Molybdenum Atomic Layer Deposition Using MoF6 and Si2H6 as

Menno Bouman and Francisco Zaera. The Journal of Physical Chemistry C 2016 120 (15), 8232-8239. Abstract | Full Text HTML | PDF | PDF w/ Links...
1 downloads 0 Views 1MB Size
ARTICLE pubs.acs.org/cm

Molybdenum Atomic Layer Deposition Using MoF6 and Si2H6 as the Reactants D. Seghete,† G.B. Rayner, Jr.,^,† A.S. Cavanagh,‡ V.R. Anderson,† and S.M. George*,†,§ †

Department of Chemistry and Biochemistry, ‡Department of Physics, and §Department of Chemical and Biological Engineering, University of Colorado, Boulder, Colorado 80309, United States ABSTRACT: Mo ALD has been demonstrated by fluorosilane elimination chemistry using MoF6 and Si2H6 as the reactants. The nucleation and growth characteristics of Mo ALD were investigated using a variety of in situ and ex situ techniques in both high vacuum and viscous flow reactors. Quartz crystal microbalance (QCM) and X-ray reflectivity (XRR) investigations showed that Mo ALD has significant growth rate of 500-600 ng/cm2 per cycle or 6-7 Å per cycle for temperatures between 90 and 150 C. The large growth rates could result from extra Mo deposition by MoF6 f Mo þ 3F2 that may be facilitated by the very exothermic reaction of MoF6 with silicon-containing surface species. The QCM studies revealed that the Mo ALD surface chemistry is self-limiting. The QCM and Auger electron spectroscopy (AES) studies indicated that Mo ALD nucleates very rapidly on Al2O3 ALD surfaces and reaches the linear growth regime after only 4-5 ALD cycles. Oscillatory behavior for the total mass gain and individual mass gains was observed versus ALD cycle number during the nucleation region. The AES studies revealed that Mo films grown in a high vacuum reactor do not contain silicon impurities. In contrast to the AES results, Rutherford backscattering spectroscopy (RBS) analysis showed that Mo ALD films grown in a viscous flow reactor contain ∼16 at % Si impurities. X-ray photoelectron spectroscopy (XPS) analysis confirmed the presence of silicon and showed that varying temperature, precursor dose and purge parameters did not lower the Si impurities significantly. Glancing incidence X-ray diffraction (GIXRD) studies indicated that Mo ALD films were nanocrystalline. The Si impurities may exist at grain boundaries or amorphous Mo silicides as a result of Si2H6 decomposition during the highly exothermic fluorosilane elimination reaction. Fourier transform infrared (FTIR) analysis revealed that MoFx surface species are reduced to metallic Mo during the Si2H6 exposure. Because of its rapid nucleation rate, Mo ALD films could serve as ultrathin continuous conducting films or as adhesion layers for other metal ALD systems on oxide surfaces. KEYWORDS: molybdenum, atomic layer deposition, nucleation, fluorosilane elimination, molybdenum hexafluoride, disilane

I. INTRODUCTION Atomic layer deposition (ALD) is based on sequential selflimiting surface reactions and has become a well-established thin film deposition technique.1,2 The binary nature of ALD has led to a multitude of compounds deposited by ALD, such as various oxides and nitrides.3 The deposition of single component metal films by ALD has been a more difficult endeavor. Many metals have yet to be deposited using ALD techniques. The metals that can be deposited using ALD can be divided into three distinct chemical mechanism categories: combustion, reduction, and fluorosilane elimination.1 Metal ALD using combustion chemistry is characterized by the oxidation of the organic ligands of the organometallic metal precursor to produce CO2 and H2O as products. Ru and Pt ALD are examples of this combustion chemistry.4-6 Reduction chemistry for metal ALD is based on reactants that reduce the organometallic metal precursor. Cu ALD and Pd ALD using H2 and formalin as reducing agents, respectively, are examples of metal ALD using reduction chemistry.7,8 Fluorosilane elimination for metal ALD is characterized by the reaction of metal fluorides with silicon precursors, such as SiH4 and Si2H6. The driving force for this mechanism is the formation r 2011 American Chemical Society

of the very stable Si-F bond. The first fluorosilane elimination ALD system was W ALD.9 The overall equation for W ALD growth from WF6 and Si2H6 was determined to be10,11 2WF6 þ 1:5Si2 H6 f 2W þ SiHF3 þ 3:5H2 þ 2SiF4 þ HF ð1Þ This reaction is highly exothermic and has a heat of reaction of ΔH = -331 kcal and Gibbs free energy of ΔG = -360 kcal.12 The W ALD system has been characterized extensively and its growth rate, surface chemistry, and nucleation are well understood.9-11,13-16 W ALD grows linearly with a growth rate of 3-7 Å per cycle for temperatures ranging from 90 to 325 C and various Si2H6 exposures.15 W ALD growth onto an Al2O3 surface was shown to proceed through an inhibited nucleation mechanism.11,13 W ALD has been used as a very smooth reflective layer in X-ray multilayer Bragg mirrors.17,18 W ALD is also widely employed in Received: June 16, 2010 Revised: October 31, 2010 Published: March 11, 2011 1668

dx.doi.org/10.1021/cm101673u | Chem. Mater. 2011, 23, 1668–1678

Chemistry of Materials

ARTICLE

semiconductor fabrication as a nucleation layer for W chemical vapor deposition (CVD).19 Molybdenum is a metal that has similar chemical and physical properties to tungsten.20 In this work, Mo ALD is investigated using fluorosilane elimination chemistry with MoF6 and Si2H6 as the reactants. One possible simple overall reaction for Mo ALD can be9 MoF6 þ Si2 H6 f Mo þ 2SiHF3 þ 2H2

