Suppression of interfacial layers in ZrO₂/TiN capacitors by atomic layer deposition using ligand-engineered Zr precursors for scalable DRAM

Abstract

As dynamic random-access memory continues to scale down, the feasible physical thickness of the capacitor dielectric layer continuously decreases, thus, controlling the low-k interfacial layer formed at the ZrO₂ dielectric/TiN electrode interface is becoming crucial. The interfacial layer reduces the capacitance density and increases the leakage current density, and both of them contribute to the degradation of the overall properties of the capacitor. In this study, two precursors were compared: the commonly used Cp-based Zr precursor, Cp–Zr(NMe₂)₃ [Cp–Zr] and the novel MePrCp–Zr(NMe₂)₃ [MePrCp–Zr] precursor, with the two terminal hydrogens of the Cp ligand substituted with Me and Pr groups. MePrCp–Zr was confirmed to suppress the formation of low-k interfacial layers such as TiO_x or TiO_xN_y at the initial ZrO₂ growth stage, owing to its higher reactivity than Cp–Zr. Furthermore, analysis of oxidation behavior using TiN and Ru bottom electrodes clearly revealed that the application of MePrCp–Zr led to improved interfacial sharpness compared to Cp–Zr. Electrical properties also confirmed enhanced interfacial properties, indicating that the equivalent oxide thickness decreased by 0.38 nm with the MePrCp–Zr precursor compared to Cp–Zr. This ligand-engineering strategy provides a scalable approach to achieving ultrathin high-k dielectrics with stable interfaces, enabling reliable capacitor integration for next-generation DRAM and advanced logic technologies.

---

New concepts

This work introduces a new approach to interface engineering in ZrO₂-based DRAM capacitors by using MePrCp–Zr, a chemically modified Zr precursor with improved properties, for reducing electrode interface degradation during atomic layer deposition. Unlike the conventional Cp–Zr precursor that requires prolonged O₃ exposure and induces oxidation of TiN electrodes, the MePrCp–Zr precursor exhibits higher reactivity, offering two advantages: first, the shorter O₃ injection time suppresses TiN surface oxidation; second, the enhanced reactivity directly suppresses the TiO_x or TiO_xN_y interfacial layers during the initial stages of ZrO₂ growth. This optimized reactivity of MePrCp–Zr minimizes interfacial layers and enhances the electrical properties of ZrO₂/TiN capacitors by improving interfacial sharpness and reducing leakage current, compared to Cp–Zr. These improvements are driven by precursor ligand modification without additional processes and enable integration into advanced DRAM structures. Furthermore, ZrO₂ films deposited with MePrCp–Zr exhibit excellent conformality on high aspect ratio structures and demonstrate suitability with both TiN and Ru electrodes, highlighting their potential for next-generation capacitor applications. By optimizing the precursor ligand structure at the molecular level, this work provides a new pathway for precise control of interfacial reactions in high-k dielectric systems, which is critical for overcoming scaling limitations in memory devices.

Introduction

Modern computing systems have evolved from simple calculators into versatile platforms that require rapid access to large volumes of data. As a result, overall system performance strongly depends on memory technologies. In particular, dynamic random-access memory (DRAM) serves as a crucial high speed memory component that supports efficient processor operation. To maintain reliable performance, DRAM capacitors should offer sufficient capacitance and low leakage current for stable data retention.^1–4 As DRAM devices continue to scale down, research has been conducted on next-generation capacitor dielectrics such as TiO₂ and SrTiO₃ to achieve sufficient capacitance. However, issues including nonideal growth behavior, impurity incorporation, and crack formation continue to restrict their effective application as dielectric materials in complex 3D-structured capacitors.^5–10 In the meantime, the physical thickness of the dielectric layer in current DRAM capacitors has reduced to below 5 nm, and as a result, the impact of the low-k layer formation at the dielectric/electrode interface has become crucial for overall dielectric properties.^11,12 In mass production of DRAM capacitors, oxide dielectric films such as ZrO₂ are deposited on TiN electrodes using atomic layer deposition (ALD), with O₃ employed as the oxygen source. The high oxidizing power of O₃, along with the high reactivity of the underlying TiN electrode, results in the formation of undesirable interfacial layers such as TiO_x or TiO_xN_y. The low-k layer not only degrades the effective capacitance density, but the interfacial layers with high defect densities also cause charge trapping, which results in deterioration of the insulating properties of the capacitors.^12–17

