ARTICLE pubs.acs.org/Langmuir
Influence of Surface Temperature on the Mechanism of Atomic Layer Deposition of Aluminum Oxide Using an Oxygen Plasma and Ozone Vikrant R. Rai,† Vincent Vandalon,‡ and Sumit Agarwal*,† † ‡
Department of Chemical Engineering, Colorado School of Mines, Golden, Colorado 80401, United States Department of Applied Physics, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands ABSTRACT: We have examined the role of substrate temperature on the surface reaction mechanisms during the atomic layer deposition (ALD) of Al2O3 from trimethyl aluminum (TMA) in combination with an O2 plasma and O3 over a substrate temperature range of 70 200 °C. The ligand-exchange reactions were investigated using in situ attenuated total reflection Fourier transform infrared spectroscopy. Consistent with our previous work on ALD of Al2O3 from an O2 plasma and O3 [Rai, V. R.; Vandalon, V.; Agarwal, S. Langmuir 2010, 26, 13732], both OH groups and carbonates were the chemisorption sites for TMA over the entire temperature range explored. The concentration of surface CH3 groups after the TMA cycle was, however, strongly dependent on the surface temperature and the type of oxidizer, which in turn influenced the corresponding growth per cycle. The combustion of surface CH3 ligands was not complete at 70 °C during O3 exposure, indicating that an O2 plasma is a relatively stronger oxidizing agent. Further, in O3-assisted ALD, the ratio of mono- and bidentate carbonates on the surface after O3 exposure was dependent on the substrate temperature.
1. INTRODUCTION Deposition of ultrathin films with controlled properties is of key interest in semiconductor technology,1 energy systems,2 and optoelectronic applications.3 Atomic layer deposition (ALD) is an ideal technique for depositing nanometer-thick films over complex nanostructures with exceptional uniformity and subnanometer-level control over the film thickness.4,5 The ALD process is based on sequential self-limiting surface reactions where growth occurs by depositing the material layer by layer.6,7 ALD is most suited for binary compounds, and remarkable progress has been made over the last two decades in developing ALD processes for a variety of metal oxides and metal nitrides where surface chemisorption of the molecular precursors is thermally activated.8 10 Among the different metal oxides that can be deposited with ALD, Al2O3 is one of the most extensively explored due to its numerous applications such as a dielectric in electronic devices,11,12 passivation layer in of Si-based solar cells,13 and water permeation barrier in organic electronics.14 Thermal ALD of Al2 O 3 using trimethyl aluminum (TMA, Al[CH3]3) and H2O serves as a benchmark for ALD processes.8 10,15 The ligand-exchange reactions for ALD of Al2O3 using TMA and H 2 O have been extensively studied: the reaction sequence involves CH3 groups as the reactive sites for the dissociative chemisorption of H2O, which results in an OH-terminated surface. In the subsequent cycle, TMA reacts with the OH groups to restore the CH3-terminated surface. CH4 is produced as the primary reaction byproduct during both halfreaction cycles.8,9,16 The surface chemistry during Al2O3 ALD has been extensively investigated using in situ transmission Fourier-transform infrared (IR) spectroscopy,8,9,16 quartz crystal microbalance (QCM),10,17,18 mass spectrometry,18,19 X-ray photoelectron spectroscopy (XPS),20 and spectroscopic ellipsometry.21,22 r 2011 American Chemical Society
Although H2O works very well as an oxidizer, several shortcomings of H2O-based ALD processes have been reported in the literature: these include long purge times due to slow desorption of excess H2O especially in high-aspect-ratio structures, and incorporation of OH groups in the films, which leads to inferior electrical properties.9,23 Therefore, recently, there is an inclination toward the use of oxidizing agents such as an O2 plasma or O3 as an alternative to H2O during ALD of metal oxides. These oxidizers have several advantages such as a low deposition temperature22 and improved electrical properties of the deposited films.22,24 Kessels and co-workers have reported a reduction in OH incorporation in Al2O3 when an O2 plasma was used instead of H2O as an oxidizer.22 Kim et al. have reported an improvement in the leakage current density, interfacial properties, and reduction in OH impurity incorporation for O3-based ALD of Al2O3 as compared to H2Obased ALD.24 Various research groups have investigated the effect of O3 and O2 plasma as oxidizers on the film quality by studying the underlying surface reactions.25 27 Surface chemistry during