ð2Þ

This reaction has a heat of reaction of ΔH = -221 kcal and Gibbs free energy of ΔG = -235 kcal.12 This reaction has a higher exothermicity per MoF6 molecule than the exothermicity per WF6 molecule during W ALD. There is very little previous work on Mo ALD. Only one previous study investigated Mo ALD growth using MoF6 and Zn as the reactants.21 The problems with this system were the low volatility of Zn, the low growth rates of 0.6 - 0.8 Å per cycle and the resulting Zn impurities in the Mo film.21 There have also been only a few studies that have reported Mo CVD using MoF6 and H2 or SiH4.22-25 The current study demonstrates that Mo ALD can be grown efficiently using MoF6 and Si2H6 reactants. The growth of Mo ALD was characterized using quartz crystal microbalance (QCM), X-ray reflectivity (XRR), glancing incidence X-ray diffraction (GIXRD), Rutherford Backscattering (RBS) and X-ray photoelectron (XPS) studies. The nucleation of Mo ALD was studied using QCM and Auger electron spectroscopy (AES) measurements. Mo ALD has similarities with W ALD and displays rapid nucleation and a large growth per cycle. Mo ALD also displays some differences compared with W ALD that may result from its extremely exothermic heat of reaction.

II. EXPERIMENTAL SECTION A. ALD Reactors. Mo ALD was studied both in a hot wall viscous flow reactor and in a cold wall high vacuum reactor. The viscous flow reactor was similar to the one described in detail elsewhere.11 The base pressure in the viscous flow ALD reactor was 1.3 Torr with a N2 gas flow of 200 sccm to the mechanical pump. The high vacuum apparatus used in this study was described in detail previously.26 Briefly, this reactor consisted of a sample introduction chamber, two reaction chambers, and an AES analysis chamber. One of the reaction chambers was used for sample cleaning using H atoms produced by the dissociation of H2 on a W filament.26 After cleaning, the samples were transferred into the second reaction chamber that was used for ALD growth using static exposures of the MoF6 and Si2H6 precursors. Transmission Fourier transform infrared (FTIR) studies of the Mo ALD surface chemistry were performed in a warm wall ALD reactor that has been described in detail previously.27,28 The walls of the reactor were heated to 90 C. The sample was independently heated to the desired reaction temperature of 120 C. The base pressure in the FTIR ALD reactor was 1.2 Torr with a N2 gas flow of 200 sccm to the mechanical pump. B. Sample Preparation and Reaction Conditions. Mo ALD was performed on n-type Si(100) wafers (SVM, Inc.) that were cut into 2.5  2.5 cm2 samples. Traces of organic impurities and large particles were removed from the substrates by rinsing with acetone, then isopropanol, and followed by drying with nitrogen. For the AES studies, 100 Å of Al2O3 ALD were grown onto the Si wafers using trimethylaluminum (TMA) and H2O in the viscous flow ALD reactor at 170 C. Since these samples were exposed to air prior to loading in the high vacuum system, the surfaces were contaminated by hydrocarbon impurities. Before Mo ALD growth, the Al2O3 surface was cleaned using exposure to atomic H.26 The conditions used for removing surface contamination were: 300 C substrate temperature, 2.4  105 L H2

exposure, and a W filament temperature of ∼2000 K. After exposure to atomic H, the C AES features were no longer detected on the Al2O3 ALD surface. Mo ALD films were grown using MoF6 (Aldrich, 99.9þ%; Alfa-Aesar 99.9þ%; Synquest Laboratories, Inc. 99þ%) and Si2H6 (Voltaix, 99.998þ%). Pneumatic valves were used to pulse the reactants into the N2 carrier gas stream in the viscous flow reactor. The pulse sequence was defined by (t1, t2, t3, t4) where t1 is the MoF6 dose time, t2 is the purge time after the MoF6 dose, t3 is the Si2H6 dose time, and t4 is the purge time after the Si2H6 dose. The times for the pulse sequences are given in seconds. The pressure was measured at the back of the reactor with a 10 Torr capacitance manometer (MKS, Andover, MA). Al2O3 ALD films in the viscous flow reactor were grown using TMA (Sigma Aldrich, 98þ%) and H2O (Fisher, Optima). The Mo ALD nucleation and growth studies in the viscous flow reactor were performed at 120 C. The AES nucleation study was performed at 130 C. The viscous flow reactor was maintained at the desired temperature using external ceramic fiber heaters (Watlow) controlled by a proportional integral derivative (PID) controller (Eurotherm, 2604). In the high vacuum reactor, the samples were heated by placing them in direct contact with a resistively heated pyrolitic BN heating element.26 A heater temperature of 210 C corresponded to a substrate temperature of 130 C.26 For the FTIR study, the Mo ALD films were coated onto SiO2 nanopowders supported in a 2  3 cm2 tungsten grid.27,28 The SiO2 nanoparticles were used as substrates because they have a high surface area that provides excellent signal-to-noise in the FTIR experiment. The tungsten grid served as both a support for the nanoparticles and a resistive electrical heater. A K-type thermocouple attached to the tungsten mesh was connected to a feedback controller to control the substrate temperature. C. Thin Film Analysis. QCM studies were performed in situ in the viscous flow reactor using a modified BSH-150 bakeable sensor head (Inficon, Inc.).29 The crystals used in the sensor were polished, goldcoated, AT-cut, 6 MHz quartz discs (Colorado Crystal Corporation). As received gold-coated quartz crystals were coated with 300 Å thick Al2O3 ALD films to create a reproducible starting surface. All QCM experiments started with the deposition of a 50 Å thick Al2O3 ALD film to ensure an identical initial substrate. The period of the crystal was recorded every 100 ms using a thin-film deposition monitor (Inficon Inc., TM400). By replacing the quartz crystal after 150 C and reaches a growth rate of ∼10 Å per cycle at 270 C. This increase in Mo ALD growth rate is attributed to MoF6 decomposition via MoF6 f Mo þ 3F2 that may be facilitated by the highly exothermic MoF6 reaction and may be enhanced at higher surface temperatures. There may also be some contribution from increased Si2H6 insertion into Si-H bonds or Si2H6 decomposition at higher temperatures and additional MoF6 reaction with these additional Si surface species.14,15 The surface roughnesses of the Mo ALD films was also measured using XRR after 50 ALD cycles at the various substrate temperatures. The root-mean-square (rms) surface roughnesses were 8, 9, 13, 14, and 23 Å for film thicknesses of 298, 277, 332, 375, and 501 Å at growth temperatures of 90, 120, 150, 180, and 270 C, respectively. A roughness that is equal to the square root of the film thickness is expected for a random CVD process.35 The roughness is much lower than the square root of the film thickness at the lower temperatures of 90 and 120 C. The roughness at 270 C is comparable to the square root of the film thickness. The surface roughness is consistent with the MoF6 f Mo þ 3F2 decomposition reaction contributing more to the Mo deposition at the higher temperatures. C. Nucleation of Mo ALD on Al2O3 ALD Surfaces. The nucleation of Mo ALD on Al2O3 ALD surfaces at 120 C was studied using both QCM and AES investigations. Figure 6 shows the mass gains for the MoF6 and Si2H6 exposures for 30 cycles of Mo ALD at 120 C using the pulse sequence of (1.5,