In previous studies, numerous approaches have been attempted to minimize the detrimental interfacial layer. First, a two-step oxygen source process using H₂O and O₃ was developed for ZrO₂ ALD, with H₂O minimizing the initial formation of TiO_x or TiO_xN_y and subsequent O₃ exposure providing high crystallinity and low impurity content in the remaining film.¹⁸ Another strategy involved the modification of the top electrode stack from a single TiN electrode to a TiN/Ru structure, with Ru contacting the ZrO₂ layer. This strategy enabled top TiN to function as an oxygen scavenger, thereby reducing the TiO_x or TiO_xN_y interfacial layers formed at the ZrO₂/TiN bottom interface.¹⁹ In addition, Ar plasma treatment on the dielectric layer was proposed to decrease TiO_x or TiO_xN_y interfacial layers on the TiN surface by ion bombardment. This treatment successfully reduced the targeted interfacial layer while maintaining the oxygen content in the dielectric films, thereby preserving a low trap density.²⁰ These approaches either sacrifice the quality of ZrO₂ during the initial stages of growth or focus on the reduction of TiO_x or TiO_xN_y formed after the deposition of high-purity ZrO₂ films. Beyond that, fundamental suppression of in situ TiO_x or TiO_xN_y formation during high-quality ZrO₂ growth is essential for achieving a ZrO₂/TiN capacitor structure with superior interfacial and bulk properties. Lee et al. inserted buffer layers such as Ta₂O₅, ZnO and TiO₂ at the interface to block oxygen diffusion and inhibit interfacial layer formation.²¹ This approach results in changes in the energy diagram and dielectric thickness in precisely fabricated capacitor structures, necessitating additional optimization engineering.²² Thus, it is most desirable to form a high-quality dielectric film on the unoxidized TiN surface only by improving the dielectric film growth process itself, without inserting additional layers.

Previous studies have reported that replacing tetrakis(ethylmethylamino)zirconium with cyclopentadienyl (Cp)-based precursors enhances the thermal stability of the precursor, thereby enabling higher deposition temperatures and improving film crystallinity and electrical properties.^23–26 Despite these advantages, the removal of Cp ligands requires sufficient O₃ exposure, which results in unavoidable damage to the underlying TiN electrode. Consequently, optimized precursors with adequate reactivity must be employed to enable stable ZrO₂ film growth with less TiN damage. Since O₃ exposure during the dielectric growth is essential for the high crystallinity and low residue of the ZrO₂ thin film, the Zr precursor is required to be properly oxidized by ligand exchange while protecting the underlying layer. S. Park et al. proposed a method to enhance the reactivity of a Cp-based Zr precursor with O₃ during the ALD process by substituting the hydrogen atoms on the Cp ligands with deuterium. Deuterium substitution is an approach that tunes reactivity by inducing changes in molecular vibrational properties and weakens the bond dissociation energy between the central Zr atom and both the Cp and amine ligands. Although using deuterium is not feasible for the mass production of precursors, this suggests a potential direction for precursor ligand design.²⁷ Alternative approaches to modifying reactivity can also be considered, as altering the ligand structure may lead to more fundamental changes.

In this study, to suppress surface oxidation of the TiN electrode caused by the O₃ reactant required for ZrO₂ formation, the terminal hydrogen atoms of the conventional Cp–Zr(NMe₂)₃ [Cp–Zr] were substituted with Me and Pr groups, resulting in MePrCp–Zr(NMe₂)₃ [MePrCp–Zr]. Me and Pr denote methyl and propyl, respectively. The following differences by ligand modification in precursor reactivity were investigated in comparison with the Cp–Zr precursor. The oxidation states at the ZrO₂/TiN interface induced by using the two different precursors were carefully analyzed. The MePrCp–Zr precursor exhibited higher reactivity, thereby effectively reducing the formation of TiO_x or TiO_xN_y interfacial layers. Consequently, Pt/ZrO₂/TiN metal/insulator/metal (MIM) capacitors were fabricated, and their electrical properties were evaluated to investigate the improvement in device performance depending on the ALD precursor ligands.

