Herein, we describe an atomic layer deposition (ALD) system that is optimized for the growth of thin films on high-surface-area, porous materials. The system incorporates a moveable dual-zone furnace allowing for rapid transfer of a powder substrate between heating zones whose temperatures are optimized for precursor adsorption and oxidative removal of the precursor ligands. The reactor can both be evacuated, eliminating the need for a carrier gas during precursor exposure, and rotated, to enhance contact between a powder support and the gas phase, both of which help us to minimize mass transfer limitations in the pores during film growth. The capabilities of the ALD system were demonstrated by growing La2O3, Fe2O3, and LaFeO3 films on a 120 m2 g−1 MgAl2O4 powder. Analysis of these films using scanning transmission electron microscopy and temperature-programmed desorption of 2-propanol confirmed the conformal nature of the oxide films.
I. INTRODUCTION
Atomic layer deposition (ALD) is widely applied in the microelectronics industry for the fabrication of thin films.1,2 The process typically includes four primary steps: (1) exposure of the substrate to a gaseous precursor containing the element that is to be deposited; (2) purging of the sample, usually with an inert gas, to remove excess precursor and by-products; (3) exposure of the adsorbed precursor to a second reactant to remove the precursor ligands; and (4) purging. This cycle is repeated as many times as necessary to produce a film of the desired thickness. The main advantage of ALD is that it can produce pin-hole-free, conformal films with great precision. A major disadvantage is that growth rates are typically only 0.01–0.1 nm/cycle;3,4 so, many cycles are required to produce films even as thin as 10 nm. To circumvent this problem, ALD equipment for microelectronics applications is designed to cycle rapidly between the four steps.
There has been much interest in using ALD to prepare films on high-surface-area, porous materials for applications ranging from catalyst synthesis5–7 to coating of ceramic powders,8 due to the fact that conformal films can be formed on almost any surface. Note, however, that for high-surface-area materials, there is a limit to how thick a film one can grow by ALD without filling the pores. For example, a 1 nm thick film of a material with a density of 5 g/cm3 will increase the sample mass by 50% if the support has a surface area of 100 m2/g. For this reason, it is usually not practical or needed to deposit films much thicker than about 1–2 nm on high-surface-area materials. This allows the requirement for rapid cycling to be relaxed somewhat since many growth cycles are not required. However, since gas-phase diffusion is a relatively slow process, equipment to perform ALD on porous powders must be designed to take this into account and avoid concentration gradients in the sample during growth.9–11 Various designs have been implemented to deposit uniformly on powders,12 including fluidized-bed reactors13,14 and rotary-drum reactors.15–17 Also, ALD is usually performed at much higher pressures when the substrate is a powder.18
A second major issue with ALD on porous powders is encountered when ligand removal is difficult. On flat surfaces, removal of hard to oxidize organic ligands can be accomplished using either oxygen plasmas or ozone but these reactants are ineffective on porous materials due to facile oxygen radical recombination and ozone decomposition on the external surfaces of the powder.19–22 For the ALD synthesis of high-surface-area catalysts in our laboratories, we have performed the ligand-removal step by physically removing the sample from the precursor-adsorption apparatus and then oxidizing the ligands in the air in a furnace at higher temperatures.23 This has allowed us to prepare a wide range of interesting catalytic materials;24,25 however, the repeated transfer of powders between the precursor-adsorption setup and the furnace is obviously not practical for larger-scale manufacturing.
In the present work, we demonstrate a rotating bed, dual heating zone system that can be used to coat a large quantity of the powder sample in a relatively short period of time. We also demonstrate the applicability of this system by growing a 0.58-nm film of LaFeO3 on a 120-m2/g MgAl2O4 (MAO) powder, a material that has previously been shown to exhibit interesting properties when used as support for Pt catalysts.26
II. DESCRIPTION OF THE ALD SYSTEM
A. Overview of the ALD system
A schematic of the powder ALD system used in this study is shown in Fig. 1 and the supplementary material.29 A salient feature of this system is a rotating, quartz glass, reaction tube that is 90 cm in length and 6 cm in diameter, positioned in a moveable, two-zone, tube furnace. The inlet side of the reactor tube is interfaced to a high-temperature dosing manifold that has a series of valves connected to small glass tubes that are used to contain the ALD precursors. The dosing manifold is also configured to allow air or other gaseous oxidants to be admitted into the reactor tube. The downstream end of the reactor is connected to a rotary-vane, vacuum pump, which is used to evacuate the system and eliminates the need for a carrier gas during precursor exposure. A thermocouple pressure gauge is positioned at the outlet to monitor reactor pressure, and the vacuum pump can be isolated from the rest of the system by a stainless-steel valve.
