after deposition of HfO₂. These results are consistent with the Fermi level remaining unpinned. In addition, annealing effects are studied. At temperatures of 300 °C and above, ordered oxide structures are seen in STM which form rows in the [-110] direction. However, lower annealing temperatures of 200 °C and below are preferable for good STS results. Hafnium oxide, evaporated via electron beam deposition, likely creates some O₂ and HfO, which may react in an undesirable way with the semiconductor surface. For this reason, a method is also proposed to protect the surface during e-beam deposition via a CO₂ protecting layer at low temperature (90 K), which does not appear to perturb the surface.

A combined scanning tunneling spectroscopy (STS) and Kelvin probe force microscopy (KPFM) study was performed on InAs(001)-(4×2) to elucidate the surface electronic properties since this surface is a leading candidate for III-V MOSFETs. KPFM provides higher energy resolution than STS, which is critical for studies of materials like InAs which have small band gaps (Eg=0.354 eV). Amplitude modulation (AM) mode KPFM provides especially high energy resolution (10 meV) and is free of tip induced band bending because of the low applied voltage (70 meV). STS spectra of InAs(001)-(4×2) consistently show pinning, with the surface Fermi level near the conduction band for both n-type and p-type samples even on nearly defect free surfaces. Using KPFM, the work functions for both n-type and p-type clean InAs(001)-(4×2) surfaces is 4.3 eV consistent with surface pinning. Using the electron counting rule, indium dimers in the trough are sp² hybridized having a completely empty dangling bond. However the indium dimer atoms are positioned in a sp³ tetrahedral configuration; these strained bonding sites might be responsible for the pinning. If the pinning is solely due to the (4×2) reconstruction, it is expected that other reconstructions without indium dimers would be unpinned. KPFM was also performed on cross-sectional InAs, which is defect free and has no indium dimers. These results were consistent with the bulk values with a work function difference between n-type and p-type of 0.49 eV. By eliminate the buckled indium dimer states with a passivation layer the InAs(001)-(4×2) surface could become unpinned.

Self-assembled III-V nanowire heterostructures could be key components in many future optoelectronic devices [1], for example solar cells [2]. To realize photovoltaics from these structures variable p- and n-type doping along the nanowires are a fundamental prerequisite. The active component in solar cells, the pn-junction, has been grown axially in InP nanowires with different p- and n-type doping levels. However determining specific doping levels, effects of the nanowire surfaces and junction abruptness and band alignment across the interface with any precision is very difficult. Recently we have shown that scanning tunneling spectroscopy on nanowires with high resolution is possible [3] and we are now combining this with synchrotron based photoemission methods.

We have examined InP nanowires with up to two axial pn-junctions with Spectroscopic PhotoEmission and Low Energy Electron Microscopy (SPELEEM), X-ray Photoelectron Spectroscopy (XPS) as well as Scanning Tunneling Microscopy/Spectroscopy (STM/S). These techniques have given us the possibility to probe not only the structure of the nanowires but also the electrical properties (such as doping level) with high lateral resolution.

With our different setups we can probe the local density of states, atomic scale structure and work function differences along the wires. We can clearly distinguish between the different n- and p-type parts of the nanowires with both the scanning probe as well as with the synchrotron radiation based techniques. Both surface and the inner regions of the wires can to some extent be probed by varying photon energies in SPELEEM or modifying the surface for STM. This gives us the opportunity to understand the device at many different levels and improve its future quality.

[1]. L. Samuelson et al., Physica E 2004, 21, 560-567

[2]. M. T. Borgström et al., Nanotechnology 2008, 19, 445602

Finding a good passivant for Ge surface is critical for fabricating a Ge-based MOSFET device. Recent studies have shown that GeON or GeO₂ interfacial layers can partially passivate the Ge/high-k dielectric interface and improve the electrical properties of the device. Introducing N (GeO_xN_y or Ge₃N₄) suppresses the Ge outdiffusion from the passivation layer into the high-k oxide layer at elevated temperatures, thereby reducing the post annealing density of interface states between Ge and high-k gate oxide. To minimize the density of interface states, the GeO_xN_y or Ge₃N₄ must be formed with a minimal dangling bond density, which is challenging in a thermal oxidation or nitridation process. To investigate the bonding and electronic structures of Ge-N and Ge-O surface species, in-situ scanning tunneling microscopy (STM) and scanning tunneling spectroscopy (STS) experiments were performed after oxidation and nitridation. Direct nitridation was carried out on Ge(100) using an electron cyclotron resonance plasma source, both at room temperature and at 500^oC. The nitridation at room temperature generated nitride sites, O sites (from trace water) and Ge adatoms which pin the surface Fermi level. The Ge adatoms are created because both O and N displace Ge surface atoms in order to bond at high coordination sites. These Ge adatoms can be removed by high temperature annealing. Nitridation at 500^oC produced a highly ordered Ge-N structure on the surface without O sites or Ge adatoms, but the Fermi level of the n-type surface was still pinned near the valence band probably due to the surface defects caused by plasma damage. Oxidation of Ge(100) was studied using a differentially pumped H₂O dosing system and the results were compared with our previous study on O₂ dosing of Ge(100). The H₂O dosed surface showed dark –OH adsorption sites with very few Ge adatoms, while the O₂ dosed surface had the equal densities of Ge adatoms and O sites. Annealing the H₂O/Ge(100) surface to 300^oC induces formation of bright Ge oxide sites which are slightly taller than Ge adatoms. However, both H₂O and O₂ dosing form GeO sites which are observed in STS to pin the Fermi Level. DFT calculations are being performed to determine the ordered nitride structure. In addition, the e-beam or ALD deposition of Ge₃N₄ or GeO₂ are being studied since they may form passivation layers without Ge displacement, plasma damage, and GeO, thereby unpinning the Fermi level.