Enhanced crystallization and dielectric properties of atomic layer deposited SrTiO₃ thin films on Ru electrode by inserting GeO₂ interfacial layer

Abstract

This study investigates the effects of a GeO₂ insertion layer between the SrTiO₃ (STO) dielectric layer and the Ru thin film bottom electrode on the crystallization behavior and associated electrical properties of the STO layer. A GeO₂ film as thin as 0.6 nm feasibly suppresses abnormal STO growth and ensures uniform Sr/Ti stoichiometry across the entire STO film thickness by blocking the oxygen exchange between the Ru and growing STO film. Furthermore, the GeO₂ insertion layer decreases the crystallization temperature of the STO film by ∼100 °C, enhancing film quality and facilitating crystallization through the diffusion of Ge atoms. The decreased thermal budget for fabrication enhances the surface smoothness of the STO layer, resulting in a void-free film. Finally, a Pt/RuO₂/STO/GeO₂/Ru capacitor is fabricated and optimized by fine-tuning the GeO₂ layer thickness and annealing temperature, achieving a minimum equivalent oxide thickness of 0.41 nm while ensuring a low leakage current density (<10⁻⁷ A cm⁻² at 0.8 V), even at a lowered annealing temperature. This highlights the excellent low-temperature compatibility and scalability of STO with GeO₂ insertion, making it a promising candidate for next-generation dynamic random access memory capacitor technologies.

---

New concepts

This work demonstrates a transformative approach by introducing an ultrathin (0.6 nm) GeO₂ interfacial layer, which fundamentally redefines interfacial engineering in atomic layer deposition. Unlike conventional barriers such as TiO₂ or Al₂O₃, which require considerably thicker films and increase interfacial defects or processing complexity, GeO₂ uniquely suppresses unwanted oxygen diffusion with minimal material addition. This precise oxygen-blocking capability ensures uniform Sr/Ti stoichiometry and lowers the crystallization temperature by ∼100 °C, thereby enabling high-quality, dense STO films at lessened thermal budgets. The novelty of this research becomes evident as scaling in memory device technology demands ultra-thin, high-dielectric materials processed under low-temperature conditions to prevent damage to adjacent components. By eliminating the need for a pre-crystallized seed layer, this approach simplifies fabrication while addressing persistent issues of abnormal film growth, carbon impurity incorporation, and defect formation. Moreover, the diffusion of Ge ions into the STO matrix acts as a crystallization enhancer, which provides a new pathway for modulating film properties at the atomic scale. Overall, this study not only offers a practical solution to current challenges in DRAM capacitor fabrication but also contributes critical insights into low-temperature crystallization mechanisms. It paves the way for advanced materials design in semiconductor applications, highlighting the pivotal role of innovative interfacial modifications in overcoming the limitations of existing research.

Introduction

Current-generation dynamic random access memory (DRAM) has been continuously scaled down to achieve higher bit density while maintaining efficient operation. As DRAM scaling decreased the capacitor area within the unit cell, maintaining sufficient cell capacitance using ZrO₂-based dielectric materials has become challenging due to their relatively low dielectric constant (30–40).^1–4 To meet the stringent capacitance requirements for DRAM with sub-10 nm design rules, research into higher-k (k > 100, where k is the dielectric constant) materials, such as TiO₂ and SrTiO₃ (STO), has become essential to secure sufficient cell capacitance and sensing margin.^5–8 Among the candidates, STO with a cubic perovskite structure has been regarded as the ultimate higher-k electric layer for next-generation DRAM capacitors due to its high bulk material k-value of ∼300.^9–12

Despite its higher-k properties, the relatively small bandgap energy (∼3.2 eV) of STO compared to ZrO₂ (∼5.5 eV) makes it more susceptible to leakage current, thereby compromising charge retention in the capacitor.^13–16 However, using the high work function materials, such as Ru (4.7 eV) or RuO₂ (5.2 eV), as bottom electrodes for STO to mitigate the leakage current results in the non-ideal atomic layer deposition (ALD) growth of STO film,^17–19 where ALD is an essential process for uniform deposition of the film on the extreme three-dimensional configuration of the DRAM capacitor. The interfacial RuO_x layer formed at the surface of the Ru substrate during the oxidative ALD process is reduced back to Ru during subsequent Sr or Ti precursor injections, since the Ru–O bonding strength is significantly weaker than that of Sr–O and Ti–O.^20,21 The unintended additional oxygen supply from reduced RuO_x disrupts the desired self-saturated ALD reaction during the initial deposition stage, leading to the overgrowth of a non-stoichiometric film, particularly with Sr-rich STO, due to the more substantial oxygen-scavenging property of Sr compared to Ti. Since the electrical and crystallization properties of STO are highly dependent on the cation stoichiometry, it is crucial to control the stoichiometry precisely by preventing abnormal growth.²²

The uncontrolled overgrowth can also result in carbonate formation due to incomplete ligand-exchange reactions, introducing defects such as nanoscale cracks or voids through low-density film deposition.^20,23 To address these issues, the authors’ group previously developed a two-step deposition process, which involves initially depositing and annealing a thin (3–5 nm) STO seed layer for crystallization, followed by subsequent deposition of the main STO layer, which resulted in the in situ crystallization of the main layer at the ALD temperature of 370 °C.²³ However, since the two-step process increases fabrication complexity and introduces risks of contamination from the ex situ processing step, obtaining defect-free STO films without employing the two-step method by effectively suppressing abnormal initial growth is highly desirable.

Several strategies have been explored to achieve high-density STO films without defects by regulating the reaction chemistry between the STO and Ru layers during the ALD process. Sr(demamp)(tmhd)₂ (tmhd = 2,2,6,6-tetramethyl-3,5-heptanedione, demamp = 1-{[2-(dimethylamino)ethyl](methyl)amino}-2-methylpropanol,) precursor with low reactivity was introduced to limit the reaction between Sr and RuO_x. However, the relatively long O₃ injection time required for the complete removal of the bulky (demamp)(tmhd)₂ ligand led to the formation of volatile RuO₄ at the bottom Ru electrode and subsequent degradation of its electrical properties.²⁴ Moreover, the difficulty in synthesizing the bulky Sr precursor would hinder mass production. The SrO overgrowth was also controlled by using a highly reactive Ti(Me₅Cp)(OMe) precursor (Me = methyl, Cp = cyclopentadienyl group), which enhanced the initial reactions between the Ti precursor and the RuO_x layer before the SrO deposition step.²⁵ The initial TiO₂ sub-cycles layer sufficiently covered the Ru substrate before Sr precursor injection, thus limiting the reaction between Sr precursor and RuO_x. However, the initial overgrowth of SrO and TiO₂ still existed, and the degree changed as the thin film grew, making precise atomic-level compositional control challenging. An alternative approach was to lower the STO ALD process temperature from 370 °C to 230 °C, which limited the oxygen diffusion from RuO_x. However, the low process temperature hindered the complete removal of ligands from the Sr and Ti precursors, which degraded the dielectric properties of the STO layer (k ∼49).²⁶

A different approach was inserting an oxygen-blocking layer between the STO and Ru layers to inhibit oxygen diffusion from RuO_x. Lee et al. inserted a 3 nm-thick TiO₂ layer and 1 nm-thick Al₂O₃ layer between Ru and STO.^20,27 Although these layers effectively blocked the oxygen supply from RuO_x, the capacitors suffered from an increase in equivalent oxide thickness (t_ox) due to the lower dielectric constants of the barrier oxides (Al₂O₃ ∼ 9, rutile TiO₂ ∼ 80) compared to STO (∼173).^28–30 The minimum t_ox values that meet the DRAM leakage criterion were 1.14 nm at a physical thickness of 9.9 nm (k ∼ 34) and 0.57 nm at a physical thickness of 10 nm (k ∼ 56) for STO with Al₂O₃ and TiO₂ barriers, respectively. These larger oxide thicknesses severely degrade capacitance and undermine DRAM scaling goals, underscoring the urgent need for a new, atomically thin barrier material.