ARTICLE

Figure 6. QCM mass gains for each reactant exposure for 30 Mo ALD cycles on an Al2O3 ALD surface at 120 C using a pulse sequence of (1.5, 60, 6, 30).

60, 6, 30). During the first three cycles, the mass gains for the MoF6 exposures and the mass losses for the Si2H6 exposures are both increasing. The total MGPC reaches a maximum during the third ALD cycle. Subsequently, there is a decrease of both the mass gain for the MoF6 exposure and mass loss for the Si2H6 exposure during the fourth cycle. The mass gains and losses are approximately constant versus ALD cycle number after five ALD cycles. There is a distinct oscillating behavior of the mass gains and losses for the individual reactants and the total MGPC during the first several cycles in Figure 6. This oscillating behavior is very similar to the behavior observed earlier for W ALD on Al2O3 ALD surfaces.11 A portion of this behavior is indicative of an island growth mechanism for Mo ALD on Al2O3 ALD surfaces known as “substrate-inhibited growth of type 2”.36 The island growth mechanism is characterized by: formation of nucleation centers; radial growth of nucleation centers into islands; coalescence of islands to form a continuous film; and, finally, steadystate growth. Numerical simulations have predicted that ALD growth via island nucleation leads to a MGPC maximum before reaching a steady state value.37,38 The maximum in MGPC is correlated with the maximum in surface area that is reached when the islands just begin to coalesce. The maximum in the MGPC after only 3 ALD cycles in Figure 6 suggests that there is a very high density of nucleation centers for Mo ALD. For comparison, W ALD did not reach a maximum MGPC until after 8 ALD cycles.11 The maximum MGPC measured for Mo ALD for the third ALD cycle was 661 ng/cm2 and the steady state MGPC for Mo ALD was 526 ng/cm2. The ratio between the maximum MGPC and the steady state MGPC is 661/526 = 1.26. In good agreement, the predicted ratio between the maximum MGPC and the steady state MGPC for nucleation and growth based on island growth for an amorphous film is estimated to be ∼1.2 by numerical modeling.37 The oscillatory behavior of the mass gains and losses in Figure 6 is defined primarily by the pronounced decrease in mass gain for MoF6 and mass loss for Si2H6 in the fourth ALD cycle. This behavior is not expected from simple island growth and coalescence. A similar oscillation was observed for the nucleation of W ALD on Al2O3 ALD surfaces.11,39 This oscillation may be related to recrystallization of the underlying Mo metal film and the subsequent renucleation of additional Mo islands on top of the underlying Mo metal film.11 This behavior is 1672

dx.doi.org/10.1021/cm101673u |Chem. Mater. 2011, 23, 1668–1678

Chemistry of Materials

ARTICLE

Figure 7. QCM mass gains for 8 Mo ALD cycles grown on an Al2O3 ALD surface at 120 C. A series of 6 MoF6 minidoses and 10 Si2H6 minidoses were employed using a pulse sequence of (6(0.5, 90), 10(1, 30)).

Figure 9. Normalized Si LMM, Mo MNN and O KLL AES signal intensities versus the number of Mo ALD cycles grown on an Al2O3 ALD surface at 130 C. The AES experiments were performed in a high vacuum chamber.

