Forum Article www.acsami.org
Inhibiting Metal Oxide Atomic Layer Deposition: Beyond Zinc Oxide Matthew D. Sampson,† Jonathan D. Emery,†,‡ Michael J. Pellin,† and Alex B. F. Martinson*,† †
Material Science Division, Argonne National Laboratory, 9700 South Cass Avenue, Lemont, Illinois 60439, United States Department of Materials Science and Engineering, Northwestern University, 2220 Campus Drive, Evanston, Illinois 60208, United States
‡
S Supporting Information *
ABSTRACT: Atomic layer deposition (ALD) of several metal oxides is selectivity inhibited on alkanethiol self-assembled monolayers (SAMs) on Au, and the eventual nucleation mechanism is investigated. The inhibition ability of the SAM is significantly improved by the in situ H2-plasma pretreatment of the Au substrate prior to the gas-phase deposition of a long-chain alkanethiol, 1-dodecanethiol (DDT). This more rigorous surface preparation inhibits even aggressive oxide ALD precursors, including trimethylaluminum and water, for at least 20 cycles. We study the effect that the ALD precursor purge times, growth temperature, alkanethiol chain length, alkanethiol deposition time, and plasma treatment time have on Al2O3 ALD inhibition. This is the first example of Al2O3 ALD inhibition from a vapor-deposited SAM. The inhibitions of Al2O3, ZnO, and MnO ALD processes are compared, revealing the versatility of this selective surface treatment. Atomic force microscopy and grazingincidence X-ray fluorescence further reveal insight into the mechanism by which the well-defined surface chemistry of ALD may eventually be circumvented to allow metal oxide nucleation and growth on SAM-modified surfaces. KEYWORDS: atomic layer deposition, self-assembled monolayers, metal oxides, alkanethiols, selective deposition, aluminum oxide, zinc oxide, manganese oxide
■
INTRODUCTION
Multiple studies have now explored routes to chemically inhibiting ALD growth through the use of self-assembled monolayers (SAMs).2,6−12 Long-chain alkylsilanes,12−17 phosphonates,8,10,11 and thiols6 are the most commonly employed SAM precursors. Although alkylsilanes form well-ordered, tightly packed SAMs on a variety of substrates, to date the inhibition properties of these SAMs have been proven against a limited set of metal oxide (HfO2 and ZrO2)18,19 or metal (Ru and Pt)12,13 ALD processes, and the inhibition delays have been modest. Bent and co-workers have recently reported successes in inhibiting ZnO ALD7−10 and polyurea molecular layer deposition (MLD)11 on patterned Cu/Si substrates for the selective deposition of dielectric films. These studies have reported the use of both alkanethiol7 and octadecylphosphonic acid8 SAMs as inhibition layers. Bent and co-workers have
Atomic layer deposition (ALD) has emerged as a powerful tool for the atomically precise design and synthesis of thin films of oxides, nitrides, sulfides, and pure metals.1−3 In contrast to physical deposition methods, ALD is grounded in well-defined surface chemistry, which affords the opportunity for selflimiting and therefore precise deposition, which is also conformal. Furthermore, the chemical specificity of ALD surface reactions should allow, in theory,4 for highly selective substrate growth. For example, the rapid chemical reaction of a specific metal precursor with one surface termination (e.g., hydroxyl)5 may be severely inhibited or even excluded with another (e.g., alkane) as in more traditional solution chemistry. However, perhaps in part due to the aggressive, and therefore unselective, chemistry of some of the most popular ALD precursors [e.g., trimethylaluminum (TMA) and diethylzinc (DEZ)], this prospect remains largely unrealized. Although ALD science has made remarkable strides in depositing uniformly everywhere, there is great room for improvement in selective, inhibited growth. © XXXX American Chemical Society
Special Issue: Hupp 60th Birthday Forum Received: January 30, 2017 Accepted: March 23, 2017
A
DOI: 10.1021/acsami.7b01410 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Forum Article
ACS Applied Materials & Interfaces
used as received. Prior to Au thin-film deposition, glass slides were sonicated twice in acetone and then isopropyl alcohol before being blown dry under flowing N2. Au films (50−100 nm) were prepared by sputtering (Denton Desk II, Ar atmosphere) a Au target (99.99%, Denton Vacuum) on glass slides. Higher-purity gold ingots (99.999%; Lesker; see Table S1) were also thermally evaporated for some experiments without a statistically significant improvement in inhibition. Si substrates were sonicated twice in acetone and then isopropyl alcohol before being blown dry under flowing N2. For vapor-phase SAM loading, ethanethiol was used at room temperature, 1-octanethiol was heated to 50 °C, and DDT was heated to 115 °C for delivery from a standard ALD precursor cylinder (Swagelok, 50 mL). Organometallic ALD precursors tested in this