This study introduced a new oxygen-blocking barrier material, GeO₂, as an alternative to conventional barrier materials. The oxygen diffusion coefficient in GeO₂ is 4.2 × 10² pm² s⁻¹ at 350 °C, substantially lower than that of TiO₂ (7.1 × 10³ pm² s⁻¹), which suggests a superior ability to block oxygen diffusion even with a thinner thickness than TiO₂ (3–4 nm).^31,32 Although the oxygen diffusion coefficient of Al₂O₃ is even lower (3.5 × 10¹ pm² s⁻¹),³³ a film thicker than 2–3 nm in the previous study was necessary to ensure a sufficient oxygen barrier property. This finding suggests that oxygen diffusivity alone does not determine the barrier performance of the device, and the interaction between the barrier oxides and the underlying Ru (RuO_x) layer is also crucial.³⁴ In this regard, GeO₂ demonstrated a unique capability to form an ultra-thin and conformal layer via ALD using a newly developed Ge precursor with moderate reactivity toward Ru. Additionally, the relatively weak Ge–O bond, compared to Ti–O and Al–O, induced the fluent diffusion of Ge atoms into STO after annealing.²¹ This Ge diffusion could promote the low-temperature crystallization of STO by enhancing atomic mobility and decreasing the formation of an interfacial low-k layer. As the capacitor integration in DRAM fabrication typically is conducted at the back-end-of-line (BEOL) stage, the processing temperatures below 600 °C can mitigate the risk of metal-line recrystallization, grain growth, and agglomeration, which can compromise structural integrity and device reliability. Thus, decreasing the annealing temperature below 600 °C is essential to ensure compatibility with existing structures while maintaining overall manufacturing yields and device performance. Consequently, inserting an ultra-thin GeO₂ layer as an oxygen barrier improved the STO film quality and electrical properties even without adopting a seed STO layer, making it a promising dielectric layer for next-generation DRAM capacitors.

Results and discussion

ALD behavior of GeO₂ and SrTiO₃

First, ALD process for GeO₂ thin films was established at a process temperature of 350 °C. The Ge(NMePh)(NMe₂)₃, (Ph = phenyl group, as shown in Fig. S1a, ESI†) was used as the Ge precursor, with O₃ as the oxygen source. Fig. 1(a) illustrates the changes in Ge layer density in a GeO₂ film with 50 ALD cycles on a Ru substrate as a function of precursor injection/purge and O₃ gas injection/purge times. When varying the Ge precursor injection time, a self-limiting growth of the GeO₂ layer was achieved after 5 seconds. Similarly, the self-limiting growth behavior was achieved with durations of 10 s, 2 s, and 10 s for the Ge precursor purge, O₃ injection, and O₃ purge, respectively. Hence, one ALD cycle for GeO₂ was set as 5–10–2–10 s for Ge precursor injection/purge and O₃ injection/purge, respectively. Fig. 1(b) illustrates the variations in Ge layer density on Ru and Si substrates as a function of substrate temperature. Here, the ALD sequence was identical to the abovementioned case, and the number of deposition cycles was 50. The Ge layer density was almost independent of the substrate temperature, ranging from 230 to 390 °C, indicating that the temperature range was in an ALD window. Previous studies have shown that STO deposition at lower temperatures (e.g., 230 °C) can result in degraded dielectric properties due to incomplete ligand removal. Therefore, a higher process temperature (e.g., 350 °C) is essential to obtain high-quality STO films.²⁰ Besides, the Sr precursor partially decomposed at temperatures above 350 °C.^20,26 Therefore, a process temperature of 350 °C for the GeO₂ film was employed in this study to enable consecutive ALD processes for the GeO₂ and STO layers in a given ALD chamber, facilitating the fabrication of STO/GeO₂ multilayer structures without severe thermal decomposition of precursors.

Fig. 1 Changes in Ge layer density of 50 cycles as functions of (a) precursor injection, precursor purge, reactant injection and reactant purge time on Ru substrates, and (b) substrate temperatures of Ru and Si substrates. (c) Changes in Ge layer density as a function of the number of Ge–O deposition cycles on Ru and Si substrate. (d) Ge 3d XP spectrum of 14 nm-thick GeO₂ film grown on Ru substrate.

Fig. 1(c) illustrates the variation in GeO₂ layer density on Ru and Si substrates as a function of the number of deposition cycles. On the Si substrate, linear growth was observed from the onset of growth, with a growth per cycle (GPC) of 14.45 ng cm⁻² cycle⁻¹ (0.49 Å cycle⁻¹), indicating no incubation cycles or excessive growth. In contrast, on the Ru substrate, the initial overgrowth occurred during the first 10 cycles, where the linear growth with a comparable GPC of 14.50 ng cm⁻² cycle⁻¹ (0.49 Å cycle⁻¹) was observed after five cycles. This overgrowth was attributed to the reduction of the interfacial RuO_x layer, as mentioned above, which was formed during the O₃ injection step and supplied additional oxygen atoms during the subsequent Ge precursor injection step. This oxygen-scavenging effect was attributed to the lower oxidation potential of RuO₂ compared to GeO₂ (standard Gibbs free energy of formation at 600 K, ΔG_f,600 K: GeO₂ −251, RuO₂ −201 kJ mol⁻¹).²¹ Such extra oxygen supply would enhance the initial adsorption of the Ge precursors, thereby increasing deposition amounts. However, the degree of the overgrowth was insignificant. Once the GeO₂ layer sufficiently covered the Ru surface, the oxygen-supplying effect from the substrate diminished, resulting in a growth rate similar to that on Si substrates. Fig. 1(d) shows the Ge 3d XP spectrum of 14 nm-thick GeO₂ film grown on Ru substrate at 350 °C. The peak position of the deconvoluted XP spectrum was located at ∼33.0 eV, corresponding to Ge⁴⁺, which confirms that the fully oxidized GeO₂ film was well-formed without any suboxides, such as GeO or Ge₂O₃.³⁵ Using the established ALD process, a GeO₂ layer was inserted between the STO layer and the Ru substrate as an oxygen-blocking barrier to suppress the abnormal growth of the STO film.

Next, the oxygen barrier property of the GeO₂ layer during the STO film growth on top was examined. Fig. 2(a) shows the variation in Sr and Ti layer densities as a function of the inserted GeO₂ thickness. Each film was deposited using a Ti : Sr sub-cycle ratio of 4 : 1 (as shown in Fig. S1b, ESI†), comprising 16 super-cycles. Without the GeO₂ insertion layer, both Sr and Ti exhibited high deposition amounts due to the above-mentioned oxygen-scavenging effects of Sr and Ti atoms. However, when even one cycle of the GeO₂ layer (∼0.3 nm) was introduced at the bottom interface, the deposition amounts of Sr and Ti decreased abruptly. They became saturated when the GeO₂ layer reached a thickness of 0.7 nm, corresponding to three deposition cycles of GeO₂. This result is highly improved as compared to the previously reported barriers by the authors’ group, where 3–4 nm-thick TiO₂ and 2–3 nm-thick Al₂O₃ layers were required to achieve saturated Ti and Sr deposition amounts, respectively.²⁷

Fig. 2 (a) Changes in Sr and Ti layer density in STO after 16 super-cycles on a GeO₂ barrier, as a function of the GeO₂ barrier thickness. (b) AFM images of Ru and Ru substrates covered with 0.3–0.5 nm-thick thin barriers. (c) t_ox–t_phy plots for the capacitor with a Pt/RuO₂/STO/oxygen barrier/Ru structure. The results for TiO₂ and Al₂O₃ barriers were reproduced with permission.²⁷ Copyright 2012, The Royal Society of Chemistry.

Fig. 2(b) shows the surface morphology of Ru substrate, TiO₂, Al₂O₃, and GeO₂ films (with thicknesses of 0.3–0.5 nm for the barrier films) using atomic force microscopy (AFM) images and measured root-mean-square (RMS) surface roughness (R_q). As TiO₂ and Al₂O₃ readily reduced surface RuO_x owing to their more substantial oxygen binding energies compared to RuO₂ (ΔG_f,600 K: TiO₂ −833.9, Al₂O₃ −1487 kJ mol⁻¹), excessive adsorption of the Ti or Al precursors would occur on the Ru film substrate, resulting in rougher surfaces with R_q values of 1.00 nm for TiO₂ and 1.18 nm for Al₂O₃.^25,36 Such rough surfaces may expose the uncovered Ru region, and they may also result in excessive adsorption of precursors and coalescence of the barrier films. Furthermore, when the subsequent STO layer was deposited onto these rough surfaces, the chemisorption sites for Sr and Ti precursors increased, potentially leading to excessive growth of STO. Therefore, TiO₂ and Al₂O₃ barriers should be sufficiently thick (∼3 nm) to fully cover the Ru surface without pinholes or protrusions, thereby minimizing oxygen supply from RuO_x and the excess adsorption of Sr and Ti precursors. Although GeO₂ also scavenged oxygen from RuO_x, it did so to a lesser extent due to the weaker oxygen binding energy of GeO₂ (ΔG_f,600 K: −251 kJ mol⁻¹). The surface was smooth, with a roughness of 0.79 nm, comparable to that of the bare Ru substrate. Since the Ge atom is bonded to amine groups in the Ge(NMePh)(NMe₂)₃ precursor, a strong interaction is promoted between the lone pairs of nitrogen and the charge of the highly conductive Ru metal. This resulted in a charge exchange interaction between the Ge precursor and Ru metal, allowing GeO₂ to form a flat and well-covered blocking layer with fewer cycles.³⁷ Consequently, GeO₂ exhibited a more effective blocking effect with a thinner layer compared to TiO₂ or Al₂O₃ due to its more extensive coverage and smoother surface morphology.