Experimental section/methods

Film preparation

A traveling-wave-type reactor (ULTECH, SPACE-T) with a 6-in-wafer scale was used to deposit the ZrO₂ dielectric films at 300 °C, for the ALD process. Tris(dimethylamino)(cyclopentadienyl)zirconium (CpZr(NMe₂)₃) and tris(dimethylamino)(1-methyl-3-propylcyclopentadienyl)zirconium (MePrCpZr(NMe₂)₃, developed by UP Chemical Co.), were employed as Zr precursors. High-density O₃ (200 g m⁻³, 500 sccm) was generated using an ozone generator (OZONE TECH, CN-1) and supplied as the oxidant for Zr precursors, serving as the oxygen source for oxide film growth. A 100 nm TiN film was first deposited by DC magnetron sputtering on a 100 nm SiO₂/Si substrate for the bottom electrode. After the deposition of the ZrO₂ film, a 60 nm Pt film was deposited as the top electrode. The 60 nm Pt top electrode was patterned through a metal shadow mask (nominal 0.3 mm hole diameter, with the exact area remeasured using an optical microscope) to distinguish the cells from one another.

Film analysis

The thickness of ZrO₂ was measured using single-wave ellipsometry (Gaertner Scientific Corporation, L116D), and the deposition amount of ZrO₂ was monitored using X-ray fluorescence (XRF, Nayur, NDA-200). The Ru areal density was also measured using XRF to determine the etching rate of Ru. The step coverage of ZrO₂ on the trench substrate was confirmed using transmission electron microscopy (TEM, JEOL/CEOS, JEM-2100F) with an operating voltage of 200 keV. Chemical composition and binding structures were analyzed using X-ray photoelectron spectroscopy (XPS, Thermo Fisher Scientific, NEXSA G2) with an Al Kα (1486.6 eV) monochromatic source (sampling area diameter of 200 µm, step size of 0.05 eV). All binding energies were referenced to the standard C 1s peak associated with the C–C bond (284.5 eV) for spectrum calibration. The atomic composition of the films and impurity levels were analyzed using Auger electron spectroscopy (AES, ULVAC-PHI, PHI-710) depth profiling. The AES etch rate was set at 30 Å s⁻¹, with SiO₂ as the reference material. To prevent air-induced oxidation of ZrO₂ for AES analysis, a 60 nm Au top electrode was deposited via thermal evaporation, and a 4 nm Al₂O₃ layer was deposited using trimethylaluminum (Al(CH₃)₃) and H₂O. Both layers served as protective capping layers. The oxygen distribution within the ZrO₂/TiN stack was analyzed using time-of-flight secondary ion mass spectrometry (ToF-SIMS, IONTOF GmbH, M6) to examine the oxygen contents at the ZrO₂/TiN interface. The morphology and microstructure of the ZrO₂ films were analyzed by field emission scanning electron microscopy (SEM, HITACHI, SU8010). The surface morphology and root-mean-square (RMS) roughness of the ZrO₂ films were examined using atomic force microscopy (AFM, Park Systems, NX10). The crystalline structures of the ZrO₂ films were determined using glancing angle X-ray diffraction (GAXRD, Rigaku, SmartLab). The sheet resistance of the oxidized TiN films was measured using a 4-point probe (AiT, CMT-100A). The electrical properties were investigated by measuring the capacitance density–voltage (C–V) and the current density–voltage (J–V) at room temperature using an HP 4284A LCR meter (at 10 kHz) and an HP 4145B semiconductor parameter analyzer, respectively. The top electrode was biased while the bottom electrode was grounded during the electrical tests.

Results and discussion

ALD behaviors of Cp–Zr and MePrCp–Zr

Fig. 1(a) and (b) present the chemical structure of Cp–Zr and MePrCp–Zr precursors, respectively. Using the commercially available Cp–Zr as a reference, the new MePrCp–Zr precursor was synthesized by substituting the terminal hydrogen atoms on the Cp ligand with Me and Pr groups to increase the electron density. Fig. 1(c) and (d) show the film thickness as a function of precursor and O₃ injection time, using Cp–Zr and MePrCp–Zr precursors, respectively. The basic process sequence for both precursors consisted of precursor injection–Ar purge–O₃ injection–Ar purge. ALD saturation behavior was evaluated by varying one of these four process steps. In terms of precursor injection time, the MePrCp–Zr precursor reached saturation 3 seconds earlier than the Cp–Zr precursor. Similarly, for the O₃ injection time, the MePrCp–Zr precursor achieved saturation 2 seconds earlier. The film growth by Cp–Zr and MePrCp–Zr precursors exhibited activation energies (E_a) of 0.106 eV and 0.074 eV, respectively, confirming that MePrCp–Zr exhibits higher surface reactivity and faster ALD kinetics, as demonstrated in Fig. S1(a) and (b), SI. These results indicate that MePrCp–Zr enables more sufficient chemisorption within a shorter injection time compared to Cp–Zr based on both saturation behavior and activation energy analysis. Both Cp–Zr and MePrCp–Zr showed an O₃ purge saturation time of 5 s, and longer purges caused no additional increase in growth per cycle or film thickness. The precursor and oxidant pulse/purge settings were confirmed to eliminate any parasitic CVD contribution. The corresponding saturation times for precursor purge and O₃ purge are provided in Fig. S1(c) and (d), SI. Furthermore, as shown in Fig. S1(e) and (f), SI, the GPC of both precursors increased with precursor and O₃ injection time and reached a constant value beyond the saturation point, confirming the self-limiting ALD growth behavior.

