Thin Solid Films 517 (2008) 155–158
Contents lists available at ScienceDirect
Thin Solid Films j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / t s f
Suppression of Ge–O and Ge–N bonding at Ge–HfO2 and Ge–TiO2 interfaces by deposition onto plasma-nitrided passivated Ge substrates S. Lee a, J.P. Long a, G. Lucovsky a,⁎, J. Lüning b a b
North Carolina State University, Raleigh, NC 27695-8202, USA Stanford Synchrotron Radiation Laboratory, Menlo Park, CA 94025, USA
a r t i c l e
i n f o
Available online 22 August 2008 Keywords: Ge/dielectric interfaces Remote plasma nitridation of Ge Band edge defects Remote plasma deposition Thermal annealing
a b s t r a c t A study of changes in nano-scale morphology of thin films of nano-crystalline transition metal (TM) elemental oxides, HfO2 and TiO2, on plasma-nitrided Ge(100) substrates, and Si(100) substrates with ultrathin (∼0.8 nm) plasma-nitrided Si suboxide, SiOx, x b 2, or SiON interfacial layers is presented. Near edge X-ray absorption spectroscopy (NEXAS) has been used to determine nano-scale morphology of these films by Jahn–Teller distortion removal of band-edge d-state degeneracies. These results identify a new and novel application for NEXAS based on the resonant character of the respective O K1 and N K1 edge absorptions. Their X-ray energy difference of N 150 eV is critical for this approach. © 2008 Published by Elsevier B.V.
1. Introduction A significant problem in the development of field effect transistors on Ge substrates is the formation of defective interfacial transition regions with Ge–O bonds between Ge substrates and gate dielectrics [1]. One solution is to deposit a thin Si layer on the Ge, and then form a passivating/protective SiON interfacial layer; however this adds to the equivalent oxide thickness [2]. An alternative approach for eliminating interfacial Ge–N, as well as Ge–O bonds is presented. A sacrificial interfacial GeN layer is formed on the Ge substrate, protecting the substrate surface from oxidation during film deposition. The Ge–O and Ge–N bonds are removed from this layer by a post deposition anneal in Ar at 800 °C leaving high-κ HfO2 and TiO2 thin film dielectrics in direct bonding contact with the regrown epitaxial Ge substrate with no detectable Ge-nitride or Ge-oxide transition region. This new and novel application for NEXAS relies critically on the resonant character of the X-ray absorption, and on the large energy difference of N150 eV between the O K1 and N K1 edge threshold X-ray energies of ∼ 530 eV and 400 eV. 2. Experimental methods Nano-crystalline HfO2 and TiO2 films ∼ 2 to 6 mn thick were deposited at 300 °C by remote plasma-enhanced chemical-vapordeposition onto remote plasma-nitrided Ge substrates, and compared with HfO2 and TiO2 films deposited onto Si substrates with ∼ 0.6–0.8 nm thick SiON interfacial layers [1,2]. The combination of
⁎ Corresponding author. E-mail address:
[email protected] (G. Lucovsky). 0040-6090/$ – see front matter © 2008 Published by Elsevier B.V. doi:10.1016/j.tsf.2008.08.099
resonant atom-specific O K1 and N K1 absorptions is an effective way to study buried interfaces, and in particular relationships between bonding in oxide dielectrics, and interfacial transition regions. 3. Experimental results Fig. 1(a) shows the N K1 spectrum for remote plasma assisted nitridation of Ge (100) substrates used for deposition of HfO2 and TiO2 films. Based on on-line Auger electron spectroscopy preformed prior to film deposition, the GeN thickness is 0.8 ± 0.1 nm. Films of HfO2 and TiO2 ∼2 to 6 nm thick were deposited onto these nitrided Ge(100) substrates at 300 °C, and then subjected to an ex-situ one minute 800 °C rapid thermal anneal in Ar. Fig. 1(b) shows N K1 spectra for buried interfaces for 2 nm and 6 nm thick HfO2 films annealed at 800 °C. These spectra indicate significantly reduced interfacial Ge–N bonding after the 800 °C anneal. The reductions are essentially the same even though the 6 nm film is three times thicker as the 2 nm film. Fig. 2 shows O K1 spectra for HfO2 (a) 2 nm thick, and (b) 6 nm thick on Ge(100) substrates, providing comparisons between films asdeposited at 300 °C, and after the 800 °C anneal in Ar. The O K1 spectrum for the as-deposited 2 nm HfO2 film in Fig. 2(a) on the GeN interface shows broad spectral features that are different for the 800 °C annealed 2 nm film which the GeN layers have been removed. The spectrum for the annealed film shows a band-edge resolved doublet, and three other relative sharp features. There are similar qualitative differences for O K1 spectra for 6 nm films, as-deposited, and after annealing at 800°C, and these are displayed in Fig. 2(b). The features in the O K1 spectra in Fig. 2 are O 2p π, 532–535 eV, and O 2p σ, 536–539 eV, 542 eV and 544 eV molecular orbital (MO) states localized on the O-atoms; these MO states are mixed with atomic states of the Hf atoms in symmetry adapted MO final states [4].
