The introduction of atomic layer deposition (ALD), to the microelectronics industry has introduced a large number of new possible materials able to be deposited in layers with atomic thickness control. One such material is the high-κ oxide HfO2; thermally stable and ultrathin HfO2 films deposited by ALD are a significant contender to replace SiO2 as the gate oxide in capacitor applications. We present a mechanistic study of the first deposition cycle of HfO2 on the Si(111) surface using tetrakis(dimethylamido) hafnium (TDMAHf) and water as precursors using operando ambient pressure x-ray photoelectron spectroscopy. Here, we show that the hydroxylation of the clean Si(111) surface by residual water vapor, resulting in a 0.3 monolayer coverage of hydroxyls, leads to instantaneous full surface coverage of TDMAHf. The change in the atomic ratio of Hf to C/N found during the first deposition half-cycle, however, does not match the assumed immediate ligand loss through reaction with surface hydroxyls. One would expect an immediate loss of ligands, indicated by a Hf:N ratio of approximately 1:3 as TDMAHf deposits onto the surface; however, a Hf:N ratio of 1:3.6 is observed. The partial hydroxylation on the Si(111) surface leads to binding through the TDMAHf ligand N atoms resulting in both N and CH3 being found remaining on the surface post water half-cycle. Although there is evidence of ligand exchange reactions occurring at Si–OH sites, it also seems that N binding can occur on bare Si, highlighting the complexity of the substrate/precursor reaction even when hydroxyls are present. Moreover, the initial low coverage of Si–OH/Si–H appears to severely limit the amount of Hf deposited, which we hypothesize is due to the specific geometry of the initial arrangement of Si–OH/Si–H on the rest- and adatoms.

Atomic layer deposition (ALD) as a technique has revolutionized the semiconductor and microelectronics industry by enabling the deposition of atomically resolved layers on complex topologies.1 ALD achieves this precision growth by cyclically flowing two or more precursors over the substrate onto which the layer should be deposited.2 Each precursor is kept separate from the other in the gas phase, which enables deposition control since the adsorption and reaction of the individual precursors on the surface is self-limiting. This process results in the growth of a highly uniform film with a high degree of control over thickness. In a particular ALD process, the choice of precursor pairing significantly impacts the resulting structure and properties of the film. Similarly, the selected substrate can also dictate the structure and properties of the overall film. In addition to the resulting film properties, the interface itself is exceedingly important in multilayer devices where the interface is often the limiting factor in device quality. With layers within devices now required to be as thin as 1 nm, understanding the reactions at the interface in terms of substrate preparation is of even greater importance.3 

The combination of high-κ metal oxides and Si has garnered considerable attention for applications and research in microelectronics. The miniaturization of transistors and capacitors necessitates a solution to the current leakage problem of traditional SiO2 capacitors at reduced sizes. High-κ metal oxides are primary candidates due to their thermal stability and the ability to form ultrathin films.4–6 HfO2, in particular, is of interest because of its permittivity of approximately 20 and a wide bandgap (5.5 eV), making it valuable not only in nanosemiconductor technology but also in optical uses such as antireflective coatings for mirrors and beam-splitters.

In the ALD of HfO2, using an alkylamido complex as the metal precursor and water as the oxygen source, it is typically assumed that the ALD mechanism is based on a ligand exchange reaction.7–9 This reaction follows a general formula: MLx+ Surf-OH(surf) → Surf-O-MLx−1 + LH, where M represents the metal ion, Surf the current surface, and L represents the metal complex ligand (number = x). However, this formula does not consider the initial reactions with the substrates.