Fig. 1 The schematic chemical structure of (a) Cp–Zr and (b) MePrCp–Zr precursors. Variations in the ZrO₂ films' thickness as a function of (c) precursor injection time and (d) O₃ injection time, both on Si substrates. Schematic diagrams illustrating the difference in the initial ALD growth of ZrO₂ films on TiN substrates using (e) Cp–Zr and (f) MePrCp–Zr precursors. (g) Thickness uniformity of ZrO₂ films deposited by ALD within a 6-inch wafer for Cp–Zr and MePrCp–Zr precursors. (h) Comparison of the growth behavior of ZrO₂ films deposited with Cp–Zr (injection time: 6 s–10 s–5 s–5 s) and MePrCp–Zr precursors (injection time: 3 s–5 s–3 s–5 s) on Si and TiN substrates.

Fig. 1(e) and (f) illustrate the chemical reactions of Cp–Zr and MePrCp–Zr precursors, respectively, during the initial growth stages. In each precursor injection process, MePrCp–Zr is suggested to more effectively reduce the unstable bonding states of the TiO_x or TiO_xN_y layers formed on the TiN electrode compared to Cp–Zr. Moreover, during the O₃ injection process, MePrCp–Zr appears to promote faster formation of the ZrO₂ film compared to Cp–Zr. Thus, MePrCp–Zr is believed to enable ZrO₂ saturation to be achieved with a shorter precursor and O₃ injection time compared to Cp–Zr. Interfacial reaction behavior is consistent with the experimental results presented in Fig. 1(c) and (d). In a previous study, replacing the hydrogen atoms of the Cp ligand with substituents such as Me and Pr increased the electron density at the metal center, thereby reducing the bond dissociation energy between the Cp ligand and the metal atom.²⁸ This result suggests that a similar mechanism could occur in this study, involving the substitution of hydrogen atoms at the Cp ligand terminal positions by Me and Pr. This substitution is likely closely related to a reduction in the bond dissociation energy between the Cp ligand and the central Zr atom. The higher reactivity of MePrCp–Zr compared to Cp–Zr is consistent with the experimental results presented in Fig. 1(c) and (d). For this reason, MePrCp–Zr with a shorter O₃ injection time than Cp–Zr can suppress the oxidation of the TiN electrode while simultaneously reducing the unstable TiO_x or TiO_xN_y layers back into TiN. These combined effects of higher reducing power and shorter O₃ saturation time lead to less formation of the low-k TiO_x or TiO_xN_y interfacial layer at the ZrO₂/TiN interface, thereby improving electrical properties.

Fig. 1(g) shows the uniformity of Cp–Zr and MePrCp–Zr precursors deposited on a 6-inch diameter Si wafer. On a 4-inch wafer, Cp–Zr and MePrCp–Zr exhibited uniformities of 94% and 97%, respectively, and on a 6-inch wafer, the values indicated to 97% and 99%. The MePrCp–Zr precursor demonstrated superior uniformity equivalent to that of the Cp–Zr precursor, indicating its potential for stable application as an insulator in DRAM capacitor structures. Fig. 1(h) presents the variation in the film thickness for the number of ZrO₂ ALD cycles on both Si and TiN substrates. For all precursors, the film thickness increased linearly with the number of cycles, indicating stable film growth behavior in the ALD process. Both Cp–Zr and MePrCp–Zr precursors exhibit higher growth per cycle (GPC) values on TiN substrates compared to Si substrates. The increased GPC on TiN is attributed to stronger chemical interactions between the precursors and the TiN surface, as the electrode promotes interactions more effectively than the covalent-bonded Si substrate with inherently lower surface reactivity.²⁷ The Cp–Zr and MePrCp–Zr precursors both revealed short incubation delays, with the MePrCp–Zr precursor showing 10-cycle shorter incubation delay than the Cp–Zr precursor. This difference in the initial growth behavior is associated with surface reaction kinetics and precursor adsorption behavior, which could affect the interfacial properties of the film.²⁹ This trend is consistent with the results shown in Fig. 1(e) and (f), as the higher reactivity of MePrCp–Zr enables more efficient chemisorption onto the electrode.