As noted above, diffusion of precursors molecules and oxidants within a bed of powder can be kinetically limiting.10 To minimize these effects and enhance the mixing of the gas and solid phases, the quartz tube can be rotated. This was accomplished by mounting each end of the reaction tube on rotating seals and using a small electric motor connected to the downstream side of the reactor assembly using a rubber belt and pulley system as shown in Fig. 1. A variac connected to the motor allows the speed of rotation to be adjusted.
B. Dual heating zone furnace
A key feature of the apparatus is that the furnace in which the reactor tube is positioned is equipped with two, 20-cm long, heating zones with independent temperature controllers as shown schematically Fig. 2 and in the supplementary material.29 The temperature of the “adsorption zone” is optimized for precursor adsorption, typically 100–400 °C, while the other is a “hot zone” that is maintained at a higher temperature (up to 1150 °C) that is used for oxidation and removal of the precursor ligands. The furnace is on a movable carriage that allows the portion of the reactor tube that contains the powder substrate material on which the film is being grown to be rapidly positioned in the desired heating zone.
For this system, a typical ALD cycle consists of (1) positioning the portion of the reactor tube containing the powder support material in the adsorption zone, (2) evacuation of the reactor, (3) exposure of the sample to the precursor vapor for a set amount of time, (4) evacuation of the reactor, (5) move the oven to position the sample in the hot zone, and (6) exposure to the oxidant. Of course, some delay time must be included when moving the sample between heating zones to allow thermal equilibrium to be established. This cycle is repeated as many times as needed to produce a film of the desired thickness. For the growth of mixed oxide films, the precursor molecule that is exposed to the sample in each cycle can be varied as needed to obtain the desired oxide stoichiometry.
III. DEMONSTRATION INSTRUMENT CAPABILITIES
A. ALD growth of La2O3 and Fe2O3
To demonstrate the capabilities of the ALD apparatus, thin films of La2O3 and Fe2O3 were grown on a 120 m2 g−1 MAO substrate using tris(2,2,6,6-tetramethyl-3,5-heptanedionato) lanthanum [La(TMHD)3, Strem Chemicals, Inc.] and ferrocene [Fe(Cp)2, Sigma-Aldrich] as the precursors. In these studies, to facilitate sample handling and gravimetric determination of growth rates, the MAO support powder was initially pressed into wafers that were ∼0.2 mm thick, and then ∼3 g of the pellets were positioned in the middle of the glass reactor tube. After placing the reactor tube in the ALD system, the exposed portions of the tube and the dosing manifold were covered by heating tapes and heated to 260 °C. This was done to prevent precursor condensation on these surfaces. The adsorption and hot zones of the tube furnace were then adjusted to 300 and 500 °C, respectively. In order to achieve sufficient vapor pressure, the tubes containing the ferrocene and La(TMHD)3 precursors were heated to 250 and 130 °C, respectively.
For a deposition cycle, the reactor tube was initially positioned with the sample in the 300 °C adsorption zone and evacuated to 7 × 10−3 Torr using the rotary-vane pump. Then, after the vacuum pump was isolated from the reactor, the sample was exposed to the vapor of one of the precursors and then the valve between the precursor tube and the reactor was closed allowing additional time for the precursor to adsorb on the substrate. As will be discussed below, the exposure time was sufficiently long to ensure saturation coverage of the adsorbed precursor, which is necessary to obtain self-limiting growth of the oxide film. Following precursor exposure, the reactor was evacuated for 10 min to remove any excess precursor [Fig. 2(b)]. During this time, the tube furnace was moved to position the sample in the center of the 500 °C hot zone. After isolating the reactor from the vacuum pump, air was admitted into the reactor at 1 atm for 10 min [Fig. 2(c)], which was found to be sufficient to combust the ligands and completely oxidize the adsorbed precursor. After completing the oxidation step, the oven was moved to reposition the sample in the adsorption zone and then the reactor was evacuated for 10 min to remove the excess air and ligand combustion products. This sequence was repeated to obtain the desired film thickness.
