Article pubs.acs.org/JPCC
Surface Oxide Characterization and Interface Evolution in Atomic Layer Deposition of Al2O3 on InP(100) Studied by in Situ Infrared Spectroscopy W. Cabrera,† M. D. Halls,⊥ I. M. Povey,§ and Y. J Chabal*,† †
Department of Material Science and Engineering, The University of Texas at Dallas, 800 W. Campbell Road, Richardson, Texas 75080-3021, United States ⊥ Schrödinger Inc., 5820 Oberlin Dr. Ste. 203, San Diego, California 92121, United States § Tyndall National Institute, University College Cork, Lee Maltings, Dyke Parade, Cork, Ireland ABSTRACT: Combined in situ IR measurements and firstprinciples calculations of InP(100) surfaces reveal that mild annealing (∼300 °C), typically needed for atomic layer deposition, leads to the formation of InP-derived surface hydrophosphate species (both PO and P−OH sites). The initial interaction of trimethylaluminum at 300 °C results in the formation of P−O−Al linkages through covalent and dative bonding by reaction with surface hydroxyls. During subsequent ALD cycles to deposit Al2O3, an interfacial layer composed of P−O−Al bonds (1140 cm−1) is formed, requiring approximately seven cycles for completion. Similar chemical transformations are observed on hydrofluoric acid and ammoniumsulfide treated [HF/(NH4)2S] surfaces but to a lesser degree since the oxide thickness is reduced, requiring only approximately three cycles to fully complete the interfacial layer. Initially, the ALD growth of Al2O3 is slower on the HF/(NH4)2S-treated InP(100) surface than on the native oxide surface due to a lower density of hydroxyl groups. However, this slow growth leads to a denser film, highlighting the importance of the chemical composition of the initial InP(100) substrate.
I. INTRODUCTION With the aim of increasing device performance, III−V semiconductors are under investigation as alternative channel materials to replace silicon in metal-oxide-semiconductor fieldeffect transistors (MOSFETs).1,2 One of the major issues currently hindering the adoption of III−V channels is the high density of interfacial defects stemming from the poor quality of the native oxide/III−V interface that causes instability in electrical devices.3 Strategies to improve the performance of high-k/III−V devices include a focus on surface preparation and procedures to remove the native oxide and chemically passivate the exposed surface prior to high-k deposition.3−5 Despite significant challenges, development efforts have resulted in improved performance of III−V semiconductorbased electrical devices. In particular, treatment of III−V substrates with (NH4)2S has led to significant improvements in the interfacial properties.6−8 O’Connor et al.6 showed that treating In0.53Ga0.47As with 10 vol. % (NH4)2S for 10 min effectively reduces the frequency dispersion for both n- and ptype devices. This work was complemented by an in-depth atomic force microscope (AFM) and X-ray photoelectron spectroscopy (XPS) study to optimize the (NH4)2S treatment of In0.53Ga0.47As substrates.9 Gu et al.10 observed that the concentration of (NH4)2S during wet cleaning of the InP barrier layer is critical. They also found that treatment with 10% © 2014 American Chemical Society
(NH4)2S gives the lowest subthreshold swing (SS) of 96 mV/ dec. These studies illustrate the importance of surface chemistry in determining interfacial structure that impacts the performance of electrical devices. In recent years, advanced device fabrication has increasingly relied on atomic layer deposition (ALD) for conformal and well-controlled deposition of high-k dielectric thin films on silicon.11−13 ALD has proven critical for III−V semiconductors as well for high-k deposition,14,15 giving rise to interesting phenomena. For instance, the initial exposure of TMA during the ALD-growth of Al2O3 on III−V surfaces leads to a reduction in the native oxides, known as “self-cleaning” or “clean-up” effect.14,16 Moving the high-k/III−V interface away from the channel region by adding an InP(100) barrier layer on In0.53Ga0.47As buried channels has resulted in higher transconductance and lower subthreshold with an increase in electron mobility.17−19 Such observations underscore the influence of the high-k/III−V interface on important electrical properties such as off-state performance.10 Yet, little has been done to understand the nature of InP barrier surface structure and the deposition process of high-k dielectrics on InP, Received: December 19, 2013 Revised: February 21, 2014 Published: February 24, 2014 5862
dx.doi.org/10.1021/jp412455y | J. Phys. Chem. C 2014, 118, 5862−5871
The Journal of Physical Chemistry C
Article
main findings are that (1) the chemical composition of the cleaned InP surfaces is modified upon annealing prior to any ALD process, (2) TMA reacts with specific groups at the surface, forming an interfacial layer composed of covalent POAl and dative PO···Al linkages after Al2O3 deposition, and (3) the densification of the Al2O3 film depends on surface preparation (i.e., the initial surface composition).