ALD of Al2O3 using TMA and O3 has been studied by George and coworkers who have reported both OH groups and formates as the reactive sites for TMA chemisorption.25 Kessels and co-workers, on the other hand, have emphasized a surface chemistry for O2 plasmaassisted ALD that is similar to H2O-based ALD.27 Although nonaqueous oxidizing agents have been previously investigated for the ALD of various metal oxides, the corresponding reaction mechanisms are not universal28 and, therefore, are not completely understood. For O3-based ALD of Al2O3, Goldstein et al. have proposed a reaction pathway where O inserts into Received: March 28, 2011 Revised: October 24, 2011 Published: November 14, 2011 350
dx.doi.org/10.1021/la201136k | Langmuir 2012, 28, 350–357
Langmuir
ARTICLE
Figure 2. Schematic of the IRC illustrating the total internal reflection of the IR beam at the film crystal interface. The chemisorbed species are probed by the evanescent electric field that decays exponentially into vacuum. each flat face of the IRC, greatly enhancing the signal-to-noise ratio. Because of the refractive index mismatch between ZnSe (n ≈ 2.4) and the ALD-deposited Al2O3 films (n ≈ 1.63), total internal reflection of the IR beam occurred at the film crystal interface (see Figure 2). Therefore, the surface species were probed by the evanescent field that decayed exponentially into vacuum. This IR setup was also sensitive to the gas-phase species due to the ∼25-cm-long IR beam path through the chamber (see Figure 1). The ZnSe IRC was heated using two 100-W resistive cartridge heaters (Watlow Firerod). A K-type thermocouple was mounted directly behind the IRC on the opposite face of the substrate block to measure the temperature. Because the heat distribution through the substrate stage was expected to be uniform, the thermocouple gave a reasonably accurate measure of the substrate temperature, which was controlled within (1 °C. All IR spectra were collected with a spectral resolution of 4 cm 1 and were averaged over 1000 scans and reported without any baseline correction. All of the spectra were recorded as difference spectra; that is, a fresh background spectrum was collected before every half-reaction cycle. Therefore, any increase in IR absorbance was due to the freshly adsorbed species onto the surface, while a decrease in the absorbance was due to the consumption of the species already present on the surface due to reactions with the incoming precursor. TMA was contained in a bubbler maintained at room temperature (∼19 °C) and was delivered into the surface analysis chamber without a carrier gas. The vapor pressure of TMA at 19 °C, calculated using the Antoine equation, was ∼8.68 Torr.35 The flow rate of TMA into the chamber was controlled using a low-flow-coefficient needle valve. TMA delivery lines were heated and maintained at ∼80 °C to prevent precursor condensation. TMA was pulsed for 15 s followed by a 180-s O2 (99.999%) purge, because O2 was unreactive with TMA or the substrate surface at the temperature of the experiment.21,36 The chamber was subsequently evacuated to the base pressure. For O2 plasma-assisted ALD, an inductively coupled plasma (ICP) source was fabricated from a 1/8 in. outer diameter (OD) Cu tube coiled around a 1 in. OD quartz tube. The plasma source was powered at 100 W and 13.56 MHz through a matching network using a radio frequency (rf) power supply (Advanced Energy RFPP 5S). The Cu coil was maintained at ∼10 °C by flowing cooling water through the tube to avoid any rf-induced heating. O3 was generated by flowing O2 through an in-line coronadischarge-based O3 generator (OzoneServices OL80). The pressure at the inlet of the O3 generator was kept at ∼760 Torr and produced ∼6 wt % O3. The pressure downstream of the O3 generator was regulated using a medium-flow needle valve. O2 plasma and O3 were pulsed for 60 and 180 s, respectively, followed by a 120-s O2 purge. The purge step after the oxidation cycle was simply initiated by turning off the power to the O3 generator or the ICP source. The solenoid valves and the RF power supply were controlled using Labview. To ensure that the half reactions were studied on reproducible Al2O3 surfaces, a fresh ZnSe IRC was initially coated with ∼15 20 nm of Al2O3 deposited from TMA and O2 plasma, and, thereafter, about 10 ALD cycles of TMA and O2 plasma or O3 were performed to prepare the surface at the corresponding substrate temperature. The prepared
Figure 1. Schematic of the surface analysis chamber equipped with in situ ATR-FTIR spectroscopy. The arrows indicate the IR beam path from the FTIR spectrometer through the IRC to the MCT-A detector.