To ensure conformal film growth on the porous support, the precursor exposure time must be sufficiently long to allow for diffusion into the pores and saturation of the surface. To determine the optimal precursor exposure time for this particular support, we first measured the oxide growth rate per ALD cycle as a function of La(TMHD)3 and Fe(Cp)2 exposure times. These data are shown in Figs. 3(a) and 3(b). Note that in both cases, the growth rate increased as the exposure time was increased from 6 to 9 min and then remained constant at 0.017 and 0.030 nm/cycle for La2O3 and Fe2O3, respectively, assuming that the films have the density of the corresponding bulk oxides.27 This behavior is consistent with a self-limiting growth mechanism which is an inherent characteristic of ALD.28
In these growth experiments, the utilization of the precursor, i.e., the fraction of the precursor admitted into the system that reacted on the substrate per growth cycle, for the 9 min exposure time was found to be close to 50%. Note, however, that since the reactor has a fixed volume, the utilization will be a function of sample size. In the growth studies reported here, we used a relatively small sample size of 3 g. The ALD system is capable of handling powder samples in excess of several hundred grams. These larger sample sizes would increase the ratio of reactive surface area to reactor volume, which would lead to a higher precursor utilization per deposition cycle.
Note that due to the large sample size, the exposure times required here per ALD cycle are much longer than those needed in our previous studies of the growth of these oxides on similar high-surface-area supports using our smaller adsorption ALD systems,23,26,27 but the ALD cycle time is still greatly reduced by being able to do the oxidation step in the same apparatus. It should also be noted that the exposure time required for self-limiting growth will be a function of the amount, porosity, and surface area of the support material.
After determining the required precursor exposure time, growth curves, i.e., the number of cations deposited as a function of the number of ALD cycles, were measured for the deposition of La2O3 and Fe2O3. In these experiments, the amount of deposited oxide was determined gravimetrically after 1, 3, and 5 ALD cycles. The growth curves for the La and Fe cations are displayed in Figs. 4(a) and 4(b). These data show constant growth rates of 4.18 × 1017 La atoms/m2-cycle and 1.25 × 1018 Fe atoms/m2-cycle, which corresponds to oxide film growth rates of 0.017 nm/cycle for La2O3 and 0.030 nm/cycle for Fe2O3 (see the supplementary material29). Note that these rates are close to those that we obtained for the growth of these oxides on MAO using our smaller ALD adsorption apparatus.23
B. Growth of perovskite LaFeO3 thin films on MgAl2O4
To further demonstrate the capabilities of the ALD apparatus, confirm the formation of conformal oxide films that cover the support, and demonstrate control of the oxide composition, we also investigated the growth of the mixed oxide LaFeO3 on the MAO support. To facilitate good mixing of the two cations and to compensate for the fact that the Fe2O3 deposition rate is approximately three times that of La2O3, for growth of the LaFeO3 film, we used a super-cycle consisting of three La2O3 ALD cycles followed by one Fe2O3 ALD cycle. We have previously shown using our smaller ALD systems that this approach produces sufficient mixing of the La and Fe cations to allow the formation of a single-phase perovskite film.23 A LaFeO3 film was then grown on the MAO support using seven of these super-cycles followed by annealing the sample in air for 24 h at 800 °C. This latter step was used to ensure the transformation of the as-grown mixed oxide film into the desired perovskite phase. This procedure produced a film that was approximately 0.59 nm thick, assuming it had the same density as bulk LaFeO3.
X-ray diffraction (XRD) was used to demonstrate that this ALD process resulted in the formation of a single-phase, LaFeO3 perovskite film (see the supplementary material29). The XRD pattern for the as-grown LaFeO3 is displayed in Fig. 5. The XRD pattern of the bare MgAl2O4 support and a reference spectrum for LaFeO3 are also included in the figure for comparison. The XRD pattern for the bare MgAl2O4 sample contained peaks at 31.7, 37.2, and 45.2° 2θ, which are consistent with the cubic-spinel structure of this material. In addition to these support peaks, the XRD pattern of the LaFeO3/MAO sample contained peaks at 22, 32, 40, 46, 52, and 58° 2θ, which align with those in the reference spectrum for LaFeO3, thereby confirming the formation of a single-phase film with the perovskite structure.