particularly the chemistry involved during the initial ALD nucleation process (heterodeposition). For instance, nothing is known about the dependence of surface chemical composition on temperature, even though the importance of the surface stability and composition for ALD growth has been recognized.20−24 Interestingly, an (NH4)2S InP(100) surface shows similar behavior as a clean or oxidized InP surface showing thermal stability up to 450 °C before observing an increase in metallic In.24,25 Adelmann et al.26 used synchrotron radiation X-ray photoelectron spectroscopy (SRPES) to analyze the surface compositional changes on InP(100) surfaces after ex situ annealing at 300 °C and followed the surface evolution during Al2O3 ALD using TMA. The P 2p, In 3d5/2 and In 4s spectra revealed higher binding energy oxidation states assigned to P5+, P(2±Δ)+, In3+, and In0, respectively. The PES measurements of the oxidized InP(001) surface are consistent with InPO4 , Inx(HPO4)y, and/or In2O3, without clear assignment. SRPES analysis of the effect of the first TMA half-reaction suggests the formation of an AlPO4 layer, which is consistent with the reduction in the P(2±Δ)+ and P5+ chemical states. Concurrently, a decrease in the In3+ and increase in the In0 states are observed upon TMA exposure. Similar but less marked results are observed for (NH4)2S-treated surfaces. In contrast, Brennan et al.27 suggest, from in situ X-ray photoelectron spectroscopy, that after annealing at 300 °C for 30 min, part of the native oxide is transformed from InPO4 into In(PO3)3. In their model, this InPO4-rich oxide is converted during the initial TMA exposure into an oxygen-rich oxide composed of In(PO3)3 and P2O5 species associated with an increase in the phosphorus oxide chemical states. This increase is consistent with an oxygen transfer from indium oxides to phosphorus during indium-oxide decomposition, an evolution that has been termed the “clean up” effect on III−V surfaces.16 Further development of surface treatments to enable interface engineering using InP(100) substrates requires an in-depth understanding of the local chemical structure of the starting surface. While previous studies provide critical information related to the elemental composition of the chemically prepared InP surfaces, there is scarce knowledge of the precise chemical composition (i.e., bonding configurations) after annealing for both native oxide (solventcleaned) and (NH4)2S-treated (chemically modified) surfaces. The chemical evolution of the surface upon the initial interaction of TMA with InP(100) surfaces is also unknown. Therefore, it is important to characterize the precise chemical nature of the native oxide surface at all levels of the ALD process for growth of metal oxides. In this study, in situ Fourier transform infrared spectroscopy (FTIR) is used to characterize the initial surfaces and to monitor the ALD nucleation and growth during ALD of Al2O3 using TMA and D2O on degreased (native oxide) and HF/ (NH4)2S-treated InP(100) substrates. The evolution of the surfaces during heating to the required temperature for ALD (i.e., 300 °C) is initially monitored, then the chemical modification induced by the interaction of TMA with the annealed surfaces (first half ALD cycle) is characterized to gain mechanistic insight into the details of the surface structure and interfacial layer formation. The local chemical structure of the InP(100) surfaces (i.e., chemical bond formation at the surface) is confirmed by vibrational band assignments supported by First-principles calculations of vibrational frequencies and reaction energetics for the TMA nucleation reaction. The
II. EXPERIMENTAL SECTION A. Materials and Methods. In this study, undoped and double-side polished InP(100) (ρ ≈ 0.1 Ω-cm, AXT) wafers are used. Two different types of surfaces are prepared for analysis: one with epi-ready oxide, degreased with acetone, isopropanol, and deionized (DI) water (1 min for each step);28 the other with sulfur passivation (minimal oxide), prepared by immersion first in HF (5 vol. %) for 1 min and then in (NH4)2S (10 vol. %) for 10 min.10 A final rinse is done in DI water followed by drying in a N2 gas stream to remove any remaining water. HF/(NH4)2S treated samples are immediately loaded (