Conformal atomic layer deposition (ALD) inside macroscopic nanoporous solids with aspect ratios greater than 103 can require ALD reactant exposures on the order of 103 Torr-s or greater. For some ALD chemistries, such large exposures raise the concern of non-self-limiting deposition. In the case of ZnO ALD from diethylzinc (DEZ) and H2O, exposures in the 10–103 Torr-s range have resulted in metallic Zn deposition at typical temperatures used for ZnO ALD on wafers (e.g., ∼180 °C). This Zn deposition can be suppressed by lowering the deposition temperature, but this slows H2O desorption and, thus, can necessitate impractically long purge times. In this work, we use static-dose ALD with DEZ and H2O exposures >104 Torr-s to deposit ZnO inside Al2O3 nanoparticle compacts (NPCs) with 50.5 ± 0.3% porosity, 100 nm NP diameter, 1.55 ± 0.05 mm thickness, and an aspect ratio of 7800 ± 200 (based on the half-thickness), and we explore a novel approach to the deposition temperature, T: T is cycled between 160 °C (for H2O purges) and 120 °C (for all other steps). For comparison, we also deposit ZnO with T held constant at 120 or 160 °C. Whereas the T = 160 °C process results in Zn metal deposition and nonuniform infiltration, the temperature-cycled process yields apparently self-limiting ZnO deposition at a growth per cycle (GPC) of ∼2.1 Å/cyc, forming an electrically conductive ZnO network that is uniform throughout the thickness of the NPC, with the exception of some ZnO depletion near the NPC surfaces, possibly due to the (unoptimized) long DEZ purge time. The T = 120 °C process produces similar results, although the GPC is slightly elevated, indicating diminished removal of H2O and/or OH during purges. We employ scanning electron microscopy with energy-dispersive x-ray spectroscopy, x-ray diffractometry, electrical resistivity measurements, and ALD chamber pressure analysis in our comparison of the three ALD processes.

Atomic layer deposition (ALD) is an advantageous technique for coating the internal surfaces of porous networks of nanoparticles (NPs) and thereby modifying their optical, electronic, and/or structural properties. In principle, the self-limiting deposition half-reactions can produce a conformal coating with uniform thickness throughout the internal surface of an NP network, provided open porosity and sufficient exposure of the entire surface to the gas-phase reactants. The minimum required exposure depends on the NP network porosity, ϕ, and aspect ratio (AR): the ratio of the NP network thickness, L, to the average pore diameter, dpore.1 For NP thin films with porosity between ∼0.3 and ∼0.5 and AR of ∼100 or less (L between ∼100 nm and ∼1 μm and dpore between ∼5 and ∼20 nm), exposure to ∼1 Torr or less of reactive gas for ∼1 s or less typically suffices. Such NP thin films have been coated via ALD with a variety of metal oxides, including Al2O3,2–8 HfO2,5 and ZnO.4,6–8 The latter coating is particularly interesting because of its semiconducting properties and its ability to facilitate bandlike transport in NP networks.6,7

The application space for nanocomposites comprised of NPs with ALD coatings could be expanded considerably by scaling up the ALD substrates from NP thin films to macroscopic NP networks with L on the order of 1 mm and AR on the order of 104. This requires significantly greater reactant exposures. Recently, Gayle et al.9 demonstrated one cycle of Al2O3 ALD, using trimethylaluminum (TMA) and H2O as precursors, throughout the internal surface of a silica aerogel monolith with ϕ = 0.95 and AR > 60 000; the aerogel monolith had dpore ≈ 20 nm and L ≈ 1.25 mm, where in this case, L is half the thickness of the monolith, which the authors used to calculate the AR because reactants were allowed to diffuse into both the top and the bottom. Complete infiltration, confirmed by scanning electron microscopy (SEM) with energy-dispersive x-ray spectroscopy (SEM-EDS), was achieved via a sequence of 12 TMA doses constituting a total exposure >104 Torr-s.

In the case of ZnO ALD from diethylzinc (DEZ) and H2O, the results of scaling up exposure have depended on the deposition temperature. At 80 °C, Ingale et al.10 deposited up to three cycles of ZnO onto ∼2 g of mesoporous SiO2 powder with a specific surface area, s, of 505 m2/g, and their x-ray diffraction (XRD), transmission electron microscopy (TEM), x-ray fluorescence, and powder mass gain data were consistent with ZnO deposition throughout the powder bed at a growth per cycle (GPC) of ∼1 Å/cyc. At 177 °C, Ferguson et al.11 deposited ZnO on ZrO2 NP powder with s = 20.2 m2/g (dNP = 50 nm), and they obtained a GPC of ∼2.2 Å/cyc, according to TEM measurements of the ZnO coating thickness. Moreover, they observed evidence of non-self-limiting deposition during the DEZ exposure. Increasing the exposure up to a maximum of 900 Torr-s and performing in situ Fourier-transform infrared spectroscopy, they observed a positive correlation between exposure and background absorbance, which they attributed to conductive Zn deposition via DEZ decomposition. Zn metal was, in fact, detected in ZnO ALD coatings via XRD by Libera et al. in their study of ALD on 1 g batches of mesoporous silica gel powder with s = 100 m2/g.12 Depositing ten cycles of ZnO and maintaining a DEZ exposure of 22.5 Torr-s, they varied the deposition temperature from 150 to 200 °C and observed Zn metal peaks that became detectable at 155 °C and steadily increased in intensity with increasing temperature. This increase in Zn peak intensity was accompanied by a gradual change in the color of the coated silica from white at 150 °C to black at 200 °C. The authors also varied the DEZ exposure while maintaining a constant deposition temperature of 200 °C, and they detected no Zn metal by XRD or color change for exposures <10 Torr-s. This is consistent with the results of studies of ZnO ALD infiltration of NP thin films in which the deposition temperature was 180 °C, but DEZ exposure was <1 Torr-s, and no Zn metal was detected.4,6–8

While these ZnO ALD studies have demonstrated that Zn deposition during large DEZ exposures can be minimized by reducing the deposition temperature, the application of this strategy to deposition of electrically conductive ZnO coatings on macroscopic NP networks is complicated by two factors. First, for deposition temperatures below ∼170 °C, the electrical resistivity of ALD-grown ZnO often increases with temperature due to the temperature dependences of stoichiometry, grain size, and orientation.13,14 Second, the rate of H2O desorption from the ZnO surface during purging decreases with temperature, exhibiting a typical exponential dependence.15 This translates to an exponential temperature dependence of the required H2O purge time, which is a significant consideration for the relatively extensive ZnO ALD infiltration processes (potentially ∼100 cycles) required for high-conductance coatings.

In this work, we investigate DEZ/H2O ALD infiltration of porous networks of NPs with AR ∼100 times greater than that of a typical NP thin film coated with ZnO in previous studies.6–8 This is a useful case study because of both the challenge of suppressing Zn deposition and the opportunity to evaluate infiltration via electrical measurements of the conductive ZnO coating. Specifically, we deposit 20 cycles of ZnO on cylindrical Al2O3 NP compacts (NPCs) with L = 0.78 ± 0.02 mm, dpore ≈ dNP = 100 nm, and AR = 7800 ± 200 (defining L as half the thickness of a nanoparticle compact (NPC), because ALD precursors enter through both the top and the bottom faces). Although this AR is lower than those of some AMs coated with Al2O3 or ZnO by ALD,9,16 the NPCs have a significantly lower porosity of 50.5 ± 0.3% (vs 95+% for the AMs), which translates to lower reactant diffusivity. For both DEZ and H2O, we expose the nanoparticle compacts (NPCs) to a reactant partial pressure of ∼9 Torr for 1920 s (32 min) for an exposure of ∼17 000 Torr-s. In an effort to understand whether Zn metal deposition can be suppressed while maintaining a reasonable total ALD run time, we explore a novel approach to the deposition temperature: in each ALD cycle, the temperature is 120 °C for all steps except for the H2O purge, during which it is ramped up to 160 °C and then back down to 120 °C before the subsequent DEZ exposure. For comparison, we carry out two additional ZnO ALD runs at the same conditions but with constant temperatures of 120 and 160 °C.

