Atomic layer deposition (ALD) of mixed oxides has attracted increasing research attention in recent years due to its excellent capability of film composition tuning. This in turn highlights the importance of understanding the underlying surface chemistry which dictates how a film of desired composition is achieved. In this work, the authors examined the ability of atomic layer deposition to precisely control the film thickness and composition by studying the growth behavior of SnxTi1−xOy thin films deposited from an alkylamide Ti(IV) precursor, a β-diketonate Sn(II) precursor, and ozone. A set of samples with various compositions were deposited by controlling the ALD cycle ratio (ALDCR) of tin oxide/titanium oxide using our custom-built, warm-wall reactor. Both alloy- and laminate-type of growths were attempted by changing numbers of ALD subcycles while maintaining the cycle ratio. Growth rates, calculated based on the thicknesses measured by spectroscopic ellipsometery and x-ray reflectivity, showed a deviating pattern from that of linear interpolation using binary ALD processes, marked by an almost constant ∼0.06 nm/cycle. Film composition, determined by x-ray photoelectron spectroscopy, exhibited a concave upward dependence on ALDCR. The chemisorption density of each precursor was determined by x-ray reflectivity, and a linearly ALDCR-dependent decrease was observed. Structural analysis using x-ray diffraction showed a transition from anatase to SnO2 rutile when Sn content in the film was varied from 0 to 1, for O2 annealed samples. At ∼17 at. % Sn, a mixture of anatase and rutile phases was found. Other factors, such as surface roughness and surface chemical species, were examined in the attempt to account for the decreased chemisorption.

Atomic layer deposition (ALD) is a useful thin-film deposition technique based on gas (precursor)–solid (substrate) phase surface reactions. Due to its self-terminating nature, ALD excels in precise thickness control down to monolayer level and great film conformity over high aspect ratio structures.1 As the needs for materials with novel, improved, or more tunable properties grow, ALD's capability of composition tuning has been utilized to fabricate composite materials with properties that may be tailored to demands. For example, numerous film properties, such as surface roughness,2 refractive index,3 and band gap,4,5 have been modified systematically by controlling the doping level or concentration of the constituent components.

Mixed ALD films are generally achieved by combining two or more binary ALD processes, i.e., copulsing of the precursors followed by oxidizer6 or alternating the sequences of precursor A/oxidizer exposures and precursor B/oxidizer exposures.4,7 It is known that after each binary metal-oxide ALD cycle, the deposited layer thickness is typically less than that of a full monolayer.1 In other words, it takes more than one ALD cycle to completely deposit a monolayer of the oxide. Utilizing this characteristic with a careful surface chemistry control, well blended alloy growth may be obtained using fewer alternating cycles for each metal species, whereas laminate growth can result from extended cycles from each precursor,7 as shown in Fig. 1. Other types of film structures, such as alternating patterns and graded materials can also be engineered through ALD process design.

Fig. 1.

(Color online) Strategy for ternary ALD process: (a) Alternating the exposures of finely spaced sequences of each binary ALD to generate well-mixed alloy or solid solution film, showing here an atomic layer deposition cycle ratio of 0.4. Each vertical bar represents a pulse step in ALD reaction. Purging steps were omitted in the drawing. O, ozone pulse; Sn, Sn(acac)2 pulse; and Ti, TDEAT pulse. (b) Nanolaminates were achieved by extending the number of cycles for each binary ALD process while maintaining the 0.4 atomic layer deposition cycle ratio. The dots signify the omitted subcycles.

Fig. 1.

(Color online) Strategy for ternary ALD process: (a) Alternating the exposures of finely spaced sequences of each binary ALD to generate well-mixed alloy or solid solution film, showing here an atomic layer deposition cycle ratio of 0.4. Each vertical bar represents a pulse step in ALD reaction. Purging steps were omitted in the drawing. O, ozone pulse; Sn, Sn(acac)2 pulse; and Ti, TDEAT pulse. (b) Nanolaminates were achieved by extending the number of cycles for each binary ALD process while maintaining the 0.4 atomic layer deposition cycle ratio. The dots signify the omitted subcycles.

Close modal

The key to a successful mixed oxide ALD lies in the control of surface chemistry,8 which largely depends on the selections of precursor/oxidizer. There have been many reports on deviations in growth patterns and film compositions during mixed oxide ALD, when compared to the linear combinations of its constituent binary ALD processes.4,7,9,10 These discrepancies were studied, and possible causes were attributed to surface behaviors of “incompatible” precursors that lead to etching effect,7 incomplete ligand elimination,5 interrupted nucleation,10 and transfer of ligands.11 It is of crucial importance to investigate the growth patterns of multicomponent ALD films, as their properties rely on the deposition process, especially those making use of precursors with different functional groups as ligands.

