Article pubs.acs.org/Langmuir
Atomic Layer Deposition, Characterization, and Growth Mechanistic Studies of TiO2 Thin Films Mikko Kaipio,† Timothee Blanquart,*,† Yoann Tomczak,† Jaakko Niinistö,† Marco Gavagnin,‡ Valentino Longo,§ Heinz D. Wanzenböck,‡ Venkateswara R. Pallem,∥ Christian Dussarrat,∥ Esa Puukilainen,† Mikko Ritala,† and Markku Leskela†̈ †
Laboratory of Inorganic Chemistry, Department of Chemistry, University of Helsinki, P.O. Box 55, FI-00014 Helsinki, Finland Institute for Solid State Electronics, Vienna University of Technology, A-1040 Vienna, Austria § Department of Applied Physics, Technische Universiteit Eindhoven, P.O. Box 513, 5600 MB Eindhoven, The Netherlands ∥ Air Liquide Research & Development, DRTC, 200 GBC Drive, Newark, Delaware 19702, United States ‡
ABSTRACT: Two heteroleptic titanium precursors were investigated for the atomic layer deposition (ALD) of titanium dioxide using ozone as the oxygen source. The precursors, titanium (N,N′-diisopropylacetamidinate)tris(isopropoxide) (Ti(Oi Pr) 3 (N i Pr-Me-amd)) and titanium bis(dimethylamide)bis(isopropoxide) (Ti(NMe2)2(OiPr)2), exhibit self-limiting growth behavior up to a maximum temperature of 325 °C. Ti(NMe2)2(OiPr)2 displays an excellent growth rate of 0.9 Å/cycle at 325 °C while the growth rate of Ti(OiPr)3(NiPr-Me-amd) is 0.3 Å/cycle at the same temperature. In the temperature range of 275−325 °C, both precursors deposit titanium dioxide in the anatase phase. In the case of Ti(NMe2)2(OiPr)2, high-temperature X-ray diffraction (HTXRD) studies reveal a thickness-dependent phase change from anatase to rutile at 875− 975 °C. X-ray photoelectron spectroscopy (XPS) indicates that the films have high purity and are close to the stoichiometric composition. Reaction mechanisms taking place during the ALD process were studied in situ with quadrupole mass spectrometry (QMS) and quartz crystal microbalance (QCM).
■
INTRODUCTION Atomic layer deposition (ALD) is a gas-phase thin-film deposition technique which has excellent repeatability and is easily scalable to large substrates and large batches.1,2 Thin films deposited using ALD are uniform and highly conformal, even when grown on demanding three-dimensional structures. These unique advantages are derived from the specific surface chemistry of ALD processes. Studying the reaction mechanisms taking place at each step in an ALD process allows a deeper understanding of the influence of precursors and growth parameters on the deposition characteristics.3 In situ methods are essential to performing such studies as they allow the monitoring of certain process properties directly during the ALD growth cycles. Among these techniques, quadrupole mass spectrometry (QMS) and quartz crystal microbalance (QCM) have been used to provide qualitative and quantitative information on numerous ALD processes. An extensive review regrouping in situ studies of reaction mechanisms during ALD processes has recently been written by Knapas and Ritala.3 Titanium dioxide (TiO2) is a low-cost, nontoxic material which exists principally in three crystalline phases known as rutile, anatase, and brookite, of which the rutile phase is thermodynamically the most stable one. Because of its lower thermodynamic stability, anatase can be transformed into rutile at high enough temperatures. In the case of very small grains of © 2014 American Chemical Society
anatase, however, surface energy effects become dominant and can outweigh bulk thermodynamics, due to the higher surface free energy of rutile.4 A critical size of the anatase particles (11− 45 nm) can be assigned, below which the phase transition to rutile does not occur at normal transition temperatures.4 On the other hand, it has been reported that large anatase grains undergo rutile transformation more slowly than finer ones.5 Therefore, the anatase-to-rutile transition and anatase grain growth can be considered to be competing phenomena. Very pure TiO2 thin films in the rutile phase are of a particular interest for microelectronic applications due to their high dielectric constant.6−9 On the other hand, pure anatase and anatase−rutile phase mixtures are well-known photocatalysts,10−13 and studies on atomic-layer-deposited photocatalytic TiO2 thin films have been carried out.14,15 In TiO2 thin films deposited with ALD, it is common that anatase, or a mixture of anatase and rutile, is the crystalline phase in the as-deposited state.16 If a film does not exhibit the desired phase after deposition, then it can possibly be transformed by suitable postdeposition annealing. Besides temperature, the pressure and dosage of the ALD precursors can play a part in determining the Received: March 17, 2014 Revised: May 31, 2014 Published: June 4, 2014 7395