Porous Al2O3 NPCs were pressed in the ambient atmosphere using an 11.4 mm diameter tungsten carbide die set (REFLEX Analytical Corporation) from BMA15 deagglomerated alpha alumina nanopowder with a specific surface area, s, of 15 m2/g (Baikowski) in a 12 ton manual hydraulic press (Carver Inc.). The NP diameter, dNP = 100 nm, is derived from s = 15 m2/g, which is the manufacturer’s spec based on the Brunauer–Emmett–Teller (BET) method. A powder mass of 0.320 ± 0.005 g was pressed under 481 MPa and held for <1 min before the load was released and the compact was ejected. The final NPCs are 1.55 ± 0.05 mm thick and 11.43 ± 0.02 mm in diameter as measured by electronic calipers. The initial porosity, ϕi, is 50.5 ± 0.3%; therefore, the pore diameter, dpore, is approximately equal to dNP. The powder compaction procedure was optimized to produce intact NPCs, but some mm-scale cracks were observed (see Fig. S1 in the supplementary material44).

The Al2O3 NPCs were conformally coated with ZnO using an ALD NanoSolutions (Forge Nano) custom RX-series reactor with an approximately horizontally oriented ∼20 cm L × 10 cm ID cylindrical deposition chamber with a LoPro isolation valve (MKS) between the chamber and the vacuum pump. Within the ALD chamber, the NPCs were placed in a rectangular quartz boat with a rounded bottom so that only the bottom circular edge of an NPC contacted the boat, leaving a gap with a maximum height of ∼1 mm between the NPC’s bottom face and the quartz surface, as shown in Fig. 1. All NPCs were weighed before and after ALD to determine the mass of ZnO deposited. During ZnO ALD, the deposition chamber pressure was measured with two heated (100 °C) Baratron 728A capacitance manometers (MKS) with full scale ranges of 10 and 1000 Torr. Because the true base pressure is not precisely known (due to drift of the ALD-coated gauges), it is simply set equal to 0 Torr at 120 °C; that is, a constant offset of 0.26 Torr is applied to all pressure data. This offset does not affect the pressure trace analysis in this work, which is based on changes in pressure rather than absolute pressures.

FIG. 1.

(a) Schematic of ZnO ALD at deposition temperature T inside an Al2O3 nanoparticle compact (NPC) with NP diameter dNP, half-thickness L, and initial porosity, ϕi. (b) Summary of the three ALD temperature schedules. A “dose” includes both the pulse step and the hold step.

FIG. 1.

(a) Schematic of ZnO ALD at deposition temperature T inside an Al2O3 nanoparticle compact (NPC) with NP diameter dNP, half-thickness L, and initial porosity, ϕi. (b) Summary of the three ALD temperature schedules. A “dose” includes both the pulse step and the hold step.

Close modal

Diethylzinc (DEZ, min. 95%, Strem Chemicals) and de-ionized H2O were used as ALD precursors. Both were contained in stainless steel cylinders heated to 37 °C with vapor flow rates controlled by metering valves. Infiltration of the NPCs was accomplished via static dosing; that is, precursor vapor was held in the deposition chamber under zero-flow conditions for an extended time to allow diffusion into the NPC. An ALD run consisted of 20 cycles, and a cycle consisted of the following ∼160 min sequence: 10–14 s DEZ pulse, 1920 s (32 min) DEZ hold, 2880 s (48 min) purge, 8–11 s H2O pulse, 1920 s (32 min) H2O hold, 2880 s (48 min) purge. The DEZ and H2O pulse times were varied manually to improve the consistency of the exposure pressures, which were generally ∼9 Torr but inevitably fluctuated, possibly due to imperfect electronic valve control and/or variation in the temperature of small unheated portions of the precursor lines (see Fig. S2 in the supplementary material44). During pulse and hold steps, the chamber isolation valve was closed. During hold steps, ALD valves were also closed to isolate the chamber from the precursor manifold. According to pressure versus flow studies, the total isolated deposition volume is 2.17 l, including contributions of other small volumes (e.g., 1/4 in. and KF25 tubing) in addition to the volume of the main cylindrical chamber. During purge steps, the chamber isolation valve was opened, and N2 was fed through the chamber in the following sequence: 30 s of no flow, 30 s of 10 SCCM flow through half of the manifold (metal or oxidant half, whichever was used for the preceding exposure), 2790 s of 10 SCCM flow through both halves (20 SCCM total), 30 s of no flow.

The deposition chamber was heated by an insulating heater jacket using temperature feedback from a thermocouple mounted on the chamber’s exterior to obtain deposition temperatures, T, of 120 or 160 °C. T is defined as the approximate temperature (rounded to the nearest 10 °C) of the interior surface of the quartz boat containing the NPC samples at steady state under 20 SCCM N2 flow, as measured during temperature calibration with an additional thermocouple in direct contact with the boat. Among calibration runs with the same chamber exterior temperature, measured deposition temperatures varied by approximately ±5 °C, apparently due to variations in thermocouple placement and pressure. Three 20-cycle ALD runs were carried out: (1) with T held constant at 120 °C, (2) with T held constant at 160 °C, and (3) with T cycled between 120 and 160 °C; that is, in each cycle, T was held at 120 °C for all steps except for the H2O purge, during which T was increased to 160 °C and then decreased back to 120 °C before the subsequent DEZ pulse. Each ALD run was preceded by an H2O half-cycle to hydroxylate the Al2O3 surfaces, and this half-cycle was preceded by at least 2 h of NPC degassing in the ALD chamber at the initial T (120 or 160 °C). During all three runs, the main cylindrical chamber inner wall temperature was approximately equal to the quartz boat temperature, T, but the small portions of the deposition volume outside of the main cylindrical chamber (e.g., 1/4 in. and KF25 tubing) were heated by heating tape to external surface temperatures ranging from 100 to 130 °C, which are expected to correspond to internal surface temperatures ranging from ∼90 to ∼120 °C. Some surfaces, however, were not directly heated (e.g., quartz and sapphire windows on the cylindrical chamber) and may have had internal surface temperatures below 90 °C.

The designs of these ALD processes are based on theoretical studies by Yanguas-Gil and Elam17,18 and a subsequent theoretical and experimental study by Cendejas et al.8 of the ALD exposure and purge times required for uniformly coating nanostructured substrates. For details, see Note S1 in the supplementary material44 and the references cited therein.1,8,15,17–26 Briefly, our hold times, thold, are determined by
t hold = L pore 2 D γ ,
(1)
where Lpore is the average length of a pore, accounting for tortuosity, D is the diffusion coefficient (Knudsen coefficient in this case), and γ is the number of reactant molecules inside the nanostructure per reaction site. Our purge times, tpurge, are determined by
t purge = ln ( 1 / θ f ) ν e E des / ( R T ) ,
(2)
where θf is the fractional coverage of adsorbed H2O, the pre-exponential ν is 1013 s−1,21 and Edes is the H2O desorption energy, which is assumed to be independent of coverage. For a target θf of 5 × 10−3, which Cendejas et al. found to be necessary for uniform infiltration,8  Eq. (2) yields tpurge = 360 min at T = 120 °C and tpurge = 11 min at T = 160 °C. We choose a time of 48 min to ensure that θf < 5 × 10−3 can be attained in the 120/160 and 160 °C processes.