Films of mixed SnO2 and TiO2 are attractive materials for gas sensing and photo-catalysis;12 they exhibit improved performance compared to each binary oxide alone13–16 and have been studied using other film synthesis techniques such as sol-gel17 and sputtering.18 SnO2 and TiO2 possess similar lattice parameters when both are present in rutile phases; thus, the formation of solid solution or alloy is possible.12 This may provide a means to achieve rutile TiO2 or control the percentage of the rutile structure in a film at relatively lower temperatures, which could be opposite to the generally observed tendency of anatase formation for TiO2 ALD films even after annealing at 900 °C (this work).

To extend the knowledge in mixed metal-oxide ALD processes, we investigated the growth behavior of ALD of a set of SnxTi1−xOy films with various atomic compositions using tetrakis(diethylamino)titanium (TDEAT), tin(II) acetylacetonate [Sn(acac)2], and ozone. With different ligands in the precursors, the growth behavior and film composition are found to deviate from interpolations using parameters from the known binary ALD processes. Film structural change, a potential driving force behind film compositional alteration, was examined using XRD.

ALD reactions were carried out in our custom-built, warm-wall (120 °C) reactor that is equipped with two precursor lines and an oxidizer line that can be connected to a water bubbler, an ozone generator, or any gas cylinder of choice. The operating and base pressures were maintained at ∼500 and ∼20 mTorr, respectively, with ultrahigh purity N2 as the purging and carrier gas. Both commercially available precursors, TDEAT (Sigma Aldrich) for Ti and Sn(acac)2 (Sigma Aldrich) for Sn, were kept in stainless steel bubblers and heated up to 65 and 70 °C, respectively, during deposition. P-type Si(100) wafer was used as the substrate, and cleaned according to Radio Corporation of America standard cleaning procedure (SC-1), followed by hydrofluoric acid (HF) treatment, right before each deposition.

Before ternary oxide ALD, each binary process was established and reported elsewhere.19,20 Temperature of 200 °C was selected from the wide overlapping window (175–275 °C) as the codeposition temperature. A general scheme of ternary ALD is depicted in Fig. 1, showing possible strategies to obtain alloy/solid solution and nanolaminates. Film composition is controlled by changing the ALD cycle ratio (ALDCR) of SnOx, and the sample set used in this study is listed in Table I.

Table I.

ALD conditions and film thicknesses for the samples used in this study.

Sample IDDeposition conditionsALD cycle ratio SnOx/(TiO2 + SnOx)Actual film thickness (nm)
[700 c TiO2] × 1 0.00 34.1 
[1 c SnOx, 9 c TiO2] × 50 0.10 29.9 
[1 c SnOx, 4 c TiO2] × 150 0.20 40.8 
[2 c SnOx, 3 c TiO2] × 120 0.40 45.3 
[1 c SnOx, 1 c TiO2] × 300 0.50 34.1 
[3 c SnOx, 2 c TiO2] × 80 0.60 26.7 
[2 c SnOx, 1 c TiO2] × 150 0.67 24.3 
[3 c SnOx, 1 c TiO2] × 180 0.75 39.9 
[6 c SnOx, 1 c TiO2] × 80 0.86 31.1 
10 [8 c SnOx, 1 c TiO2] × 60 0.89 27.3 
11 [350 c SnOx] × 1 1.00 32.1 
[6 c SnOx, 9 c TiO2] × 40 0.40 46.0 
[8 c SnOx, 12 c TiO2] × 35 0.40 38.7 
[10 c SnOx, 15 c TiO2] × 28 0.40 41.7 
[20 c SnOx, 30 c TiO2] × 14 0.40 41.2 
Sample IDDeposition conditionsALD cycle ratio SnOx/(TiO2 + SnOx)Actual film thickness (nm)
[700 c TiO2] × 1 0.00 34.1 
[1 c SnOx, 9 c TiO2] × 50 0.10 29.9 
[1 c SnOx, 4 c TiO2] × 150 0.20 40.8 
[2 c SnOx, 3 c TiO2] × 120 0.40 45.3 
[1 c SnOx, 1 c TiO2] × 300 0.50 34.1 
[3 c SnOx, 2 c TiO2] × 80 0.60 26.7 
[2 c SnOx, 1 c TiO2] × 150 0.67 24.3 
[3 c SnOx, 1 c TiO2] × 180 0.75 39.9 
[6 c SnOx, 1 c TiO2] × 80 0.86 31.1 
10 [8 c SnOx, 1 c TiO2] × 60 0.89 27.3 
11 [350 c SnOx] × 1 1.00 32.1 
[6 c SnOx, 9 c TiO2] × 40 0.40 46.0 
[8 c SnOx, 12 c TiO2] × 35 0.40 38.7 
[10 c SnOx, 15 c TiO2] × 28 0.40 41.7 
[20 c SnOx, 30 c TiO2] × 14 0.40 41.2 