dx.doi.org/10.1021/la500893u | Langmuir 2014, 30, 7395−7404
Langmuir
Article
boats inside the reactor at 35 and 65 °C, respectively. Ozone was produced from 99.999% O2 (Oy AGA Ab) with an ozone generator (Wedeco Ozomatic modular 4 HC Lab Ozone) and pulsed into the reactor using solenoid and needle valves. Nitrogen (>99.999%, generated with Nitrox UHPN 3000-1) was used as a carrier and purge gas. Si(100) (Okmetic, Finland) pieces with an approximate size of 5 × 2 cm2 were used as substrates. The growth rate as a function of the deposition temperature was examined in the range of 250−375 °C using the pulsing sequence metal precursor/purge/O3/purge with pulse lengths of 0.7/1.2/0.7/1.2 s, respectively. The self-limiting growth of TiO2 was determined by studying the growth rate as a function of the metal precursor pulse length. In order to test the conformality of the films, they were grown on substrates with deep trenches at 275 °C using pulse lengths of 5/10/5/10 s, respectively. The opening diameters and depths of the trenches were approximately 115 nm and 6.75 μm, resulting in an aspect ratio of approximately 60:1. The cross section was investigated with a Zeiss NEON 40 EsB CrossBeam system allowing high-resolution FE-SEM imaging of the film morphology and thickness along the trench walls. The thickness and crystallinity of the TiO2 thin films were evaluated by X-ray reflectivity (XRR) and X-ray diffraction (XRD) using a Panalytical XPert Pro MPD X-ray diffractometer. The thickness of rough films (root-mean-square roughness around 6 nm) was determined by measuring and fitting their reflectance spectra using a Hitachi U2000 UV-vis spectrophotometer with a wavelength range of 370−1100 nm and a fitting program developed by Ylilammi and Rantaaho.32 High-temperature XRD (HTXRD) measurements were made under nitrogen (99.999%, further purified with Entegris 35KF-I-4R inert gas purifier) at temperatures ranging from 25−1075 °C using an AntonPaar HTK1200N oven. The heating was done in 50 °C steps, and an XRD measurement was made at each temperature. The chemical composition of the films was determined by X-ray photoelectron spectroscopy (XPS), performed on a Thermo Scientific K-Alpha KA1066 spectrometer using a monochromatic Al Kα X-ray source (hν = 1486.6 eV). Photoelectrons were collected at a takeoff angle of 60°, and a 400-μm-diameter X-ray spot was used in the analysis. Samples were neutralized using a flood gun to correct for differential or nonuniform charging and sputtered with argon ions for 30 s at 500 eV in order to remove surface carbon contamination. High-resolution XPS scans were obtained for the Ti 3d, O 1s, C 1s, N 1s, and Si 2p electrons at a pass energy of 50 eV. Surface morphology was studied using a MultiMode V atomic force microscope (AFM) equipped with a NanoScope V controller (Bruker). Samples were measured in tapping mode using a phosphorus-doped silicon probe (RTESP) delivered by Bruker with a scanning frequency of 0.5 Hz. Several scan images were recorded from different parts of the samples to check the uniformity before measuring the roughness over a scanning area of 2 × 2 μm2. Image processing and data analysis were performed with NanoScope software version 7.30. Roughness values were calculated as root mean square (rms) values. The refractive indices and absorption coefficients of the films were measured with an M-2000D (1.25−6.5 eV) spectroscopic ellipsometer (J.A. Woollam) on a goniometric stage at an incident angle of 75°. The in situ measurements were performed with a specially modified F-120 ALD reactor (ASM Microchemistry Ltd). The pressure inside the reactor was approximately 3 mbar, and the total area of the soda lime glass substrates was approximately 3500 