Fig. 2 shows a cross-sectional TEM image of the ZrO₂ film deposited at 300 °C using the MePrCp–Zr precursor and O₃ oxygen source on a trench-patterned wafer. The opening size and depth of the trench structure are 150 nm and 2.7 µm (aspect ratio 18 : 1), respectively, and the ZrO₂ ALD process was performed for 24 cycles. As shown in Fig. 2(a), the ZrO₂ film was conformally formed throughout the entire trench structure. The enlarged images at the top opening, sidewalls, and bottom regions in Fig. 2(b)–(e) clearly confirm that the 1.6 nm ultra-thin ZrO₂ films were uniformly and continuously deposited with 100% step coverage even on the high aspect ratio trench structure without any film thinning or pin-hole formation.

Fig. 2 The cross-sectional TEM image of the ZrO₂ thin film deposited by ALD on the trench-patterned wafer using MePrCp–Zr and O₃ as the precursor and oxidant, respectively. The trench structure opening size and depth were 150 nm and 2.7 µm, respectively. (a) Overall view of the trench structure with deposited ZrO₂, and magnified images of (b) top and top sidewall regions, (c) middle sidewall region, (d) bottom sidewall region, and (e) bottom region.

Surface morphology of ZrO₂ (MePrCp–Zr) and crystallization properties of ZrO₂ (Cp–Zr and MePrCp–Zr)

Fig. 3(a) and (b) present the results of SEM and AFM analyses performed to evaluate the surface morphology and surface roughness of the 5 nm ZrO₂/TiN structure using the MePrCp–Zr precursor. As shown in Fig. 3(a), the ZrO₂ film exhibited uniformly formed dense grains without defects such as voids or cracks. In addition, the AFM analysis in Fig. 3(b) revealed a low root-mean-square roughness (R_q) of 0.7 nm, indicating the formation of a smooth surface. The low R_q value reflects smooth and continuous film growth, suggesting efficient nucleation behavior with negligible incubation delay, consistent with the growth behavior observed in Fig. 1(h).

Fig. 3 (a) SEM and (b) AFM images of 5 nm ZrO₂ films deposited using the MePrCp–Zr precursor on the TiN substrate. (c) GAXRD patterns of 15 nm ZrO₂ films deposited using Cp–Zr and MePrCp–Zr precursors on the TiN substrate.

Fig. 3(c) presents a comparison of the GAXRD patterns of 15 nm ZrO₂ films deposited on a TiN substrate using Cp–Zr and MePrCp–Zr precursors. Regardless of the precursor, diffraction peaks corresponding to (101)_t and (111)_c at 30.5°, (200)_c and (110)_t at 34.8°–35.4°, (112)_t and (220)_c at 50.7°–51.1°, (103)_t at 59.7°, and (211)_t and (311)_c at 60.6° were observed in all deposited ZrO₂ films. The subscripts t and c denote the tetragonal and cubic crystal structures, respectively. Both precursors were confirmed to achieve sufficient crystallization in the tetragonal or cubic ZrO₂ phases. At around 30.5°, 50°, and 60°, the diffraction angle difference between the tetragonal and cubic phases is too small to clearly distinguish the individual phases.^30–32 Nevertheless, peaks from multiple crystal planes were observed, confirming that both precursors form well-crystallized films.

Improved interfacial properties and electrode stability with MePrCp–Zr

Fig. 4(a) and (b) present the N 1s spectra of 2 nm Cp–ZrO₂/TiN and MePrCp–ZrO₂/TiN structures, respectively. To evaluate the oxidation state of the TiN surface, the N 1s peak was deconvoluted into TiN (N³⁻, Ti–N) and TiON (N³⁺, Ti–O–N) components. In both Cp–Zr and MePrCp–Zr samples, the Ti–N peak was observed at 395.9 eV and the Ti–O–N peak at 401.8 eV, which are consistent with the reported values (Ti–N: 395.9 eV, Ti–O–N: 401.8 eV), with no significant energy shift observed.³³ For the Cp–Zr sample, the relative ratios of TiN and TiON were 51.2% and 48.8%, respectively, whereas for the MePrCp–Zr sample, the ratios were 60.9% and 39.1%, indicating that a higher TiN fraction was maintained. This result suggests that the MePrCp–Zr precursor effectively suppresses the oxidation of the TiN electrode, thereby reducing the conversion of TiN to TiO_xN_y phases.

