Investigation of electrically active defects at the interface of high-k dielectrics and compound semiconductors

As silicon based devices in integrated circuits reach the fundamental limits of dimensional scaling there is growing research interest in the use of high electron mobility channel materials, such as indium gallium arsenide (InGaAs), in conjunction with high dielectric constant (high-k) gate oxides, for Metal-Oxide-Semiconductor Field Effect Transistor (MOSFET) based devices. The motivation for employing high mobility channel materials is to reduce power dissipation in integrated circuits while also providing improved performance. One of the primary challenges to date in the field of III-V semiconductors has been the observation of high levels of defect densities at the high-k/III-V interface, which prevents surface inversion of the semiconductor. The work presented in this PhD thesis details the characterization of MOS devices incorporating high-k dielectrics on III-V semiconductors. The analysis examines the effect of modifying the semiconductor bandgap in MOS structures incorporating InxGa1-xAs (x: 0, 0.15. 0.3, 0.53) layers, the optimization of device passivation procedures designed to reduce interface defect densities, and analysis of such electrically active interface defect states for the high-k/InGaAs system. Devices are characterized primarily through capacitance-voltage (CV) and conductance-voltage (GV) measurements of MOS structures both as a function of frequency and temperature. In particular, the density of electrically active interface states was reduced to the level which allowed the observation of true surface inversion behavior in the In0.53Ga0.47As MOS system. This was achieved by developing an optimized (NH4)2S passivation, minimized air exposure, and atomic layer deposition of an Al2O3 gate oxide. An extraction of activation energies allows discrimination of the mechanisms responsible for the inversion response. Finally a new approach is described to determine the minority carrier generation lifetime and the oxide capacitance in MOS structures. The method is demonstrated for an In0.53Ga0.47As system, but is generally applicable to any MOS structure exhibiting a minority carrier response in inversion.

Simulation of multigate SOI transistors with silicon, germanium and III-V channels

In this work by employing numerical three-dimensional simulations we study the electrical performance and short channel behavior of several multi-gate transistors based on advanced SOI technology. These include FinFETs, triple-gate and gate-all-around nanowire FETs with different channel material, namely Si, Ge, and III-V compound semiconductors, all most promising candidates for future nanoscale CMOS technologies. Also, a new type of transistor called “junctionless nanowire transistor” is presented and extensive simulations are carried out to study its electrical characteristics and compare with the conventional inversion- and accumulation-mode transistors. We study the influence of device properties such as different channel material and orientation, dimensions, and doping concentration as well as quantum effects on the performance of multi-gate SOI transistors. For the modeled n-channel nanowire devices we found that at very small cross sections the nanowires with silicon channel are more immune to short channel effects. Interestingly, the mobility of the channel material is not as significant in determining the device performance in ultrashort channels as other material properties such as the dielectric constant and the effective mass. Better electrostatic control is achieved in materials with smaller dielectric constant and smaller source-to-drain tunneling currents are observed in channels with higher transport effective mass. This explains our results on Si-based devices. In addition to using the commercial TCAD software (Silvaco and Synopsys TCAD), we have developed a three-dimensional Schrödinger-Poisson solver based on the non-equilibrium Green’s functions formalism and in the framework of effective mass approximation. This allows studying the influence of quantum effects on electrical performance of ultra-scaled devices. We have implemented different mode-space methodologies in our 3D quantum-mechanical simulator and moreover introduced a new method to deal with discontinuities in the device structures which is much faster than the coupled-mode-space approach.

