Quantification of dopants in nanomaterial by SEM / EDS

Abstract

It is long known that doping is a key element in the development of modern semiconductor technology for applications in electronic, nano-electronics, optoelectronics and photonics. Doping allows modifications in the electrical conductivity of semiconductors that depend on the type, quantity, distribution and activity of added dopants. Characterizing doping is therefore essential for the understanding and improvement of electrical and optical properties of semiconductors in order to produce reliable and performant electronic and optical devices. The increasingly reduced dimensions of semiconducting devices as well as the development of new nanomaterial-based devices require the characterization (dopant type and activity) and quantification (concentration and spatial distribution) of low-levels of dopants at the nano-scale. This represents a real challenge that is not fully achieved today by existing techniques which are globally expensive, time-consuming, difficult to implement in an industrial context (STEM, APT, electron holography), spatially unresolved (SIMS), not very sensitive/accurate as yet (EDS and EELS in TEM) and, for most of them, not fully quantitative. To date, the state of the art for conventional TEM/EDS (Electron Dispersive Spectrometry in Transmission Electron Microscope) is limited to the detection of a few dopants (As, P) in a few materials (Si, SiGe) with detection limit around 1020 at/cm3 and precision of ±20-30% (Sevanton et al., 2009). We report here the detection and quantification by SEM/EDS (SEM: Scanning Electron Microscope) of dopant concentrations as low as 5 1018 at cm-3 with a precision and detection limit around 1018 at cm-3. Such a large improvement in detection sensitivity can be achieved at low voltage (< 8 kV) using the experimental Flat Quad 5060F annular detector from Bruker that equips the Ultra55 Zeiss SEM of the Minatec's PFNC (INAC). This detector belongs to the new generation of silicon drift detectors (SDD) which are composed of four bean-shaped silicon diodes arranged in a ring around a central hole for the electron beam passage (Fig. a). It is positioned a few millimeters above the sample (Fig. b), a geometry which results in a much wider solid angle (up to 1.2 sr) compared to traditional detectors (<0.1 sr), allowing a higher counting rate at any operating conditions (up to 1000 kcps). The major difficulty for quantifying low level concentrations by EDS (even at low voltage) remains to extract a low intensity signal from a relatively high background signal. This is particularly true for Mg in GaN, the K-line (1.25 keV) of this element being very close to the L-line (1.19 keV) of Ga (Figs. c-d). To overcome this problem, a new method has been developed which is based on two innovations: 1) the use of specific windows that act as X-ray filters, allowing a large enhancement of the signal to noise ratio in the energy range corresponding to the X-ray line of the analyzed dopants, and 2) the development of a new analytical procedure for removing background based on spectrum normalization to pure reference spectrum (Fig. e). Results obtained on P-doped Ge 2D layers and Mg-, Si-doped GaN 2D layers show good consistency with SIMS analyses, even for the lowest concentrations of dopants (see Fig. f for Mg dopant in GaN). The technique was applied for quantifying low level of dopants (down to 1019 at/cm3) in various types of Ge, GaN, and AlGaN nanomaterials. Results will be presented and discussed at the conference.