As film thickness requirements get thinner and more stringent, the interface has a greater impact on the resulting film properties.3,10 We have recently demonstrated that the substrate plays a highly active role in the ALD chemistry and resulting interface properties.11–16 Irrespective of whether the substrate plays an active role or not in the deposition of HfO2, there is a commonality between the ALD chemistries of the tetrakis-dimethylamido metal precursors, namely, that protons are needed to enable the release of a dimethylamido ligand (DMA) in its entirety as dimethylamine (DMA). For instance, the initial ALD of HfO2 on an ideal SiO2 substrate without surface hydroxyls from tetrakis-dimethylamido hafnium (TDMAHf) and water proceeds by a H transfer reaction mechanism between the ligands of two physisorbed neighboring TDMAHf molecules during the first metal half-cycle, instead of a ligand exchange reaction with the surface.15 

Typically, ALD reactions and ALD chemistry are studied in ex situ or in vacuo experiments, wherein the surface is probed, at best, after every ALD half-cycle. Such studies provide good information about the static surface pre- and postreaction but do not probe the intricacies of the in situ interface/precursor reaction. It is crucial, however, to understand how the reaction progresses to truly understand what hinders or enhances the interface quality. To achieve in situ analysis, we use a specially adapted ambient pressure x-ray photoelectron spectroscopy (APXPS) setup, allowing for real-time tracking of the species evolution during precursor deposition. This method has already proven to elucidate unexpected reactions in the first few ALD cycles, building the understanding of initial ALD reaction mechanisms.11–16 

In this study, we examine the first cycle of HfO2 ALD on a partially hydroxylated Si(111) surface, using TDMAHf and water as precursors. We have successfully established the initial substrate/precursor interactions during the first cycle and investigated the role of moisture contamination in these interactions. The present study is part of a long-term project that considers the HfO2 ALD chemistry on various Si surfaces, including the thermally oxidized Si(111) surface15 and the clean Si(111) surface (present work) as well as other oxide-covered and clean Si facets and H-terminated Si surfaces (unpublished).

The experiment was performed at the APXPS end station of Sorbonne Université at the TEMPO beamline of the SOLEIL synchrotron in St-Aubin, France. A differential pumping stage separates the back-filled analysis chamber from the SPECS Phoibos 150 NAP electron energy analyzer equipped with a 3D (x, y, t) delay-line detector. The aperture of the cone between the analysis chamber and differential pumping stage of the analyzer has a diameter of 0.3 mm, and the sample was kept at a working distance of approximately twice the aperture to maintain maximal transmission and minimize the pressure gradient between sample and cone. All time-resolved measurements were recorded in snapshot mode in order to shorten acquisition time. During each precursor deposition, time-resolved data were recorded covering five core levels (Si 2p, C 1s, N 1s, Hf 4f, and O 1s). One measurement cycle took 13 s to complete, including instrument “deadtime.” All core-level spectra were recorded using a photon energy of 700 eV and resolution of 0.1 eV. The sample was moved continuously under the x-ray beam during the measurement to minimize the influence of photon-induced effects, commonly called “beam damage,” on the surface spectra. All core-level spectra were energy referenced to the Si 2p3/2 bulk peak at 99.3 eV. The IGOR Pro software was used for data analysis. Voigt functions were employed to deconvolute the peaks. Shirley, line, or polynomial type backgrounds, where appropriate, were removed from all spectra. In time-resolved spectra (of one core level) with more than one component, the Voigt profiles of a particular component were constrained to have the same full width half-maximum (FWHM) in all spectra.

A highly phosphorus-doped n-type Si(111) wafer with a resistivity of 10−3 Ohm cm (nd ≈ 2 × 1019 atoms/cm3) was used as the substrate for ALD. Degassing of the wafer was done at 600 °C for 8–12h (no chemical pretreatment was used). After degassing, the sample was cleaned from native oxide by flash annealing at 1100 °C. The surface temperature during the experiment was kept at 280 °C by using direct current heating. The sample temperature was measured using a pyrometer.

TDMAHf (99.99%, Strem Chemicals) and H2O (degassed via freeze-pump-thaw cycles) were dosed into the analysis chamber using a specialized delivery system extending into the analysis chamber to 3 cm from the sample surface. The pressure of the TDMAHf precursor was approximately 10−2 mbar. This relatively low pressure was chosen to allow for slowing of the kinetics of the surface reaction and extending observation time. It is, however, still within the range of standard reactor operation for ALD using metal alkylamido complexes.17–19 All precursor dosing was controlled via computerized valves (see Fig. S1 in the supplementary material31). ALD precursors were pulsed into static vacuum, and no pumping was carried out until evacuation of the precursor (or water). The pressure was allowed to increase from the initial precursor pulse.