Figure 8. Auger electron spectroscopy (AES) spectra of Mo ALD growth on an Al2O3 ALD surface at 130 C. AES results are shown for the initial Al2O3 ALD surface and after the Si2H6 and MoF6 doses during the 2nd and 5th Mo ALD cycles.

not fully understood and indicates the complexity of metal ALD nucleation. The Mo ALD nucleation was also studied using a series of minidoses to define the MoF6 and Si2H6 exposures. Figure 7 displays the results for the first 8 ALD cycles using minidoses of MoF6 and Si2H6. Each ALD cycle contained six MoF6 minidoses and ten Si2H6 minidoses. The pulse sequence was (6(0.5, 90), 10(1, 30)). Each vertical bar represents the mass change corresponding to each minidose. The MoF6 minidoses lead to mass gain and the Si2H6 minidoses lead to mass loss. Figure 7 indicates that only two multidose ALD cycles are necessary for Mo ALD to reach an approximately reproducible growth regime on Al2O3 ALD surfaces. As expected from the earlier QCM results in Figure 3, the mass loss from the Si2H6 minidoses occurs primarily on the first Si2H6 minidose after two complete ALD cycles. Figure 7 shows that there is an unexpected large mass loss after the ninth Si2H6 minidose during the second complete ALD cycle. This large pronounced mass loss after the ninth Si2H6 minidose is always followed by the maximum MoF6 mass gain. This

puzzling behavior was reproducible and may be related to the oscillatory mass gains observed during the first 3-4 ALD cycles in Figure 6. These unexpected phenomena indicate the richness of the initial nucleation process. Figure 8 shows the AES spectra that were collected during Mo ALD nucleation on Al2O3 ALD surfaces at 130 C. These AES measurements were performed after each Si2H6 and MoF6 exposure and were recorded at separate positions on the sample surface. No AES measurements were repeated at the same position. The MoF6 exposures were 6  105 L (10 mTorr  60 s) and the Si2H6 exposures were 1.5  106 L (25 mTorr  60 s). The Al2O3 starting surface was exposed to atomic H prior to the Mo ALD deposition. The bottom spectrum in Figure 8 displays the initial AES spectrum of the Al2O3 surface. This AES spectrum shows only two Al features at 35 and 51 eV and a dominant O feature at 503 eV. For these AES studies, the first dose onto the Al2O3 ALD surface was Si2H6. Figure 8 shows the AES feature at 90 eV that is attributed to the Si LMM signal during the second ALD cycle after the Si2H6 exposure. A reduction in the O and Al AES signal intensities is also observed after this Si2H6 dose. During the second ALD cycle after the MoF6 exposure, the AES spectrum shows a further attenuation of the Al and O AES signals. Additionally, new features appear at 186 and 219 eV that are consistent with the Mo MNN AES signal. The AES spectra during the fifth ALD cycle are presented at the top of Figure 8. After the exposure to Si2H6 during the fifth ALD cycle, the Si AES peak reappeared at 90 eV. This peak was qualitatively larger than the Si AES peak observed during the second ALD cycle. During the fifth ALD cycle after the MoF6 exposure, the O AES signal was reduced to an insignificant value. The Mo AES peaks also grew in comparison with the Mo AES peaks measured during the second ALD cycle. A more detailed view of Mo ALD nucleation on Al2O3 ALD surfaces at 130 C is obtained by observing the Si, Mo and O AES signals for seven consecutive ALD cycles. Figure 9 shows the normalized Si LMM, Mo MNN and O KLL Auger signal intensities versus the number of Mo ALD cycles. A rapid attenuation 1673

dx.doi.org/10.1021/cm101673u |Chem. Mater. 2011, 23, 1668–1678

Chemistry of Materials

ARTICLE

Figure 11. Mass changes after the MoF6 and Si2H6 reactions and after the complete Mo ALD cycle versus the MoF6 purge time at 120 C. The pulse sequence was (1, x, 6, 30). Figure 10. Mo ALD film thickness versus the number of Mo ALD cycles. Mo film thickness was determined from Mo/O AES signal ratios using the Cumpson method. Results are shown for the 1st (circles) and 7th (triangles) set of AES spectra.

of the O KLL signal was observed over 2-3 ALD cycles. This behavior suggests that a continuous Mo ALD film is depositing on the Al2O3 ALD surface. Similar AES studies of the W ALD chemistry using WF6 and Si2H6 have shown that the attenuation of the underlying O KLL signal from the initial Al2O3 ALD surface occurs only after 10 W ALD cycles.16 In agreement with the QCM results, this behavior indicates that Mo ALD nucleation is much faster than W ALD nucleation on the Al2O3 ALD surface. Although the O KLL signal is rapidly attenuated, Figure 9 also shows that a small O KLL signal persists after 7 ALD cycles. This O KLL signal is attributed to a low H2O partial pressure in the high vacuum chamber. Figure 9 also shows an increase in the Si LMM signal after each Si2H6 exposure and an increase in the Mo MNN signal after each MoF6 exposure. After each MoF6 exposure, the Si AES signal is reduced to insignificant levels. This disappearance of the Si AES signal suggests the complete removal of the Si species from the surface. After each Si2H6 exposure, the Mo AES signal is reduced by approximately 50%. This reduction is caused by the shadowing of the underlying Mo ALD film by the Si species on the surface. Figure 9 also reveals that the oscillating Si and Mo signals display reproducible behavior after the second ALD cycle. This behavior suggests that the Mo ALD has reached the steady state region after only 2-3 ALD cycles. The Mo MNN and O KLL signal intensities can be used to quantify the Mo film thickness during Mo ALD on the initial Al2O3 ALD surface.26 Modeling of the AES signals is based on exponential attenuation of Auger electrons traveling through the Mo film.26 The model assumes a thin, uniform Mo film adsorbed on an infinitely thick substrate. Following the Cumpson method,40 AES signals are defined for the Mo film and the underlying Al2O3 substrate. The Mo AES signal intensity for a Mo film thickness, d, is defined according to: I Mo ðdÞ=σMo ¼ 1 - exp½ - d=ðλMo cos θÞ

ð9Þ

where σMo is the Mo AES sensitivity factor, λMo is the attenuation length and θ is the angle between the AES detector axis and the surface normal. Similarly, the O AES substrate signal intensity

with an overlayer film of thickness, d, is defined as: I Ox ðdÞ=σOx ¼ exp½ - d=ðλOx cos θÞ