Prior to characterization (and during characterization in the case of XRD and electrical resistivity measurements), all ZnO-coated NPCs were exposed to the ambient atmosphere. XRD measurements were performed with Cu Kα1 radiation using a rotating anode generator and a high-resolution powder diffractometer (Rigaku Corporation) at 50 kV and 200 mA. 2θ was scanned from 20° to 80° at 1°/min with a 0.02° step size. The NPCs were affixed to low-background quartz XRD sample holders with adhesive tape, and one exterior surface was scanned. The NPCs were then fractured to expose the internal surfaces and rescanned along the exposed surfaces. No significant differences in XRD patterns were observed between the internal and external surfaces. The patterns were indexed using ICDD PDF cards for Al2O3, ZnO, and Zn (#04-002-5941, #04-001-7297, #00-004-0831, respectively). Phase quantities and grain size were determined via Rietveld refinement and Halder–Wagner crystallite size analysis in Jade 9 software (MDI). The whole-pattern fitting procedure was iterated until the residual stabilized at a value within one percentage point of 15% for all three samples. This procedure yielded corundum grain sizes between 90 and 100 nm, in agreement with the Al2O3 NP size from BET, dNP = 100 nm.

For electrical resistivity measurements, the top and bottom surfaces of the NPCs were coated with blanket films deposited by e-beam evaporation at 10−6 Torr and nominally consisting of 20 nm of Ti and 200 nm of Au. Following metal deposition, prisms approximately 6 × 3 × 1.6 mm were cut out of the NPCs for measurement in order to avoid sample edge effects. With the as-deposited metal films serving as contact pads, two-point resistivity was measured through the thickness of the samples using a standard probe station with one probe on the top of the sample and the other on the conductive sample stage. Resistance measured at 10 V was converted to volumetric resistivity using the areas of the metal films, which were somewhat reduced by flaking during sample cutting.

NPCs were prepared for SEM by sectioning, mounting in epoxy, grinding, and polishing, followed by coating with a conductive layer (see Note S2 in the supplementary material44 for details). SEM imaging and characterization was performed in a Quattro S ESEM (ThermoFisher Scientific) equipped with an EDAX x-ray detector (AMETEK Inc.) for energy-dispersive x-ray spectroscopy (EDS). EDS was used to collect maps and line scans of the prepared NPC cross sections at 30 kV accelerating voltage, with a 63 nA probe current, 50 μm aperture, 10 mm working distance, and 0.96 μS amp time. Data collection and analysis were performed using APEX software (ametek). Maps were collected through the thickness of the NPC cross section, an area approximately 1.5 × 1 mm with a 200 μS dwell time over 32 frames. Line scans of 100 points through the thickness of the NPCs, along the approximate center line, were collected with a 100 μm width and 20 ms dwell time per point over 512 frames. Imaging in backscattered electron mode was also performed.

X-ray photoelectron spectroscopy (XPS) data were collected using a Thermo Scientific Nexsa surface analysis system with a monochromatic Al Kα source (1486.7 eV), 125 μm spot size, and flood gun for charge compensation. Spectra were obtained before a 1 keV monatomic Ar ion gun etch and after 15 s and 30 s of etching to remove adventitious carbon. Survey spectra were acquired using a 1 eV step size and 200 eV pass energy. High-resolution scans of the Al 2p, C 1s, O 1s, and Zn 2p peaks were collected using a 0.1 eV step size and 20 eV pass energy. Peak fits were performed using Thermo Avantage software with convolutions of Gaussian–Lorentzian line shapes and Shirley backgrounds. Spectra were referenced to the C 1s peak of adventitious carbon at 284.8 eV.

Based on the ZnO ALD studies discussed in Sec. I, our DEZ exposure of 32 min at ∼9 Torr is expected to result in Zn metal deposition at T = 160 °C.11,12 On the other hand, at T = 120 °C, our H2O purge time of 48 min is expected to result in additional ZnO deposition due to the reaction between residual H2O and DEZ during the subsequent DEZ exposure.8 Therefore, the temperature-cycled 120/160 °C process, which is designed to suppress both potential side reactions, is expected to yield the least deposited mass. A comparison of the weights of the Al2O3 NPCs (two for each run) before and after ZnO ALD confirms this hypothesis: the 120/160, 120, and 160 °C percentage mass gains of the NPCs are 26.4 ± 0.1%, 28.4 ± 0.1%, and 36.4 ± 1.6%, respectively, with ± values representing sample-to-sample variability. Provisionally assuming uniform coating of the NPCs (evaluated in Sec. III C), these mass gains correspond to reductions in the porosity from an initial value of ϕi = 50.5 ± 0.3% to final values of ϕf = 36.7 ± 0.1%, 30.6 ± 1.4%, and 38.0 ± 0.1%, respectively. Converting porosity reduction to ZnO coating thickness (averaging the results obtained from a spherical NP model and a spherical pore model4), the average ALD growth per cycle (GPC) for the 120/160 °C process is 2.14 ± 0.01 Å/cyc, which lies at the upper end of the range typically observed in conventional ZnO ALD on flat substrates.14,25 The GPCs for T = 120 °C and T = 160 °C are 2.37 ± 0.02 and 3.48 ± 0.26 Å/cyc, respectively, with ± values again representing sample-to-sample variability in mass gain.

Qualitative and basic quantitative understanding of the deposition reactions that give rise to these different GPCs can be obtained from the ALD chamber pressure traces. Plots of p(t) over the first five cycles of the three ZnO ALD runs are shown in Fig. 2, and the remaining 15 cycles are shown in Fig. S2 in the supplementary material.44 During a DEZ or H2O pulse, p sharply rises to ∼9 Torr as the reactant is admitted into the chamber. During a subsequent DEZ or H2O hold, when all valves are closed and the chamber is isolated, p gradually undergoes a net increase or decrease, respectively, due (in part) to the deposition half-reaction. The full deposition reaction is
Zn ( C 2 H 5 ) 2 + H 2 O ZnO + 2 C 2 H 6 .
(3)
FIG. 2.

(a)–(c) ALD chamber pressure, p, during the first five cycles of ZnO ALD inside Al2O3 NPCs: (a) T held at 120 °C, (b) T held at 160 °C, (c) T cycled between 120 and 160 °C. In all ALD runs, ZnO ALD is preceded by an H2O half-cycle. All 20 cycles are shown in Fig. S1 (Ref. 44). The ∼1 Torr spikes and dips during purges are due to periodic pressure drops in the N2 feed line that occurs when the connected glovebox draws N2. (d) Temperature of the interior surface of the quartz boat measured during a ZnO ALD test run performed under the same conditions as (c). In this work, T is defined as the approximate boat temperature at steady state under 20 SCCM N2 flow.

FIG. 2.