Nanoporous InP: anodic formation and growth mechanism in aqueous electrolytes

Porous layers can be formed electrochemically on (100) oriented n-InP substrates in aqueous KOH. A nanoporous layer is obtained underneath a dense near-surface layer and the pores appear to propagate from holes through the near-surface layer. In the early stages of the anodization transmission electron microscopy (TEM) clearly shows individual porous domains that appear to have a square-based pyramidal shape. Each domain appears to develop from an individual surface pit which forms a channel through this near-surface layer. We suggest that the pyramidal structure arises as a result of preferential pore propagation along the <100> directions. AFM measurements show that the density of surface pits increases with time. Each of these pits acts as a source for a pyramidal porous domain. When the domains grow, the current density increases correspondingly. Eventually the domains meet, forming a continuous porous layer, the interface between the porous and bulk InP becomes relatively flat and its total effective surface area decreases resulting in a decrease in the current density. Current-time curves at constant potential exhibit a peak and porous layers are observed to form beneath the electrode surface. The density of pits formed on the surface increases with time and approaches a plateau value. Porous layers are also observed in highly doped InP but are not observed in wafers with doping densities below ~5 × 1017 cm-3. Numerical models of this process have been developed invoking a mechanism of directional selectivity of pore growth preferentially along the <100> lattice directions. Manipulation of the parameters controlling these curves shows that the fall-off in current is controlled by the rate of diffusion of electrolyte through the pore structure with the final decline in current being caused by the termination of growth at the pore tips through the formation of passivating films or some other irreversible modification of the pore tips.

Anodic behavior of InP: film growth, porous structures and current oscillations.

We review our recent work on the anodization of InP in KOH electrolytes. The anodic oxidation processes are shown to be remarkably different in different concentrations of KOH. Anodization in 2 - 5 mol dm-3 KOH electrolytes results in the formation of porous InP layers but, under similar conditions in a 1 mol dm-3 KOH, no porous structure is evident. Rather, the InP electrode is covered with a thin, compact surface film at lower potentials and, at higher potentials, a highly porous surface film is formed which cracks on drying. Anodization of electrodes in 2 - 5 mol dm-3 KOH results in the formation of porous InP under both potential sweep and constant potential conditions. The porosity is estimated at ~65%. A thin layer (~ 30 nm) close to the surface appears to be unmodified. It is observed that this dense, near-surface layer is penetrated by a low density of pores which appear to connected it to the electrolyte. Well-defined oscillations are observed when InP is anodized in both the KOH and (NH4)2S. The charge per cycle remains constant at 0.32 C cm-2 in (NH4)2S but increases linearly with potential in KOH. Although the characteristics of the oscillations in the two systems differ, both show reproducible and well-behaved values of charge per cycle.

Pitting and porous layer formation on n-InP anodes

Surface pitting occurs when InP electrodes are anodized in KOH electrolytes at concentrations in the range 2 - 5 mol dm-3. The process has been investigated using atomic force microscopy (AFM) and the results correlated with cross-sectional transmission electron microscopy (TEM) and electroanalytical measurements. AFM measurements show that pitting of the surface occurs and the density of pits is observed to increase with time under both potentiodynamic and potentiostatic conditions. This indicates a progressive pit nucleation process and implies that the development of porous domains beneath the surface is also progressive in nature. Evidence for this is seen in plan view TEM images in which individual domains are seen to be at different stages of development. Analysis of the cyclic voltammograms of InP electrodes in 5 mol dm-3 KOH indicates that, above a critical potential for pit formation, the anodic current is predominantly time dependent and there is little differential dependence of the current on potential. Thus, pores continue to grow with time when the potential is high enough to maintain depletion layer breakdown conditions.

Formation and characterization of porous InP layers in KOH Solutions

Porous InP layers were formed electrochemically on (100) oriented n-InP substrates in various concentrations of aqueous KOH under dark conditions. In KOH concentrations from 2 mol dm-3 to 5 mol dm-3, a porous layer is obtained underneath a dense near-surface layer. The pores within the porous layer appear to propagate from holes through the near-surface layer. Transmission electron microscopy studies of the porous layers formed under both potentiodynamic and potentiostatic conditions show that both the thickness of the porous layer and the mean pore diameter decrease with increasing KOH concentration. The degree of porosity, estimated to be 65%, was found to remain relatively constant for all the porous layers studied.

Comparison of oscillatory behavior on InP electrodes in KOH solutions

The observation of current oscillations under potential sweep conditions when an n-InP electrode is anodized in a KOH electrolyte is reported and compared to the oscillatory behavior noted during anodization in an (NH4)2S electrolyte. In both cases oscillations are observed above 1.7 V (SCE). The charge per cycle was found to increase linearly with potential for the InP/KOH system but was observed to be independent of potential for the InP/(NH4)2S system. The period of the oscillations in the InP/KOH was found to increase with applied potential. In this case the oscillations are asymmetrical and the rising and falling segments have a different dependence on potential. Although the exact mechanism is not yet know for either system, transmission electron microscopy studies show that in both cases, the electrode is covered by a thick porous film in the oscillatory region.