Al C and C H bonds of chemisorbed TMA to primarily produce OH groups and formates on the surface.25 On the other hand, for O2-plasma-assisted ALD of Al2O3, Kessels and co-workers21,27 have proposed a combustion-like reaction mechanism. In this mechanism, CO, CO2, and H2O are generated due to the combustion of the CH3 ligands in chemisorbed TMA: only OH groups were reported as the reactive sites. These OH groups were consumed during the subsequent TMA cycle after releasing CH4 into the gas phase.21,27 In this Article, we have elucidated the reaction mechanisms for ALD of Al2O3 from TMA using both an O2 plasma and O3 as oxidizers through real-time in situ attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy.28 34 We show that the reaction mechanisms for both oxidizers are combustion-like, which produces OH groups and carbonates as the reactive sites. The carbonates are stable during prolonged O3 exposure, but appear as thermally stable reactive intermediates during the O2 plasma cycle. The surface concentration of carbonates, however, does not affect TMA adsorption because it readily chemisorbs on both reactive sites. The growth per cycle data at different temperatures have been correlated with the ligand-exchange reactions at that temperature to understand how the surface chemistry affects film growth and the corresponding flim properties.
2. EXPERIMENTAL SECTION a. Surface Analysis Chamber with in Situ ATR-FTIR Spectroscopy Setup. The surface chemistry during ALD of Al2O3 from TMA and O2 plasma/O3 was studied at different substrate temperatures in a cold-wall vacuum chamber (see Figure 1). The chamber was evacuated to a base pressure of ∼10 3 Torr by a mechanical pump. A leak rate of ∼0.5 mTorr/min was measured by isolating the chamber from the vacuum pump and monitoring the pressure rise on a 2-Torrrange capacitance manometer (MKS 122A). For in situ diagnostics of the surface species and the reaction products during TMA and oxidation half-reaction cycles, the chamber was equipped with an in situ real-time ATR-FTIR spectroscopy setup (see Figure 1). In this setup, an IR beam from an FTIR spectrometer (Thermo 6700) was focused on to the beveled edge of a trapezoidal-shaped infrared-transparent ZnSe internal reflection crystal (IRC) with the dimensions 50 mm 10 mm 1 mm and the short faces beveled at 45°. The IR beam was incident onto the flat face of the IRC at an angle that is greater than the critical angle for total internal reflection. The beam was internally reflected 25 times on 351
dx.doi.org/10.1021/la201136k |Langmuir 2012, 28, 350–357
Langmuir
ARTICLE
Figure 3. Schematic of the high-throughput hot-wall ALD reactor.
Figure 5. IR difference spectrum showing (a) TMA half-reaction cycle for PEALD of Al2O3 at 150 °C. (b f) Temporal evolution of the surface species during O2 plasma exposure of the CH3-terminated surface, indicating the formation and subsequent consumption of metal carbonates in the 1800 1350 cm 1 region. the optical data were fitted with a Cauchy model over the wavelength range of 300 1300 nm.37
3. RESULTS AND DISCUSSION a. Temperature Dependence of O2 Plasma-Assisted ALD of Al2O3. The initial sets of experiments were performed in the
surface analysis chamber equipped with the ATR-FTIR spectroscopy setup. To perform each half-reaction cycle under selflimiting conditions, we first determined the saturation dose for each precursor. IR difference spectra were collected after every incremental increase in precursor dose measured in exposure time with the background spectrum collected before the first dose. The corresponding integrated absorbance change was calculated for the CH3 stretching vibrations in the 3000 2800 cm 1 region and was normalized with the saturated value for the longest precursor exposure. Figure 4 shows the temporal evolution of the normalized integrated absorbance in the 3000 2800 cm 1 region for TMA and O2 plasma half-reaction cycles at 150 °C. The solid lines in Figure 4 are exponential fits to the experimental data, representing first-order reaction kinetics. The inset shows the integrated area in the 3000 2800 cm 1 region for the IR spectrum recorded after 15 s of TMA exposure. From