pMOS transistor with embedded SiGe: Elastic and plastic relaxation issues A. Hikavyy a, ⁎, N. Bhouri a , R. Loo a , P. Verheyen a , F. Clemente a , J. Hopkins b , R. Trussell b , M. Caymax a a IMEC, Kapeldreef 75, 3001 Leuven, Belgium b Bede X-ray metrology, Belmont Business Park Durham DH1 1TW, UK abstract article info Available online 22 August 2008 Keywords: SiGe Relaxation pMOS Local SiGe layer thickening next to spacers in the embedded planar Si pMOS transistors showing an improved performance comparing to Si reference was found. Such SiGe growth behavior was simulated using a mask with different window sizes and studied by various techniques. It was found that although Nomarski spectroscopy shows a decrease of misfit dislocations linear density with window size shrinkage, suggesting perfectly strained layers, high resolution X-ray diffraction (HRXRD) and Raman investigations show a dramatic increase of relaxation for windows smaller than 5 × 5 μm 2 for all investigated samples. It is suggested that this is because of local thickening at the window edges (similar to thickening next to spacers in devices), which is due to elastic relaxation caused by the convex corners of the recessed areas. © 2008 Elsevier B.V. All rights reserved. 1. Introduction Recessed strained SiGe in the source/drain regions of planar Si pMOS devices is a well-known technique to enhance pMOS drive current. Recently, we demonstrated a 40% improvement of pMOS drive current in comparison to Si reference devices. Scanning electron microscope (SEM) inspection of these devices showed a local thickening of epitaxially grown SiGe layer next to spacers (Fig. 1) which cannot be explained by loading effects, because the thickening was not seen at the STI sidewalls [1]. In some cases difference between the SiGe layer thickness next to the spacer and the target thickness can be a few tenth of nm, which is remarkable. One can easily expect relaxation in these regions especially at high Ge concentrations. This effect is not quite understood though is rather common and frequently observed. Its presence motivated us to investigate proper- ties of SiGe epilayers in areas where similar local SiGe thickening was seen. In order to do this we used a mask with different window sizes where relaxation of SiGe could be studied as well. In this article we present results of optical inspections, HRXRD, Raman and SEM investigations of SiGe selectively epitaxially grown (SEG) layers on patterned Si wafers and argue that elastic relaxation might cause local thickening of SiGe layers. 2. Experimental To simulate areas next to the spacer a mask with open Si windows sized from 270 × 270 down to 0.2 × 0.2 μm 2 with or without Si recess was used. For epitaxial layer growth we use a standard ASM Epsilon 2000 production epi reactor. H 2 is used as carrier gas. Selective Epitaxial SiGe Growth was carried out at reduced pressure (20 Torr) using Dichlor- osilane and Germane as Si and Ge source gases, respectively. HCl is added to the gas mixture to maintain selectivity. High temperature bake (1050°C) (HTB) or combination of HF dip with a low temperature bake (85 °C) (LTB) was used for native oxide removal prior SiGe deposition. All layers have a nominal Ge content of ∼ 22.5%. The layer thickness was varied between 80 and 200 nm. Window size influence on SiGe relaxation was studied by optical microscope inspection of the traces left by interfacial misfit dislocations (windows in the range of 10×5– 270 × 270 μm 2 ), HRXRD (0.3 × 0.3–270 × 270 μm 2 windows) and Raman spectroscopy (10 × 0.5–270 × 270 μm 2 ). HRXRD measurements were performed using a BedeMetrix™-L X- ray tool fitted with a Microsource™ micro-focus source, a Scribe- View™ optic and a channel-cut beam conditioning crystal. The beam cross-section at the sample position was less than 100 μm×100 μm giving a footprint of about 200 μm×100 μm for the 004 reflection. Micro-Raman experiments were performed in backscattering geometry with a Dilor XY system using the 457.9 nm laser line of an argon ion laser with an output power of 20 mW. The spot size on the sample was ∼ 1 μm. A silicon reference sample and plasma lines from the laser were used for Raman shift calibration. The fit-error (sdev) on the frequency is about 0.02 cm - 1 . It must be pointed out that both HRXRD and Raman were measured over a number of similar windows and an average value for relaxation was deduced. 3. Results and discussion After a defect etch of samples in a Schimmel solution to reveal the misfit dislocations, their density was determined with Nomarski microscopy. We defined linear dislocation density as the number of dislocation that cross one edge of a window divided by its length. It must Thin Solid Films 517 (2008) 113–116 ⁎ Corresponding author. Tel.: +32 16 28 11 20; fax: +32 16 28 17 06. E-mail address: Andriy.Hikavyy@imec.be (A. Hikavyy). 0040-6090/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2008.08.107 Contents lists available at ScienceDirect Thin Solid Films journal homepage: www.elsevier.com/locate/tsf