Anodic oxidation of InP in KOH electrolytes

The anodic behavior of InP in 1 mol dm-3 KOH was investigated and compared with its behavior at higher concentrations of KOH. At concentrations of 2 mol dm-3 KOH or greater, selective etching of InP occurs leading to thick porous InP layers near the surface of the sustrate. In contrast, in 1 mol dm-3 KOH, no such porous layers are formed but a thin surface film is formed at potentials in the range 0.6 V to 1.3 V. The thickness of this film was determined by spectroscopic ellipsometry as a function of the upper potential and the measured film thickness corresponds to the charge passed up to a potential of 1.0 V. Anodization to potentials above 1.5 V in 1 mol dm- 3 KOH results in the growth of thick, porous oxide films (~ 1.2 µm). These films are observed to crack, ex-situ, due to shrinkage after drying in ambient air. Comparisons between the charge density and film thickness measurements indicate a porosity of approximately 77% for such films.

Electrochemical pore formation on InP in alkaline solutions

The surface properties of InP electrodes were examined following anodization in (NH4)2S and KOH electrolytes. In both solutions, the observation of current peaks in the cyclic voltammetric curves was attributed to selective etching of the substrate and a film formation process. AFM images of samples anodized in the sulfide solution, revealed surface pitting and TEM micrographs revealed the porous nature of the film formed on top of the pitted substrate. After anodization in the KOH electrolyte, TEM images revealed that a porous layer extending 500 nm into the substrate had been formed. Analysis of the composition of the anodic products indicates the presence of In2S3 in films grown in (NH4)2S and an In2O3 phase within the porous network formed in KOH.

Growth and characterization of anodic Films on InP in KOH and (NH4)2S

The current-voltage characteristics of InP were investigated in (NH4)2S and KOH electrolytes. In both solutions, the observation of current peaks in the cyclic voltammetric curves was attributed to the growth of passivating films. The relationship between the peak currents and the scan rates suggests that the film formation process is diffusion controlled in both cases. The film thickness required to inhibit current flow was found to be much lower on samples anodized in the sulphide solution. Focused ion beam (FIB) secondary electron images of the surface films show that film cracking of the type reported previously for films grown in (NH4)2S is also observed for films grown in KOH. X-ray and electron diffraction measurements indicate the presence of In2O3 and InPO4 in films grown in KOH and In2S3 in films grown in (NH4)2S.

Self-healing thermal annealing: surface morphological restructuring control of GaN nanorods

With advances in nanolithography and dry etching, top-down methods of nanostructuring have become a widely used tool for improving the efficiency of optoelectronics. These nano dimensions can offer various benefits to the device performance in terms of light extraction and efficiency, but often at the expense of emission color quality. Broadening of the target emission peak and unwanted yellow luminescence are characteristic defect-related effects due to the ion beam etching damage, particularly for III–N based materials. In this article we focus on GaN based nanorods, showing that through thermal annealing the surface roughness and deformities of the crystal structure can be “self-healed”. Correlative electron microscopy and atomic force microscopy show the change from spherical nanorods to faceted hexagonal structures, revealing the temperature-dependent surface morphology faceting evolution. The faceted nanorods were shown to be strain- and defect-free by cathodoluminescence hyperspectral imaging, micro-Raman, and transmission electron microscopy (TEM). In-situ TEM thermal annealing experiments allowed for real time observation of dislocation movements and surface restructuring observed in ex-situ annealing TEM sampling. This thermal annealing investigation gives new insight into the redistribution path of GaN material and dislocation movement post growth, allowing for improved understanding and in turn advances in optoelectronic device processing of compound semiconductors.