Film thicknesses were measured using spectroscopic ellipsometry (SE) (J.A. Woollam Co., Inc., model M44) and x-ray reflectivity (XRR) with an incident beam of Cu Kα 0.1542 nm radiation (Philips X'pert). SE data were fitted in wvase software using a Cauchy model in the visible region, and XRR data were fitted using genx, an open source program written in python that employs differential evolution algorithm.

Elemental compositions of the films were determined using an x-ray photoelectron (XP) spectrometer (Kratos AXIS-165, Kratoz Analytical, Ltd., UK). Considering the surface sensitivity of the technique, Rutherford backscattering (RBS) spectroscopy that made use of a general ionex tandetron accelerator (model 5510, General Ionex, Co.) of selected films was also performed as a second proof. Grazing incidence x-ray diffraction (GIXRD) patterns of as-deposited and annealed films were obtained using a diffractometer (Philips X'pert, PANalytical, BN, Co.) configured with a Cu 0.1542 nm emission line.

The compositions of films were controlled by varying SnOx/(TiO2 + SnOx) ALDCRs. Samples 1–11 in Table I follow the trend of increasing Sn content, while maintaining the most finely spaced sequences of each binary ALD cycles, so that alloy films may result. Samples A–D were deposited at a fixed ALDCR of SnOx/(TiO2 + SnOx) of 0.4, each binary ALD cycles increased by factors of 2, 3, 4, and 9, respectively; thus, film structures that resemble those of nanolaminates were expected.

Figure 2(a) shows schematically how layers of ALD cycles were stacked for each sample. The theoretical thicknesses were represented as heights of bars. They were derived by linearly combining binary ALD growth rates (GRs), assuming each as-deposited layer preserves its thickness in ternary ALD as it does in the binary process. Actual thicknesses of films were calculated as the average of SE and XRR results, shown in Fig. 2(b). The percent difference in thickness using these techniques was less than 5%, indicating good agreement and reliable thickness determination. Figure 2(c) presents the comparison between the actual thickness and the theoretical thickness. It can be seen clearly that the actual thicknesses deviate from theoretical ones, indicative of different growth behavior during the ternary ALD process.

Fig. 2.

(Color online) (a) Theoretical thicknesses of samples 1–11. Each bar is a stack of horizontal short lines. A short line represents thickness deposited per ALD subcycle. TiO2 subcycles are shown in black, and those of SnOx in gray. (b) Actual thicknesses of the films measured using SE and XRR. (c) Comparison between theoretical thicknesses and actual thicknesses.

Fig. 2.

(Color online) (a) Theoretical thicknesses of samples 1–11. Each bar is a stack of horizontal short lines. A short line represents thickness deposited per ALD subcycle. TiO2 subcycles are shown in black, and those of SnOx in gray. (b) Actual thicknesses of the films measured using SE and XRR. (c) Comparison between theoretical thicknesses and actual thicknesses.

Close modal

To further investigate the deviating growth pattern, the term GR is defined as thickness grown per cycle, and calculated using the following formula:

GR=FilmthicknessTotalnumberofALDsubcycles.

Figure 3(a) shows growth rates for binary ALD processes, and linear least square fittings were used to generate GRTiO2 = 0.05 nm/cycle, and GRSnOx = 0.09 nm/cycle—nearly twice as much as GRTiO2. From these curves, we see no sign of prolonged nucleation period, and the perfect linearity signifies the typical ALD growth characteristic—the repeatable, well reactivated surface after each ALD cycle.

Fig. 3.

(Color online) (a) Linear dependence of binary ALD film thicknesses on number of ALD cycles. (b) Ternary ALD GRs as a function of ALDCR. (c) GRs of ternary ALD as a function of total number of subcycles, ALDCR is fixed at 0.4. (d) GRs of thin and thicker ternary ALD films at two cycle ratios.

Fig. 3.