cm2. Argon (Oy AGA Ab, 99,999%) was used as the carrier gas. The gas species present in the reactor during the ALD cycle were detected with a Hiden HAL/3F 501 RC QMS with a Faraday cup detector and an ionization energy of 70 eV. The pressure in the QMS chamber was around 1 × 10−5 mbar, obtained by differential pumping through a 100 μm orifice. The mass changes on the substrate were recorded using a Maxtek TM 400 QCM tool with a sampling rate of 20 Hz. In the in situ QMS measurement, the background arising from the fragmentation of the precursors had to be evaluated and then subtracted from the QMS data measured during the ALD cycles for the latter to be relevant. Five reference pulses were applied to assess the contribution of the reactant to the integrated intensities measured for a specific m/z signal. The reference pulses were launched before the ALD cycles for the oxygen precursor and after the ALD cycles for the metal precursor.
crystalline phase of the films.17,18 For accurate control of the deposited phase as well as to gain knowledge of the phase transitions occurring during postdeposition annealing, the anatase to rutile phase transition is of a particular interest in ALD of TiO2. Ozone is an efficient source of oxygen in oxide ALD, notably together with organometallic precursors but also when using metal precursors coordinated through oxygen or nitrogen donors.3 In some cases, especially when films are deposited on large batches of high-aspect-ratio substrates, ozone is preferred to water because the latter suffers from long purging times.19,20 Ozone is a strong oxidizer and can therefore offer an advantage in the deposition of oxides from metal precursors that are less reactive to water. Ozone combusts organic surface groups in highly exothermic reactions, and as a consequence, carbon and nitrogen impurities are released from the film into the gas phase as oxides,3 possibly having a purifying effect on the resulting film.21−24 In some cases, the use of ozone also results in a higher film growth rate.25,26 Examples of titanium precursors which offer better growth rates with ozone than with water are amides Ti(NMe2)4, Ti(NEt2)4, and Ti(NEtMe)4 and heteroleptic Ti(OiPr)2(thd)2.27,28 An investigation of the water ALD processes of novel titanium precursors Ti(NMe2)2(OiPr)2 and Ti(OiPr)3(NiPr-Me-amd) has been published recently by our group.29 It was noted that both precursors are thermally highly stable over a large temperature range and that the Ti(NMe2)2(OiPr)2/H2O process exhibits a high growth rate. Although the films deposited using the water process were of high purity, it was speculated that ozone could be helpful in depositing even purer films. In the present study, we examine the growth characteristics of these precursors with ozone and verify the high-film purity. Also, postdeposition annealing studies were done on TiO2 thin films with different thicknesses to investigate crystallization behavior. The heteroleptic precursors in the current study contain alkoxide, alkylamide, and amidinate ligands. Several in situ studies on homoleptic precursors with alkoxide ligands have been conducted in the past.3 However, only a few mechanistic studies using QMS and QCM on heteroleptic precursors with ligands coordinated through nitrogen have been carried out.3 The reaction mechanism is more complex as several types of ligands compete during the metal precursor adsorption. Moreover, combustion reactions occurring during an ozonebased process remain poorly understood, and very few in situ studies have been conducted on such processes.3 To the best of our knowledge, the only in situ QMS and QCM mechanistic study of a metal amidinate ALD process is our recent study on the water processes of the precursors presented in this study.30 However, temperature-programmed desorption (TPD) and Xray photoelectron spectroscopy (XPS) have been used to examine copper(I) acetamidinates in connection with ALD processes.31 In the current study, an investigation of the novel Ti(NMe2)2(OiPr)2/O3 and Ti(OiPr)3(NiPr-Me-amd)/O3 ALD processes in situ with QCM and QMS aims to gain new insight into the adsorption and combustion chemistry in ozone-based ALD processes and the reactivity of heteroleptic precursors.
