AVS2010 Session TF-TuA: ALD/CVD: Surface Chemistry and Fundamentals

In this presentation, the authors will elucidate the surface reaction mechanisms during the atomic layer deposition (ALD) of aluminum oxide from trimethyl aluminum (TMA) in conjunction with O₃ and an O₂ plasma. The deposition mechanism was explored over a substrate temperature range of 70–200 °C using in situ attenuated total reflection Fourier transform infrared spectroscopy. Our IR data show that both –OH groups and carbonates are formed on the surface during the oxidation cycle. Gas-phase IR data indicate that oxidizer-assisted combustion of methyl ligands in chemisorbed TMA produces CO₂ and H₂O, which react simultaneously on the Al₂O₃ surface to produce carbonates. The origin of the –OH groups was attributed to the reaction of the uncombusted methyl ligands with gas-phase H₂O. While the type of surface sites are common to both oxidizing agents, in the case of O₂-plasma-assiated ALD, surface carbonates are simply reactive intermediates, which completely decompose upon prolonged plasma exposure. The ratio of carbonates to –OH groups is strongly dependent on the oxidizing agent, and its dose in case of plasma-assisted ALD. Surface reactions such as chemisoption of TMA, formation of –OH groups and Al₂O₃ were pseudo first order. On the other hand, the kinetic behavior of the carbonates suggests a series reaction of the type, A (CH₃) → B (carbonates) → C (Al₂O₃). Although carbonate sites contribute to Al₂O₃ growth, their contribution was determined to be insignificant.

Preparation of inorganic micro/mesoporous materials has attracted considerable attention because of their function in catalytic, separations, and other applications. Although many approaches are known to synthesize porous materials, methods to form mesoporous materials with pre-determined shape and morphology are not readily known. Sub-surface deposition, recently observed during atomic layer deposition (ALD) on polymer substrates, provides a potential method to transform polymers from fully organic solids into organic-inorganic hybrid and micro/mesoporous materials. Moreover, the transformation allows polymers with well defined micro and nano-structure, such as polymer fiber matrices, to maintain their shape and structure to form new replica materials. In this presented work, we apply this process to synthesis of Al-O/organic hybrid micro fibers by gas phase infiltration of tri-methyl aluminum and water alternatively into polyesters, such as polybutylene terephthalate (PBT). Through in-situ infrared analysis, we show that TMA acts as a strong Lewis acid and attacks the nucleophilic ester groups in polyesters to insert Al-O bonds into the polymer chains. The organic components are removed by post annealing to produce a micro/mesoporous alumina fiber. Surface area, pore volume, and pore size distribution of the porous alumina fibers were tested by nitrogen adsorption/desorption experiments, and surface areas exceeding 400 m²/g were obtained. SEM was used to track the morphology change along the process, and cross-section TEM images of annealed samples confirmed the formation of porous structures.

In this work we explore the effect that alkyl alcohols (ROH) have on the saturation growth rate during the ALD of metal oxides. The traditional dosing sequence for metal oxide ALD is: M/O/M/O… where M is the metal precursor and O is the oxygen source. We find that by dosing organic molecules prior to dosing the metal precursor (e.g. ROH/M/O…) we can modify the surface chemistry and control the saturation growth rate. We will present results describing the effect of alkyl alcohols (R=Me, Et, iPr, and Bu) using H₂O as the oxygen source and the metal precursors Ti(iPr)₄ for TiO₂ ALD, TMA for Al₂O₃ ALD, and DEZ for ZnO ALD. Furthermore, we demonstrate this effect in the ALD of doped metal oxides.

Our results show that the ROH/M/H₂O sequence causes a substantial reduction in the growth per cycle for all of the ALD systems studied. For instance, the growth per cycle reduces from 0.31 to 0.06 Å/cycle in the case of TiO₂ ALD using MeOH/Ti(iPr)₄/H2O, and from 1.2 to 0.4 Å/cycle in the case of Al₂O₃ ALD using MeOH/TMA/H₂O at 200°C.