156
S. Lee et al. / Thin Solid Films 517 (2008) 155–158
qualitative differences in the Eg feature. These are attributed to changes in unit cell morphology as a function of film thickness, and are similar to results for 6 nm thick phase separated Hf silicates in Ref. [3]. Fig. 3(a) is for an as-deposited and 900 °C annealed 2 nm thick film HfO2 on a SiON interface on a Si substrate. The O K1 spectra display broad features similar to those for the as-deposited 2 nm thick HfO2 films in Fig. 2(a)[3]. The absence of J–T degeneracy splitting for the band-edge Hg Eg features in these films is attributed to a suppression of coherent π-bonding [3,5]. This incoherent π-bonding in 2 nm thick HfO2 as-deposited and annealed films on SiON is consistent with homogenous nucleation of ∼2 nm nano-grains with a length scale below the threshold level for coherent π-bonding [3]. The incoherent π-bonding after the 900 °C anneal is due to a dimensional constraint associated with the 2 nm film thickness. In marked contrast, coherent π-bonding with a J–T distortion is observed in annealed 6 nm thick film on SiON in Fig. 3(b) where nano-grains are N3.5 nm; this nanograin size is not constrained by the 6 nm film thickness [3]. In both the as-deposited films, the incoherent π-bonding is a kinetic limitation associated with the low growth temperature. Finally, based on these
Fig. 1. N K1 spectra for remote plasma assisted nitridation of Ge (100) substrates; (a) Ge substrate nitridation for all Ge samples; (b) buried interfaces for 2 nm and 6 nm thick HfO2 films after 800 °C anneal.
The lowest energy Hf features for these nano-crystalline HfO2 films between 532 and 535 eV have contributions from Hf Eg anti-bonding π-states, and the peak between 536 and 539 eV from Hf T2g antibonding σ-states, and features at ∼542 and 544 eV have contributions Hf 6s A1g states, and Hf 6p T1u states [3,4]. The scale of nanocrystallinity, i.e., grain size, has been established by the relative intensities of the features in these O K1 spectra and has been addressed in detail in Refs. [3,5]. The broader spectral features in the as-deposited films in Fig. 2(a) and (b) are attributed to a less than 2 nm nano-crystallites grain size, whilst the sharper features in 800 °C annealed films are attributed to nano-crystallite grains N3.5 to 4 nm [3,5]. This aspect of nanocrystallinity is based on the splitting of the band-edge Eg state into a doublet, and is attributed to a cooperative Jahh-Teller distortion which removes the degeneracy. This requires a nano-scale grain size N3.5 nm [3,5]. This length scale has been established by systematic studies of HfO2 and ZrO2 films on SiON interfaces using Si(100) substrates [3]. Additionally, the as-deposited film 2 nm and 6 nm films display
Fig. 2. O K1 spectra for HfO2 on Ge(100) substrates: comparisons between as-deposited at 300 °C, and after an 800 °C 1 minute anneal in Ar: (a) 2 nm thick, and (b) 6 nm thick.