(a)–(c) ALD chamber pressure, p, during the first five cycles of ZnO ALD inside Al2O3 NPCs: (a) T held at 120 °C, (b) T held at 160 °C, (c) T cycled between 120 and 160 °C. In all ALD runs, ZnO ALD is preceded by an H2O half-cycle. All 20 cycles are shown in Fig. S1 (Ref. 44). The ∼1 Torr spikes and dips during purges are due to periodic pressure drops in the N2 feed line that occurs when the connected glovebox draws N2. (d) Temperature of the interior surface of the quartz boat measured during a ZnO ALD test run performed under the same conditions as (c). In this work, T is defined as the approximate boat temperature at steady state under 20 SCCM N2 flow.

Close modal
The behavior of the chamber pressure, p, during the DEZ and H2O hold steps can be understood through the following formulation of the half-reactions:10,27
| ( OH ) x + Zn ( C 2 H 5 ) 2 | O x Zn ( C 2 H 5 ) 2 x + x C 2 H 6
(4)
and
| O x Zn ( C 2 H 5 ) 2 x + H 2 O | OZn ( OH ) x + ( 2 x ) C 2 H 6 ,
(5)
where |– denotes a surface species, and x can in principle vary from 0 to 2. In previous studies of ZnO ALD, x has been found to be ∼1 for ALD on silica powder at 80 °C,10 1.37 for ALD on a quartz crystal microbalance (QCM) at 177 °C,28 and ∼1.4–1.6 for ALD on a QCM in the 80–250 °C temperature range.27  Figure 2 suggests that x is indeed greater than 1 in our ZnO ALD process: in the DEZ half-reaction [Eq. (4)], the number of moles of C2H6 produced exceeds the number of moles of reactant consumed (p increases), while in the H2O half-reaction [Eq. (5)], the opposite occurs (p decreases). We focus our analysis on DEZ rather than H2O, because the latter is more likely to undergo substantial multilayer adsorption, particularly on the non-sample surfaces within the chamber that are ∼90 °C or colder (see Sec. II B), which complicates the interpretation of the change in p.

Figure 3(a) shows the change in p during a DEZ hold step, δp, zooming in on the representative example of cycle 15. We assume that two processes contribute to the change in p during a DEZ hold: the deposition half-reaction [Eq. (4)] and a net leak of gas out of the chamber. The assumption of a net outward leak is based on a series of tests that revealed that the system’s dominant leak is typically across the pump valve that is used to isolate the chamber during hold steps. In the case of the temperature-cycled (120/160 °C) process, δp increases for the first ∼10 min of the DEZ hold, plateauing at ∼450 mTorr before slowly decreasing for the remainder of the hold, attaining an average rate of decrease of ∼2 mTorr/min over the last ∼10 min before the pump step. We interpret the change in the slope of δp from positive to negative as a transition from a reaction-dominated regime to a leak-dominated regime. Whether the diffusion-reaction process has saturated is difficult to determine from δpt) both in principle (the expected shape of the δp curve is not obvious, given the different leak rates of DEZ and C2H6 molecules) and in practice (leak rate can vary slightly due to inconsistent valve seating, and our slope analysis is limited by the 10 mTorr resolution of the recorded pressure data) see Note S3 in the supplementary material.44 Regardless, it is clear that gas generation due to reaction slows down significantly. In the case of T = 120 °C, δp qualitatively follows the same pattern as that of the 120/160 °C process but with quantitative differences: δp attains its maximum later (after ∼15 vs 10 min), the maximum is higher (∼500 vs 450 mTorr), and the terminal slope (TS, average slope over the last 10 min) is higher (approximately −1vs −2 mTorr/min). For T = 160 °C, the behavior of δp is quite different: δp rises throughout the hold, attaining a positive TS of ∼80 mTorr/min and reaching ∼2.7 Torr before the pump step. The higher maxima of the T = 120 °C and T = 160 °C δpt) curves are consistent with increased deposition relative to the 120/160 °C process. The higher TSs suggest that for T = 120 °C and T = 160 °C, gas-generating processes—subtle for T = 120 °C and dramatic for T = 160 °C—are still ongoing at the end of the DEZ hold that are absent or diminished in the 120/160 °C case.

FIG. 3.

ALD chamber pressure data and analysis for T held at 120 °C (blue, intermediate Δp), T held at 160 °C (red, highest Δp), and T cycled between 120 and 160 °C (green, lowest Δp). For details, see Note S3 in the supplementary material (Ref. 44). (a) Pressure change, δp, during a representative DEZ hold step (cycle 15) obtained by subtracting the pressure at the start of the hold step (∼9 Torr) from the total pressure. The noisier portion of the T = 160 °C data was obtained from the 1000 Torr gauge; all other data were obtained from the 10 Torr gauge. (b) Total increase in the number of moles of gas over the course of the DEZ hold step, Δng, (derived from Δp) vs cycle number. (c) Terminal slope (TS, the average pressure slope over the last 10 min of the DEZ hold step) vs cycle number.

FIG. 3.

ALD chamber pressure data and analysis for T held at 120 °C (blue, intermediate Δp), T held at 160 °C (red, highest Δp), and T cycled between 120 and 160 °C (green, lowest Δp). For details, see Note S3 in the supplementary material (Ref. 44). (a) Pressure change, δp, during a representative DEZ hold step (cycle 15) obtained by subtracting the pressure at the start of the hold step (∼9 Torr) from the total pressure. The noisier portion of the T = 160 °C data was obtained from the 1000 Torr gauge; all other data were obtained from the 10 Torr gauge. (b) Total increase in the number of moles of gas over the course of the DEZ hold step, Δng, (derived from Δp) vs cycle number. (c) Terminal slope (TS, the average pressure slope over the last 10 min of the DEZ hold step) vs cycle number.

Close modal

Figures 3(b) and 3(c) show the differences in gas generation among the three ALD runs across all 20 cycles. For Fig. 3(b), the net change in pressure over the course of the hold, Δp, is converted to a net change in the number of moles of gas due to the DEZ half-reaction, Δng. This conversion involves correcting for the leak of gas across the stop valve as well as accounting for the reaction progress that occurs during the pulse step (see Note S3 in the supplementary material44). Setting aside the first five ALD cycles, which are assumed to be complicated by ZnO nucleation and the transition from deposition on Al2O3 to deposition on ZnO, and examining cycles 6 through 20, we observe a consistent pattern: Δng and TS are lowest for the 120/160 °C process, slightly higher for T = 120 °C, and much higher for T = 160 °C. We tentatively attribute the increases in Δng over the course of earlier cycles to increases in effective coating thickness deposited per cycle [as often observed in early cycles of many ALD processes, including ZnO ALD (Ref. 27)] and the decreases in Δng over the course of later cycles to reductions of the total mass deposited per cycle due to diminishing NPC surface area. By this logic, the terminal slope follows the same pattern because the rate of gas generation during the last 10 min (large for T = 160 °C and small for the other runs) is correlated with the total quantity of gas generated.

Analysis of the relationship between Δng and GPC provides additional insight into the differences among the three ALD runs. GPC, Δ n g ¯ (cycle-averaged Δng in units of μmol/cyc), and related quantities are summarized in Table I. Assuming for the sake of argument that only ZnO is deposited in all three ALD runs, we convert GPC into Δ n s ¯ (cycle-averaged moles of solid deposited, assumed to be ZnO, in units of μmol/cyc), which can be directly compared to Δ n g ¯. To facilitate the comparison of the three runs, we define a parameter, y ¯, that quantifies the net gain in gas moles per gain in solid moles,
y ¯ = 1 + Δ n g ¯ Δ n s ¯ .
(6)
TABLE I.