Previous studies in the literature indicate that ROH reacts with basic sites on the metal oxide surface. Alcohol deprotonation followed by metal oxygen heterolytic bond formation leads to the formation of alkoxide functional groups bound to metal cations. To investigate this process, we performed in situ mass spectrometry and quartz crystal microbalance studies during the ROH/M/H2O dosing sequence. We discovered that the ROH adsorbed on the surface desorbs intact as ROH during the subsequent water pulse, but no alcohol is released during the metal precursor pulse. Furthermore, the reduction of the growth rate per cycle is not affected by purge times, suggesting that the ROH molecules bond strongly to the metal oxide surface. Finally, no reduction in growth per cycle is observed using the dosing sequence: ROH/H₂O/TMA/H₂O. This finding suggests that the ROH and H₂O are able to displace one another, and signifies an almost complete elimination of the alkoxide groups during the water pulse. This observation agrees with the many reports of successful metal oxide ALD using metal alkoxide and water.

The ability to tune the saturation growth rate by modifying the surface chemistry can be of great utility for the ALD of doped materials where a homogenous distribution of dopants at a low concentration is desired.

Most processes for TiO₂ atomic layer deposition (ALD) utilize water or other oxidants that can oxidize some substrates of interest. To avoid this oxidation, waterless or oxidant-free surface chemistry can be used that involves titanium halides and titanium alkoxides. This waterless surface chemistry approach for metal oxide ALD was originally proposed by the University of Helsinki group (M. Ritala et al., Science 288, 319 (2000)). In this study, TiO₂ ALD was accomplished using titanium tetrachloride (TiCl₄) and titanium tetraisopropoxide (TTIP). In situ Fourier transform infrared (FTIR) studies revealed that the mechanism for TiO₂ ALD using TiCl₄ and TTIP changed with temperature. At high temperatures between 250 and 300 ºC, the isopropoxide species after TTIP exposures quickly underwent beta-hydride elimination to produce TiOH species on the surface. The observation of propene by quadrupole mass spectrometry confirmed the beta-hydride elimination reaction pathway. The TiCl₄ exposure then reacted with the TiOH species to deposit TiCl_x species on the surface. At low temperatures between 125 and 200 ºC, the isopropoxide species remained after TTIP exposures to react with TiCl₄. However, this reaction was much less efficient than the reaction of TiCl₄ with TiOH species. Quartz crystal microbalance (QCM) studies were also used to monitor TiO₂ ALD at low and high temperatures. The QCM studies measured low TiO₂ growth rates of ~3 ng/cm² at a low temperature of 150°C. In contrast, much higher TiO₂ growth of ~15 ng/cm² were observed at a higher temperature of 250°C under similar reaction conditions. X-Ray reflectivity measurements determined that TiO₂ ALD using TiCl₄ and TTIP at 250°C produced a growth rate of 0.5-0.6 Å per cycle. X-Ray photoelectron studies also confirmed TiO₂ film growth with a chlorine contamination of less than 0.5 at%. This waterless TiO₂ ALD process using TiCl₄ and TTIP should be valuable for preventing substrate oxidation during TiO₂ ALD on oxygen-sensitive substrates such as cobalt.

Magnesium-zinc oxide (Mg_xZn_(1-x)O) ternary films are an interesting class of alloy materials in which the band gap can be tuned by adjusting the Mg doping concentration. Consequently, Mg_xZn_(1-x)O has been widely studied for application in the fields such as electronics, optics, photoelectronics, and solar cells. Mg_xZn_(1-x)O thin films have been fabricated through a variety of methods including chemical vapor deposition, physical vapor deposition, molecular beam deposition, and ALD. Although there have recently been a few reports describing ALD Mg_xZn_(1-x)O for application in photvoltaics, there has been no detailed study of the growth mechanism and properties of the ALD Mg_xZn_(1-x)O thin films.

In this work, the ALD Mg_xZn_(1-x)O was systematically explored with different doping concentration of Mg by using diethyl zinc (DEZ), bis-cyclopentadienyl-magnesium (Cp₂Mg) and H₂O as the precursors. The growth mechanism was investigated using quartz crystal microbalance and quadrupole mass spectrometry measurements. The growth rate of the Mg_xZn_(1-x)O alloy films was determined using spectroscopic ellipsometry. The crystal structures of the films before and after thermal treatment were analyzed using x-ray diffraction. In addition, the optical properties of the Mg_xZn_(1-x)O with different Mg concentrations were analyzed using UV-vis absorption spectroscopy and the electrical properties were evaluated using mercury probe current-voltage measurements. The thermal stability of the conductivity and structure of the Mg_xZn_(1-x)O films were studied as well. The system will be compared with Al doped ZnO system fabricated by ALD, to illustrate the conduction mechanism in doped ZnO synthesized by ALD process.