Fig. 2(c) shows equivalent oxide thickness (t_ox) – physical oxide thickness (t_phy) plots for the metal–insulator–metal (MIM) capacitors with a Pt/RuO₂/STO/oxygen barrier/Ru structure, where a 3 nm-thick TiO₂, 1 nm-thick Al₂O₃, or 0.6 nm-thick GeO₂ layer was employed as an oxygen barrier. The post-deposition annealing of the STO films was conducted at 650 °C. The bulk dielectric constant (k_bulk) was estimated from the slope of the best linear-fitted graph (t_ox = tⁱ_ox + 3.9 × t_phy/k_bulk), and the interfacial t_ox value (tⁱ_ox) was determined from the y-intercept, corresponding to the low-k and interfacial layers in the capacitor. The results for STO with TiO₂ and Al₂O₃ barriers were reproduced from a previous study, which employed a two-step process to suppress leakage current.²⁷ In contrast, the present study employs a one-step process to crystallize STO without a seed layer. Previous studies demonstrated that the leakage current property can be improved through a two-step process by suppressing cracks and shrinkage in the STO films.^24,38 The device performance estimated from the leakage current density (J–t_ox) graph was enhanced by the two-step process due to the significant suppression of the leakage current, as shown in Fig. S2 (ESI†). However, Fig. 2(c) shows that the bulk dielectric constant of the main STO layer deteriorated to ∼127 with a two-step process, compared to a one-step process (k ∼ 199), possibly due to the degraded crystallinity of the main STO layer resulting from cracks and contamination introduced in the seed STO layer using ex situ annealing. Therefore, a one-step STO process would be advantageous for achieving a high-k STO film if the high leakage current issue could be overcome by improving film quality. In the case of two-step processed STO capacitors with TiO₂ and Al₂O₃ oxygen barriers, the suppressed overgrowth resulted in higher k_bulk (152–173) compared to two-step processed STO directly deposited on Ru (∼127), as the barriers mitigated abnormal growth during the initial growth stage of the seed STO layers. However, these barriers also induced significantly higher tⁱ_ox values (∼0.90 nm and ∼0.48 nm for Al₂O₃ and TiO₂, respectively) due to their intrinsically low dielectric constants compared to the STO layer (Al₂O₃: ∼9, rutile TiO₂: ∼80), which hindered the t_ox scaling in DRAM capacitors. In contrast, STO capacitors with GeO₂ barrier showed the highest k_bulk (∼210) while maintaining a comparable tⁱ_ox value (∼0.17 nm) to the STO directly deposited on Ru substrate, despite the low dielectric constant of GeO₂ (∼4–13).^39,40 Although the GeO₂ layer slightly increased tⁱ_ox, it effectively prevented the more significant tⁱ_ox degradation caused by initial Sr-rich growth and interface roughening in STO films without oxygen-blocking barrier, resulting in a minimal net impact on tⁱ_ox. This performance was further attributed to the formation of a diffusive GeO₂ film rather than an independent layer between the STO and Ru, as discussed later. Although a thicker GeO₂ film might provide even higher oxygen-blocking capabilities, it would also degrade the overall capacitance, which is undesirable. Therefore, the optimal GeO₂ thickness to block oxygen diffusion was determined to be 0.6 nm in this study, adopted for further evaluations. In the following sections, STO films with a 0.6 nm-thick GeO₂ insertion layer at the bottom are referred to as ‘b-Ge-STO’, while STO films without any insertion layer are referred to as ‘pristine STO’.

The influence of the 0.6 nm-thick GeO₂ insertion layer on the growth behaviors of SrO, TiO₂, and STO films on Ru substrates was further examined in detail. Fig. 3(a) illustrates the changes in Ti layer densities in TiO₂ films deposited on bare Ru substrates and those covered with a 0.6 nm-thick GeO₂ layer. TiO₂ directly deposited on Ru exhibited initial overgrowth due to the additional oxygen supply from the RuO_x. Once the Ru surface was sufficiently covered with TiO₂ (∼20 cycles), self-saturated growth was achieved with a GPC of 9.7 ng cm⁻² cycle⁻¹ (0.38 Å cycle⁻¹). In contrast, TiO₂ deposited on Ru covered with a 0.6 nm-thick GeO₂ showed self-saturated growth with a similar rate of 9.4 ng cm⁻² cycle⁻¹ (0.37 Å cycle⁻¹) from the initial stage, demonstrating that the GeO₂ layer effectively blocked the additional oxygen supply from the Ru (or RuO_x layer) substrate to TiO₂.

Fig. 3 Changes in (a) Ti and (b) Sr layer densities on bare Ru and Ru covered with a 0.6 nm-thick GeO₂ as functions of TiO₂ and SrO deposition cycles, respectively. Changes in (c) Sr and Ti layer densities and (d) cation composition of pristine STO and b-Ge-STO as a function of the number of STO super-cycles.

Fig. 3(b) compares Sr layer densities in SrO films deposited on bare Ru substrates and Ru substrates covered with a 0.6 nm-thick GeO₂ layer, highlighting the differences when using H₂O and O₃ as oxygen source gases. The Ru surface was oxidized to RuO_x only when the highly reactive O₃ was used, whereas no significant oxidation occurred when H₂O was used.⁶ As a result, consistent SrO growth on bare Ru (GPC: 195 ng cm⁻² cycle⁻¹, 4.9 Å cycle⁻¹) and 0.6 nm GeO₂/Ru (GPC: 210 ng cm⁻² cycle⁻¹, 5.2 Å cycle⁻¹) without initial overgrowth was observed regardless of the presence of a GeO₂ insertion layer when H₂O was used. The higher GPC in the steady state when using H₂O instead of O₃ was attributed to the partial thermal decomposition of the Sr(ⁱPr₃Cp)₂ precursor at 350 °C, catalyzed by water adducts remaining on the precursor-adsorbed surface.²⁰ On the other hand, initial overgrowths caused by oxygen supply from the Ru surface were observed during the first five cycles when O₃ was used. In contrast, the 0.6 nm-thick GeO₂ layer suppressed severe overgrowth of SrO grown with O₃.

Fig. 3(c) shows changes in Sr and Ti layer densities in STO films grown on bare Ru (pristine STO) and Ru covered with a 0.6 nm-thick GeO₂ (b-Ge-STO) as a function of STO super-cycles. The y-intercepts of Sr and Ti layer densities represent the excessive incorporation at the initial stage. It should be noted that the oxygen supply from the bottom electrode during the SrO ALD subcycle in pristine STO growth still occurred despite using H₂O as the oxygen source for SrO because the previous TiO₂ ALD subcycle had utilized O₃ as the oxygen source, which had previously oxidized the Ru surface. The y-intercepts of Sr and Ti layer densities decreased from 1.14 and 0.33 μg cm⁻² (Sr and Ti layer densities for pristine STO, respectively) to 0.38 and 0.18 μg cm⁻² (Sr and Ti layer densities for b-Ge-STO, respectively), demonstrating the oxygen barrier effect of the GeO₂ layer. The constant growth rate per supercycle (GPSC) of Sr and Ti, calculated from the slope of each plot, was comparable between pristine STO and b-Ge-STO, indicating that the 0.6 nm-thick GeO₂ layer did not affect the growth behavior of STO beyond the initial stage.

Fig. 3(d) compares the cation composition ratio between Sr and Ti in pristine STO and b-Ge-STO films as a function of the STO supercycle number. Since the cation stoichiometry of STO significantly affects its dielectric and crystalline properties, precise control of stoichiometry is essential for achieving high-quality STO films, particularly in very thin layers.²² In the case of pristine STO, a Sr-rich composition (Sr ratio of 63 at%) was obtained during the first few super-cycles, subsequently saturating at around a Sr ratio of 56 at% as the number of super-cycles increased. In contrast, b-Ge-STO maintained a Sr ratio of 52–54 at% from the initial super-cycle stage, indicating that the almost stoichiometric composition of b-Ge-STO was achieved without substantial influence from the Ru substrate. These results suggested that an ultra-thin 0.6 nm-thick GeO₂ bottom insertion layer for the STO layer could be a viable solution to achieve uniform stoichiometry between Sr and Ti by suppressing the abnormal growth of the SrO and TiO₂ during the initial growth stage.

