Nano-Cleaning of Ge(100) Surface: A STM Study

Abstract

Recently there has been a major quest for new materials with high carrier mobility to substitute for silicon in CMOS semiconductor devices. Since Ge has higher hole and electron mobility compared to silicon, it is a good candidate for development of a new channel material. One of the obstacles in using Ge as a channel material is the high interface trap density between Ge and Ge native oxide. Air exposed Ge surfaces have a high density of defects and contaminants, but, in order to make optimal semiconductor devices, nearly perfect bonding between each unit cell and the gate oxide layer is required. Although there are many methods available for cleaning the Ge surface, the effectiveness of each of these methods highly depends on the cleanliness of the processing chambers. After cleaning, the Ge surface is typically functionalized with OH groups via water (H2O) or hydrogen peroxide (HOOH) during atomic layer deposition of the gate oxide. This OH functionalized surface ideally provides a high density of reactive sites for precursor nucleation. We have studied the effect of a very small amount of hydrocarbon in the processing chambers, and its effect on both the clean Ge surface and the OH functionalized surface since this may increase the density of interface traps and limit Equivalent Oxide Thickness (EOT) scaling. In-situ cleaned Ge surfaces as well as HOOH dosed surfaces have been studied after exposure to hydrocarbon contaminants with x-ray electron spectroscopy (XPS) and scanning tunneling microscopy STM). An Argon ion source sputtering system was employed for in-situ cleaning of the Ge surface. After exposure to trace hydrocarbon contaminants, two different nanoscale features were observed by STM on the Ge and HOOH/Ge surfaces. Figure 1 shows one type of contamination denoted as carbon nanoclusters. Nanocluster features were observed on the clean Ge surface and were strongly adherent. The Ge surface was exposed to trace hydrocarbon in the load lock chamber with base pressure of 2x10-8 Torr at 25C and transferred to UHV chamber for sputter cleaning and subsequent annealing at 700°c. The nominally clean Ge surface has 2% coverage of nanoclusters on the surface. The XPS studies confirm that these features are carbon. These carbon nanoclusters are about 0.3-0.5nm in height and 2-4nm in diameter. As shown in Fig 2, a distinctly different feature is observed on the Ge-OH terminated surface denoted as carbon nanoflakes. In contrast to nanoclusters, nanoflakes were only observed on the Ge surfaces dosed with hydrogen peroxide. To form the Ge-OH terminated surface, clean Ge was dosed with the vapor from 30% HOOH/H2O for 45 sec at 25C in the load lock with a base pressure of 2x10-8 Torr. There is about 6% coverage of nanoflakes on the hydrogen peroxide dosed Ge surface. These nanoflakes are about 0.3- 0.5 nm in height and 2- 4nm in diameter. There are two possible mechanism of nanocluster and nanoflake formation [1, 2]: (a) deposition of nanoparticles from the gas phase or (b) sticking of hydrocarbon molecules from the gas phase and particle formation on the surfaces. The distinct size distribution of the particles on the clean Ge(100) vs HOOH/Ge(100) is most consistent with the carbon nanoclusters and nanoflakes forming via molecular hydrocarbon deposition with subsequent particle formation on the surface. The results show the need for fast ALD reactions to insure that after in-situ cleaning and Ge-OH functionalization, the surface is only briefly exposed to background hydrocarbon.