■
EXPERIMENTAL SECTION
The films were deposited using a commercial flow-type F-120 ALD reactor (ASM Microchemistry Ltd) with an operating pressure of 5−10 mbar. Ti(NMe2)2(OiPr)2 and Ti(OiPr)3(NiPr-Me-amd) (both from Air Liquide) were transported from a glovebox into the reactor using a syringe to minimize exposure to air and evaporated from open glass 7396
dx.doi.org/10.1021/la500893u | Langmuir 2014, 30, 7395−7404
Langmuir
Article
Figure 1. Growth rate of TiO2 films as a function of temperature using (a) Ti(OiPr)3(NiPr-Me-amd) and (b) Ti(NMe2)2(OiPr)2 as the titanium precursor. The insets depict the growth rate as a function of the metal precursor pulse length at 325 °C.
Figure 2. Left: HTXRD patterns of a 65 nm TiO2 film deposited with Ti(NMe2)2(OiPr)2 at 325 °C. Right: XRD patterns at 825−1075 °C.
(NMe2)2(OiPr)2/O3 processes as a function of the deposition temperature. The growth rate of the Ti(OiPr)3(NiPr-Me-amd)/ O3 process stays fairly constant at approximately 0.3 Å/cycle from 275 to 350 °C, whereas the growth rate of the Ti(NMe2)2(OiPr)2/O3 process increases from 0.3 to 0.9 Å/ cycle when increasing the deposition temperature from 250 to 350 °C. The insets in Figure 1 display the growth rates at 325 °C as a function of the titanium precursor pulse length. Self-limiting growth modes were confirmed at 325 °C as the growth rates saturated to constant values when increasing the precursor pulse lengths. With Ti(OiPr)3(NiPr-Me-amd), a short pulse time of 0.3 s was enough to saturate the surface, while Ti(NMe2)2(OiPr)2 needs a 0.6 s pulse. The saturated growth rates were approximately 0.3 and 0.9 Å/cycle with Ti(OiPr)3(NiPr-Meamd) and Ti(NMe2)2(OiPr)2, respectively. As found in our earlier study on the same precursors, they decompose at temperatures of 350 °C and above.29 The effect of decomposition is slightly more pronounced in the case of Ti(NMe2)2(OiPr)2, whose growth rate is almost tripled from 350 to 375 °C (Figure 1b), while in the case of Ti(OiPr)3(NiPr-Meamd) the growth rate doubles in the same temperature range (Figure 1a).
Accordingly, the following pulsing sequence was used for the Ti(NMe2)2(OiPr)2-O3 process: · 5 × (20 s O3 pulse/80 s purge) · 10 × (20 s Ti(NMe2)2(OiPr)2 precursor pulse/30 s purge/20 s O3 pulse/30 s purge) · 5 × (20 s Ti(NMe2)2(OiPr)2 precursor pulse/80 s purge) The Ti(NMe2)2(OiPr)2-O3 process was investigated at 275 and 325 °C. The pulse sequence for the Ti(OiPr)3(NiPr-Me-amd)-O3 process was the following: · 5 × (20 s O3 pulse/80 s purge) · 10 × (20 s Ti(OiPr)3(NiPr-Me-amd) precursor pulse/30 s purge/20 s O3 pulse/30 s purge) · 5 × (20 s Ti(OiPr)3(NiPr-Me-amd) precursor pulse/80 s purge) The Ti(OiPr)3(NiPr-Me-amd)-O3 process was investigated at 275 and 350 °C. The quantities of observed byproducts released during the metal precursor and ozone pulses were calculated as percentage values. That is, the total amount of byproducts released during both pulses corresponds to 100% of a specific m/z value, and the percentages reported for a given pulse are relative to this amount.