ð10Þ

In eq 10, σOx is the O AES sensitivity factor and λOx is the attenuation length. The ratio of the Mo and O AES peak intensities in eq 9 and eq 10 yields the following relationship ln½ðI Mo =σMo Þ=ðI Ox =σOx Þ - ½ðλMo =λOx Þ - 1=2 3 ½d=ðλMo 3 cos θÞ

- ln 2 ¼ ln½sinh½d=ð2 3 λMo 3 cos θÞ

ð11Þ

Equation 11 can be used to determine the Mo film thickness, d, from the Mo and O AES signals, sensitivity factors and attenuation lengths. The sensitivity factors are σMo = 0.34 for the Mo MNN AES signal at ∼186 eV and σOx = 0.50 for the O KLL AES signal at ∼503 eV for a primary beam energy of 3 keV.41 A semiempirical equation has been developed for estimating attenuation lengths for electrons with kinetic energy between 50 and 2000 eV.42 The attenuation lengths used for Mo Auger electrons at 186 eV and O Auger electrons at 503 eV, were λMo = 6.2 Å and λOx = 11.0 Å, respectively. These attenuation lengths were determined using an atomic number Z = 42 for Mo and an overlayer lattice parameter of R = 0.31470 nm.20 The thickness of the Mo ALD layer can be obtained from the ratio of the Mo MNN and O KLL AES signals. The AES spectra collected at a primary beam energies of 2 and 3 keV were analyzed to determine the Mo/O AES signal ratios. For these experiments, a total of seven consecutive spectra were collected after each Mo ALD reaction. These multiple AES spectra were performed to determine the stability and reproducibility of the AES results, evaluate differences between primary beam energies of 2 and 3 keV, and obtain integrated electron intensity versus energy over the entire energy spectrum for normalization. The Mo/O AES signal ratios from the first and the seventh set of AES spectra recorded with a primary beam energy of 3 keV were utilized in eq 7. The obtained Mo film thicknesses versus number of ALD cycles are presented in Figure 10. The initial growth rate of Mo ALD on the Al2O3 ALD surface at 130 C can be calculated from a linear fit of the Mo film thickness versus number of ALD cycles for the first 4 ALD cycles. As shown in Figure 10, the growth rate of Mo ALD on the Al2O3 ALD surface ranges from 5.7 to 7.4 Å/cycle for the seventh and first AES scans, respectively. The only difference between the two 1674

dx.doi.org/10.1021/cm101673u |Chem. Mater. 2011, 23, 1668–1678

Chemistry of Materials

Figure 12. Difference FTIR spectra after 4 Si2H6 minidoses during the 2nd Mo ALD cycle. All spectra are referenced to the spectrum after the last MoF6 dose.

AES scans was the time between recording the first and seventh AES spectra. These growth rates are in excellent agreement with the growth rates of 6-7 Å per cycle obtained by the QCM and XRR studies at temperatures between 90 and 150 C. For cycles after the fourth ALD cycle, the Cumpson method fails to produce a reliable film thickness because the O AES signal is strongly attenuated by the Mo ALD overlayer. The AES analysis estimates a total Mo ALD film thickness of 22 Å after 4 ALD cycles. For a similar experiment in the viscous flow reactor shown in Figure 6, the QCM measured a cumulative mass of 1781 ng/cm2 deposited after the first 4 ALD cycles. Using the XRR density for Mo ALD of 8.8 g/cm3 at 120 C, the equivalent thickness after 4 ALD cycles is 20.2 Å. The agreement between the QCM and AES analysis is very good. D. Effect of MoF6 Purge Time and Reduction of MoFx Surface Species. Given the unusual behavior following the first MoF6 minidose in Figure 3, additional investigations explored the effect of MoF6 purge time on the Mo ALD growth rate. Figure 11 shows the variation in mass changes measured at 120 C for the MoF6 and Si2H6 reactions and for the whole entire Mo ALD cycle versus the MoF6 purge time. Each data point in Figure 11 was obtained in the steady state growth region using a pulse sequence of (1, x, 6, 30). Figure 11 shows that mass gained during the MoF6 exposure decreases slightly versus the MoF6 purge time. This behavior is consistent with desorption of either physisorbed MoF6 or reaction products. Concurrent with the slight decrease in the MGPC after the MoF6 exposure, the MGPC after the Si2H6 exposure increases slightly. The net result is that the total mass gain per Mo ALD cycle does not vary with the MoF6 purge time. The growth of Mo ALD was also studied using transmission FTIR spectroscopy on SiO2 nanoparticles. The nanoparticles provide a high surface area that facilitates the observation of surface species. In addition, FTIR spectroscopy can also observe the deposition of a conducting metal because the metal leads to background infrared absorption. Figure 12 shows FTIR difference spectra after four consecutive Si2H6 minidoses during the second Mo ALD cycle after a MoF6 exposure. All four spectra shown in Figure 12 are referenced to the FTIR spectrum after the previous MoF6 exposure. Figure 12 reveals positive absorbance features in the Si-H stretching and in the Si-H scissors region of the spectrum. The

ARTICLE

Figure 13. GIXRD scan of a Mo ALD film with a thickness of 300 Å grown at 90 C on an Al2O3 ALD film with a thickness of 55 Å on a Si wafer.

