Engineering High-k/SiGe Interface with ALD Oxide for Selective GeO_x Reduction

Abstract

Suppression of electronic defects induced by GeO x at the high-k gate oxide/SiGe interface is critical for implementation of high mobility SiGe channels in CMOS technology. Theoretical and experimental studies have shown that a low defect density interface can be formed with an SiO xrich interlayer on SiGe. Experimental studies in literature indicates better interface formation with Al 2 O 3 in contrast to HfO 2 on SiGe however the mechanism behind this is not well understood. In this study, the mechanism of forming a low defect density interface between Al 2 O 3 /SiGe is investigated using atomic layer deposited (ALD) Al 2 O 3 insertion into or on top of ALD HfO 2 gate oxides. To elucidate the mechanism, correlations are made between the defect density determined by impedance measurements and the chemical and physical structure of the interface determined by high resolution scanning transmission electron microscopy and electron energy loss spectroscopy (STEM-EELS). Compositional analysis reveals an SiO x rich interlayer for both Al 2 O 3 /SiGe and HfO 2 /SiGe interfaces with insertion of Al 2 O 3 into or on top of the HfO 2 oxide. The data is consistent with the Al 2 O 3 insertion inducing decomposition of the GeO x from the interface to form an electrically passive, SiO x rich interface on SiGe. This mechanism shows that nanolaminate gate oxide chemistry cannot be interpreted as resulting from a simple layer by layer ideal ALD process because the precursor or its reaction products can diffuse though the oxide during growth and react at the semiconductor interface. This result shows that in scaled CMOS, remote oxide ALD (oxide ALD on top of the gate oxide) can be used to suppress electronic defects at gate-oxide semiconductor interfaces by oxygen scavenging.

Understanding the Mechanism of Electronic Defect Suppression Enabled by Nonidealities in Atomic Layer Deposition

“…17 The 75% decrease in D it for ozone dosing into Al 2 O 3 /SiGe devices is consistent with both ozone and TMA dosing reducing interfacial GeO x but using different yet complementary chemical processes. The TMA reduces GeO x by gettering the oxygen from the SiGeO x interface to form Al 2 O 3 throughout the entire ALD process, 5 and it is hypothesized that the ozone promotes GeO x outdiffusion and eventually sublimation to form a Si-rich interface. Therefore, two distinct processes take place when ozone is inserted during Al 2 O 3 ALD: (1) D it reduction with ozone and (2) oxygen scavenging with remote oxide (TMA) gettering.…”

“…As previously documented, multiple devices on the same wafer were probed to define standard D it error analysis and verify the repeatability . It is shown that the typical standard error is 3.9%; therefore, relative D it variation as low as 10% among different processing conditions can be reliably distinguished 5 .…”

“…In the present study, comprehensive analysis of the effect of reactive oxidant exposure during ALD oxide deposition is studied with impedance measurements correlated with STEM-EELS and photon energy dependent PES analysis to elucidate the defect reduction mechanism with ozone insertion. The impact of ozone exposure during ALD oxide deposition on the SiGe/high-k oxide interface is investigated with a 5 . In this study, by correlating the two advance metrologies, STEM-EELS and energy-resolved PES, with multifrequency impedance spectroscopy, the mechanism for reduction of surface states during the ALD process is elucidated for the key new channel material in CMOS technology.…”

“…It is shown that the typical standard error is 3.9%; therefore, a relative D it variation as low as 10% among different processing conditions can be reliably distinguished. 5 The structures and the compositions of the MOSCAP devices and interfaces were studied using electron transparent specimens (<50 nm) prepared from device cross sections with a FEI Scios focused ion beam eV −1 cm −2 , and the D it increases when ozone is introduced prior to HfO 2 deposition in (b) and decreases when ozone is introduced after few cycles of HfO 2 deposition as seen in (c)− (e). Note that (a), (j) and (k) were previously reported results.…”

“…Note that (a), (j) and (k) were previously reported results. 5 Inset drawings illustrate the device structure along with the description above for a given graph. Except for the device in (m), all devices were treated with sulfur passivation prior to ALD oxide growth after native oxide removal.…”