■
RESULTS AND DISCUSSION Film Growth. Figures 1a,b depict the growth rates of TiO2 films grown with the Ti(OiPr)3(NiPr-Me-amd)/O3 and Ti7397
dx.doi.org/10.1021/la500893u | Langmuir 2014, 30, 7395−7404
Langmuir
Article
crystalline growth of very thin films. However, this cannot be the whole reason for the threshold, as films of comparable thickness deposited using different precursors and on similar substrates have different phase-change temperatures (see below). TiO2 films deposited with Ti(OiPr)3(NiPr-Me-amd) did not exhibit similar phase change behavior. Films thinner than 52 nm did not show a rutile phase at 925−975 °C. However, a weak rutile reflection (2θ ≈ 27.3°) appears at 975 °C in a 52-nm-thick film (data not shown). A high stability of the anatase phase was also observed in the case of TiO2 thin films deposited using the Ti(OiPr)3(NiPr-Me-amd)/H2O process: a 65-nm-thick film deposited at 325 °C stayed as anatase when annealed at 800 °C for 30 min in N2.29 It is unclear why the films deposited with Ti(OiPr)3(NiPr-Meamd) do not change phase at similar annealing temperatures as the films of comparable thickness deposited with Ti(NMe2)2(OiPr)2. There are several parameters which may affect the phase-transition behavior, including film purity, roughness, and the size of anatase crystals.4 According to XPS studies the films deposited with the two titanium precursors are of similar purity (Table 1).
Similar growth rate behavior was also noticed in the precedent water process studies on the same precursors,29 except that the growth rate of the Ti(OiPr)3(NiPr-Me-amd)/O3 process is about 0.15 Å/cycle lower than that of the corresponding H2O process, while the growth rate of the Ti(NMe2)2(OiPr)2/O3 process is approximately 0.15 Å/cycle higher than in its H2O process. The higher growth rate of the Ti(NMe2)2(OiPr)2/O3 process as compared to that of the Ti(OiPr)3(NiPr-Me-amd)/O3 process is most likely due to the large amidinate ligand in Ti(OiPr)3(NiPrMe-amd), which requires a large area on the surface. The variation of the growth rates of the water and ozone processes at 325 °C look very similar as a function of both temperature and precursor pulse length. It seems then that with these metal precursors the nature of the oxygen precursor is not as important as the nature of the metal precursor ligands themselves. Film Properties. In order to determine the crystal structure of the films, XRD was used to investigate films deposited at various temperatures. When using Ti(OiPr)3(NiPr-Me-amd), the TiO2 films had the anatase structure at 275 °C and above, whereas at a deposition temperature of 250 °C they were amorphous. Therefore, the onset temperature of the crystallization in this process is in the range of 250−275 °C. With Ti(NMe2)2(OiPr)2 the films were in the anatase phase at 250 °C and higher temperatures. With the purpose of investigating possible thickness-dependent phase-change behavior, HTXRD studies were carried out with films of different thicknesses. Figure 2 depicts the HTXRD patterns of a 65-nm-thick film deposited with Ti(NMe2)2(OiPr)2 at 325 °C. An anatase to rutile transition begins at 875 °C, as indicated by the vanishing anatase peak at 2θ ≈ 25.5° and the appearance of the rutile peak at 2θ ≈ 27.3°. An additional but weak rutile peak at 2θ ≈ 43.7° appears at 925 °C. Phase-pure rutile is observed at 1075 °C. Figure 3 depicts the XRD data, measured at 925 °C, of TiO2 thin films of different thicknesses (23, 37, and 65 nm) deposited
Table 1. XPS Compositional Analysis of TiO2 Films Deposited from Ti(NMe2)2(OiPr)2 and Ti(OiPr)3(NiPr-Meamd) at Various Temperatures atom % precursor i
Ti(NMe2)2(O Pr)2
Ti(OiPr)3(NiPr-Meamd) a
deposition temp (°C)
Ca
Ti
O
Ti/O
375 325 275 375 325 275