Improved film properties of b-Ge-STO

Fig. 4 shows the X-ray reflectivity (XRR) spectra of as-deposited and 650 °C-annealed pristine STO and b-Ge-STO films. The parameters used for simulating the spectra (solid red lines) to fit the experimental data (black dots) are noted in the figures. The total deposition amounts of STO in the two films estimated by X-ray fluorescence (XRF) spectroscopy were nearly identical: 4.11 μg cm⁻² for pristine STO and 4.08 μg cm⁻² for b-Ge-STO. The higher film density of as-deposited b-Ge-STO (4.13 g cm⁻³) compared to as-deposited pristine STO (3.79 g cm⁻³) resulted in a more compact and well-ordered structure with a thinner thickness (12.9 nm for b-Ge-STO vs. 14.0 nm for pristine STO). Since the ALD process for b-Ge-STO required a longer deposition time to achieve the same amount of STO (i.e., 12 super-cycles for pristine STO vs. 17 super-cycles for b-Ge-STO), the extended exposure to the 350 °C allowed more time for atomic arrangement, resulting in a denser and more uniform configuration.²⁶ After post-deposition annealing at 650 °C, the densities of the pristine STO and b-Ge-STO films increased to 4.63 g cm⁻³ and 4.73 g cm⁻³, respectively. At the same time, their thicknesses decreased from 14.0 to 10.9 nm and from 12.9 to 10.1 nm, respectively, due to crystallization and densification. The lower densification with film shrinkage in b-Ge-STO, due to its higher as-deposited density compared to pristine STO, would be advantageous for obtaining defect-free STO thin films.

Fig. 4 XRR spectra of as-deposited and 650 °C-annealed (a) pristine STO and (b) b-Ge-STO films.

The chemical composition of the pristine STO and b-Ge-STO films was analyzed in more detail. Fig. 5(a) and (b) show the time-of-flight secondary ion mass spectrometer (ToF-SIMS) depth profile of 12–13 nm-thick pristine STO and b-Ge-STO annealed at 650 °C. Due to the lack of a quantitative standard in ToF-SIMS analysis for STO matrix, raw ion-intensity signals could not be directly converted into absolute concentrations, and thus, all depth profiles were interpreted qualitatively. The Sr profile of the pristine STO film in Fig. 5(a) shows a higher Sr content at the interface between the STO and Ru layers than the bulk STO region. This Sr overgrowth was more pronounced than Ti due to a more substantial oxygen-scavenging property of Sr. On the other hand, Fig. 5(b) shows an almost uniform Sr profile throughout the b-Ge-STO film from the bottom interface to the bulk region due to the oxygen barrier effect of the GeO₂ layer.

Fig. 5 ToF-SIMS profiles of (a) 13.3 nm-thick pristine STO and (b) 12.6 nm-thick b-Ge-STO annealed at 650 °C. (c) O 1s XP spectra of 10 nm-thick pristine STO and b-Ge-STO annealed at 650 °C.

The carbon depth profiles of the two samples also showed notable differences. The pristine STO films underwent an overgrowth at the initial growth stage, resulting in the incorporation of ligands and associated carbon residues, particularly in the bottom interfacial region. In contrast, it was significantly suppressed in the case of b-Ge-STO, as the oxygen-blocking layer mitigated the overgrowth. The comparison of O 1s XP spectra in Fig. 5(c) can also provide further evidence of decreased carbon impurities. Each spectrum was deconvoluted into contributions from SrTiO₃ (529.2 eV), SrCO₃ (531.1 eV), and sub-oxides. Since surface contamination by air exposure was removed by Ar⁺ ion etching before XPS analysis, the relative SrCO₃ fraction reliably reflects the influence of carbon impurities within STO films. The SrCO₃ mole fraction, determined by the area ratio of the SrCO₃ to STO peaks, was significantly decreased from 18.2% for pristine STO to 4.8% for b-Ge-STO. These decreased carbon impurities enhanced the quality of STO films, as carbon impurities can introduce gap states within the STO bandgap, which may increase leakage currents and degrade high-k dielectric properties.^41,42 These results suggest that the GeO₂ bottom insertion layer not only improves the stoichiometric uniformity between Sr and Ti throughout the STO film but also suppresses impurity incorporation, which is crucial for achieving high-quality STO films.

Low-temperature crystallization of b-Ge-STO

The crystallization behavior of pristine STO and b-Ge-STO was examined. Fig. 6(a) shows the glancing incident angle X-ray diffraction (GIXRD) spectra of 12 nm-thick pristine STO and b-Ge-STO films annealed at various temperatures (400–650 °C), with an incident X-ray beam angle of 2°. Up to 400 °C, amorphous states of the STO films were observed regardless of the presence of a GeO₂ insertion layer due to insufficient thermal energy. At 475 °C, the STO (110) peak, observed at ∼32.4° (JCPDS#350734), was only present in b-Ge-STO, indicating that crystallization occurs at a lower temperature than in pristine STO. While the peak intensity of STO continued to increase up to 650 °C, b-Ge-STO showed comparable peak intensities at 575 and 650 °C, suggesting that complete crystallization of b-Ge-STO was already achieved at 575 °C.

Fig. 6 (a) STO (110) plane peaks of GIXRD spectra of 12 nm-thick pristine STO and b-Ge-STO annealed at 400–650 °C. (b) The RMS surface roughnesses of as-deposited and annealed pristine STO and b-Ge-STO films. (c) SEM images of pristine STO and b-Ge-STO annealed at 575 and 650 °C (left panel: high magnification (200k), and right panel: low magnification (50k)).

Fig. 6(b) compares the R_q roughness values of 12 nm-thick STO layers, as measured by AFM analysis, with Fig. S3 (ESI†) showing the corresponding AFM images. Even in the as-deposited amorphous state, pristine STO exhibited a high roughness value (R_q = 1.39 nm) due to accelerated growth in the initial growth stage, which induced roughening and grain coalescence. In contrast, b-Ge-STO exhibited a smooth surface with an R_q of only 0.40 nm due to the suppression of overgrowth on the smooth GeO₂ barrier layer. Although grain growth during post-deposition annealing resulted in a slight increase in roughness compared to the as-deposited state, the stabilized deposition of b-Ge-STO ensured that it maintained a low roughness value (R_q = 0.67 nm).

Fig. 6(c) shows the surface morphology of pristine STO and b-Ge-STO films annealed at 575 °C and 650 °C, as observed through planar scanning electron microscopy (SEM) images with high (×200k) and low (×50k) magnifications. As expected from Fig. 6(a), pristine STO annealed at 575 °C exhibited a partially crystallized structure, and complete crystallization with uniform grain growth was observed at 650 °C. The lower density of the amorphous region compared to the crystalline region resulted in variations in film thickness, leading to increased surface roughness (R_q = 1.57 nm). Furthermore, the shrinkage of pristine STO film by densification during post-deposition annealing would be more pronounced due to the lower film density of as-deposited pristine STO, which led to the formation of cracks and voids at the STO film surface after annealing, as demonstrated by the low-magnification SEM image in Fig. 6(c). In contrast, b-Ge-STO, even when annealed at 575 °C, exhibited uniform crystallinity throughout the entire film, resulting in a smoother surface (R_q = 0.70 nm) with minimal film defects, such as cracks and voids. The promoted crystallization at lower temperatures in b-Ge-STO can be further examined by analyzing the grain size of fully crystallized films annealed at 650 °C. Fig. 6(c) shows that the b-Ge-STO has a smaller average grain size (d_avg: 73 nm) compared to pristine STO (d_avg: 165 nm). This difference was attributed to the higher density of nano-crystallites in b-Ge-STO, resulting from the uniformly stoichiometric STO matrix during the early stages of crystallization, which limited lateral grain growth by increasing competition among adjacent nuclei.⁴³

Fig. 7 shows the crystal structures of pristine STO and b-Ge-STO annealed at 575 °C, as observed through cross-sectional high-resolution transmission electron microscope (HRTEM) images. The fast Fourier transform (FFT) patterns of the marked regions (A–C) are shown in the insets of Fig. 7(a) and (b). Pristine STO exhibited remaining amorphous phases (region B) without any reflections in the FFT pattern, while showing partially crystallized phases (region A). In contrast, the entire b-Ge-STO film was fully crystallized (region C). The d-spacing values of the (110) plane (d₁₁₀) for both samples, calculated from inverse FFT patterns, were identical as 2.75 Å, affirming that the GeO₂ insertion layer induced no significant lattice distortion. The low-magnification (×100k) TEM images in the lower panels of Fig. 7(a) and (b) demonstrate a more uniform and smoother b-Ge-STO than the pristine STO film.