Mass gain measured by weighing Al2O3 NPCs before and after ZnO ALD; final porosity of the NPCs, ϕf (initial porosities are ϕi = 50.5 ± 0.3%), and ALD growth per cycle (GPC) derived from mass gain; average gain per cycle in the number of moles of solid, Δ n s ¯, assuming the solid is ZnO; average increase per cycle in the number of moles of gas over the course of the DEZ hold step, Δ n g ¯, derived from pressure data (see Fig. 3); gas-solid reaction product ratio, y ¯. The GPC ± values represent variability within pairs of NPCs coated in each ALD run, which is not applicable to Δ n s ¯ and Δ n g ¯, because they represent the total moles of solid deposited on and gas produced by a pair of NPCs. The measurement uncertainties of these quantities are approximately ±0.1, and, thus, the measurement uncertainty of y ¯ is approximately ±0.001; see note S3 in the supplementary material (Ref. 44).

Sample (°C)Mass gain (%)Final porosity, ϕf (%)GPC (Å/cyc) Δ n s ¯ (μmol/cyc) Δ n g ¯ (μmol/cyc) y ¯ = Δ n g ¯ / Δ n s ¯ + 1
120 28.4 ± 0.1 36.7 ± 0.1 2.37 ± 0.02 150 59.3 1.395 
160 36.4 ± 1.6 30.6 ± 1.4 3.48 ± 0.26 217 191 1.880 
120/160 26.4 ± 0.1 38.0 ± 0.1 2.14 ± 0.01 139 50.8 1.365 
Sample (°C)Mass gain (%)Final porosity, ϕf (%)GPC (Å/cyc) Δ n s ¯ (μmol/cyc) Δ n g ¯ (μmol/cyc) y ¯ = Δ n g ¯ / Δ n s ¯ + 1
120 28.4 ± 0.1 36.7 ± 0.1 2.37 ± 0.02 150 59.3 1.395 
160 36.4 ± 1.6 30.6 ± 1.4 3.48 ± 0.26 217 191 1.880 
120/160 26.4 ± 0.1 38.0 ± 0.1 2.14 ± 0.01 139 50.8 1.365 
If we assume that deposition is governed by Eqs. (4) and (5), then y ¯ is equal to x ¯, the average over 20 cycles of the reaction parameter, x,
y ¯ = 1 + Δ n C 2 H 6 ¯ Δ n Zn ( C 2 H 5 ) 2 ¯ Δ n ZnO ¯ = 1 + ( x ¯ 1 ) Δ n Zn ( C 2 H 5 ) 2 ¯ Δ n Zn ( C 2 H 5 ) 2 ¯ = x ¯ .
(7)
If other deposition reactions also occur, then
y ¯ = x ¯ + c ,
(8)
where c is a correction factor due to contributions from these reactions.

For T = 120 °C, T = 160 °C, and T cycled between 120 and 160 °C, y ¯ is 1.395, 1.880, and 1.365, respectively, with a measurement uncertainty of approximately ±0.001 (see Note S3 in the supplementary material44). In the 120 and 120/160 °C cases, in which there is no reason to doubt the pure-ZnO assumption, y ¯ can be considered a measure of ethane generation during the DEZ exposure via hydrogenation of DEZ’s ethyl ligands. In light of this, the slightly higher y ¯ for T = 120 °C is consistent with the expectation of more residual H2O, which could increase c by reacting directly with DEZ or monoethylzinc (MEZ) via surface or gas-phase reactions and/or could increase x ¯ by enhancing DEZ adsorption and facilitating DEZ-OH reactions.25,26 Additionally or alternatively, the higher y ¯ could be due to a higher surface density of chemisorbed OH, which in general is another plausible consequence of lower purge T.29 In any case, the difference in y ¯ between the 120 and 120/160 °C cases should be interpreted cautiously, both because the difference is small, and because it is not possible from these data alone to distinguish between the effects of H2O and OH. The significantly greater y ¯ for T = 160 °C, on the other hand, is consistent with the prediction that the deposited coating is, in fact, not pure ZnO but rather contains metallic Zn. Assuming that the main byproduct of Zn deposition is also C2H6,25,26 this deposition is expected to lead to high c in Eq. (4) due to both greater C2H6 generation (2 moles of C2H6 generated per mole of DEZ consumed) and the fact that the pure-ZnO approximation leads to an underestimate of Δ n s ¯ (due to an overestimate of molar mass).

Photographs of the top surfaces of NPCs before and after ZnO ALD are shown in Fig. 4. For T = 120 °C, ZnO ALD causes a slight yellowing of the surface and interior of the white NPC, which is expected for nonstoichiometric ZnO1−x containing oxygen vacancies,30,31 and indeed the n-type conductivity of ALD-grown ZnO has been shown to be associated with oxygen vacancies.32–35 In stark contrast, the T = 160 °C ZnO ALD process changes the NPC surface and interior color to charcoal gray. For T cycled between 120 and 160 °C, the NPC interior is approximately the same color as for T = 120 °C, but the top surface is light gray.

FIG. 4.

Photographs of the top surfaces of Al2O3 NPCs: (a) before ALD, (b) after ZnO ALD with T held at 120 °C, (c) after ZnO ALD with T held at 160 °C, (d) after ZnO ALD with T cycled between 120 and 160 °C. For (b) and (c), the interior of the ZnO-coated Al2O3 NS is similar in color to the top surface. For (d), the gray color is observed only on the surface, and the interior is approximately the same color as the top surface of (b).

FIG. 4.

Photographs of the top surfaces of Al2O3 NPCs: (a) before ALD, (b) after ZnO ALD with T held at 120 °C, (c) after ZnO ALD with T held at 160 °C, (d) after ZnO ALD with T cycled between 120 and 160 °C. For (b) and (c), the interior of the ZnO-coated Al2O3 NS is similar in color to the top surface. For (d), the gray color is observed only on the surface, and the interior is approximately the same color as the top surface of (b).

Close modal

As shown in Fig. 5, the material phases expected from the three ALD runs are confirmed by XRD patterns obtained from measurements of the top surfaces of the NPCs. All three samples produce narrow Al2O3 (corundum) peaks and broad ZnO (zincite) peaks, as expected for relatively large Al2O3 NPs with thin ZnO coatings. For T = 160 °C, clear metallic Zn peaks are also present. The XRD-derived weight percentages and crystallite sizes of ZnO and Zn for all three samples are shown in Table II. For T = 120 °C and T cycled between 120 and 160 °C, the zincite weight percentages (23.3 ± 0.2% and 17.7 ± 0.1%, respectively) and crystallite sizes (6.0 ± 1.1 nm and 5.1 ± 1.1 nm, respectively) from XRD are correlated with the coating weight percentages (28.4 ± 0.1% and 26.4 ± 0.1%, respectively) and thicknesses (4.73 ± 0.03 and 4.28 ± 0.03 nm, respectively) inferred from the mass gains. The discrepancy between zincite weight and coating weight suggests that a significant fraction of the coating is amorphous, which is unsurprising given that the coating is very thin and not lattice-matched with the substrate. The correlation between zincite weight/size and coating weight/thickness is not maintained for T = 160 °C—the zincite weight percentage and crystallite size are not larger, despite the significantly higher coating weight percentage of 36.4 ± 1.6%—and evidently the larger mass gain is predominantly due to metallic Zn, which is consistent with the dramatic darkening of the NPC. The light gray color of the top surface of the 120/160 °C sample, however, currently remains unexplained, despite surface characterization of all three samples via XPS (see Figs. S3 and S4 in the supplementary material44).