(Color online) (a) Linear dependence of binary ALD film thicknesses on number of ALD cycles. (b) Ternary ALD GRs as a function of ALDCR. (c) GRs of ternary ALD as a function of total number of subcycles, ALDCR is fixed at 0.4. (d) GRs of thin and thicker ternary ALD films at two cycle ratios.

Close modal

To better understand the binary oxide ALD growth, the as-deposited film densities were examined using XRR and given in Table II. Both films exhibited lower densities compared to the bulk, crystalline oxides. The decrease in density for as-deposited ALD films has been reported,1 as some of the as-deposited ALD films tend to be amorphous and loosely packed. From our results, the as-deposited TiO2 film showed ∼80% of bulk anatase density. However, the as-deposited SnOx film had a drastic decrease in density compared to bulk SnO2 rutile or SnO, suggesting a loosely stacked structure inside. The crystallization of the SnOx film starts at 500 °C,19 which is much higher than the deposition temperature used.

Table II.

As-deposited film densities of metal atoms in binary oxides determined using XRR.

SpeciesFilm density (mol/cm3)Bulk density (mol/cm3)Atom/nm2 deposited per ALD subcycle
Sn in SnOx 0.0239 0.0461 (SnO2); 0.0479 (SnO) 1.58 
Ti in TiO2 0.0404 0.0473 (anatase) 1.19 
SpeciesFilm density (mol/cm3)Bulk density (mol/cm3)Atom/nm2 deposited per ALD subcycle
Sn in SnOx 0.0239 0.0461 (SnO2); 0.0479 (SnO) 1.58 
Ti in TiO2 0.0404 0.0473 (anatase) 1.19 

During one binary metal-oxide subcycle, 1.58 atoms/nm2 of Sn and 1.19 atoms/nm2 of Ti were deposited onto the surface. It is valid to hypothesize that during chemisorption step, 1.19 TDEAT and 1.58 Sn(acac)2 precursor molecules adsorbed in 1 nm2 area, assuming that all surface-bound precursors are converted to the corresponding oxides. Ti in TDEAT carries four alkylamide ligands, and Sn in Sn(acac)2 carries two β-diketonate ligands. Thus, it is more likely that TDEAT would experience a larger steric hindrance during chemisorption and end up with less adsorbed molecules.

Ternary ALD GR as a function of normalized ALDCR is presented in Fig. 3(b). Compared with the monotonically increasing GRs obtained using interpolation, the actual ternary ALD GRs remain a somewhat constant value of ∼0.06 nm/cycle with small fluctuations across all ratios. The deviations from the interpolated GR during ternary ALD have been reported before. Elam and George6 deposited ZnO/Al2O3 alloy films using diethyl zinc, trimethyl aluminum (TMA) and H2O. They discovered reduced GRs for samples across all ALDCRs and discussed the possible factors affecting the GR, such as nanocrystal size, change of surface morphology, and number of surface hydroxyl groups. With the help of in situ quartz crystal microbalance, they found two major events happening on the surface that led to the drop in ternary GRs: (1) etching of ZnO by TMA; (2) prolonged nucleation on each as-deposited oxide surface. The authors discussed multiple potential driving forces behind these two events, i.e., exothermally favored formation of Zn(CH3)2 and ZnAl2O4 spinel promoting the decrease of Zn content and relative acidities of ZnO and Al2O3 contributing in surface hydroxyl group deactivation. They also observed an increase in ZnO GR after renucleation. The evidences they presented indicate that multiple factors come into play during multielement ALD processes.

Mullings et al.4 also discovered the reduced GRs during zinc tin oxide ALD, and they attributed the reduction to prolonged nucleation and used an analytical model to fit the GR reducing factor. They also looked at GR as function of total number of subcycles in one super-cycle, called bilayer period, and found a GR restoration with increased bilayer periods. This restoration is expected if prolonged nucleation is the major factor affecting GR. To check if prolonged nucleation applies in our case, we fixed the ALDCR at 0.4, and increased the subcycles by n factor, shown in Fig. 3(c). That is, ALDCR of SnOx/(TiO2 + SnOx) = 2:3, 4:6, 6:9 and so on up to 40:60. The resulting GRs were nearly constant and in accordance with the GR having ALDCR = 2:3. This indicates prolonged nucleation on each oxide surface was not pronounced, which also agrees with the growth rate pattern shown in Fig. 3(b). If prolonged nucleation was pronounced, then the GRs would decrease across all ALDCR. Instead, at ALDCR of 0.1, 0.2, and 0.4, the actual GRs were similar to, or even higher than the interpolated GRs. Figure 3(d) showed growth rates for thin and thicker ALD films for two ALDCRs. Similar GRs were observed for both 600-cycle and 1800-cycle films, suggesting the similar growth behaviors hold for lengthened cycles.