Fig. 7 Cross-sectional high-resolution TEM (HRTEM) images of (a) pristine STO, and (b) b-Ge-STO annealed at 575 °C (Insets: FFT patterns of marked region in (a), (b), upper panels: high magnification (400k), and lower panels: low magnification (100k)). (c) C–V curves of Pt/RuO₂/575 °C-annealed pristine STO or b-Ge-STO/Ru MIM capacitors.

Fig. 7(c) shows the capacitance density–voltage (C–V) curves of RuO₂/pristine STO or b-Ge-STO/Ru-structured MIM capacitors after post-deposition annealing at 575 °C. The thickness of the STO layer was ∼13 nm for both capacitors. The dielectric constant of b-Ge-STO, calculated from the capacitance density at 0 V, was ∼113, implying the high quality of perovskite-structured STO. In contrast, pristine STO showed a lower dielectric constant (∼78) due to the presence of low-k amorphous regions. These results suggest that incorporating a GeO₂ bottom insertion layer enables the sufficient crystallization of STO films at lower annealing temperatures, significantly decreasing the thermal budget required during fabrication. This approach can ensure sufficient dielectric performance, making it a promising strategy for advanced capacitor applications.

The promoted crystallization in b-Ge-STO might be attributed to several factors. As the non-stoichiometry of STO elevated the crystallization temperature, the pristine STO with a Sr-rich composition required higher temperatures to crystallize.²² Furthermore, the as-deposited pristine STO would require greater atomic diffusion to transition from a randomly ordered amorphous state to an ordered lattice. In contrast, the denser structure of as-deposited b-Ge-STO (Fig. 4) would lower the energy barrier for lattice alignment.⁴⁴ Moreover, the role of inserted Ge (GeO₂) layers within STO should be considered, as Ge ions at the interface diffuse into the STO layer after post-deposition annealing, as shown in the Ge profile of b-Ge-STO (Fig. 5(b)). Previous studies have reported that metal or metal oxide additives with low melting points can facilitate crystallization by acting as a network modifier during the crystal growth process.^45–47 These additives enhanced the long-range rearrangement of ions and increased atomic mobility within the amorphous phase, which is essential for crystallization.⁴⁸ As the melting point of GeO₂ is significantly lower (1115 °C) than that of STO (2080 °C), Ge ions can modify the local electronic structure in the STO matrix, weakening atomic bonding energies and thereby enhancing atomic mobility.^49,50 These structural and electronic alterations collectively decrease the nucleation barrier, effectively decreasing the crystallization temperature compared to pristine STO.

Auger electron spectroscopy (AES) analysis was conducted to elucidate the diffusion behavior of Ge ions in STO layers in more detail. Fig. 8 shows the Ge AES depth profiles of as-deposited and annealed b-Ge-STO films with thicknesses of 11 and 23 nm. In the as-deposited state, Ge ions were predominantly concentrated at the inserted position. As the annealing temperature increased to 575–650 °C, Ge ions gradually diffused further into the STO layer due to the increased diffusivity of Ge and the dissociation of Ge from GeO₂ at higher temperatures. Since the amount of Ge ions remained consistent (two cycles of GeO₂) regardless of STO thickness, the diffused Ge ions primarily remained in the bottom region of the STO layer, as shown in Fig. 8(b), suggesting that the impact of the Ge ions would be diminished in the top region of the STO layer. Therefore, the effect of Ge ions as crystallization agents on STO would be more pronounced in thinner films, as later data confirms.

Fig. 8 AES depth profile of Ge atoms in (a) 11 nm-thick and (b) 23 nm-thick b-Ge-STO films.

The crystallization behavior of the STO at a low temperature of 575 °C was examined for various STO thicknesses. The minimum STO thickness required for crystallization at 575 °C was determined through GIXRD and t_ox measurements, as shown in Fig. 9. Fig. 9(a) and (b) show GIXRD spectra of pristine STO and b-Ge-STO films with various thicknesses annealed at 575 °C, respectively. The incident angle of the X-ray beam was 2°, ensuring that the crystallinity could be assessed across the entire film thickness. The diffraction peaks corresponding to STO (110) at ∼32.4°, (111) at ∼40.0°, and (200) at ∼46.5° were first detected at 8.5 nm for pristine STO and 6.4 nm for b-Ge-STO, indicating more efficient crystallization in b-Ge-STO. Fig. 9(c) and (d) show the t_ox–t_phy plots for pristine STO and b-Ge-STO films annealed at 575 and 650 °C. Films thinner than the minimum crystallization thickness remained amorphous, resulting in relatively large t_ox values with a dielectric constant of ∼20.⁵¹ Once t_phy exceeded the critical thickness for crystallization, t_ox abruptly decreased due to the formation of the crystalline phase. This transition occurred at 7.9 nm for pristine STO and 6.4 nm for b-Ge-STO after annealing at 575 °C, further confirming that the GeO₂ insertion layer facilitates crystallization at lower temperatures, in agreement with the GIXRD results. Even when pristine STO was crystallized, its k_bulk value was ∼82, significantly lower than that of pristine STO annealed at 650 °C (∼199) due to residual low-k amorphous regions at 575 °C, as shown in Fig. 7(a). In contrast, the k_bulk value of b-Ge-STO annealed at 575 °C was ∼210 for thicknesses up to ∼17 nm, identical to that of b-Ge-STO annealed at 650 °C (∼210). The slightly larger tⁱ_ox of 575 °C-annealed b-Ge-STO than 650 °C-annealed b-Ge-STO might be due to a higher density of Ge at the interface caused by less diffusion, as shown in Fig. 8. Meanwhile, the t_ox of b-Ge-STO annealed at 575 °C increased to levels comparable to pristine STO at t_phy values above 17 nm. As the thickness increased, the influence of Ge ions diminished since the Ge incorporation in the STO matrix decreased in the top region of the STO layer, rendering them insufficient to serve as effective agents for promoting overall crystallization in thicker films.

Fig. 9 (a) and (b) GIXRD spectra (the incident angle of X-ray was 2°), and (c) and (d) t_ox–t_phy plots for pristine STO ((a) and (c)), and b-Ge-STO ((b) and (d)) films of various thicknesses annealed at 575 °C.

Effect of GeO₂ thickness on crystalline and electrical properties of STO

GIXRD analysis with various incident X-ray beam angles was conducted to examine the effect of the interfacial GeO₂ layer and diffused Ge ions on the crystalline behavior of STO films. Fig. 10 shows the peak position and area of the STO (110) plane for 12 nm-thick STO films deposited on a bare Ru and Ru covered with 1–4 cycles of GeO₂. All samples were annealed at 650 °C to assess the crystalline properties of fully crystallized STO, eliminating any influence from residual amorphous states. Fig. S5 (ESI†) shows the GIXRD spectra and the deconvolution of the STO (110) plane peaks obtained through Gaussian fitting. Since the depth of interest in GIXRD measurements can be controlled by adjusting the incident beam angle (ω), angles of 0.2 and 2° were used to analyze the top region of the STO layer excluding the influence of the GeO₂ insertion layer and entire region of the STO layer including the bottom interface region, respectively. The calculated X-ray penetration depths in STO thin films based on ω were 2.9 nm at 0.2° and 115.5 nm at 2° (details for the calculations are provided in ESI,† Fig. S5).

Fig. 10 Peak position and peak area of STO (110) plane peak extracted from Gaussian fitting of GIXRD spectra for STO on bare Ru and Ru covered with 1–4 cycles of Ge–O (Fig. S5, ESI†). The incident angles of the X-ray were (a) 0.2° and (b) 2°. All specimens were annealed at 650 °C.