The XPS data obtained during the first ALD half-cycle of HfO2 on a cleaned Si(111) surface from TDMAHf are presented in Fig. 1. Time-resolved APXPS enables the tracking of the surface species evolution. Although the analysis chamber was evacuated to ultrahigh vacuum (UHV) conditions prior to the experiment, the O 1s core-level spectra (see Figs. S2 and S3 in the supplementary material31) indicate that the initial highly reactive Si(111) surface is partially oxidized, with both SiOH and Si–O–Si species present in the O 1s spectra at 532.9 and 531.4 eV, respectively.20 Hence, residual oxygen species are present and are oxidizing the highly reactive surface. The total oxygen coverage was estimated to be 0.3 of a monolayer (ML) prior to the TDMAHf pulse. This estimation was calculated by modeling the bulk- and oxide-related components of the Si 2p core level and comparing to our experimental data. The results correspond well with literature values of Si(111) low-pressure studies of the adsorption of H2O and O on the Si(111)(7 × 7) surface, where the adsorption primarily occurs at rest- and adatom sites,21–23 which make up approximately 0.3 ML of the surface [1 ML is defined as 1 O atom per the simplest Si(111) unreconstructed surface unit cell].

FIG. 1.

(a)–(f) Time-resolved data of the first metal half-cycle. All spectra have been calibrated and normalized to the Si 2p Si bulk peak (99.3 eV). The color scales for (a)–(c) and (e) are 10% of the Si 2p spectra scaling in (d). From left to right, the O1s, N1s, C1s, Si2p, and Hf4f core-level spectra are shown. Time 0 is the time of the initial introduction of the precursor into the system. Panel (f) shows the overall pressure in the sample environment during exposure. (g)–(i) Fitted O 1s spectra taken at the indicated points pre-exposure (−2.5–0), 0–2, 6–8, 12–14, and 18–20 min (or postexposure). The so-derived spectra were further normalized by dividing by the intensity by the number of summed spectra. (g) The initial Si–OH O 1s peak, in purple, is indistinguishable from noise during the TDMAHf exposure, only returning once the chamber has been evacuated. This is evidence of an ongoing oxidation of the Si substrate in addition to the formation of Si–O–Hf. (h) N1s spectra fitted with two Voigt functions representing N from the original DMA ligand of the precursor (blue) and N, which is forming a linking bond to Si in the substrate and the Hf metal center (green). (i) C1s spectra fitted to three Voigt functions representing C in the original DMA ligand (blue), CH3 groups bound to the Si substrate (green), and C contamination originally found on the Si substrate (orange). The raw data are represented by brown + markers and the fit envelope is in black.

FIG. 1.

(a)–(f) Time-resolved data of the first metal half-cycle. All spectra have been calibrated and normalized to the Si 2p Si bulk peak (99.3 eV). The color scales for (a)–(c) and (e) are 10% of the Si 2p spectra scaling in (d). From left to right, the O1s, N1s, C1s, Si2p, and Hf4f core-level spectra are shown. Time 0 is the time of the initial introduction of the precursor into the system. Panel (f) shows the overall pressure in the sample environment during exposure. (g)–(i) Fitted O 1s spectra taken at the indicated points pre-exposure (−2.5–0), 0–2, 6–8, 12–14, and 18–20 min (or postexposure). The so-derived spectra were further normalized by dividing by the intensity by the number of summed spectra. (g) The initial Si–OH O 1s peak, in purple, is indistinguishable from noise during the TDMAHf exposure, only returning once the chamber has been evacuated. This is evidence of an ongoing oxidation of the Si substrate in addition to the formation of Si–O–Hf. (h) N1s spectra fitted with two Voigt functions representing N from the original DMA ligand of the precursor (blue) and N, which is forming a linking bond to Si in the substrate and the Hf metal center (green). (i) C1s spectra fitted to three Voigt functions representing C in the original DMA ligand (blue), CH3 groups bound to the Si substrate (green), and C contamination originally found on the Si substrate (orange). The raw data are represented by brown + markers and the fit envelope is in black.