Scanning transmission electron microscopy-energy dispersive x-ray spectroscopy was used to help verify that the ALD procedure produced a conformal LaFeO3 film on the MAO support. A STEM image of the LaFeO3/MAO sample is shown in Fig. 6 along with EDS elemental maps for Al, La, and Fe from the indicated region. Note that the La and Fe EDS maps are essentially identical to that of Al from the support, which is consistent with LaFeO3 forming a continuous film on the MAO rather than being present in the form of small particles.
To further corroborate the growth of a conformal LaFeO3 film on the MAO support, 2-propanol temperature-programmed desorption and thermogravimetric analysis (TPD-TGA) measurements were conducted. For these experiments, the sample was placed in a vacuum microbalance, exposed to 2-propanol vapor at room temperature, followed by evacuation for 1 h to remove any weakly adsorbed reactant and then heated in a vacuum while monitoring the mass and desorbing products using a mass spectrometer. Figure 7(a) displays the TPD data for the bare MAO support, which shows an initial 2-propanol coverage of 580 μmol/g, which corresponds to 2.9 × 1018 molecules per m2, which is close to that expected for monolayer coverage. Upon heating, ∼300 μmol/g of the alcohol desorbs unreacted between 300 and 500 K, with the remainder undergoing dehydration on Lewis acid sites to produce propene which appears in the TPD data as a sharp peak at 510 K.24
The 2-propanol TPD data for the LaFeO3/MgAl2O4 sample shown in Fig. 7(b) are significantly different than that from the bare support. Note that the amount of 2-propanol adsorbed decreased by 150–430 μmol/g, which is consistent with a decrease in the surface area associated with the ALD film growth. More importantly, the reaction products were also different. For the LaFeO3/MgAl2O4 sample, a mixture of propene (m/e 41) and acetone (m/e 43) was produced between 500 and 625 K with the majority of the product being produced at 570 K. Note that the large propene peak at 510 K, which is indicative of reaction on the support, is largely absent. This result is consistent with the LaFeO3 film covering the majority of the MgAl2O4 surface, confirming film growth within the pores.
IV. SUMMARY
In this paper, we have described an ALD apparatus that is optimized for the growth of conformal oxide films on high-surface-area materials. Distinguishing features of the apparatus include the ability to evacuate the sample prior to precursor exposure eliminating the need for a carrier gas, and a moveable furnace with dual heating zones that are optimized for precursor adsorption and oxidative removal of organic ligands. The evacuation and rotating reactor capabilities effectively eliminate mass transfer limitations encountered in conventional flow ALD systems, which can prevent the growth of conformal films on highly porous substrates. The dual-zone movable furnace facilitates the use of optimal temperatures for both precursor adsorption and oxidation, thereby allowing for a range of organometallic precursors with hard to oxidize ligands to be used.
The capabilities of the apparatus were demonstrated through the growth of thin films of La2O3, Fe2O3, and LaFeO3 on a high-surface-area MAO support. Self-limiting growth was demonstrated for ALD of these oxides with a precursor utilization of ∼50%, with higher utilizations being expected for sample sizes larger than those used here. While the system described here can only handle samples of modest size (several hundred grams), the technology could easily be scaled up, making ALD a practical method to tailor the surface composition of high-surface-area materials that have applications as heterogeneous catalysts, as adsorbents, and for use in electrodes, just to name a few.
ACKNOWLEDGMENTS
This work was supported by the Department of Energy, Office of Basic Energy Sciences, Chemical Sciences, Geosciences and Biosciences Division under Grant No. DE-FG02-13ER16380. The STEM work was carried out at the Singh Center for Nanotechnology, part of the National Nanotechnology Coordinated Infrastructure Program, which is supported by the National Science Foundation (NSF) under Grant No. NNCI-1542153.
AUTHOR DECLARATIONS
Conflict of Interest
The authors have no conflicts to disclose.
Ethics Approval
Ethics approval is not required.
DATA AVAILAILITY
The data that support the findings of this study are available from the corresponding authors upon reasonable request.