Dilute nitride semiconductors : band structure, scattering and high field transport

The substitution of a small fraction x of nitrogen atoms, for the group V elements in conventional III-V semiconductors such as GaAs and GaSb strongly perturbs the conduction band of the host semiconductor. In this thesis we investigate the effects of nitrogen states on the band dispersion, carrier scattering and mobility of dilute nitride alloys. In the supercell model we solve the single particle Hamiltonian for a very large supercell containing randomly placed nitrogen. This model predicts a gap in the density of states of GaNxAs1−x, where this gap is filled in the Green’s function model. Therefore we develop a self-consistent Green’s function (SCGF) approach, which provides excellent agreement with supercell calculations and reveals a gap in the DOS, in contrast with the results of previous non-self-consistent Green’s function calculations. However, including the distribution of N states destroys this gap, as seen in experiment. We then examine the high field transport of carriers by solving the steadystate Boltzmann transport equation and find that it is necessary to include the full distribution of N levels in order to account for the small, low-field mobility and the absence of a negative differential velocity regime observed experimentally with increasing x. Overall the results account well for a wide range of experimental data. We also investigate the band structure, scattering and mobility of carriers by finding the poles of the SCGF, which gives lower carrier mobility for GaNxAs1−x, compared to those already calculated, in better agreement with experiments. The calculated optical absorption spectra for InyGa1−yNxAs1−x and GaNxSb1−x using the SCGF agree well with the experimental data, confirming the validity of this approach to study the band structure of these materials.

A tight-binding analysis of the electronic properties of III-nitride semiconductors

This thesis divides into two distinct parts, both of which are underpinned by the tight-binding model. The first part covers our implementation of the tight-binding model in conjunction with the Berry phase theory of electronic polarisation to probe the atomistic origins of spontaneous polarisation and piezoelectricity as well as attempting to accurately calculate the values and coefficients associated with these phenomena. We first develop an analytic model for the polarisation of a one-dimensional linear chain of atoms. We compare the zincblende and ideal wurtzite structures in terms of effective charges, spontaneous polarisation and piezoelectric coefficients, within a first nearest neighbour tight-binding model. We further compare these to real wurtzite structures and conclude that accurate quantitative results are beyond the scope of this model but qualitative trends can still be described. The second part of this thesis deals with implementing the tight-binding model to investigate the effect of local alloy fluctuations in bulk AlGaN alloys and InGaN quantum wells. We calculate the band gap evolution of Al1_xGaxN across the full composition range and compare it to experiment as well as fitting bowing parameters to the band gap as well as to the conduction band and valence band edges. We also investigate the wavefunction character of the valence band edge to determine the composition at which the optical polarisation switches in Al1_xGaxN alloys. Finally, we examine electron and hole localisation in InGaN quantum wells. We show how the built-in field localises the carriers along the c-axis and how local alloy fluctuations strongly localise the highest hole states in the c-plane, while the electrons remain delocalised in the c-plane. We show how this localisation affects the charge density overlap and also investigate the effect of well width fluctuations on the localisation of the electrons.

Carrier momentum relaxation in highly doped polar semiconductors and semiconductor heterostructures

Highly doped polar semiconductors are essential components of today’s semiconductor industry. Most strikingly, transistors in modern electronic devices are polar semiconductor heterostructures. It is important to thoroughly understand carrier transport in such structures. In doped polar semiconductors, collective excitations of the carriers (plasmons) and the atoms (polar phonons) couple. These coupled collective excitations affect the electrical conductivity, here quantified through the carrier mobility. In scattering events, the carriers and the coupled collective modes transfer momentum between each other. Carrier momentum transferred to polar phonons can be lost to other phonons through anharmonic decay, resulting in a finite carrier mobility. The plasmons do not have a decay mechanism which transfers carrier momentum irretrievably. Hence, carrier-plasmon scattering results in infinite carrier mobility. Momentum relaxation due to either carrier–plasmon scattering or carrier–polar-phonon scattering alone are well understood. However, only this thesis manages to treat momentum relaxation due to both scattering mechanisms on an equal footing, enabling us to properly calculate the mobility limited by carrier–coupled plasmon–polar phonon scattering. We achieved this by solving the coupled Boltzmann equations for the carriers and the collective excitations, focusing on the “drag” term and on the anharmonic decay process of the collective modes. Our approach uses dielectric functions to describe both the carrier-collective mode scattering and the decay of the collective modes. We applied our method to bulk polar semiconductors and heterostructures where various polar dielectrics surround a semiconducting monolayer of MoS2, where taking plasmons into account can increase the mobility by up to a factor 15 for certain parameters. This screening effect is up to 85% higher than if calculated with previous methods. To conclude, our approach provides insight into the momentum relaxation mechanism for carrier–coupled collective mode scattering, and better tools for calculating the screened polar phonon and interface polar phonon limited mobility.