AVS2009 Session EM-WeA: Contacts, Interfaces, and Defects in Semiconductors

GaN devices promise advantages over other compound semiconductors including higher power amplification, increased linearity, and less temperature dependent degradation. But today current collapse, slow switching transients, and poor gain profiles are present and may be related to interface trap densities. To shed some light, we analyzed the interfaces found in GaN HEMTs and correlated results with amplifier performance characteristics and first principle atomic and electronic structure simulations.

MOCVD grown GaN Schottky diodes were mesa isolation etched, and KOH etched to remove Ga residue and surface defects. Then Ti/Al/Ti/Au Ohmic and Ni/Au Schottky contacts were deposited and annealed.

These blocked 340V in the off-state. KOH reduced surface roughness and improved the on-state performance (150mA at 3.5V in good devices). The undoped GaN layer had a free carrier concentration of 5x10¹⁶ cm^-3 from CV measurements. The reverse bias exhibits a soft breakdown due to an initial depletion followed by a slower field spreading. For three devices, the forward (2V) and reverse leakage (-1 V) currents were (A) 45mA, 3.4x10^-10A, (B) 17mA, 2.0x10^-9A, and (C) 4 mA, 3.6x10^-8A. Comparing the ideality factors for the three devices over a voltage range of 0 to 0.5 V, device (A) exhibited no bumps between 0.1 and 0.4 V (n = 1.04 at 0.2 V) and a smooth transition into the series resistance region, device (B) exhibited a bump at 0.4V with a peak n= 4.3. Device (C) exhibited a bump at 0.1V with a peak n=2.08.

These ideality factor bumps were seen with GaAs Schottky diodes and linked to interface trap densities. Our conductance measurements gave corresponding trap densities 0.3 eV from the band edge of (C) 1.02x10¹², and (B) 7.9x10¹¹ cm^-2 eV^-1, and for (A) much less. Also, the interface trap time constants are (C) 97 ns and (B) 81.9 ns and for (A) much longer.

Our first-principle simulation model of a planar bonded metal-semiconductor interface included an inhomogeneity at the interface and semiconductor interface defect interactions with H atoms and OH radicals during KOH etching. We assume that N antisite defects are common defects in GaN causing point defect pinning. After KOH etching these defects are passivated by two hydrogen atoms forming complexes H-N(Ga)-H and OH-N(Ga)-OH. The complexes do not have dangling bonds and do not participate in hybridization with extended states of a metal. Formation of the complexes could improve the interface state associated electrical properties of the Schottky diode, smooth the GaN (0001) surface and remove oxides residues.

Vendor provided GaN HEMTs were similarly analyzed and will be discussed.

A combination of simulations and experiments often provides a powerful approach to address scientific and engineering problems. In this presentation, we describe two examples of research on contacts to semiconductors in which we use simulations to develop methods to analyze data collected from Schottky barrier contacts. In the first example, we present an approach to accurately extract the Schottky barrier height from an axial contact to a semiconductor nanowire. Modification of the method usually used to analyze current-voltage (I-V) data from microscale planar Schottky diodes is necessary because of the influence of the semiconductor nanowire surface adjacent to the Schottky barrier contact. Band-bending at this surface influences the shape of the depletion region at the metal/semiconductor interface and must be controlled. We accomplish this control with a wrap-around gate. By analyzing I-V data generated using a commercial device simulator, we identify a method to treat the data to accurately extract the Schottky barrier height. Using the same approach that worked well for the simulated data, we next analyze Schottky barrier contacts to silicon nanowires that we nanofabricated with wrap-around gates. In another investigation, we examine approaches to extract the Richardson constant from planar Schottky barrier contacts that contain nanoscale inhomogeneities. Using Tung’s model for inhomogeneous Schottky barrier contacts, we generate simulated I-V data for contacts with inhomogeneities of different sizes, densities, and departures from the homogeneous background barrier height. Then we compare various published approaches for extracting the Richardson constant, identifying their strengths and weaknesses. This section of the presentation concludes with the recommendation of a new approach for treating experimental data, and the approach is demonstrated for Schottky barrier contacts to wide band gap semiconductors.

A major challenge in Nanotechnology is the incorporation of single nano-devices into larger integrated circuits. Although work on individual (and non-integrated) nano-structures such as molecules is intense, the question of their electrical connection with more than two probes (such as conventional SPM experiments) remains an open question. Established nano-lithography techniques such as EBL and FIB seem to not satisfy requirements for ultra clean and defined contact structures at the atomic scale. Traditional instrumentation for analysis is fundamentally limited: How to cover the dimensional range of an integrated circuit (mm) down to the atomic scale of a single molecule device and at the same time to have an adequate integrated navigation system? To meet these requirements, we have established and being advancing a new approach integrating SPM technology with high resolution electron microscopy: (1) Bridging dimensions by combined SEM (down to below 3nm resolution) and STM operation at the atomic scale; (2) Rapid SEM navigation of four local STM probes; (3) Individual probe fine positioning by high resolution STM imaging; (4) STM based probe approach for "soft-landing" of sharp and f-ragile probes and controlled electrical contact for transport measurements. To open a route for fundamental evaluation of the potential of single molecule devices, this instrumental technology is employed to establish electrical connection for local transport measurements. As a model system, we have chosen Au nano-islands on MoS₂. These islands represent contact pads, each electrically connected by an individual STM probe. As good band gap (approx. 1.3eV transverse gap) semiconductor, MoS₂ has the potential to sufficiently decouple those nano-structures electrically at low voltage. Those Au triangular nano-islands have a lateral size of typically 10-30nm and form an "atomically" ultra clean and defined metal-semiconductor interface. We present measurements that prove (1) SEM based navigation and STM based electrical contacting with a tip radius in the 10nm range: (2) reproducible Schottky like IV properties for the individual STM tip/Au nano-island/substrate contact; (3) surface conductance measurements with variable inter-island distance down to 17nm; (4) comparison with surface conductance measurements of the bare MoS2 substrate. We also show that the individual STM probe can be employed under SEM to manipulate those Au nano-islands [1] with high precision in order to generate arbitrary multi probe planar contact configurations.

[1]: J.S. Yang, D. Jie, N. Chandrasekar and C. Joachim J. Vac. Sci. Tech. B, 25, 1694 (2007).