Close modal

Previous theoretical studies on the reaction of water with the reconstructed Si(111) surface conclude that the surface hydroxyls are predominantly located at rest-atoms and the proton at both ad- and rest-atoms. “Rest-atoms” refer to atoms that retain their original place within the Si(111) lattice structure, while “adatoms,” short for adsorbed atoms, are atoms that have been displaced from their original lattice position and “adsorbed” onto the surface during reconstruction. Considering the TDMAHf molecule as a sphere of radius 5 Å and the location of the hydroxyls at rest-atoms within the Si(111) (7 × 7) surface reconstruction, the bulky TDMAHf molecule is spatially limited to two molecules per (7 × 7) unit cell (see Fig. S4 in the supplementary material31). This restricted arrangement gives a maximal TDMAHf coverage of 0.3 ML.

Upon exposure of the surface to the TDMAHf precursor (t = 0), we immediately observe signals relating to the intact amido ligands of the TDMAHf precursor in the N 1s, C 1s, and Hf 4f core levels at 398.8, 286.1, and, 16.7 eV, respectively (cf. Fig. 1).14–16 This is accompanied by a significant reduction in the original Si–OH O 1s signal [Fig. 1(g)]. In addition to TDMAHf precursor-related signals, we also observe an N 1s signal at 397.7 eV, which most likely arises from a Si–N(CH3)–Hf(DMA)3 surface complex [cf. Fig. 1(h)]. Studies on the decomposition of trimethyl amine on Si(111) show the loss of a methyl group and subsequent attachment to the Si surface through nitrogen, giving both Si–N–(CH3) and Si–CH3 groups.24,25 Indeed, we see an indication of methyl bound to Si at 284.9 eV. At t = 6 min, there is an increase of the Si–CH3 component at 284.9 eV in the C 1s spectra [Fig. 1(i)] and of the Si–N component in the N 1s spectra [Fig. 1(h)]. These signals continue increasing until the end of the exposure at t = 18 min.

Analysis of the Hf 4f spectra shows that 0.3 monolayers of TDMAHf has been deposited within the first 30 s, calculated with respect to a maximal packing density of TDMAHf on the surface (0.06 molecules per Å2). This corresponds to a 1:4.7 ratio between surface-bound TDMAHf and initial silicon/oxygen surface species. As shown in Fig. 2, the Hf 4f intensity does not exhibit a significant increase during the rest of the exposure, indicating no further deposition of the precursor after the initial exposure. A comparison of the Hf 4f and N 1s intensity gives an estimate of 3.7 ± 0.1 N atoms per Hf atom, following the methodology set out by D’Acunto et al.15 This amount is significantly more than the expected maximum of 3 N atoms (or less) per Hf atom. Throughout the deposition, the Hf 4f 7/2 signal moves to a higher binding energy from 16.7 to 17.1 eV (cf. Fig. 2), which has been an indication Hf–O bonds or evidence of bidentate-bonded TDMAHf.15,16,26 We expect the movement of the Hf 4f binding energy to continue to 18.0 eV, which is seen at the start of the H2O pulse after 4h of characterization.

FIG. 2.

Left axis shows the evolution of the Hf 4f7/2 peak binding energy over the first metal half-cycle, and the right axis shows the overall intensity of the peak; no statistically significant broadening of the peak was observed during the exposure so data has been fitted to one Voigt doublet.

FIG. 2.

Left axis shows the evolution of the Hf 4f7/2 peak binding energy over the first metal half-cycle, and the right axis shows the overall intensity of the peak; no statistically significant broadening of the peak was observed during the exposure so data has been fitted to one Voigt doublet.

Close modal

The O 1s signal broadens and increases over the first metal cycle [Fig. 1(g)]. This increase and broadening is a feature of further oxidation of the Si surface. In-depth studies of Si(111) oxidation show, depending on the extent and location of oxidation of the Si surface, significantly different O 1s binding energies.15,27 Although the snapshot mode is not sensitive enough to detect the oxidation in the Si 2p core level, the survey scans pre- and postexposure at UHV conditions confirm that Si has oxidized (see Fig. S2 in the supplementary material31). The silicate species (Si–O–Si, Si–O–Hf, and Si–O–C) are retained on the surface, along with expected N 1s and C1s related to both the precursor DMA- and Si-bonded (Si–C, Si–N, etc.) species. The Hf signal is also shifted toward a higher binding energy, 17.5 eV, in comparison to the end of the TDMAHf exposure. This is evidence of hydroxylation occurring on the Hf metal center.