Ternary GRs can be better understood in the context of film compositions. Figures 4(a) and 4(b) show XP spectra of tin titanate film deposited at 2:(3 + 2) ALDCR and the corresponding RBS spectra. The results from XPS and RBS agree well. Figures 4(c) and 4(d) shows the plot of film composition as a function of normalized ALDCR for Sn and Ti. The theoretical composition values according to rule of mixtures were calculated using the following equations:

CalculatedSnat.ratio=ρSn×%SnOx[ρSn×%SnOx+ρTi×(100%%SnOx)]=%SnOx[%SnOx+(ρTi/ρSn)×(100%%SnOx)],
(1)
Calculated Ti at. ratio=1calculated Sn atomic ratio,
(2)

where ρSn and ρTi are densities of Sn and Ti atoms deposited during each ALD cycle for pure SnOx and TiO2 films, namely, 1.19 atoms/nm2 and 1.58 atoms/nm2, respectively, and %SnOx is ALDCR of SnOx/(TiO2 + SnOx) × 100%.

Fig. 4.

(Color online) (a) XP spectra and (b) RBS spectra of the ternary film deposited at ALDCR = 2:(3 + 2). (c) Sn atomic ratio and (d) Ti atomic ratio as a function of normalized ALDCR.

Fig. 4.

(Color online) (a) XP spectra and (b) RBS spectra of the ternary film deposited at ALDCR = 2:(3 + 2). (c) Sn atomic ratio and (d) Ti atomic ratio as a function of normalized ALDCR.

Close modal

A slight deficit in Ti atomic ratio is observed in Fig. 4(d) for all ALDCRs. However, the trend of decrease does not resemble the one shown in the growth rate [Fig. 3(b)]. From Eq. (1), we see that the key term determining film composition is (ρTiSn), namely, the ratio of atoms deposited per unit area per ALD subcycle for Ti and Sn. This ratio is 1.19/1.58 = 0.75 from the pure binary ALD processes. The more concave upward shape of actual Sn ratio in Fig. 4(a) and the more concave downward shape of actual Ti ratio in Fig. 4(b) suggest that, during ternary ALD, a lowered (ρTiSn)ternary resulted. (ρTiSn)ternary was found to be 0.57 after fitting the film compositional data [Fig. 5(b) inset]. In order to find out the details about this decrease in ratio, the densities of all composite films were measured by XRR and ρTi and ρSn were plotted individually in Fig. 5(a).

Fig. 5.

(a) Chemisorption densities of precursor as function of ALDCR and film content. (b) Density ratio during ternary ALD process, and the fitted ratio using experimental composition data (inset).

Fig. 5.

(a) Chemisorption densities of precursor as function of ALDCR and film content. (b) Density ratio during ternary ALD process, and the fitted ratio using experimental composition data (inset).

Close modal

Both ρTi and ρSn showed a declining trend, which means as Sn cycles increase, less atoms were deposited during each ALD subcycle. The pure TiO2 film yielded a ρTi of 1.19 atoms/nm2. After the introduction of 1 Sn subcycle for every 9 subcycles of Ti, this number was lowered to 1.02, ∼15% of reduction. On the other hand, ρSn was 1.58 atoms/nm2 from the pure SnOx film, the incorporation of Ti ALD subcycles first enhanced the chemisorption of Sn(acac)2 by increasing the ρSn to 1.89 atoms/nm2, then ρSn gradually dropped as ALDCR increased. Before ALDCR of 0.5, ρSn was about the same as or higher than that from pure SnOx, thus preserving or exceeding the GR from pure binary process. ρTi, albeit less than that from binary process for all ALDCRs, had less impact on ternary GR since its growth rate from pure TiO2 was smaller (0.05 nm/cycle). This explains why ternary GRs were comparable to or larger than those interpolated ones at ALDCR less than 0.5 [Fig. 3(b)]. ALDCR of 0.4 is an interesting point where both ρTi and ρSn were recovered to some extent, thus at this ALDCR the ternary GR was about the same as predicted.

As Sn content further increased in the film, i.e., ALDCRs ≥ 0.5, the atoms deposited from each ALD subcycle further lessened. The lowest ρTi and ρSn were 0.51 and 0.77 atoms/nm2, respectively, occurred at ALDCR ∼ 0.89. The large discrepancy between the actual GRs and the predicted ones after ALDCR of 0.4 observed in Fig. 3(b) is likely a result of such low densities.