FIG. 5.

X-ray diffraction patterns of ZnO-coated Al2O3 NPCs for T held at 120 °C (blue, top pattern), T held at 160 °C (red, middle pattern), and T cycled between 120 and 160 °C (green, bottom pattern). Reference powder diffraction patterns are for Al2O3 (corundum), ZnO (zincite), and metallic Zn.

FIG. 5.

X-ray diffraction patterns of ZnO-coated Al2O3 NPCs for T held at 120 °C (blue, top pattern), T held at 160 °C (red, middle pattern), and T cycled between 120 and 160 °C (green, bottom pattern). Reference powder diffraction patterns are for Al2O3 (corundum), ZnO (zincite), and metallic Zn.

Close modal
TABLE II.

Weight percent and average crystallite size of ZnO and Zn metal, derived from XRD patterns via Rietveld refinement and The Halder–Wagner method, in ZnO-coated Al2O3 NPCs; electrical resistivity from vertical I-V measurements of current through the thicknesses of ZnO-coated Al2O3 NPCs with Ti–Au blanket electrodes deposited on the top and bottom faces. The wt. % and crystallite size ± values are XRD pattern fitting errors, and the ρ ± values are measurement uncertainties based on the uncertainties of the Ti–Au electrode areas due to flaking.

Sample (°C)ZnO (zincite) properties from XRDZn (metallic) properties from XRDElectrical resistivity, ρ (Ω-cm)
Wt. %Crystallite size (nm)Wt. %Crystallite size (nm)
120 23.3 ± 0.2 6.0 ± 1.1 Not detected — 1.3 ± 0.2 × 107 
160 17.1 ± 0.2 6.0 ± 1.5 12.5 ± 0.7 57 ± 42 8.3 ± 1.0 × 103 
120/160 17.7 ± 0.1 5.1 ± 1.1 Not detected — 1.2 ± 0.1 × 107 
Sample (°C)ZnO (zincite) properties from XRDZn (metallic) properties from XRDElectrical resistivity, ρ (Ω-cm)
Wt. %Crystallite size (nm)Wt. %Crystallite size (nm)
120 23.3 ± 0.2 6.0 ± 1.1 Not detected — 1.3 ± 0.2 × 107 
160 17.1 ± 0.2 6.0 ± 1.5 12.5 ± 0.7 57 ± 42 8.3 ± 1.0 × 103 
120/160 17.7 ± 0.1 5.1 ± 1.1 Not detected — 1.2 ± 0.1 × 107 

As shown in Table II, the electrical resistivities, ρ, of the NPCs from vertical two-point measurements (I-V measurements of current through the NPC thicknesses) are qualitatively consistent with the presence of metallic Zn in the 160 °C sample: the 160 °C sample’s resistivity of ∼8 × 103 Ω-cm is significantly lower than the resistivities of the 120 and 120/160 °C samples, ∼1 × 107 Ω-cm for both. Given that resistivity is likely highly sensitive to small variations among samples (cracking, adventitious carbon, etc.) and only one NPC was measured for each ALD condition, the very close match of the 120 and 120/160 °C samples should be interpreted cautiously; nevertheless, the more-than-three-order-of-magnitude difference between these samples and the 160 °C sample is significant. Considering the morphology of the ZnO coatings, it is not surprising that all three samples have much higher resistivity than typical ALD-grown ZnO thin films used for electrical measurements,13,14 which generally have thickness of the order of 100 nm and resistivity of the order of 10−2 Ω-cm. Factors contributing to the high resistivity of our NPCs include the ZnO coating’s very small thickness and crystallite size (smaller than a typical free electron mean free path in ZnO36), likely high amorphous fraction,37 and high tortuosity and low volume fraction.38 Regardless of the magnitude of resistivity, however, the measurable vertical current (at least ∼300 nA at 10 V, six orders of magnitude larger than the noise floor) reveals that the ALD coatings form percolating networks that span the thicknesses of the NPCs.

Cross-sectional SEM-EDS elemental maps and line scans, shown in Fig. 6, confirm that the ALD coatings infiltrate the full thicknesses of the NPCs. For each of Figs. 6(d)6(f), Kα x-ray counts are normalized to the total counts summed over all elements and all positions. The SEM images and Al Kα maps and line scans (and O Kα maps shown in Fig. S5 in the supplementary material44) reveal that the density of the Al2O3 varies slightly, with a ∼0.3–0.5 mm band of higher density located near the bottom of each NPC cross section, likely originating from nonuniform packing of the Al2O3 powder beds prior to pressing. Nevertheless, for T = 120 °C and T cycled between 120 and 160 °C, aside from the regions within ∼0.2 mm of the top and bottom surfaces, the normalized Zn Kα signal is essentially constant throughout the thicknesses. The normalized Zn Kα counts are 12% higher in the 120 °C sample, in agreement with the 10% higher fractional mass gain.

FIG. 6.

Cross-sectional SEM-EDS of ZnO-coated Al2O3 NPCs. (a)–(c) SE images, Al Kα maps, and Zn Kα maps for (a) T held at 120 °C, (b) T held at 160 °C, and (c) T cycled between 120 and 160 °C. The 160 °C sample cross section is incomplete due to sample damage during transport and handling. (d)–(f) Corresponding Al Kα, Zn Kα, O Kα, and C Kα line scans. Alignment with the images and maps in panels (a)–(c) is approximate.

FIG. 6.

Cross-sectional SEM-EDS of ZnO-coated Al2O3 NPCs. (a)–(c) SE images, Al Kα maps, and Zn Kα maps for (a) T held at 120 °C, (b) T held at 160 °C, and (c) T cycled between 120 and 160 °C. The 160 °C sample cross section is incomplete due to sample damage during transport and handling. (d)–(f) Corresponding Al Kα, Zn Kα, O Kα, and C Kα line scans. Alignment with the images and maps in panels (a)–(c) is approximate.

Close modal

XPS measurements of the top and interior surfaces of the NPCs (Figs. S3 and S4 in the supplementary material44) indicate that the slight decreases in Zn kα signal within ∼0.2 mm of the top and bottom surfaces are due to actual decreases in ZnO coating thickness. A possible cause of this ZnO depletion is the desorption of Zn-containing species during DEZ purges, analogous to, for example, desorption of Ti-containing species during TiO2 ALD from TiCl4 and H2O at high temperatures.39 While signs of significant desorption are not typically observed in ZnO ALD at moderate temperatures, our DEZ purge time of 48 min, which is not optimized and simply set equal to the H2O purge time, is unusually long. Additionally, the very high curvature of a ZnO coating deposited on 100 nm pores could facilitate desorption. One mechanism to consider is a curvature-induced increase in the barrier to the reaction between DEZ and OH (Ref. 25) such that some deposited Zn remains in the form of unreacted DEZ, which would be relatively weakly bound and likely to desorb compared to the typical surface species, MEZ.

For T = 160 °C, the notable decrease in Zn concentration near the center of the sample is consistent with non-self-limiting metallic Zn deposition: less Zn is expected farther from the NPC surfaces, where the time-averaged DEZ partial pressure is lower. Another factor that could contribute to a Zn gradient is a slightly slower reaction-diffusion process due to slightly fewer moles of DEZ; DEZ pulse pressure was held constant across runs, so there are ∼10% fewer moles at T = 160 °C than at T = 120 °C. In any case, the disproportionate increase in Al Kα signal that accompanies the Zn Kα decrease is likely due to a combination of a normalization effect and an x-ray reabsorption effect, given that the Zn L edge (∼1.2 keV) lies just below the Al K edge (∼1.6 keV) so that Al Kα counts can be somewhat enhanced by a reduction in Zn concentration.