Figure 14. RBS spectrum of a Mo ALD film with a thickness of 600 Å film grown at 90 C on an Al2O3 ALD substrate with a thickness of 20 Å deposited on a glassy C substrate.

features at 2287 and 2135 cm-1 are correlated to absorptions in the Si-H stretching region resulting from SiHxFy and SiHx species, respectively.43,44 The larger absorption peak at 2135 cm-1 compared with 2287 cm-1 suggests a larger concentration of SiHx species compared with SiHxFy species on the surface. This suggests that the majority of the surface species have x = 3 for reaction B in eq 6 during the second ALD cycle on the SiO2 surface in the nucleation region. The surface species could change following nucleation in the linear growth region. The feature at 950 cm-1 is attributed to the SiH2 scissors mode that is expected after a Si2H6 reaction that produces SiH2F* or SiH3* species.45 The most prominent feature observed by the spectra in Figure 12 is the rapid rise of the background absorbance after each Si2H6 minidose. This behavior is attributed to the reduction of MoFx surface species to metallic Mo. Si2H6 is a strong reducer and reduces the MoFx surface species with an oxidation state of Moxþ to Mo0. The production of metallic Mo increases the background infrared absorption resulting from Drude absorption. The FTIR experiments could only be conducted for a few ALD cycles because the background Drude absorption rapidly saturated the infrared absorption. The lack of absorption at 1000-1250 cm-1 in the difference spectrum corresponds with absorption by the SiO2 nanoparticles. 1675

dx.doi.org/10.1021/cm101673u |Chem. Mater. 2011, 23, 1668–1678

Chemistry of Materials

ARTICLE

Table 1. Atomic Composition of Mo ALD Films Grown at Various Deposition Conditions Measured by XPS Analysis dosing conditions sample

MoF6

XPS elemental analysis (atomic %) Si2H6

temp

Mo

Si

O

1

1 s/120 s

2 s/20 s

120 C

78.25

21.75

2

1 s/120 s  3

2 s/20 s  3

120 C

77.36

19.67

2.96

3

1 s/30 s

8 s/20 s

120 C

79.33

18.52

2.08

4

1 s/120 s

8 s/20 s

90 C

69.32

30.68

5

1 s/120 s

8 s/20 s

120 C

70.82

21.85

7.33

6

1 s/120 s

8 s/20 s

150 C

79.65

17.67

2.69

E. Crystallinity, Silicon Impurities and Electrical Resistivity. The crystalline structure of the Mo ALD films grown in the viscous flow reactor was investigated ex situ by GIXRD. Figure 13 shows the GIXRD scan obtained for a Mo ALD film with a thickness of 300 Å deposited at 90 C. This Mo ALD film was grown on an Al2O3 ALD film with a thickness of 55 Å on a Si wafer. The largest XRD peak observed at 40.6 corresponds to the Mo (110) plane.46 The broad shoulder feature from 50 to 70 results from the amorphous Al2O3 substrate. The lack of XRD peaks at 46 and 43 indicates that the MoSi2 and Mo5Si3 crystalline phases, respectively, are not present.47,48 The full width at half-maximum (fwhm) for the Mo (110) peak is 5.92. This broad diffraction peak is characteristic of small Mo crystallites. Using the Scherrer formula, the measured fwhm corresponds to an average Mo crystalline diameter of 1.5 nm.49 The crystalline structure of the Mo ALD film is similar to the previously reported crystallinity of W ALD films.9 The composition of the Mo ALD films grown in the viscous flow reactor was also determined ex situ by Rutherford backscattering spectroscopy (RBS) and X-ray photoelectron spectroscopy (XPS). Figure 14 shows the RBS spectrum of a Mo ALD film with a thickness of 600 Å film deposited at 90 C. This Mo ALD film was grown on an Al2O3 ALD film with a thickness of 20 Å film deposited on a glassy carbon substrate. The RBS fitting analysis indicates that the bulk of the Mo ALD film contains Si impurities at 16.7 atomic percent. In addition, no O, F or C impurities are present in the bulk of the Mo ALD film. Given that the GIXRD results observe the presence of Mo nanocrystallites in the Mo ALD film, the Si impurities in the Mo ALD film are either believed to be segregated atomic Si at grain boundaries or result from amorphous MoSi2 or Mo5Si3. XPS was used to further investigate the effects of reaction conditions on the Si impurities in the Mo ALD films grown in the viscous flow reactor. Table 1 shows the composition of various Mo ALD films grown at three different temperatures, and at various MoF6 and Si2H6 exposure and purge times. All samples used in the XPS analysis consisted of 50 cycles of Mo ALD deposited on an Al2O3 ALD surface on a Si wafer. The elemental atomic percents shown in Table 1 were obtained from XPS spectra taken after the top layer was removed by sputtering. Some Mo ALD samples showed oxygen impurities after sputtering. Because the RBS analysis did not confirm the presence of oxygen, these oxygen impurities may be explained by preferential sputtering where O atoms are maintained at the surface throughout the sputtering process. The XPS analysis reveals silicon at even higher atomic percentages than the RBS analysis. The silicon impurity concentration in the Mo ALD films grown in the viscous flow reactor decreases with increasing temperature. Decreasing the Si2H6