Pt ALD using thermal chemistry has nucleation difficulties and leads to the deposition of Pt nanoclusters. In contrast, Pt ALD using O₂ plasma nucleates much more readily and spectroscopic ellipsometry (SE) studies are consistent with continuous Pt films (H.C.M. Knoops et al., Electrochem. Solid-State Lett. 12, G34 (2009)). However, SE measurements alone were insufficient to characterize the early stages of the Pt ALD process. In this investigation, we have examined Pt ALD with MeCpPtMe₃ and O₂ plasma as the reactants using SE, X-ray reflectivity (XRR), X-ray photoelectron spectroscopy (XPS) and scanning electron microscopy (SEM) studies versus the number of ALD cycles. Analysis of the XRR and XPS results indicates that plasma Pt ALD on Al₂O₃ ALD substrates has a short nucleation delay. The nucleation delay is followed by a brief period of rapid Pt ALD film growth of 0.1 nm per cycle that is followed by a slower steady-state Pt ALD film growth of ~0.05 nm per cycle. During the Pt ALD nucleation and growth on the Al₂O₃ ALD substrate, SEM images show that the Pt film morphology evolves from isolated nanoclusters to worm-like nanostructures and finally to a conformal film at a Pt film thickness of approximately 7 nm. Nucleation and growth of Pt ALD on W ALD substrates led to very different results. In this case, a H₂ plasma was used instead of an O₂ plasma to prevent oxidation of the W ALD substrate. XRR and XPS studies revealed that Pt ALD with MeCpPtMe₃ and H₂ plasma as the reactants on W ALD substrates nucleated immediately and a continuous and conformal Pt ALD film was formed at a Pt ALD thickness of less than 5 nm. These results indicate that Pt ALD can be tuned to produce Pt nanoclusters or a continuous and conformal ultrathin Pt film using either thermal or plasma Pt ALD.

Thermal decomposition pathways of triethylaluminum ((C₂H₅)₃Al, TEAl) were investigated in a custom, up-flow, cold-wall CVD reactor . The extent of homogeneous decomposition of TEAl in N₂ as well as with added NH₃ was measured using in situ Raman spectroscopy measurement. The results of Density Functional Theory (DFT) calculations were used to assist in assignment of the observed Raman shifts to the decomposition products TEAl:NH₃, DEAlH, TEAl:NH₃ H₂N-AlH-NH-AlH₂, H₂Al-NH₂, MEAlH, MEAlH-AlH₂ and DEAl-AlH₂ as well as the estimating the rates of selected pathways. For the case of thermal decomposition of TEAl with N₂ carrier gas, the species H₂Al–NH₂ was observed . This is believed to result from the reaction of ammonia with the product of β–hydride elimination. This species is a possible reactant for AlN formation. Raman shifts of 600, 1989, 2025, 2580, 2835, 2849, 2900, 2918, 2939, and 3173 cm^-1 were recorded for TEAl in N₂, while shifts of 452, 1462, 1525, 1639, 2580, 2853, 2841, and 2849 cm^-1 were observed for a mixture of TEAl with ammonia in N₂. In addition, four vibrational bands (930, 965, 3230, and 3334 cm^-1) for ammonia were observed with high intensity. The temperature profile along the reactor centerline was measured and this was compared to the simulated results using a custom-FEM Galerkin method. DFT calculations using B3LYP/LanL2DZ level of theory were carried out to find the optimized geometry of each intermediate and transition structure and to calculate activation energy. This methodology, which uses the results from in situ Raman spectroscopy, DFT calculations, and FEM reactor modeling, is a powerful approach to understanding thermal decomposition mechanisms.

III-V materials have been considered as alternative channel materials for future high-speed low-power digital logic applications mainly because of their higher electron mobility compared to silicon. However, the integration of III-V materials into device structures faces multiple challenges due to the lack of a high quality native oxide and the high density of traps at metal-oxide interface, resulting in Fermi level pinning.