S. Lee et al. / Thin Solid Films 517 (2008) 155–158
157
comparisons, the J–T splittings in Fig. 2(b) for annealed nm thick film on Ge is also indicative of a nano-grain size N3.5 nm, but in this case, it is not limited by film thickness. Fig. 4(a) shows O K1 spectra for nano-crystalline TiO2: as-deposited on a GeN interfacial layer, and after an 800 °C anneal and in bonding contact with the Ge. The half-width half-maximum (hwhm) for the annealed film is 0.5 eV, and for the as-deposited film, 1.0 eV. The hwhm in the 800 °C annealed films correlates a J–T term splitting and degeneracy removal; however, 0.5 eV is ∼35% less than the ∼0.67 eV hwhm for 900 °C annealed TiO2 films deposited on SiON–Si(100) substrates which also display J–T term splittings and degeneracy removal [3]. Fig. 4(b) indicates differences in the Ge–N bonding in the as-deposited and 800 °C annealed films establishing the role of interface bonding in determining nano-scale morphology. The results for TiO2 depositions parallel the results in Figs. 1 and 2 for HfO2 on Ge substrates. In addition, and more importantly, the results in Fig. 5 are a comparison of epsilon 2 (ε2) spectra, extracted from vis–VUV SE, for 5 nm TiO2 on Ge(100) and Ge (111) substrates after an 800 °C 1 minute
Fig. 4. Comparisons between 2 nm thick TiO2 on Ge(100) substrates as-deposited at 300 °C, and after an 800 °C 1 minute anneal in Ar: (a) O K1 spectra and (b) N K1 edge spectra.
anneal in Ar. Significant differences between the respective band-edge T2g two spectral features clearly indicate a Ge substrate geometry controlled alignment between the TiO2 films and the respective Ge (100) and Ge(111) substrates. In addition, the spectra of the same films as-deposited show essentially the same spectral features, consistent with the results in Fig. 4(a), and the more extensive discussion of the HfO2 spectra in Figs. 1–3. 4. Summary
Fig. 3. O K1 spectra for HfO2 on Si(100) substrates with SiON substrates as-deposited at 300 °C, and after a 900 °C 1 minute anneal in Ar (a) 2 nm, and (b) 6 nm.
The combination of resonant O K1 and N K1 edge spectra has been shown to be an effective way to study buried interfaces. The results in Figs. 1–4 demonstrate that removal of interfacial Ge–N bonding for annealed stacks allows the regrown epitaxial Ge surface to act as a template for in-plane controlled HfO2 and TiO2 grain growth to a crystallite size N3.5 nm that promotes observable J–T spittings in O K1 edge spectra. The spectral weighting factors for these splittings are different than those for annealed HfO2 and TiO2 on SiON terminated Si
158
S. Lee et al. / Thin Solid Films 517 (2008) 155–158
tional support to a regrown Ge epitaxial controlled surface grain growth morphology after 800 °C annealing. This model is also consistent with the results in Fig. 2(a) and (b) that show J–T splitting cannot be suppressed by out of the plane film thickness, and is different than in 2 nm thick HfO2 films on SiON interfaces where coherent π-bonding is film thickness determined. In summary, this paper has demonstrated the following: i) a processing technique in which a sacrificial GeN formed by a plasma process can be used to protect a Ge surface from oxidation during deposition of the transition metal oxides: HfO2 and TiO2; ii) removal of Ge–N and any residual Ge–O bonding from the interfacial transition region during an 800 °C anneal in Ar, and iii) regrowth of an epitaxial Ge layer on the Ge substrate which acts as a template for HfO2 and TiO2 with a grain size sufficient for coherent π-bonding as detected by Jahn–Teller term splitting in band-edge d-states — in HfO2 by a degeneracy removal in the band-edge Eg state, and in TiO2 by a narrowed line (the Jahn–Teller splitting is obtained by differentiation of this feature). References Fig. 5. Comparison of epsilon 2 (ε2) spectra for 5 nm TiO2 on Ge(100) and Ge (111) substrates after an 800 °C 1 minute anneal in Ar.
substrates. Finally, the results in Fig. 5 demonstrate surface symmetry determined alignment between TiO2 nano-grains and Ge(100) and Ge(111) substrates after the 800 °C anneal. These results add addi-
[1] M. Houssa, M.M. Hyens, in: M. Houssa (Ed.), High-K Gate Dielectrics, IOP, Bristol, 2004, Chap 1.1. [2] H. Niimi, G. Lucovsky, J. Vac. Sci. Technol. B 17 (1999) 2610. [3] G. Lucovsky, et al., Jpn. J. Appl. Phys. 46 (2007) 1899. [4] R.A. Wheeler, et al., J. Am. Chem. Soc. 108 (1986) 2222. [5] F.A. Cotton, Chemical Applications of Group Theory, 2nd ed., Wiley-Interscience, New York, 1971, Chap. 8.