dose time or decreasing the MoF6 purge time did not significantly affect the silicon impurity concentration. The Si impurities are believed to be produced by Si2H6 decomposition during reaction with MFx surface species. The large exothermicity of the surface reaction may facilitate the decomposition of additional Si2H6 and lead to Si CVD. In contrast to the RBS and XPS results for the Mo ALD films grown in the viscous flow reactor, the AES results for the Mo ALD films grown in the high vacuum reactor did not observe any Si impurities in the Mo ALD films after the MoF6 exposures. The Si2H6 exposures in the viscous flow reactor and the high vacuum reactor were very comparable. The typical exposure of 1.5  106 L Si2H6 for the AES experiments in the high vacuum reactor was similar to the estimated exposure of 2.2  106 L Si2H6 used in the viscous flow ALD reactor. However, there was a time difference in the two experiments. Si2H6 was introduced in the high vacuum reactor slowly via a manual leak valve and the Si2H6 exposure was static for 60 s. In the viscous flow reactor, the Si2H6 was introduced by opening a pneumatic valve to the Si2H6 reservoir at 30 Torr. The pneumatic valves produced Si2H6 pressure spikes of ∼0.4-0.6 Torr for 6-8 s in the reactor. The larger Si2H6 pressures for shorter times in the viscous flow reactor are believed to lead to much more violent reactions of Si2H6 with the MoFx surface species. The high exothermicity of this reaction could heat the surface and enable Si2H6 decomposition and lead to Si CVD. In contrast, the smaller Si2H6 pressures and longer time for the Si2H6 exposures in the high vacuum reactor may have not heated the surface. Without the extra heating, perhaps the Si2H6 is not decomposing or depositing additional silicon on the surface. The electrical resistivity of the Mo ALD films was also measured using the four point probe technique. The samples were Mo ALD films with a thickness of 300 Å grown at 120 C. The Mo ALD was deposited on Al2O3 ALD films with a thickness of 20 Å that were deposited on both glass and thermal oxides on Si wafers. The average measured resistivity of the Mo ALD film was 124 μΩ 3 cm. This resistivity is in the range of previous studies that have reported DC-sputtered and e-beam evaporated Mo films with electrical resistivities ranging from 50 to 156 μΩ 3 cm, respectively.50,51 In comparison, the resistivity for pure Mo metal is 5.4 μΩ 3 cm.20 The higher resistivity of the sputtered, e-beam evaporated and Mo ALD films is attributed to their smaller crystalline grains and more amorphous structure.

IV. CONCLUSIONS Mo ALD was demonstrated using MoF6 and Si2H6 as the reactants. The growth and film properties of Mo ALD on Al2O3 1676

dx.doi.org/10.1021/cm101673u |Chem. Mater. 2011, 23, 1668–1678

Chemistry of Materials ALD surfaces were investigated using a variety of techniques. QCM studies revealed that the MoF6 exposure produced a mass gain and the Si2H6 exposure led to mass loss. These mass changes were different than observed earlier for W ALD where a mass gain was observed after Si2H6 exposures. The Mo ALD growth rate was 6-7 Å per cycle at temperatures between 90 and 150 C. Higher Mo ALD growth rates occurred at higher temperature. These larger growth rates are believed to result from Mo deposition by MoF6 f Mo þ 3F2. The very exothermic MoF6 reaction may facilitate the endothermic MoF6 decomposition reaction. Mo ALD nucleated very rapidly on Al2O3 ALD surfaces and reached the linear growth regime after only 4-5 ALD cycles. Interesting oscillatory behavior was observed versus ALD cycle number for the total mass gains and individual mass gains after the MoF6 and Si2H6 exposures during the nucleation region. AES analysis revealed that the Mo ALD films grown in a high vacuum chamber using lower reaction pressures and longer reactant times did not contain any Si impurities. In contrast, Mo ALD films grown in the hot wall viscous flow reactor were found by RBS to contain silicon impurities at a level of ∼16 at%. The Si impurities could result from the large exothermicity of the Mo ALD reaction that may favor Si2H6 decomposition during higher reaction rates. The Mo ALD film was composed of Mo nanocrystallites. The Si impurities may either exist at the grain boundaries of nanocrystallites or in segregated amorphous MoSi2 or Mo5Si3 regions. FTIR studies showed that the MoFx surface species were reduced to metallic Mo during the Si2H6 exposure. Four point probe measurements yielded a film resistivity of 124 μΩ 3 cm. The rapid nucleation rate and large growth rate of Mo ALD could allow Mo ALD to serve as a nucleation layer for other metal ALD systems on oxide surfaces.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]. Present Addresses ^

Kurt J. Lesker Company, 1515 Worthington Ave., Clairton, PA 15025.

’ ACKNOWLEDGMENT This work was supported by the U.S. Department of Energy, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering under Award DE-FG02-06ER46348, the National Science Foundation under Grant CHE-0715552 and International Sematech. Some of the equipment used in this research was provided by the Air Force Office of Scientific Research. The RBS analysis was performed at the Nanofabrication Center (NFC) at the University of Minnesota. The NFC receives partial support from NSF through the National Nanotechnology Infrastructure Network (NNIN) program. The authors also thank Jacob Bertrand for help with the four point probe measurements. ’ REFERENCES (1) George, S. M. Chem. Rev. 2010, 110, 111. (2) Suntola, T. Thin Solid Films 1992, 216, 84. (3) Puurunen, R. L. J. Appl. Phys. 2005, 97, 121301. (4) Aaltonen, T.; Alen, P.; Ritala, M.; Leskel€a, M. Chem. Vapor Depos. 2003, 9, 45. (5) Aaltonen, T.; Rahtu, A.; Ritala, M.; Leskela, M. Electrochem. Solid-State Lett. 2003, 6, C130.