study include diethylzinc (DEZ; Sigma-Aldrich), trimethylaluminum (TMA; Sigma-Aldrich), and bis(ethylcyclopentadienyl)manganese(II) [Mn(EtCp)2; SigmaAldrich]. Millipore water (H2O; 18 MΩ, Milli-Q system) was used as the oxygen source in all experiments. ALD Instrumentation. ALD was performed in a Fiji F200 hot-wall ALD reactor (Cambridge Nanotech, Inc.). The ALD reactor is equipped with a remote plasma source connected to Ar, N2, O2, and H2 gas lines. An Ocean Optics USB4000 UV−vis is attached to the wall of the ALD reactor to monitor plasma emission. Unless stated otherwise, the H2-plasma conditions were 300 W with 30 sccm of H2 gas and 200 sccm of Ar gas flow delivered for 3 min. Unless stated otherwise, vapor-phase SAM growth was performed under a continuous Ar flow (60 sccm), resulting in an equilibrium base pressure of 0.1 Torr. A consecutive half-cycle pulse−purge sequence (t1−t2) was utilized for SAM deposition, where t1 is the alkanethiol pulse (0.1 s for ethanethiol, 4 s for 1-octanethiol, and 5 s for DDT) and t2 is the purge time (60 s). Unless stated otherwise, SAM deposition was performed via 30 consecutive half-cycles. SAM thicknesses and densities were modeled by the Cauchy fitting of ex situ variable-angle spectroscopic ellipsometric data (Woollam M-2000, 380−1000 nm). The measured film thicknesses for DDT SAMs are in agreement with previous reports23 (20.78 ± 1.91 and 18.41 ± 1.13 Å for nonplasmaand H2-plasma-treated Au, respectively). As in previous studies on a similar ALD reactor, DDT saturation occurs on the Au substrate after ∼20 consecutive half-cycles of 1 s doses.6 For depositions with H2-plasma pretreatment, there is no vacuum break prior to SAM deposition. In all depositions, there is no vacuum break prior to metal oxide ALD. A conventional metal oxide deposition “AB cycle” sequence was used (t1−t2−t3−t4), where t1 is the pulse time for the metal precursor, t3 is the pulse time for H2O, and t2 and t4 are the purge times for the metal precursor and H2O, respectively. The pulse times (t1) for DEZ, TMA, and Mn(EtCp)2 were 1, 1, and 2 s, respectively, and the pulse time (t2) for H2O was 2 s, unless otherwise noted. The purge times (t2 and t4) were 20 and 40 s, respectively, unless otherwise noted. These pulse times result in saturating and self-limiting surface reactions on silicon wafers, as demonstrated in Figures S1−S3. DEZ, TMA, and H2O reservoirs were delivered at room temperature, while DDT and Mn(EtCp)2 were held at 115 °C in order for all to be delivered under their own vapor pressure. Preparation of Template-Stripped (TS) Au. TS Au substrates on mica templates were purchased from Platypus Technologies and used as received. For TS Au on Si templates, the preparation was adapted from previous reports.24,25 Au (50−100 nm) was deposited by evaporation onto a Si wafer. No treatment or cleaning was performed on the Si wafer before deposition. For TS Au on glass, glass slides were cleaned by sonicating twice in acetone followed by once in isopropyl alcohol. The glass slides were then blown dry under flowing N2 and exposed to a 10 min UV/O3 cleaning treatment. A small drop of epoxy (Norland Optical Adhesive 83H) was placed on the treated glass slide, and the glass was carefully pressed onto the Au surface. The pressure was maintained on the glass in order to allow the epoxy to uniformly spread out to fill the area under the glass slide. This process was repeated for the number of glass slides prepared. The epoxy was cured by placing the Si substrate on a heating plate and heating overnight (12−20 h) at 150 °C. A beaker was placed on top of the heating template to provide light pressure during curing. Immediately before loading a TS Au substrate into the ALD reactor, the Au was freshly
optimized ZnO inhibition by utilizing a variety of methods, such as redosing alkanethiol SAM precursors to restore SAM quality after it begins to degrade or by using a mild liquid etchant to selectively remove ALD growth on SAMs.7,9,10 Using these procedures,9 they have inhibited ZnO ALD growth on Cu for up to 600 cycles. While this successful inhibition of ZnO is significant, there is evidence that inhibiting ZnO ALD may be a special case.6 The selective inhibition of other metal oxide ALD processes remains less often reported and significantly less successful. For example, there is only one report of inhibiting ALD Al2O3 with SAMs and it requires a 48 h solution loading of SAM onto Cu.10 Therefore, a general route to the selective inhibition of metal oxide ALD processes, especially with vacuum compatibility and reasonable speed, remains a critical challenge in electronic device fabrication and heterogeneous catalysis. In situ quartz crystal microbalance measurements have previously been undertaken to probe ALD inhibition on alkanethiol SAMs on Au.6 Vapor-deposited 1-dodecanethiol (DDT) SAMs were found to be equally, if not more, effective than