The peak positions were used to determine the evolution of film stress. Stress within thin films primarily originates from (1) intrinsic stress due to grain growth and densification, (2) extrinsic thermal stress, and (3) extrinsic stress due to diffused Ge. During the crystallization of amorphous STO, densification of the film to form a crystal structure occurred due to the lower density of the amorphous STO. It resulted in film shrinkage in both out-of-plane (thickness) and in-plane directions. The similar thickness reductions (∼22.3% for pristine STO and ∼21.4% for b-Ge-STO, as shown in Fig. 4) suggest comparable shrinkage along the out-of-plane direction. In contrast, the more considerable density increase in pristine STO compared to b-Ge-STO (22.2% versus 14.5%) indicated a more pronounced in-plane shrinkage, leading to higher residual in-plane tensile stress than b-Ge-STO. The stress in the surface region of STO was first analyzed using GIXRD with a shallow penetration depth of a few nanometers (ω = 0.2°) to exclude the effect of the bottom GeO₂ layer, as the diffused Ge ions primarily concentrated in the bottom region of STO. The change in in-plane stress can be estimated from the peak position of the (110) plane, at ∼32.4–32.7°, and the corresponding d-spacing. As the normal vector of the diffracting (110) plane was tilted by ∼16.3° from the out-of-plane direction, the in-plane tensile stress would primarily decrease the d-spacing (increase the 2θ value of the peak). In contrast, the in-plane compressive stress would increase it (decrease the 2θ value). Fig. 10(a) shows the STO (110) plane peak position and the area of STO films as a function of Ge–O cycles. The upper region of the pristine STO film underwent more severe in-plane shrinkage during densification, resulting in a higher residual tensile stress (a higher 2θ ∼32.66°) compared to b-Ge-STO. In contrast, the 2θ value gradually decreased with increasing Ge–O cycles in the case of b-Ge-STO film, indicating relaxed residual tensile stress at higher Ge–O cycles. This stress relaxation is attributed to the higher density of b-Ge-STO in its as-deposited state, which prevented severe shrinkage of the film during annealing. The residual stress within STO films, calculated based on Poisson's ratio and the Young's modulus of bulk STO, was 991 MPa for pristine STO and 802–900 MPa for b-Ge-STO (details for the calculations are provided in the ESI,† Fig. S5). More than three cycles of Ge–O efficiently suppressed the unwanted exaggerated reaction, as shown in Fig. 2(a), and relaxed tensile stress. These results suggest that a significant portion of in-plane tensile stress can be relieved by the higher density of b-Ge-STO in the as-deposited state, which inhibits the severe film shrinkage of STO during annealing. Furthermore, as the high nuclei density of b-Ge-STO led to smaller grain sizes, which shortened the average coalescence distance, the coalescence-induced tensile stress would be relaxed due to earlier grain impingement.

Extrinsic thermal stress was induced during the post-deposition annealing process due to the mismatch in thermal expansion coefficients (TEC) between the STO layer (9.4 × 10⁻⁶ K⁻¹) and the underlying Si substrate (2.6 × 10⁻⁶ K⁻¹).^52,53 The larger TEC of STO resulted in substantial tensile stress after annealing, potentially causing numerous micro-cracks and voids within the STO layer. Given the extremely thin thickness of interfacial GeO₂, the thermal stress arising from TEC mismatch would exert a similar effect on both pristine STO and b-Ge-STO layers. Thus, the relaxation of tensile stress was predominantly attributed to lower intrinsic stress induced by grain growth and densification. As the residual tensile stress has been reported to retard the crystallization of thin films, the lower crystallization temperature of b-Ge-STO was also attributed to the relaxation of such tensile stress.⁵⁴

The extrinsic stress caused by the diffused Ge ions should be analyzed by GIXRD using a different incident angle with a deeper penetration depth, as the diffused Ge ions are mainly concentrated in the bottom region of STO. Fig. 10(b) shows GIXRD spectra measured with an incident angle of 2°. The observed 2θ value of pristine STO reflected the combined effects of the overall intrinsic and extrinsic stress mentioned above. In the case of b-Ge-STO (1–4 cycles of Ge–O), the incorporation of Ge ions into the STO lattice was found to further influence the film stress. The diffused Ge ions would prefer to substitute for Ti sites (Ge_Ti) due to an identical +4 oxidation state and a slight ionic size difference of 10.1% (Ti⁴⁺ 74.5 pm, Ge⁴⁺ 67 pm, Sr²⁺ 132 pm).⁵⁵ These Ge_Ti defects would induce lattice compression (an increase in the 2θ value in the GIXRD result), particularly at the interface between the STO and GeO₂ layers, due to the smaller ionic radius of Ge compared to Ti. The 2θ value increased from 32.43° for pristine STO to 32.48° after four Ge–O insertion cycles, indicating a higher Ge_Ti concentration. The crystallinity of the STO film, as inferred from the peak area (bar graphs), was not significantly degraded by a small portion (1–2 cycles) of Ge_Ti in STO. However, a high concentration (3–4 cycles) of Ge_Ti intensified lattice distortion, resulting in a decrease in the GIXRD peak area. Consequently, it was confirmed that 1–2 cycles of Ge–O effectively released residual in-plane tensile stress with minimal lattice distortion by Ge_Ti, thereby representing the optimal insertion cycle.

Fig. 11(a) shows the t_ox–t_phy plots of MIM capacitors with Pt/RuO₂/STO/n cycles of Ge–O/Ru structure (n = 0–4). All samples were annealed at 650 °C to obtain a high-k phase within the entire STO layers. The almost identical tⁱ_ox regardless of Ge–O cycles indicated that the impact of the interfacial layer caused by the GeO₂ layer was nearly identical while the amount of diffused Ge ions increased with Ge–O cycles. However, with the Ge–O cycles over three, the high incorporation of Ge_Ti degraded the crystallinity of STO (as shown in Fig. 10), decreasing the dielectric constant of STO. In contrast, with one or two cycles of Ge–O, the dielectric constant was even larger than that of pristine STO due to the suppression of carbon residue, uniform cation stoichiometry, and conformal morphology without severe lattice distortion. The comparison of the J–t_ox curves in Fig. 11(b) demonstrates that even a single cycle of GeO₂ resulted in improved J–t_ox performance compared to pristine STO. Although pristine STO exhibited comparable t_ox with b-Ge-STO, its minimum t_ox that meet the DRAM leakage criterion was ∼0.7 nm, which was attributed to the formation of numerous leakage paths due to cracks induced by significant thermal stress during ex situ annealing without seed layer (as shown in Fig. 6(c)). The insertion of GeO₂ effectively suppressed defect formation from the decreased thermal stress and higher density in the as-deposited state, resulting in superior leakage properties. Moreover, a two-cycle (0.6 nm-thick) GeO₂ barrier provided sufficient oxygen-blocking effect, yielding the most desirable J–t_ox performance (t_ox ∼ 0.40 nm with 12 nm-thick STO film). This is particularly remarkable compared to STO films employing TiO₂ and Al₂O₃ barrier layers, which exhibited significantly larger minimum t_ox values of 0.69 nm and 1.14 nm, respectively.²⁷ Consequently, the GeO₂ interfacial barrier strategy presented in this study uniquely achieved optimal dielectric scalability and superior electrical performance simultaneously, demonstrating its potential for DRAM capacitor applications.

Fig. 11 (a) t_ox–t_phy plots and (b) J–t_ox curves for 650 °C-annealed STO on Ru covered with 0–4 cycles of GeO₂.

Electrical properties of low-temperature annealed b-Ge-STO with optimization

Fig. 12(a) shows leakage current density–voltage (J–V) characteristics, exhibiting a significant suppression in leakage current for b-Ge-STO. The 12 nm-thick b-Ge-STO exhibited remarkably low leakage current despite its fine grain size. In contrast, pristine STO suffered from higher leakage current (∼10⁻⁴ A cm⁻² at 0.8 V) through both grain and phase boundaries. Although the 20 nm-thick films were partially crystallized and showed similar t_ox (∼1 nm) regardless of GeO₂ insertion (as shown in Fig. 9(d)), b-Ge-STO still demonstrated improved leakage current performance. It was attributed to lower carbon impurities, almost no cracks, and minimized interfacial roughness between the b-Ge-STO and electrode layers. Since protrusions and crack edges on a rough interface locally concentrate the electric field and facilitate the current flow, the improved surface morphology of b-Ge-STO would greatly help leakage current suppression.^56–60 Also, the J–V curves in negative and positive bias regions were symmetric, despite the asymmetric stacking structure of the capacitors. It indicated that the GeO₂ insertion layer influenced both the bulk and interface regions in the band diagram of the capacitor. Further detailed investigations will be conducted in future studies to clarify the influence of the GeO₂ insertion layer on the conduction mechanism of STO film.

Fig. 12 (a) J–V curves of 12 and 20 nm-thick pristine STO and b-Ge-STO annealed at 575 °C. (b) J–t_ox curves of pristine STO and b-Ge-STO films annealed at 575 °C.