Interestingly, even though both ρTi and ρSn were decreasing, the ratio (ρTiSn), which the film atomic ratio depends on, did not vary much across the ALDCRs, shown in Fig. 5(b). 0.61 is the average of all measured ratios, 0.57 is the ratio extracted from fitting the compositions, and 0.75 is calculated using densities from the binary ALD processes. The consistent decreases in both ρTi and ρSn suggest that the factor contributing to reduced chemisorption takes effect on both oxide surfaces.

The decrease of chemisorption of both precursors shown in Fig. 5(a) may be the result of multiple driving forces; in this section, we present our findings on how surface roughness, and structural and compositional changes may affect the chemisorption of both precursors.

All as-deposited films of various ratios showed XRD-amorphous nature. The onset of crystallization occurred at an annealing temperature of 700 °C [Fig. 6(a)] for film with a Sn ratio of 0.6, yielding rutile phases. Figure 6(b) shows the structural evolution of films as a function of Sn content after annealing in O2 at 900 °C. Anatase, rutile, and cassiterite reference peaks were indicated using dashed lines and labeled as a, r, and c, respectively. Pure TiO2 film exhibited anatase phases, with a preferential (101) growth direction. After 17% of Sn incorporation, the anatase (101) peak intensity shrunk down by more than one third, and a new rutile (110) phase emerged, suggesting the coexistence of both phases. As Sn content further increases in the film, anatase (101) phase completely disappeared, and rutile (110) peak became predominant. At Sn ratio of 0.84, cassiterite (101) and (211) phases showed up, signifying the structural transition to cassiterite phases possessed by pure SnOx.

Fig. 6.

(Color online) (a) Film structure as function of annealing temperature. (b) Structural evolution of films with various Sn/(Sn + Ti) ratio at 900 °C.

Fig. 6.

(Color online) (a) Film structure as function of annealing temperature. (b) Structural evolution of films with various Sn/(Sn + Ti) ratio at 900 °C.

Close modal

The structural evolution reported here might affect the chemisorption in two possible ways: (1) the change in density of surface-bound hydroxyl groups due to a transition from anatase (101) to rutile (110) surface; (2) the interruption of anatase formation and the subsequent change in the growth mode. According to our experimental results and other reported work, the formation of anatase is favored for ALD of TiO2 at 200 °C,21 whose main trigger is compressive growth stress.22 The mode of anatase TiO2 growth was described as 3D island growth on the HF treated Si surface.22 The initial enhanced chemisorption of Sn(acac)2 observed at ALDCR of 1:9 (17% of Sn) was likely due to the enlarged surface area resulting from TiO2 crystalized islands or a possibly favored chemisorption reaction on the anatase (101) surface. However, the introduction of Sn tends to hinder the anatase formation process and change the growth stress, because the as-deposited SnOx was XRD-amorphous and SnO2 lattice parameters also differ greatly from those of anatase. It is reported that, on anatase (101) surface, adsorption of water was facilitated by compressive strain, but impeded by tensile strain.23 Due to the interruption of anatase formation by introducing Sn, the chemisorption of TDEAT would very likely decrease compared to the pure anatase case.

It is also important to look at the surface roughness of these samples. Rougher surfaces have larger surface areas, thus permitting more adsorption of precursors. Figure 7 shows the film surface roughness calculated from XRR,24 as a function of Sn ratio and ALDCR. Overall, the films showed smooth surfaces with slightly increased roughness at ALDCR of 1:9. The pure SnOx and TiO2 films exhibited similar roughness of ∼1.4 nm. At this roughness, the chemisorption of Sn(acac)2 on its own oxide surface (1.58 atoms/nm2) was much greater than on the mixed oxide surfaces grown at ALDCR ≥ 0.5, indicating that surface roughness alone is not the dominating factor affecting the chemisorption.

Fig. 7.

(Color online) Surface roughness measured using XRR.

Fig. 7.

(Color online) Surface roughness measured using XRR.