Near-surface ZnO depletion aside, the otherwise uniform ZnO deposition throughout the 120 and 120/160 °C samples is consistent with saturation of a self-limiting reaction-diffusion process. The temperature-cycled process is indeed expected to produce this result, but it is not obvious that the T = 120 °C process should do the same, given the differences described in Sec. III A (higher GPC, Δ n g ¯, terminal slope, y ¯). As discussed above, these differences may be due to the predicted residual H2O, a higher surface density of OH, or both. The case of incomplete H2O removal is interesting, because in this scenario, it is likely that significant H2O desorption is ongoing when the DEZ exposure begins so that a significant portion of the excess deposition could proceed via gas-phase reactions akin to conventional chemical vapor deposition (CVD). In efforts to completely infill NP thin films via ALD, such residual-H2O-induced CVD has led to pore clogging and unwanted residual porosity,8 but arguably it is possible that in our experiments, uniform infiltration could be maintained provided that we remain far from complete infilling and that the ZnO thickness contributed by CVD is small compared to the pore diameter; for T = 120 °C, the excess deposition is 0.4–0.5 nm versus ∼100 nm pores. On the other hand, if the observed excess deposition in the T = 120 °C process is due to a higher OH density, an alternative hypothesis to consider is that the infiltration uniformity is simply due to an effective lack of CVD; that is, the excess deposition proceeds (nearly) entirely via additional surface reactions rather than gas-phase reactions. This argument assumes that, on a 103 s timescale, OH desorption is effectively self-limiting at an ultimate surface OH density inversely related to temperature (due to OH desorbing mainly in the form of H2O formed by OH–OH reactions with distance-dependent barrier height),40 which would imply that the dominant effect of a colder purge could be a higher quasi-steady-state surface OH density rather than a high final H2O vapor pressure in the pores due to insufficient progress in desorption. While the temperature dependence of OH coverage of ZnO between 120 and 160 °C may in many cases be too small to explain a 10% difference in ALD GPC,23,41 desorption from a high-curvature surface over the course of very long purges may be atypical (as discussed above in the context of DEZ), and, in general, we cannot rule out significant differences in OH coverage among ALD runs without further in situ and/or ex situ characterization.

Using a 20-cycle static-dose DEZ/H2O ALD process based on a simplified 1D reaction-diffusion model, we have deposited conformal ZnO coatings throughout the pores of macroscopic Al2O3 NPCs with ϕ = 50.5 ± 0.3%, L = 0.78 ± 0.02 mm, and AR = 7800 ± 200. When T is cycled between 120 °C (for all steps except the H2O purges) and 160 °C (for H2O purges), ZnO is deposited at an average GPC of 2.14 ± 0.01 Å/cyc. The ZnO mass appears to be constant throughout the thickness of the NPC, except for some depletion within ∼0.2 mm of the top and bottom surfaces, possibly due to losses during (unoptimized) long purges. The ZnO coating percolates across the NPC and enables vertical conduction from the top surface to the bottom, although the coating’s resistivity is significantly higher than that of a typical ALD-grown ZnO thin film. The resistivity can likely be greatly reduced by increasing the ZnO thickness and/or by improving crystallinity via post-ALD thermal annealing4 or increasing the H2O purge temperature.

When T is held constant at 160 °C throughout the ALD process, the ZnO deposition is accompanied by metallic Zn deposition. When T is held constant at 120 °C, the results are similar to those of the temperature-cycled process: the ZnO coating is pure, uniform (with the same near-surface exception), and has similar resistivity. The GPC, however, is 2.37 ± 0.02 Å/cyc, and the ALD chamber pressure suggests that in each cycle, the DEZ half-reaction takes longer to saturate and generates slightly more ethane molecules per Zn atom deposited. These results are arguably consistent with the prediction that, after a 48 min H2O purge at only 120 °C, there is significant additional residual H2O that can react with DEZ in the subsequent exposure step. Further experiments are required, however, to distinguish among potential reactions contributing to high GPC in the T = 120 °C process, including gas-phase reactions between DEZ and H2O (chemical vapor deposition) and surface reactions between DEZ and OH.

Regardless of whether the 120/160 °C process mainly reduces OH or H2O coverage, these results demonstrate that ALD temperature cycling can be used to control the purge step of ZnO ALD without inducing metallic Zn deposition. Therefore, temperature cycling may be a useful strategy for reducing purge time in cases where residual H2O is more likely to be a significant obstacle to uniform ALD infiltration—e.g., when the goal is complete infilling or when particularly low deposition temperature is required. Alternative heating strategies should be explored as a path to practical purge times. For example, whereas in this work heating of the ALD chamber has been used to increase T from 120 to 160 °C over the course of a 33.5 min temperature ramp, previously an ALD-reactor-integrated flash lamp has been used to increase Si substrate temperature by up to ∼400 °C within 5 s,42 and a high-power resistive heater has been used to increase sapphire substrate temperature directly by up to ∼800 °C within 15 s to drive film crystallization during ZnO ALD.43 On the other hand, the combination of uniform infiltration and higher GPC achieved by our 120 °C process suggests that there are some cases in which the number of ALD cycles required for a target coating thickness, and, hence, the total process time can be reduced by decreasing the purge temperature. The results of the 120 °C process merit further investigation of the relationships among H2O purge time/temperature, coating thickness, and infiltration uniformity.

The work at NRL was supported by the Office of Naval Research and the NRL Base Program. Austin J. Cendejas gratefully acknowledges support of the American Society for Engineering Education and the U.S. Naval Research Laboratory postdoctoral fellowship program (ASEE-NRL PDF program).

The authors have no conflicts to disclose.