Most of the recent work on GaAs passivation has been done using atomic layer deposition (ALD) method. ALD offers precise control of film thickness, low processing temperatures, and excellent conformality. Additionally, a self-cleaning effect has been reported during Al₂O₃, HfO₂, and TiO₂ ALD on GaAs leading to the reduction or removal of GaAs oxides. Understanding the surface reactions involved in this self-cleaning effect is important in order to improve the GaAs-oxide interface, while depositing high-k dielectric films using ALD.

In this study, the ALD of trimethylaluminum (TMA, Al(CH₃)₃) and titanium tetrachloride (TiCl₄) on hydrofluoric acid (HF) treated GaAs (100) samples was conducted to investigate the similarities and differences in the surface chemistry of these precursors. After aqueous HF etching and air re-oxidation, the oxides were composed of As₂O₃ and Ga₂O. Annealing the sample desorbed volatile compounds such as As₂O₃ from the surface resulting in an As-rich (2.7:1, As:Ga ratio) surface. The precursors, TiCl₄ and TMA, reacted with As and Ga oxides and completely removing them from the surface. This is the first time that a TiCl₄ self-cleaning effect on GaAs is reported. TMA, as expected, produced a film of Al₂O₃ (2.3 monolayers thick) on top of the GaAs surface when the reaction was performed at 170°C.

In contrast to TMA, the oxygen and titanium levels remained below the XPS detection limits with TiCl₄ exposures in the temperature range from 170°C to 230°C, while a 0.03 monolayer-thick film was deposited at temperatures ranging from 89°C to 170°C. XPS results showed that in the higher temperature range, TiCl₄ reacted with the GaAs oxides on the surface producing only volatile compounds, leading to a clean, sharp GaAs surface. The proposed mechanism consists of two processes: TiO₂ formation on the surface and in-situ etching of these TiO₂ islands by chlorine atoms from breakdown of the precursor. This novel GaAs oxide cleaning method produced an oxygen-free surface, which has potential applications for GaAs integration in microelectronics and optoelectronics devices.

Cobalt-aluminum alloys are of great interest due to their unique properties such as corrosion resistance, high thermal stability, and unusual magnetic properties. In a search for appropriate precursors, we have investigated the growth of Co-Al thin films by atomic layer deposition (ALD) using CCTBA (μ²-h²-(^tBuacetylene)dicobalthexacarbonyl) and DMAH (Dimethylaluminumhydride) on H-terminated Si(111) and on 2-nm TaN films. In-situ infrared absorbance spectra show that upon the first CCTBA pulse on H/Si(111), alkynes (CC triple bonds) bound to Co₂(CO)₆ are converted to a benzene ring, as evidenced by the ring C=C stretching mode at 1475 and 1610 cm^-1. This transformation is not completely unexpected because cobalt carbonyl complexes (Co_x(CO)_y) are used to catalyze cyclotrimerization reactions in organotransition-metal chemistry. IR spectra also show that the presence of hydrogen enhances the adsorption of carbonyl groups on the H/Si(111) surface. After the first CCTBA pulse which reacts almost completely with H–Si bonds (2083 cm^-1), the amount of adsorbed CO on the surface (1970 cm^-1) is found to decrease. The subsequent DMAH pulse is effective to remove the surface carbonyl groups, leaving Al–CH₃ and/or Al–H bonds on the surface. In contrast, the adsorption of carbonyl on a TaN surface where H is absent is negligible after the first CCTBA pulse.

In all cases, CH_x ligands are not removed during CCTBA or DMAH cycles, leading to accumulation of carbon species in the film. The growth of metallic Co-Al films is hindered due to Al–O bond formation during deposition. The source of oxygen is likely associated to a surface Fischer-Tropsch (FT) process. It appears that reaction of cobalt particles with hydrogen (originating from the DMAH precursor) generates water as a by-product through a FT-like process, thus forming Al–O bonds. This formation of Al-O bonds through a FT process is greatly suppressed (~ 50%) by annealing the sample to 300 ^oC in N₂ ambient before exposure to DMAH, which removes the carbonyl group from the surface and therefore the source of oxygen. Although the growth pattern is similar for both H/Si(111) and TaN substrates, a part of Ta atoms in the original TaN films are reduced to metallic Ta⁰ during growth, according to XPS Ta 4f core level spectrum after deposition of Co-Al(O_x) films.