Close modal

Summarized in Table III are the XPS results for samples with various ALDCRs. Two types of oxygen were observed: those whose peaks located at 529.8–530.5 eV were assigned to lattice oxygen, and 531.1–531.7 eV were assigned to surface oxygen and/or oxygen defects.17,18 For pure TiO2 in polycrystalline anatase form, ∼19% oxygen was contributed from surface oxygen, very likely from the adsorbed hydroxyl groups. This number slightly increased to 23% for polycrystalline rutile TiO2 film deposited via ALD using annealed, polycrystalline SnO2 as the seed layer. Incorporation of Sn tends to increase the nonlattice oxygen percentage, at the same time oxygen peaks shift to higher binding energies. Pure SnOx film showed ∼48% surface oxygen or defects, much higher than that of pure TiO2 film. From the XPS results, the decrease in the number of surface hydroxyl groups cannot be definitively determined.

Table III.

XPS peak positions and calculated ratios of films with different Sn content.

Sn/(Sn + Ti)Binding energy (eV)OS/(OS + OL)OL at. ratio
Sn 3d5/2Ti 2p3/2Lattice oxygen (OL)Surface oxygen or defects (OS)
TiO2: anatase — 458.37 529.78 531.15 0.19 0.53 
TiO2: rutile — 458.49 529.91 531.29 0.23 0.48 
0.17 486.48 458.33 529.82 531.30 0.29 0.46 
0.31 486.56 458.34 529.82 531.30 0.44 0.41 
0.51 486.62 458.36 529.95 531.39 0.54 0.42 
0.62 486.71 458.43 530.23 531.65 0.44 0.42 
0.76 486.66 458.32 530.22 531.59 0.47 0.43 
0.84 486.72 458.39 530.38 531.72 0.41 0.42 
0.90 486.80 458.49 530.46 531.77 0.43 0.43 
0.92 486.81 458.48 530.48 531.78 0.44 0.42 
SnOx: amorphous 486.75 — 530.48 531.67 0.48 0.44 
Sn/(Sn + Ti)Binding energy (eV)OS/(OS + OL)OL at. ratio
Sn 3d5/2Ti 2p3/2Lattice oxygen (OL)Surface oxygen or defects (OS)
TiO2: anatase — 458.37 529.78 531.15 0.19 0.53 
TiO2: rutile — 458.49 529.91 531.29 0.23 0.48 
0.17 486.48 458.33 529.82 531.30 0.29 0.46 
0.31 486.56 458.34 529.82 531.30 0.44 0.41 
0.51 486.62 458.36 529.95 531.39 0.54 0.42 
0.62 486.71 458.43 530.23 531.65 0.44 0.42 
0.76 486.66 458.32 530.22 531.59 0.47 0.43 
0.84 486.72 458.39 530.38 531.72 0.41 0.42 
0.90 486.80 458.49 530.46 531.77 0.43 0.43 
0.92 486.81 458.48 530.48 531.78 0.44 0.42 
SnOx: amorphous 486.75 — 530.48 531.67 0.48 0.44 

At ALDCR ≥ 0.5, ρSn has shown a continuous, significant decrease, despite the fact that after this ratio the Sn cycles outnumber that of Ti. From the characteristics of growth pattern and film composition, we have ruled out the prolonged nucleation and etching effect as the major contributor to the decrease in density. Further evidence was given in Fig. 8, at ALDCR of 0.4, when bilayer periods increased, ρTi and ρSn were not recovered, but slightly reduced. There are other factors affecting the chemisorption of precursors, such as the reactivity of surface hydroxyl groups, which can be studied using other techniques.

Fig. 8.

Deposited atom per unit area per ALD subcycle as a function of bilayer periods at a fixed ALDCR of 0.4.

Fig. 8.

Deposited atom per unit area per ALD subcycle as a function of bilayer periods at a fixed ALDCR of 0.4.

Close modal

In this work, we studied the growth behavior and structural properties of atomic layer deposited tin titanium oxides using an alkylamide Ti precursor and a β-diketonate precursor. Excellent compositional tunability was achieved by controlling ALDCR. The observed deviating growth rates were successfully explained by change in densities of chemisorbed precursors. Chemisorption of Sn(acac)2 on TiO2 surface was first enhanced for ALDCR of 0.1 and 0.2, then retarded after ALDCR of 0.5. On the other hand, the density of chemisorbed TDEAT showed continuous decrease with increasing ALDCR, which, according to the structural analysis by XRD, was related to the interruption of anatase formation. Structural analysis has shown that all as-deposited films exhibit XRD-amorphous nature. The onset of crystallization occurred at an annealing temperature of 500 °C in N2 or O2. Mixed phases of rutile and anatase were obtained in films with 17 at. % Sn. Then a transition from TiO2 rutile to SnO2 rutile phases was observed with the increase of Sn content. The growth behavior and film composition reflect the surface chemistry that determine the characteristics of ALD process; by studying them, a better understanding of the ALD mechanism in mixed oxides can be achieved, and this will in turn facilitate the engineering of novel materials with tunable properties.