solution-deposited SAMs in inhibiting oxide ALD with much shorter loading time. Vapor-deposited alkanethiol SAMs have also proven to be effective at inhibiting Ir ALD on Cu.20 Bent and co-workers corroborate the comparison for ZnO ALD inhibition, with only 30 s of alkanethiol vapor-phase exposure showing more effective inhibition than multihour solutiondeposited SAMs.9 Although DDT SAMs on Au effectively inhibit ZnO ALD for nearly 100 cycles, they were not as effective at inhibiting other metal oxide ALD processes, such as TiO2 or SnO2 (less than 20 cycles of inhibition)6 and provided no resistance to Al2O3 ALD. Kobayashi et al. also reported uninhibited growth of Al2O3 ALD on alkanethiol SAMs;21 however, this study differs from previous studies in that the temperature of the substrate was held at 45 °C, which may be insufficient to avoid condensation, an easy target for TMA nucleation. However, even at 150 °C, ALD with TMA and H2O is uninhibited, suggesting an alternate mechanism.22 One such mechanism may be penetration of the SAM layer by the ALD precursors through defective or loosely packed alkanes with subsequent reaction at gold oxide. In order to test this hypothesis, we sought to improve upon the density and uniformity of vapor-deposited SAMs on Au. In order to provide a more pristine Au0 surface for thiol deposition, we studied the effect of an in situ H2-plasma applied to Au on the eventual SAM deposition and ALD inhibition. This approach enabled a more effective inhibition of ALD growth for a variety of metal oxide processes. A lightly optimized surface preparation was identified by screening process variables for both Au pretreatment and vapor-phase alkanethiol deposition. While inhibition was dramatically improved, all metal oxide ALD processes eventually nucleated and grew, as evidenced by spectroscopic ellipsometry. To understand the nature of nucleation, low-angle X-ray fluorescence (XRF) was introduced as an even more sensitive probe of the first signs of ALD nucleation. Atomic force microscopy (AFM) allowed for visualization of the metal oxide nucleation morphology in order to propose mechanisms for the failure of selective surface chemistry and subsequent metal oxide growth. These results go beyond inhibition of ZnO ALD to include one of the moststudied and reliable ALD processes, Al2O3.
■
EXPERIMENTAL SECTION
Materials and Preparation. The solvents 1-dodecanethiol (DDT), 1-octanethiol, and ethanethiol were obtained from Sigma-Aldrich and B
DOI: 10.1021/acsami.7b01410 ACS Appl. Mater. Interfaces XXXX, XXX, XXX−XXX
Forum Article
ACS Applied Materials & Interfaces stripped off of the Si template by scoring the perimeter around the glass slide with a razor blade and then wedging the blade under a corner of the glass slide to pry up the Au. Samples were stored bonded to the substrate until needed for ALD. Characterization. Ex situ variable-angle spectroscopic ellipsometric data were collected from 380 to 1000 nm at an incidence angle of 75° on a Woollam M-2000 ellipsometer. Atomic force microscopy (AFM) tapping-mode images were acquired in air using an Asylum Research MFP-3D microscope. The AFM tip had a radius of 28 ± 10 nm and was driven at a resonant frequency of ∼340 kHz and a scan rate of 0.5 Hz over 1−4 μm2. AFM images were analyzed using Gwyddion. All AFM images were leveled and step-line corrected. Cluster density and size analysis was performed through a segmentation analysis of the phase images. X-ray fluorescence (XRF) data for MnO samples were acquired using monochromated Mo Kα radiation produced using a Rigaku rotating-anode X-ray generator. X-ray wavelength was 0.710 Å. The X-ray beam was collimated to ∼0.05 × 6 mm2 (vertical × horizontal) and struck the sample surface at a 4° incident angle. Spectra were acquired at near-normal exit angle using a Vortex Si-drift diode detector. Mn Kα signals were calibrated to a known As Kα standard to acquire absolute atomic coverage. These data were corrected for element-specific XRF cross sections, detector efficiency, and relative absorption of characteristic X-rays in air.
■
RESULTS AND DISCUSSION
Metal Oxide Growth on H2-Plasma-Treated Au. The ALD growth of ZnO, Al2O3, and MnO on witness Si wafers with native oxide was first completed in our plasma-equipped ALD reactor (Figures S1−S3). Metal oxide growth on Au, H2-plasma-treated Si, and H2-plasma-treated Au was also studied to ensure that H2-plasma treatment alone did not significantly alter the nucleation of ALD precursors (Figure 1). H2-plasma treatment has a surprisingly small effect on ALD precursor nucleation despite the expected depletion of the surface hydroxyl population (−OH), which serves as the reactive surface moiety in the accepted mechanism for all three metal oxide ALD processes, at least for growth on itself. Still, for all three metal oxide ALD processes studied, it should be noted that the initial deposition rates (for
