Interface engineering for Ge metal-oxide–semiconductor devices

Abstract

High mobility semiconductors such as Ge with high-k gates may be required to enhance performance of future devices. One of the biggest challenges for the development of a Ge metal-oxide–semiconductor (MOS) technology is to find appropriate passivating materials and methodologies for the Ge/high-k interfaces. Germanium oxynitride is frequently used as a passivating interlayer in combination with HfO₂ and is found to be necessary for the fabrication of functional devices. However, it is also considered to be insufficient since electrical characteristics in capacitors are non-ideal and field effect transistors underperform, probably due to a the high density of interface defects. We show that alternative passivating rare earth oxide layers prepared by molecular beam deposition produce improved electrical characteristics and a significant reduction of the density of interface states. In the case of CeO₂, a thick interfacial layer is spontaneously formed containing oxidized Ge, which is considered to be the key for the observed improvements.

Introduction

Metal oxide high-k dielectrics may replace SiO₂ in the gate of future nano-electronic devices. In addition, high mobility semiconductors such as Ge [1] may replace Si in the active channel to enhance device performance. The interface between high-k materials and Ge determines critical performance characteristics of MOS capacitors and transistors. For aggressive oxide scaling with equivalent oxide thickness (EOT) well below 1 nm, avoiding interfacial layers is highly desirable. However, very often, an interfacial layer (IL) is absolutely necessary to passivate the surface. Surface treatment by ammonia is beneficial to prevent Ge diffusion [2] and improve electrical characteristics [3] in Atomic Layer Deposition (ALD) HfO₂. Also, it has been found [4] that electrical properties are improved by using O₃ precursor in ALD HfO₂. The spontaneous formation of a GeO_x IL [4] may be partly responsible for the observed improvement. These works [2], [3], [4] show that ILs formed either spontaneously or artificially benefit electrical response; on the other hand, they may present an obstacle for oxide scaling. It has been shown that Ge oxynitride ILs can be controlled to very small thickness [2], [5], which is a good sign for scaling. Indeed, works by ultra-high vacuum (UHV) ozone oxidation of ZrO₂ [6] or molecular beam deposition (MBD) of HfO₂ [5], [7] have shown that these dielectrics combined with ultrathin Ge oxynitride ILs, are scalable to EOT values well below 1 nm with low gate leakage, below 10^− 3 A/cm² at V_FB-1V and a high-k value ∼ 25 [5]. Moreover, Ge p-channel field-effect-transistors (FETs) with MBD GeON/HfO₂ gates have been demonstrated which show EOT much less than 1 nm [8] and remarkably high I_ON/I_OFF ratio ∼ 10⁵. The general consensus is that a passivating GeO_x or GeO_xN_y layer is necessary prior to HfO₂ MBD to obtain functional capacitor and transistor devices but it does not provide sufficient passivation of Ge interface. As a result, the current–voltage (CV) characteristics are non-ideal [5] exhibiting large hysteresis and frequency dispersion mainly in inversion [9], [10] and high interface density of states D_it, typically around 5 × 10¹²–10¹³ eV^− 1 cm^− 2, which cannot be cured even after H₂ or other post deposition annealing treatment. In addition, channel mobility in Ge FETs [8], is lower than expected, especially in n-channel FETs. This is a serious drawback attributed, in part, to the poor quality of the GeON passivating layer. It is generally believed that oxidation of Ge is the primary source of defects. Therefore, the effort is focused on alternative passivating materials like GeN, Si [11] or AlN [12] ultrathin layers. The main motivation is to isolate the Ge substrate from the high-k dielectric with an oxidation barrier which will also act as a diffusion barrier prohibiting Ge oxidation as well as Ge and metal ion inter-diffusion.

Here, we provide new evidence that oxidation of Ge is not necessarily harmful. Using a variety of analytical methods such as X-ray photoelectron spectroscopy (XPS), time-of-flight secondary ion mass spectroscopy (ToF-SIMS) and high resolution transmission electron microscopy (HRTEM) we have confirmed that CeO₂ deposited directly on Ge by MBD reacts strongly with the substrate producing interfacial layers which contain oxidized Ge. Despite Ge oxidation, these layers produce metal–insulator–semiconductor (MIS) capacitors with improved CV characteristics, as for example very weak frequency dispersion in depletion and inversion, small hysteresis and reduced D_it around 1 × 10¹² eV^− 1 cm^− 2. Knowing that CeO₂ is a good catalyst, spontaneous formation of GeO_x or Ce–Ge–O interfacial layers is not surprising. However, whether CeO₂-assisted catalytic oxidation of Ge plays a role in the formation of high quality passivating layers is an open question. Our work shows that other rare earth oxides also give good quality passivating layers. We investigate ceria as a model system to understand the chemistry at the interfaces with Ge, which may help us applying our methodology to other, potentially better rare earth oxide candidates.