This project was supported by the National Science Foundation (Nos. DMR 1309114, CBET 1067424, and EEC 1062943). The GIXRD and XRR measurements were carried out in the Frederick Seitz Material Research Laboratory Central Facilities at the University of Illinois at Urbana-Champaign. This work also made use of instruments in the Electron Microscopy Service (Research Resources Center, UIC).

1.
S. M.
George
,
Chem. Rev.
110
,
111
(
2010
).
2.
J. W.
Elam
,
Z. A.
Sechrist
, and
S. M.
George
,
Thin Solid Films
414
,
43
(
2002
).
3.
H.
Kattelus
,
M.
Ylilammi
,
J.
Saarilahti
,
J.
Antson
, and
S.
Lindfors
,
Thin Solid Films
225
,
296
(
1993
).
4.
M. N.
Mullings
,
C.
Hägglund
,
J. T.
Tanskanen
,
Y
,
Yee
,
S.
Geyer
, and
S. F.
Bent
,
Thin Solid Films
556
,
186
(
2014
).
5.
A. J. M.
Mackus
,
C.
Maclsaac
,
W. H.
Kim
, and
S. F.
Bent
,
J. Chem. Phys.
146
,
052802
(
2017
).
6.
M. R.
Laskar
 et al,
ACS Appl. Mater. Interfaces
8
,
10572
(
2016
).
7.
J. W.
Elam
and
S. M.
George
,
Chem. Mater.
15
,
1020
(
2003
).
8.
F.
Zaera
,
J. Mater. Chem.
18
,
3521
(
2008
).
9.
J.
Heo
,
Y.
Liu
,
P.
Sinsermsuksakul
,
Z.
Li
,
L.
Sun
,
W.
Noh
, and
R. G.
Gordon
,
J. Phys. Chem. C
115
,
10277
(
2011
).
10.
T.
Törndahl
,
C.
Platzer-Björkman
,
J.
Kessler
, and
M.
Edoff
,
Prog. Photovoltaics: Res. Appl.
15
,
225
(
2007
).
11.
C.
Murray
and
S. D.
Elliott
,
ACS Appl. Mater. Interfaces
5
,
3704
(
2013
).
12.
K.
Zakrzewska
,
Thin Solid Films
391
,
229
(
2001
).
13.
M.
Radecka
,
K.
Zakrzewska
, and
M.
Rekas
,
Sens. Actuators, B
47
,
194
(
1998
).
14.
W.
Tai
and
J. H.
Oh
,
Sens. Actuators, B
85
,
154
(
2002
).
15.
F.
Fresno
,
J. A.
Coronado
,
S.
Tudela
, and
J.
Soria
,
Appl. Catal., B
55
,
159
(
2005
).
16.
C. F.
Lin
,
C. H.
Wu
, and
Z. N.
Onn
,
J. Hazard. Mater.
154
,
1033
(
2008
).
17.
B. K.
Kaleji
,
R.
Sarraf-Mamoory
,
K.
Nakata
, and
A.
Fujishima
,
J. Sol-Gel Sci. Technol.
60
,
99
(
2011
).
18.
K.
Zakrzewska
,
M.
Radecka
,
J.
Przewoźnik
,
K.
Kowalski
, and
P.
Czuba
,
Thin Solid Films
490
,
101
(
2005
).
19.
S. K.
Selvaraj
,
A.
Feinerman
, and
C. G.
Takoudis
,
J. Vac. Sci. Technol., A
32
,
01A112
(
2014
).
20.
S.
Chang
,
S. K.
Selvaraj
,
Y.
Choi
,
S.
Hong
,
S. M.
Nakhmanson
, and
C. G.
Takoudis
,
J. Vac. Sci. Technol., A
34
,
01A119
(
2016
).
21.
R.
Xu
and
C. G.
Takoudis
,
ECS J. Solid State Sci. Technol.
1
,
N107
(
2012
).
22.
D. R. G.
Mitchell
,
D. J.
Attard
, and
G.
Triani
,
Thin Solid Films
441
,
85
(
2003
).
23.
Y.
Zhang
,
F.
Hao
,
C.
Liu
, and
X.
Chen
,
Theor. Chem. Acc.
135
,
171
(
2016
).
24.
See supplementary material at https://doi.org/10.1116/1.5004993 for XRR analysis and film composition results comparison between that obtained by RBS and XPS.

Supplementary Material