Figure 4, it is clear that a 5 s TMA dose and a 10 s O2 plasma dose were sufficient for the surface saturation. To ensure complete surface saturation, we proceeded with 15 and 70 s of TMA and O2 plasma exposure cycles, respectively. Figure 5 shows the IR spectra for the ligand-exchange reactions during one complete TMA O2 plasma ALD cycle at 150 °C. A one-to-one exchange of the surface species can be clearly observed by comparing the IR spectra for (a) 15 s TMA exposure and (f) 70 s O2 plasma exposure. In Figure 5a, an increase in absorbance in 3000 2800 and 1250 1100 cm 1 corresponds to CH3 stretching and deformation modes, respectively, in TMA. Specifically, the vibrational bands at 2942, 2896, 2831, and 1211 cm 1 were assigned to CH3 antisymmetric stretching, symmetric stretching, bending overtone, and deformation modes, respectively.16,27,38,39 The decrease in absorbance in Figure 5a
Figure 4. Temporal evolution of the normalized integrated absorbance change for the CH3 stretching vibrations (3000 2800 cm 1) during the TMA (9) and O2 plasma (O) half-reaction cycles at a substrate temperature of 150 °C. The solid lines represent exponential fits that correspond to first-order reaction kinetics. The inset shows the integrated area in the 3000 2800 cm 1 region for the CH3 stretching vibrations during the TMA cycle, recorded at 15 s of TMA exposure. surface was examined for thermal stability by collecting the IR spectra over time without any precursor exposure. These IR spectra showed no observable absorbance change in the spectral range of 4000 650 cm 1, indicating that the Al2O3 surface was stable. b. High Throughput ALD Reactor. To determine the growth per cycle, 20 50-nm-thick Al2O3 films were deposited on HF-treated SiO2-free Si wafers in a tubular hot-wall reactor. The schematic in Figure 3 shows the front view of the reactor, which shared the same precursor and gas-delivery lines as the surface analysis chamber. The reactor walls were maintained at ∼80 °C. The Si wafers were mounted on a substrate stage and resistively heated using cartridge heaters. The temperature of the substrate was measured using a K-type thermocouple and controlled within (2 °C. This reactor was also equipped with an almost identical rf ICP source for O2 plasma generation, operated at 40 W. A lower power was sufficient as compared to the main surface analysis chamber because the plasma source was much closer to the substrate stage. The ALD cycles consisted of a 2-s TMA cycle followed by a 10-s O2 purge. The O2 plasma was turned on for 10 s followed by a 2-s purge. The thickness and refractive index of the deposited Al2O3 films were characterized by ex situ spectroscopic ellipsometry (Woollam M-44): 352
dx.doi.org/10.1021/la201136k |Langmuir 2012, 28, 350–357
Langmuir
ARTICLE
Figure 6. IR difference spectra for two complete ALD cycles at 150 °C. (a) The surface was exposed to TMA for 15 s followed by a 180-s inertgas purge. (b) The TMA-chemisorbed surface in (a) was exposed to O2 plasma for 10 s, just sufficient to completely combust the methyl ligands. (c) The carbonates in (b) were completely removed in the subsequent TMA exposure cycle (see 1800 1350 cm 1 region). (d) A 70-s O2 plasma exposure of the surface in (c) completely removed the carbonates that appeared in (b) after 10 s.
over 3800 3200 cm 1 was assigned to surface OH stretching vibrations,8,27,40 42 as TMA dissociatively chemisorbed onto surface OH groups. The CH3-terminated surface was then exposed to increasing doses of an O2 plasma for a total of 70 s. The IR difference spectra recorded at these different stages of O2 plasma exposure are shown in Figure 5b f. Upon a 70-s O2 plasma exposure (see Figure 5f), CH3 groups on the surface were completely oxidized as indicated by the decrease in absorbance in the CH3 stretching and deformation regions, while OH groups were simultaneously restored on the surface. Thus, the IR spectra in Figure 5a and f are consistent with previous observations by Kessels and co-workers who reported OH groups as the reactive sites during plasmaassisted ALD in a mechanism that was very similar to H2O-based ALD.27 However, for O2 plasma exposures of