Upon exposure of the sample to H2O (10 s pulse) in the first water ALD cycle, the majority of the DMA signal from the previous cycle is lost immediately (cf. Fig. 3). In the N 1s core-level spectrum, the signal remains, although it is too weak to be accurately fitted [cf. Fig. 3(h)], and in the C 1s core level, the Si–CH3 and N–CH3 signals remain along with the original contamination [cf. Fig. 3(i)]. Due to the low intensity and lack of separation of features, the initial fitting parameters were taken from the C 1s spectra post-TDMAHf deposition ensuring the peak position remained the same. The hydroxyl O 1s signal at 532.9 eV increases, showing the hydroxylation of the Hf metal center as well as the Si substrate [note the gas phase water peak at 536.1 eV (Ref. 28)]. The Hf 4f core-level spectrum now exhibits a well-defined doublet that remains unchanged throughout the remaining H2O exposure. We assign the doublet to Si–(O/N)x–Hf–(OH)4−x in agreement with the N 1s and C 1s core-level data.

FIG. 3.

(a)–(f) Time-resolved data of the first water half-cycle. All spectra have been calibrated and normalized to the Si 2p Si bulk peak. The color scales for (a)–(c) and (e) are 10% of the Si 2p spectra scaling in (d). From left to right core levels, O1s, N1s, C1s, Si2p, and Hf4f are shown, with time 0 as the initial introduction of the precursor into the experimental chamber. Panel (f) shows the overall pressure in the sample environment during exposure. (g)–(i) O 1s, N 1s, and C 1s spectra taken at the indicated points pre-exposure (−1–0), 0–2, 6–8, 12–14, and 18–20 min (or postexposure). The so-derived spectra were further normalized by dividing by the intensity by the number of summed spectra. The N 1s signal remains unfitted due to the lack of sufficient signal that would allow us to perform an accurate fitting.

FIG. 3.

(a)–(f) Time-resolved data of the first water half-cycle. All spectra have been calibrated and normalized to the Si 2p Si bulk peak. The color scales for (a)–(c) and (e) are 10% of the Si 2p spectra scaling in (d). From left to right core levels, O1s, N1s, C1s, Si2p, and Hf4f are shown, with time 0 as the initial introduction of the precursor into the experimental chamber. Panel (f) shows the overall pressure in the sample environment during exposure. (g)–(i) O 1s, N 1s, and C 1s spectra taken at the indicated points pre-exposure (−1–0), 0–2, 6–8, 12–14, and 18–20 min (or postexposure). The so-derived spectra were further normalized by dividing by the intensity by the number of summed spectra. The N 1s signal remains unfitted due to the lack of sufficient signal that would allow us to perform an accurate fitting.

Close modal

Initially, a pristine Si (111) surface was prepared; however, a slight background pressure of water (approximately 1 × 10−7 mbar) resulted in a partially hydroxylated surface with an estimated Si–O–Si/Si–OH coverage of 0.3 ML [w.r.t. to one O atom per base unit cell of Si(111) = 1 ML]. In spite of the presence of surface oxygen and surface hydroxyls, the experiment leads to significantly different results from those found for deposition of TDMAHf on the pristine SiO2 and fully hydroxylated surfaces.