ARTICLE

(6) Aaltonen, T.; Ritala, M.; Sajavaara, T.; Keinonen, J.; Leskela, M. Chem. Mater. 2003, 15, 1924. (7) Martensson, P.; Carlsson, J.-O. J. Electrochem. Soc. 1998, 145, 2926. (8) Elam, J. W.; Zinovev, A.; Han, C. Y.; Wang, H. H.; Welp, U.; Hryn, J. N.; Pellin, M. J. Thin Solid Films 2006, 515, 1664. (9) Klaus, J. W.; Ferro, S. J.; George, S. M. Thin Solid Films 2000, 360, 145. (10) Grubbs, R. K.; Steinmetz, N. J.; George, S. M. J. Vac. Sci. Technol. B 2004, 22, 1811. (11) Wind, R. W.; Fabreguette, F. H.; Sechrist, Z. A.; George, S. M. J. Appl. Phys. 2009, 105, 074309. (12) HSC Chemistry 5.1; Outokumpu Research Oy: Pori, Finland. Values are given at 273 K. (13) Elam, J. W.; Nelson, C. E.; Grubbs, R. K.; George, S. M. Thin Solid Films 2001, 386, 41. (14) Elam, J. W.; Nelson, C. E.; Grubbs, R. K.; George, S. M. Surf. Sci. 2001, 479, 121. (15) Fabreguette, F. H.; Sechrist, Z. A.; Elam, J. W.; George, S. M. Thin Solid Films 2005, 488, 103. (16) Grubbs, R. K.; Nelson, C. E.; Steinmetz, N. J.; George, S. M. Thin Solid Films 2004, 467, 16. (17) Fabreguette, F. H.; George, S. M. Thin Solid Films 2007, 515, 7177. (18) Fabreguette, F. H.; Wind, R. A.; George, S. M. Appl. Phys. Lett. 2006, 88, 013116. (19) Luoh, T.; Su, C.-T.; Yang, T.-H.; Chen, K.-C.; Lu, C.-Y. Microelectron. Eng. 2008, 85, 1739. (20) Handbook of Chemistry and Physics, 72nd ed.; CRC Press: Boston, 1997. (21) Juppo, M.; Vehkamaki, M.; Ritala, M.; Leskela, M. J. Vac. Sci. Technol. A 1998, 16, 2845. (22) Bjorklund, K. L.; Heszler, P.; Boman, M. Appl. Surf. Sci. 2002, 186, 179. (23) Harsta, A.; Carlsson, J. O. Thin Solid Films 1990, 185, 235. (24) Lifshitz, N.; Green, M. L. J. Electrochem. Soc. 1988, 135, 1832. (25) Lifshitz, N.; Williams, D. S.; Capio, C. D.; Brown, J. M. J. Electrochem. Soc. 1987, 134, 2061. (26) Rayner, G. B.; George, S. M. J. Vac. Sci. Technol. A 2009, 27, 716. (27) Ferguson, J. D.; Weimer, A. W.; George, S. M. Thin Solid Films 2000, 371, 95. (28) Goldstein, D. N.; McCormick, J. A.; George, S. M. J. Phys. Chem. C 2008, 112, 19530. (29) Elam, J. W.; Groner, M. D.; George, S. M. Rev. Sci. Instrum. 2002, 73, 2981. (30) Porter, W. A.; Anderson, S. W. J. Vac. Sci. Technol. 1972, 9, 1472. (31) Rocklein, M. N.; George, S. M. Anal. Chem. 2003, 75, 4975. (32) Choudhary, V. R.; Rajput, A. M.; Prabhakar, B. Angew. Chem., Int. Ed. 1994, 33, 2104. (33) Choudhary, V. R.; Uphade, B. S.; Mulla, S. A. R. Angew. Chem., Int. Ed. 1995, 34, 665. (34) Itoh, N.; Wu, T. H. J. Membr. Sci. 1997, 124, 213. (35) Smith, D. L. Thin Film Deposition: Principles and Practice; McGraw-Hill: New York, 1995. (36) Puurunen, R. L.; Vandervorst, W. J. Appl. Phys. 2004, 96, 7686.  (37) Nilsen, O.; Karlsen, O. B.; Kjekshus, A.; Fjellvag, H. Thin Solid Films 2007, 515, 4527. (38) Nilsen, O.; Mohn, C. E.; Kjekshus, A.; Fjellvag, H. J. Appl. Phys. 2007, 102, 024906. (39) Sechrist, Z. A.; Fabreguette, F. H.; Heintz, O.; Phung, T. M.; Johnson, D. C.; George, S. M. Chem. Mater. 2005, 17, 3475. (40) Cumpson, P. J. Surf. Interface Anal. 2000, 29, 403. (41) Database in AugerScan 3; RBD Enterprises: Bend, OR, 2005. (42) Cumpson, P. J.; Seah, M. P. Surf. Interface Anal. 1997, 25, 430. (43) Kobayashi, N.; Goto, H.; Suzuki, M. J. Appl. Phys. 1991, 69, 1013. (44) Kobayashi, N.; Nakamura, Y.; Goto, H.; Homma, Y. J. Appl. Phys. 1993, 73, 4637. 1677

dx.doi.org/10.1021/cm101673u |Chem. Mater. 2011, 23, 1668–1678

Chemistry of Materials

ARTICLE

(45) Dillon, A. C.; Robinson, M. B.; George, S. M. Surf. Sci. 1993, 295, L998. (46) Edwards, J. W.; Speiser, R.; Johnston, H. L. J. Appl. Phys. 1951, 22, 424. (47) Schwarz, R. B.; Srinivasan, S. R.; Petrovic, J. J.; Maggiore, C. J. Mater. Sci. Eng., A 1992, 155, 75. (48) Chu, F.; Thoma, D. J.; McClellan, K.; Peralta, P.; He, Y. Intermetallics 1999, 7, 611. (49) Patterson, A. L. Phys. Rev. 1939, 56, 978. (50) Gordillo, G.; Grizalez, M.; Hernandez, L. C. Sol. Energy Mater. Sol. Cells 1998, 51, 327. (51) Schmid, U.; Seidel, H. Thin Solid Films 2005, 489, 310.

1678

dx.doi.org/10.1021/cm101673u |Chem. Mater. 2011, 23, 1668–1678