Fig. 12(b) summarizes the electrical properties of pristine STO and b-Ge-STO annealed at 575 °C through J–t_ox curves. The shift of the J–t_ox curve to a lower t_ox range in b-Ge-STO compared to pristine STO was attributed to both improved leakage current density and enhanced dielectric constants. Consequently, the achieved minimum t_ox was 0.41 nm (∼5 × 10⁻⁸ A cm⁻² at 0.8 V) while meeting the DRAM capacitor leakage current criterion (∼10⁻⁷ A cm⁻² at 0.8 V) with a physical film thickness of 11.8 nm even without adopting a two-step process. Such performance achieved at a lower annealing temperature (575 °C) was highly comparable to the minimum t_ox of 0.40 nm with 12.2 nm-thick b-Ge-STO when annealed at 650 °C. This highlights that the 0.6 nm-thick GeO₂ barrier not only improved dielectric properties but also lessened the required thermal budget and fabrication complexity without compromising electrical performance. Such low-temperature compatibility further emphasizes the advantages of employing a GeO₂ insertion layer for scalable production of defect-free and high-performance DRAM capacitors.

Experimental section/methods

Film preparation

STO and GeO₂ films were deposited at a substrate temperature of 350 °C using a traveling-wave-type ALD reactor (CN-1, Atomic Classic). Sr(ⁱPr₃Cp)₂, Ti(Me₅Cp)(OMe)₃ (synthesized by Air Liquide), and Ge(NMePh)(NMe₂)₃ (synthesized by EGTM) precursors were used as sources of Sr, Ti, and Ge, respectively (Pr = propyl, Ph = phenyl groups). The canisters of Sr, Ti and Ge precursor were heated to 110, 70, and 100 °C, respectively. O₃ was used as an oxygen source for TiO₂ and GeO₂, and H₂O was used for SrO, following the optimization process reported elsewhere.^25,38 The O₃ concentration was 220 g cm⁻³, and H₂O was cooled to 5 °C to maintain the appropriate vapor pressure. Ar gas was used as the carrier and purging gas for the precursors and reactants, with a flow rate of 200 standard cubic centimeters per minute (sccm). A super-cycle of the STO ALD process consisted of four TiO₂ and one SrO sub-cycle to achieve a stoichiometric cation composition, as shown in Fig. S1b (ESI†). The ALD sequence for SrO and TiO₂ involved precursor injection (3 s) – precursor purge (10 s) – reactant injection (2 s) – reactant purge (10 s). The ALD sequence of GeO₂ involved precursor injection (5 s) – precursor purge (10 s) – reactant injection (2 s) – reactant purge (10 s). Sputtered Ru (40 nm)/Ta (3 nm)/SiO₂ (100 nm)/Si wafer and bare Si wafer were used as substrates for GeO₂ and STO films. Rapid thermal annealing (RTA) was conducted under an N₂ atmosphere for 2 min at temperatures ranging from 400 to 650 °C after STO deposition to examine the crystallization behavior.

Film analysis

The layer densities (amount of deposited atoms per unit area) of Sr, Ti and Ge were measured via XRF (Thermo-Fisher, ARL Quant’X). The physical thicknesses and film densities were estimated using an ellipsometer (SE; M-2000, J.A. Woollam) and X-ray reflectivity (XRR) analysis (PANalytical, X’pert Pro). The crystallinity of STO films was analyzed using GIXRD (PANalytical, X’pert Pro) with Cu Kα radiation at incident X-ray beam angles of 0.2° or 2°. The chemical bonding properties of films were analyzed by XPS (Kratos, AXIS SUPRA). The surface morphology of the films was analyzed by FE-SEM (Carl Zeiss, SUPRA 55VP) and AFM (Park Systems, NX10). Grain size measurement from SEM images was conducted using ImageJ software (National Institutes of Health). The depth profiles of film components were analyzed via ToF-SIMS (ON-TOF, SIMS-5) and AES (Ulvac-PHI, PHI 700Xi). Spherical aberration corrected-TEM (JEOL, JEM-ARM200F) was used to investigate film thickness and crystal structure. For electrical characterization, 20 nm-thick RuO₂ and 50 nm-thick Pt films were sequentially deposited via sputtering on top of STO films to form the top electrodes of the metal–insulator–metal capacitor structure, using a shadow mask with a 300 μm diameter hole pattern. Leakage current and capacitance were measured using an HP 4140 picoammeter and an HP 4194A impedance analyzer at 10 kHz.

Conclusions

This study demonstrates the significant advantages of incorporating a thin GeO₂ bottom insertion layer for high-k STO films. The 0.6 nm-thick GeO₂ layer effectively passivated the Ru substrate, suppressing oxygen diffusion between the STO and Ru(O_x) layers, thereby mitigating the initial overgrowth of the STO film while maintaining the self-saturated growth of each layer. The improved film quality, including uniform Sr/Ti stoichiometry, high film density and low carbon impurities, and facilitated Ge diffusion into the STO film enabled effective crystallization at lower temperatures (575 °C). As a result, dielectric properties comparable to those obtained at higher temperatures (650 °C), with a dielectric constant of ∼210, were achieved. The insertion of GeO₂ also minimized surface morphology roughening and crack formation in the STO films, thereby improving leakage current properties in the capacitor. Furthermore, it simplified the STO deposition process by eliminating the need for a pre-crystallized seed layer, previously considered essential for crystallizing the main layer without degrading the leakage current performance. These findings highlight the potential of the GeO₂ insertion layer as an effective oxygen barrier for the scalable and low-temperature fabrication of high-performance DRAM capacitors.

Author contributions

H. Paik designed and performed the experiments and wrote the manuscript. J. Lim, H. Seo, T. K. Kim, J. H. Shin, H. Song, and D. G. Kim contributed to discussion. W. Lee reviewed the manuscript. D. S. Kwon assisted the data interpretations and edited manuscript. C. S. Hwang supervised the whole research and manuscript preparation.

Conflicts of interest

There are no conflicts to declare.

Data availability

All the relevant data are available from the corresponding authors upon request.

Acknowledgements

This work was supported by the National Research Foundation of Korea (grant no. 2020R1A3B2079882).