At the initial exposure of our surface to TDMAHf, there is an immediate uptake of TDMAHf on the surface at approximately 0.3 of a ML or 1:4.6 of TDMAHf to the initial Si–O–Si/Si–OH concentration. In addition to the TDMAHf uptake, there is a corresponding loss of all of the –OH signal in the O 1s spectra, meaning that all initial –OH sites have reacted with TDMAHf molecules. However, the ratio of N:Hf is 3.7 ± 0.1 N per Hf molecule and not the expected <3 N per Hf if one follows an idealized ligand exchange model. It is evident that not all TDMAHf molecules are reacting by a ligand exchange process (see Fig. S5 in the supplementary material31).15 The N 1s spectra show a second N 1s species at 397.7 eV, cf. Fig. 1(h), which, considering no corresponding changes were observed in the Hf 4f or O 1s lines, is attributed to TDMAHf bound through the N atom forming Si–N(CH3)-Hf(DMA)3, in the process losing a methyl group. Similar behavior has been seen for trimethylamine (TMA) and DMA species on Si(111) at temperatures up to 600 K.24,26 In the C 1s spectra, we see a corresponding Si–CH3 signal at 284.9 eV, which then increases at the same rate as the Si–N signal. For the methyl group to be removed from the DMA ligand, it can either react with a Si–H site or remain bound to the surface. The lack of Si–CH3 in the initial part of the exposure suggests that removal into the gas phase is the more favorable reaction, but since available Si–H sites are used up, reaction directly with Si is still possible.24,25 The loss of methyl from the DMA ligand is supported by the C:N ratio, which is 1.5 rather than the expected 2 for an intact DMA ligand. When the Si–C species are included in this calculation, the ratio increases to 2.2. This is slightly higher than 2, which would be expected even in the case of partial dissociation of the DMA ligand, if the fragments remain on the surface. Likely, the higher value is due to a small amount of pre-existing contamination at the same binding energy.

As the reaction proceeds, we see a slow shift of the Hf 4f line toward higher binding energy, without any statistically significant increase in the intensity or full width at half-maximum (FWHM) of the signal. The intensity of a normalized core level is directly related to the amount of that element present, when considering one monolayer or less of a given molecule. Second, a change in the FWHM of a peak may be used as an indicator of a change in the number of chemical environments present for a specific element. This leads us to conclude that there is no additional new precursor bonding with the surface after the initial deposition; moreover, the number of chemically different environments present globally remains the same throughout the observed reaction. We, therefore, hypothesize that the increase in Hf 4f binding energy is likely due to the bound precursors moving from monodentate to bidentate bonding with the surface.26,29,30

At t = 6 min, there is an increase in the O 1s spectra due to the Si–Ox environments, which, when compared to previous XPS studies of Si(111) surface oxidation, match well with the continued insertion of oxygen into the Si–Si surface bonds and the various species this creates.15,27 Hydroxyls are not detected above the noise in the O 1s spectra at this point. However, it is likely they exist as a transient species on the surface involved in the oxidation of the Si surface and potentially in the transition from mono- to bidentate bonding of the TDMAHf being released as H2 or DMA, respectively. Increases are also observed in the Si–CH3 and Si–N peaks, TDMAHf molecules are also continuing to bind to the surface through the N of the DMA ligand. Within the time frame of the exposure (18 min), there is no stabilization of the Hf 4f binding energy, which may be due to a slow reaction mechanism with the substrate. Unexpected hydroxylation of the Hf metal center by background water can be excluded since other indicators of surface hydroxylation (i.e., strong O 1s OH signal and an increase in Hf 4f signal since OH groups provide fresh reaction sites) are not present.

Upon the water pulse, hydroxylation of the Hf center is exemplified in the significant decrease of both the N and C signals prior to the H2O pulse of the first water half-cycle. A H2O pulse removes the remaining DMA signal from the N 1s and C 1s spectra, which is accompanied by a shift to a higher binding energy in the Hf 4f spectra corresponding to a fully hydroxylated Hf center and final removal of all remaining DMA ligands. The O1s signal shows a significant increase in the –OH region as expected.

The results show that very small levels of gas and vapor residuals in the vacuum environment of the ALD chamber have a very profound influence on the ALD surface chemistry. In the present experiment, the “contamination” of the Si(111) surface with water has taken place after the sample had been removed from the UHV environment for the preparation of the ALD experiment, entailing the closure of pumping lines and concomitant increase of the pressure to 10−6 mbar. From a surface science perspective, it is not surprising that a surface that is exposed to such a pressure has some adsorbates after the shortest time, given that, according to basic gas kinetics, a full coverage of a surface by residual gas molecules [with a high sticking coefficient—such as water on the Si(111) surface] could be achieved within approximately 1 s in 10−6 mbar. In essentially all types of standard ALD reactors, the vacuum conditions are not better than in the present experiment, which means that residual gas levels are not better, either. This suggests that the influence of residuals gases in general and residual water, in particular, needs to be considered very thoroughly when formulating ALD reaction schemes and designing ALD processes.