Benjamin L. Greenberg: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (equal); Validation (equal); Visualization (equal); Writing – original draft (equal). Kevin P. Anderson: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Alan G. Jacobs: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Writing – review & editing (equal). Austin J. Cendejas: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Methodology (equal); Software (equal); Visualization (equal); Writing – review & editing (equal). Jenifer R. Hajzus: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Eric A. Patterson: Formal analysis (equal); Investigation (equal); Validation (equal); Writing – review & editing (equal). James A. Wollmershauser: Conceptualization (equal); Investigation (equal); Validation (equal); Writing – review & editing (equal). Boris N. Feigelson: Conceptualization (equal); Data curation (equal); Funding acquisition (equal); Methodology (equal); Project administration (equal); Resources (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
V.
Cremers
,
R. L.
Puurunen
, and
J.
Dendooven
,
Appl. Phys. Rev.
6
,
021302
(
2019
).
2.
Y.
Liu
,
M.
Gibbs
,
C. L.
Perkins
,
J.
Tolentino
,
M. H.
Zarghami
,
J.
Bustamante
, and
M.
Law
,
Nano Lett.
11
,
5349
(
2011
).
3.
Y.
Liu
,
J.
Tolentino
,
M.
Gibbs
,
R.
Ihly
,
C. L.
Perkins
,
Y.
Liu
,
N.
Crawford
,
J. C.
Hemminger
, and
M.
Law
,
Nano Lett.
13
,
1578
(
2013
).
4.
E.
Thimsen
,
M.
Johnson
,
X.
Zhang
,
A. J.
Wagner
,
K. A.
Mkhoyan
,
U. R.
Kortshagen
, and
E. S.
Aydil
,
Nat. Commun.
5
,
5822
(
2014
).
5.
J.
Ephraim
,
D.
Lanigan
,
C.
Staller
,
D. J.
Milliron
, and
E.
Thimsen
,
Chem. Mater.
28
,
5549
(
2016
).
6.
D.
Lanigan
and
E.
Thimsen
,
ACS Nano
10
,
6744
(
2016
).
7.
B. L.
Greenberg
,
Z. L.
Robinson
,
Y.
Ayino
,
J. T.
Held
,
T. A.
Peterson
,
K. A.
Mkhoyan
,
V. S.
Pribiag
,
E. S.
Aydil
, and
U. R.
Kortshagen
,
Sci. Adv.
5
,
eaaw1462
(
2019
).
8.
A.
Cendejas
,
D.
Moher
, and
E.
Thimsen
,
J. Vac. Sci. Technol. A
39
,
012406
(
2020
).
9.
A. J.
Gayle
,
Z. J.
Berquist
,
Y.
Chen
,
A. J.
Hill
,
J. Y.
Hoffman
,
A. R.
Bielinski
,
A.
Lenert
, and
N. P.
Dasgupta
,
Chem. Mater.
33
,
5572
(
2021
).
10.
P.
Ingale
,
K.
Knemeyer
,
M.
Piernavieja Hermida
,
R.
Naumann d’Alnoncourt
,
A.
Thomas
, and
F.
Rosowski
,
Nanomaterials
10
,
981
(
2020
).
11.
J. D.
Ferguson
,
A. W.
Weimer
, and
S. M.
George
,
J. Vac. Sci. Technol. A
23
,
118
(
2005
).
12.
J. A.
Libera
,
J. W.
Elam
, and
M. J.
Pellin
,
Thin Solid Films
516
,
6158
(
2008
).
13.
B. S. B.
Sang
and
M. K. M.
Konagai
,
Jpn. J. Appl. Phys.
35
,
L602
(
1996
).
14.
T.
Tynell
and
M.
Karppinen
,
Semicond. Sci. Technol.
29
,
043001
(
2014
).
15.
D.
Gouvêa
,
S. V.
Ushakov
, and
A.
Navrotsky
,
Langmuir
30
,
9091
(
2014
).
16.
S. O.
Kucheyev
et al,
Appl. Phys. Lett.
86
,
083108
(
2005
).
17.
A.
Yanguas-Gil
and
J. W.
Elam
,
ECS Trans.
41
,
169
(
2011
).
18.
A.
Yanguas-Gil
and
J. W.
Elam
,
Chem. Vap. Depos.
18
,
46
(
2012
).
19.
A.
Haaland
et al,
Dalton Trans.
22
,
4356
(
2003
).
20.
J.
Yim
,
E.
Verkama
,
J. A.
Velasco
,
K.
Arts
, and
R. L.
Puurunen
,
Phys. Chem. Chem. Phys.
24
,
8645
(
2022
).
21.
C. E.
Nelson
,
J. W.
Elam
,
M. A.
Cameron
,
M. A.
Tolbert
, and
S. M.
George
,
Surf. Sci.
416
,
341
(
1998
).
22.
M.
Schiek
,
K.
Al-Shamery
,
M.
Kunat
,
F.
Traeger
, and
Ch.
Wöll
,
Phys. Chem. Chem. Phys.
8
,
1505
(
2006
).
23.
K.
Morishige
,
S.
Kittaka
,
T.
Moriyasu
, and
T.
Morimoto
,
J. Chem. Soc., Faraday Trans.
76
,
738
(
1980
).
24.
B.
Meyer
,
D.
Marx
,
O.
Dulub
,
U.
Diebold
,
M.
Kunat
,
D.
Langenberg
, and
C.
Wöll
,
Angew. Chem. Int. Ed.
43
,
6641
(
2004
).
25.
T.
Weckman
and
K.
Laasonen
,
J. Phys. Chem. C
120
,
21460
(
2016
).
26.
T.
Weckman
and
K.
Laasonen
,
J. Phys. Chem. C
122
,
7685
(
2018
).
27.
J.
Cai
,
Z.
Ma
,
U.
Wejinya
,
M.
Zou
,
Y.
Liu
,
H.
Zhou
, and
X.
Meng
,
J. Mater. Sci.
54
,
5236
(
2019
).
28.
J. W.
Elam
and
S. M.
George
,
Chem. Mater.
15
,
1020
(
2003
).
29.
30.
V.
Srikant
and
D. R.
Clarke
,
J. Appl. Phys.
83
,
5447
(
1998
).
31.
C. K.
Coogan
and
A. L. G.
Rees
,
J. Chem. Phys.
20
,
1650
(
1952
).
32.
F. L.
Kuo
,
M.-T.
Lin
,
B. A.
Mensah
,
T. W.
Scharf
, and
N. D.
Shepherd
,
Phys. Status Solidi
207
,
2487
(
2010
).
33.
N.
Huby
,
S.
Ferrari
,
E.
Guziewicz
,
M.
Godlewski
, and
V.
Osinniy
,
Appl. Phys. Lett.
92
,
023502
(
2008
).
34.
T.
Nguyen
et al,
J. Mater. Chem. C
9
,
4307
(
2021
).
35.
Y.
Wang
,
K.-M.
Kang
,
M.
Kim
, and
H.-H.
Park
,
J. Vac. Sci. Technol. A
36
,
031504
(
2018
).
36.
J. F.
Chang
and
M. H.
Hon
,
Thin Solid Films
386
,
79
(
2001
).
37.
D. C.
Look
,
K. D.
Leedy
,
A.
Kiefer
,
B.
Claflin
,
N.
Itagaki
,
K.
Matsushima
, and
I.
Surhariadi
,
Opt. Eng.
52
,
033801
(
2013
).
38.
B. I.
Shklovskii
and
A. L.
Efros
,
Electronic Properties of Doped Semiconductors
(
Springer
,
Berlin
,
1984
).
39.
J.
Aarik
,
A.
Aidla
,
H.
Mändar
, and
T.
Uustare
,
Appl. Surf. Sci.
172
,
148
(
2001
).
40.
O.
Sneh
and
S. M.
George
,
J. Phys. Chem.
99
,
4639
(
1995
).
41.
A.
Afshar
and
K. C.
Cadien
,
Appl. Phys. Lett.
103
,
251906
(
2013
).
42.
M. C.
Langston
,
N. P.
Dasgupta
,
H. J.
Jung
,
M.
Logar
,
Y.
Huang
,
R.
Sinclair
, and
F. B.
Prinz
,
J. Phys. Chem. C
116
,
24177
(
2012
).
43.
B. D.
Piercy
,
J. P.
Wooding
,
S. A.
Gregory
, and
M. D.
Losego
,
Dalton Trans.
51
,
303
(
2021
).
44.
See supplementary material online for cross-sectional SEM of an Al2O3 NPC with no ZnO coating, ALD chamber pressure for all 20 cycles, atomic percentages of Al, Zn, O, and C from XPS, example XPS spectra, additional cross-sectional SEM-EDS of ZnO-coated Al2O3 NSs, note on the 1D reaction-diffusion model used for ALD recipe design, note on SEM specimen preparation, and note on determination of Δng and TS from the ALD chamber pressure traces with table of representative raw pressure data and corrections.

Supplementary Material