Notes and references

1. W. Weinreich, T. Tauchnitz, P. Polakowski, M. Drescher, S. Riedel, J. Sundqvist, K. Seidel, M. Shirazi, S. D. Elliott and S. Ohsiek, J. Vac. Sci. Technol., A, 2013, 31, 01A123.
2. J.-H. Kim, V. A. Ignatova, J. Heitmann and L. Oberbeck, J. Phys. D: Appl. Phys., 2008, 41, 172005.
3. H. Kashyap, P.-C. Lee, K. Chae, M. Passlack, A. K. Yadav, K. Wong, S. Nemani, E. Yieh, J. Spiegelman, K. Cho, A. I. Khan and A. C. Kummel, J. Vac. Sci. Technol., A, 2025, 43, 022401.
4. Y. N. Thakare, U. A. Wankhade and S. N. Kale, Int. J. Electron., 2025, 1–20,  DOI:10.1080/00207217.2025.2450730.
5. M. Hiratani, T. Hamada, S. Iijima, Y. Ohji, I. Asano, N. Nakanishi and S. Kimura, A heteroepitaxial MIM-Ta₂O₅/capacitor with enhanced dielectric constant for DRAMs of G-bit generation and beyond, in 2001 Symposium on VLSI Technology. Digest of Technical Papers (IEEE Cat. No. 01 CH37184), IEEE, 2001, pp. 41–42.
6. S. K. Kim, W.-D. Kim, K.-M. Kim, C. S. Hwang and J. Jeong, Appl. Phys. Lett., 2004, 85, 4112–4114.
7. R. Waser, Nanoelectronics and information technology: advanced electronic materials and novel devices, John Wiley & Sons, 2012.
8. B. Kim, T. Kang, S. Song, C. Jung, J. Lee, S. Cheon and H. Jeon, Vacuum, 2022, 199, 110957.
9. A. Kosola, M. Putkonen, L.-S. Johansson and L. Niinistö, Appl. Surf. Sci., 2003, 211, 102–112.
10. S. W. Lee, O. S. Kwon and C. S. Hwang, Microelectron. Eng., 2005, 80, 158–161.
11. G. Giusi, M. Aoulaiche, J. Swerts, M. Popovici, A. Redolfi, E. Simoen and M. Jurczak, IEEE Electron Device Lett., 2014, 35, 942–944.
12. M. Pawlak, B. Kaczer, M.-S. Kim, M. Popovici, K. Tomida, J. Swerts, K. Opsomer, W. Polspoel, P. Favia and C. Vrancken, Appl. Phys. Lett., 2010, 97, 162906.
13. S. Saha, T. Sinha and A. Mookerjee, J. Phys.: Condens. Matter, 2000, 12, 3325.
14. K. Van Benthem, C. Elsässer and R. French, J. Appl. Phys., 2001, 90, 6156–6164.
15. A. Emeline, G. V. Kataeva, A. S. Litke, A. V. Rudakova, V. K. Ryabchuk and N. Serpone, Langmuir, 1998, 14, 5011–5022.
16. J. C. Garcia, L. M. R. Scolfaro, A. T. Lino, V. Freire, G. Farias, C. Silva, H. Alves, S. Rodrigues and E. Da Silva, J. Appl. Phys., 2006, 100, 104103.
17. D. Z. Austin, M. A. Jenkins, D. Allman, S. Hose, D. Price, C. L. Dezelah and J. F. Conley, Jr., Chem. Mater., 2017, 29, 1107–1115.
18. O. S. Kwon, S. W. Lee, J. H. Han and C. S. Hwang, J. Electrochem. Soc., 2007, 154, G127.
19. S. W. Lee, J. H. Han, O. S. Kwon and C. S. Hwang, J. Electrochem. Soc., 2008, 155, G253.
20. S. W. Lee, J. H. Han, S. Han, W. Lee, J. H. Jang, M. Seo, S. K. Kim, C. Dussarrat, J. Gatineau, Y.-S. Min and C. S. Hwang, Chem. Mater., 2011, 23, 2227–2236.
21. I. Barin and G. Platzki, Thermochemical data of pure substances, VCh, Weinheim, 1989.
22. N. Menou, M. Popovici, S. Clima, K. Opsomer, W. Polspoel, B. Kaczer, G. Rampelberg, K. Tomida, M. A. Pawlak, C. Detavernier, D. Pierreux, J. Swerts, J. W. Maes, D. Manger, M. Badylevich, V. Afanasiev, T. Conard, P. Favia, H. Bender, B. Brijs, W. Vandervorst, S. Van Elshocht, G. Pourtois, D. J. Wouters, S. Biesemans and J. A. Kittl, J. Appl. Phys., 2009, 106, 094101.
23. S. W. Lee, O. S. Kwon, J. H. Han and C. S. Hwang, Appl. Phys. Lett., 2008, 92, 222903.
24. W. Lee, W. Jeon, C. H. An, M. J. Chung, H. J. Kim, T. Eom, S. M. George, B. K. Park, J. H. Han, C. G. Kim, T.-M. Chung, S. W. Lee and C. S. Hwang, Chem. Mater., 2015, 27, 3881–3891.
25. W. Lee, J. H. Han, W. Jeon, Y. W. Yoo, S. W. Lee, S. K. Kim, C.-H. Ko, C. Lansalot-Matras and C. S. Hwang, Chem. Mater., 2013, 25, 953–961.
26. S. H. Kim, W. Lee, C. H. An, D. S. Kwon, D. G. Kim, S. H. Cha, S. T. Cho and C. S. Hwang, ACS Appl. Mater. Interfaces, 2018, 10, 41544–41551.
27. W. Lee, J. H. Han, S. W. Lee, S. Han, W. J. Jeon and C. S. Hwang, J. Mater. Chem., 2012, 22, 15037–15044.
28. H. Birey, J. Appl. Phys., 1977, 48, 5209–5212.
29. W. G. Lee, S. I. Woo, J. C. Kim, S. H. Choi and K. H. Oh, Thin Solid Films, 1994, 237, 105–111.
30. H. Fukuda, S. Maeda, K. Salam and S. Nomura, Jpn. J. Appl. Phys., 2002, 41, 6912.
31. T. Tokuda and W. D. Kingery, J. Appl. Phys., 1963, 34, 2104–2105.
32. M. Radecka and M. Rekas, J. Phys. Chem. Solids, 1995, 56, 1031–1037.
33. A. H. Heuer, J. Eur. Ceram. Soc., 2008, 28, 1495–1507.
34. A. Heuer, J. Eur. Ceram. Soc., 2008, 28, 1495–1507.
35. Q. Xie, S. Deng, M. Schaekers, H. C. Lin, M. Caymax, A. Delabie, X.-P. Qu, Y.-L. Jiang, D. Deduytsche and C. Detavernier, Semicond. Sci. Technol., 2012, 27, 074012.
36. H.-W. Son, D. Berthebaud, K. Yubuta, A. Yoshikawa, T. Shishido, K. Suzuta and T. Mori, Sci. Rep., 2020, 10, 8914.
37. D. S. Kwon, C. H. An, S. H. Kim, D. G. Kim, J. Lim, W. Jeon and C. S. Hwang, J. Mater. Chem. C, 2020, 8, 6993–7004.
38. S. H. Kim, W. Lee, C. H. An, D. G. Kim, D. S. Kwon, S. T. Cho, S. H. Cha, J. I. Lim and C. S. Hwang, Phys. Status Solidi RRL, 2019, 13, 1800557.
39. S. N. A. Murad, P. T. Baine, D. W. McNeill, S. J. N. Mitchell, B. M. Armstrong, M. Modreanu, G. Hughes and R. K. Chellappan, Solid-State Electron., 2012, 78, 136–140.
40. Y. Kamata, Mater. Today, 2008, 11, 30–38.
41. M. Choi, J. L. Lyons, A. Janotti and C. G. Van de Walle, Appl. Phys. Lett., 2013, 102, 142902.
42. N. Li and K. L. Yao, AIP Adv., 2012, 2, 032135.
43. C. Bi, Q. Wang, Y. Shao, Y. Yuan, Z. Xiao and J. Huang, Nat. Commun., 2015, 6, 7747.
44. I. Bretos, R. Jiménez, J. Ricote and M. L. Calzada, Chem. Soc. Rev., 2018, 47, 291–308.
45. K.-H. Xue, J. Celinska and C. A. Paz de Araujo, Appl. Phys. Lett., 2009, 95, 052908.
46. H. Suzuki, T. Koizumi, Y. Kondo and S. Kaneko, J. Eur. Ceram. Soc., 1999, 19, 1397–1401.
47. M.-G. Kang, S.-M. Oh, K.-H. Cho, Y.-H. Do, D.-S. Paik, B.-H. Cho, C.-Y. Kang, S. Nahm and S.-J. Yoon, J. Electrochem. Soc., 2011, 159, D9.
48. A. D. Polli, F. F. Lange and C. G. Levi, J. Am. Ceram. Soc., 2000, 83, 873–881.
49. G. S. Henderson, D. R. Neuville, B. Cochain and L. Cormier, J. Non-Cryst. Solids, 2009, 355, 468–474.
50. B. L. Phoon, C. W. Lai, J. C. Juan, P. L. Show and W. H. Chen, Int. J. Energy Res., 2019, 43, 5151–5174.
51. H. Kawano, K. Morii and Y. Nakayama, J. Appl. Phys., 1993, 73, 5141–5146.
52. M. Okaji, Int. J. Thermophys., 1988, 9, 1101–1109.
53. N. Menou, M. Popovici, K. Opsomer, B. Kaczer, M. A. Pawlak, C. Adelmann, A. Franquet, P. Favia, H. Bender, C. Detavernier, S. Van Elshocht, D. J. Wouters, S. Biesemans and J. A. Kittl, Jpn. J. Appl. Phys., 2010, 49, 094101.
54. H. Miura and N. Okamoto, J. Appl. Phys., 1994, 75, 4747–4749.
55. L. Pu, J. Wang, G. Tang and J. Zhang, J. Appl. Phys., 2015, 117, 105307.
56. J. Yang, Y.-S. Kim, J.-H. Jeon and K. Park, J. Inf. Disp., 2021, 22, 1–10.
57. G. Jegert, A. Kersch, W. Weinreich and P. Lugli, J. Appl. Phys., 2011, 109, 014504.
58. D. Suh and J.-Y. Kang, J. Vac. Sci. Technol., B:Nanotechnol. Microelectron.:Mater., Process., Meas., Phenom., 2002, 20, 717–720.
59. A. W. Groenland, R. A. Wolters, A. Y. Kovalgin and J. Schmitz, J. Nanosci. Nanotechnol., 2011, 11(9), 8368–8373.
60. E. C. Ko, W. Kang and J. H. Han, J. Alloys Compd., 2022, 895, 162579.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5mh00611b

---