In conclusion we show that, as with previous studies, the surface, its preparation, and even low-level contamination can greatly impact on the ALD reaction mechanism. In this study, we see that partial hydroxylation on the Si(111) surface leads to binding through the TDMAHf ligand N atoms, resulting in both N and CH3 being found remaining on the surface post water half-cycle. Although there is evidence of ligand exchange reactions occurring at Si–OH sites, it also seems that N binding can occur on bare Si, highlighting the complexity of the substrate/precursor reaction even when hydroxyls are present. Moreover, the initial low coverage of Si–OH/Si–H appears to severely limit the amount of Hf deposited, which we hypothesize is due to the specific geometry of the initial arrangement of Si–OH/Si–H on the rest- and adatoms. Future APXPS studies of the HfO2 ALD process, in which parameters such as deposition temperature and surface pretreatment in general and the amount of hydroxyl precoverage before the first the TDMAHf pulse, in particular, are systematically varied, will bring about further detailed insights into the exact ALD surface chemistry and how it is influenced by the initial state of the Si surface.

Considering that the majority of metal/H2O ALD reactions start with a hydroxylation step, this study brings attention to the importance of the hydroxylation itself. Even in fast pulse regimes, there is no guarantee of precursor surface reaction at hydroxyl sites only. How surfaces react with a water pulse is of utmost importance to the properties of the deposited film.

Vetenskapsrådet (VR) is acknowledged for project funding through Grant No. 2017-03871. This research used resources of the Center for Functional Nanomaterials (CFN), which is a U.S. Department of Energy Office of Science User Facility, at Brookhaven National Laboratory under Contract No. DE-SC0012704. The French synchrotron radiation facility SOLEIL is gratefully acknowledged for beamtime and their staff for assistance.

The authors have no conflicts to disclose.

Rosemary Jones: Formal analysis (equal); Methodology (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Giulio D’Acunto: Methodology (equal). Writing – review & editing (equal). Payam Shayesteh: Data curation (lead); Investigation (equal); Methodology (equal); Writing – original draft (equal). Indiana Pinsard: Data curation (equal); Investigation (equal). François Rochet: Investigation (equal). Fabrice Bournel: Investigation (equal); Methodology (equal). Jean-Jacques Gallet: Conceptualization (equal); Investigation (equal); Methodology (equal); Writing – review & editing (equal). Ashley Head: Conceptualization (equal); Data curation (equal); Investigation (equal); Methodology (equal); Writing – review & editing (equal). Joachim Schnadt: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Supervision (equal); Writing – original draft (equal); Writing – review & editing (equal).

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

1.
R. W.
Johnson
,
A.
Hultqvist
, and
S. F.
Bent
,
Mater. Today
17
,
236
(
2014
).
2.
V.
Miikkulainen
,
M.
Leskelä
,
M.
Ritala
, and
R. L.
Puurunen
,
J. Appl. Phys.
113
,
021301
(
2013
).
3.
R.
Xie
,
S. C.
Fan
,
J.
Frougier
,
M. A.
Bhuiyan
,
P.
Hashemi
,
T.
Ando
, and
A.
Reznicek
, U.S. patent 20230268388A1 (24 August 2023).
4.
W.
Jeon
,
H. S.
Chung
,
D.
Joo
, and
S.-W.
Kang
,
Electrochem. Solid State Lett.
11
,
H19
(
2008
).
5.
M. C.
Zeman
,
C. C.
Fulton
,
G.
Lucovsky
,
R. J.
Nemanich
, and
W.-C.
Yang
,
J. Appl. Phys.
99
,
023519
(
2006
).
6.
K.
Honda
,
A.
Sakai
,
M.
Sakashita
,
H.
Ikeda
,
S.
Zaima
, and
Y.
Yasuda
,
Jpn. J. Appl. Phys.
43
,
1571
(
2004
).
7.
D. M.
Hausmann
,
E.
Kim
,
J.
Becker
, and
R. G.
Gordon
,
Chem. Mater
14
,
4350
(
2002
).
8.
N. E.
Richey
,
C.
Paula
, and
S. F.
Bent
,
J. Phys. Chem
.
152
,
040902
(
2020
).
9.
S. T.
Barry
,
A. V.
Teplyakov
, and
F.
Zaera
,
Acc. Chem. Res.
51
,
800
(
2018
).
10.
R.
Xie
,
J.
Frougier
, and
K.
Cheng
, U.S. patent 11107827B2 (31 August 2021).
11.
G.
D’Acunto
et al,
ACS Appl. Electron. Mater.
2
,
3915
(
2020
).
12.
G.
D’Acunto
,
E.
Kokkonen
,
P.
Shayesteh
,
V.
Boix de la Cruz
,
F.
Rehman
,
Z.
Mosahebfard
,
E.
Lind
,
J.
Schnadt
, and
R.
Timm
,
Faraday Discuss.
236
,
71
(
2022
).
14.
A. R.
Head
,
S.
Chaudhary
,
G.
Olivieri
,
F.
Bournel
,
J. N.
Andersen
,
F.
Rochet
,
J.-J.
Gallet
, and
J.
Schnadt
,
J. Phys. Chem. C
120
,
243
(
2016
).
16.
G.
D’Acunto
et al,
J. Phys. Chem. C
126
,
12210
(
2022
).
17.
I.-K.
Oh
,
B.-E.
Park
,
S.
Seo
,
B. C.
Yeo
,
J.
Tanskanen
,
H.-B.-R.
Lee
,
W.-H.
Kim
, and
H.
Kim
,
J. Mater. Chem. C
6
,
7367
(
2018
).
18.
M. J.
Kelly
,
J. H.
Han
,
C. B.
Musgrave
, and
G. N.
Parsons
,
Chem. Mater.
17
,
5305
(
2005
).
19.
N. T.
Gabriel
,
S. S.
Kim
, and
J. J.
Talghader
,
Opt. Lett.
13
,
1958
(
2009
).
20.
A.
Namiki
,
K.
Tanimoto
,
T.
Nakamura
,
N.
Ohtake
, and
T.
Suzaki
,
Surf. Sci.
222
,
530
(
1989
).
21.
B. S.
Podolsky
,
V. A.
Ukraintsev
, and
A. A.
Chernov
,
Surf. Sci.
251–252
,
1033
(
1991
).
22.
C.
Poncey
,
F.
Rochet
,
G.
Dufour
,
H.
Roulet
,
F.
Sirotti
, and
G.
Panaccione
,
Surf. Sci.
338
,
143
(
1995
).
23.
X.
Wang
,
S.
Duan
, and
X.
Xu
,
J. Phys. Chem. C
117
,
15763
(
2013
).
24.
X.
Cao
and
R. J.
Hamers
,
J. Am. Chem. Soc.
123
,
10988
(
2011
).
25.
X.
Cao
and
R. J.
Hamers
,
Surf. Sci.
523
,
241
(
2003
).
26.
T. T.
Ngoc Van
,
D.
Jang
,
E.
Jung
,
H.
Noh
,
J.
Moon
,
D.-S.
Kil
,
S.-W.
Chung
, and
B.
Shong
,
J. Phys. Chem. C
126
,
18090
(
2011
).
27.
K.
Sakamoto
,
H. M.
Zhang
, and
R. I. G.
Uhrberg
,
Phys. Rev. B
68
,
075302
(
2003
).
28.
S.
Yamamoto
,
H.
Bluhm
,
K.
Andersson
,
G.
Ketteler
,
H.
Ogasawara
,
M.
Salmeron
, and
A.
Nilsson
,
J. Phys.: Condens. Matter
20
,
184025
(
2008
).
29.
A.
Akbari
,
J.
Hashemi
,
J.
Niskanen
,
S.
Huotari
, and
M.
Hakala
,
Phys. Chem. Chem. Phys.
17
,
10849
(
2015
).
30.
J.
Radnik
,
C.
Mohr
, and
P.
Claus
,
Phys. Chem. Chem. Phys.
5
,
172
(
2003
).
31.
See the supplementary material online for a scheme of the experimental setup, O 1s and survey XPS spectra of the initial surface, O 1s and survey XPS spectra of the initial surface, O 1s showing initial partial oxidation of the Si(111) surface, and a simplified reaction scheme.

Supplementary Material