Fig. 4 The N 1s XPS spectra of 2 nm ZrO₂/TiN samples deposited using (a) Cp–Zr and (b) MePrCp–Zr precursors. The AES depth profiles of Au/4 nm Al₂O₃/8 nm ZrO₂/TiN samples with (c) Cp–Zr and (d) MePrCp–Zr precursors. (e) Variations in the resistivity of TiN electrode as a function of ZrO₂ ALD cycle, without Zr precursor injection, depending on O₃ injection time. (f) Variations in the areal density of Ru as a function of ZrO₂ ALD cycle, with and without Zr precursor injection, represented by solid and open circles, respectively. Schematic diagrams illustrating (g) the differences in TiO_x or TiO_xN_y layer formation on TiN substrates depending on O₃ injection time, and (h) Ru layer etching on Ru substrates depending on O₃ injection time with and without Zr precursor injection using Cp–Zr and MePrCp–Zr precursors.

Fig. 4(c) and (d) provide the AES depth profiles of the Au/4 nm, Al₂O₃/8 nm and ZrO₂/TiN structures fabricated using Cp–Zr and MePrCp–Zr precursors, respectively. Only the ZrO₂/TiN interface region was focused on to investigate the effect on the bottom electrodes. As shown in Fig. 4(c), the Cp–Zr precursor exhibits oxygen impurities exceeding 5%, which remain deep into the TiN bottom electrode region. In contrast, as indicated in Fig. 4(d), the MePrCp–Zr precursor represents an oxygen impurity concentration of approximately 1% in the bottom electrode region, resulting in a significant reduction compared to the Cp–Zr case. The SIMS depth profiles also confirmed this tendency in Fig. S2, SI. The MePrCp–Zr precursor exhibited a lower oxygen intensity in the TiN region compared to the Cp–Zr precursor. These results indicate that the MePrCp–Zr precursor led to a significant suppression of the low-k TiO_x or TiO_xN_y interfacial layer formed at the ZrO₂/TiN interface compared to the Cp–Zr precursor. This improvement is attributed to the high reactivity of MePrCp–Zr, which contributes to the reduction of oxygen within the interfacial layers. In addition, the MePrCp–Zr ALD process involves shorter O₃ injection times compared to the Cp–Zr process, which is believed to contribute to the decrease of interfacial layer formation. This reaction tendency corresponds to the differences in precursor reactivity described in Fig. 1. Additionally, as shown in Fig. S3, SI, the ZrO₂ films deposited using both precursors exhibited comparably low levels of carbon impurities, detected at approximately 1.5%.

Fig. 4(e) shows the variations in resistivity of the TiN electrode as a function of the ZrO₂ ALD cycle, without precursor injection and purge steps. Based on the saturation time of each precursor ALD process, the injection conditions were 0–0–5–5 s and 0–0–3–5 s for Cp–Zr and MePrCp–Zr, respectively, consisting practically of only the O₃ injection and purge steps. As the O₃ injection time increased, the resistivity of the TiN electrode exhibited a linear increase, which directly correlates with the degradation of electrode properties due to the formation of TiO_x or TiO_xN_y interfacial layers by oxidation. Since the Cp–Zr precursor requires a longer O₃ injection step time than the MePrCp–Zr precursor, the Cp–Zr precursor case exhibited more significant degradation of the TiN electrode compared to the MePrCp–Zr precursor case. Even when excluding the direct precursor reactivity effect itself on the bottom electrode, the MePrCp–Zr precursor was confirmed to more effectively suppress the formation of TiO_x or TiO_xN_y interfacial layers than the Cp–Zr precursor by a decreased O₃ dose during the film growth. This effect is expected to significantly limit leakage mechanisms such as tunneling, thermionic emission, and Poole–Frenkel emission in MIM structure capacitors.¹⁹

The electrode surface degradation mechanism induced by O₃ was investigated using Ru electrodes. Ru undergoes etching by the formation of volatile species such as RuO₄ under O₃ exposure, enabling precise quantification of electrode degradation.^10,34,35Fig. 4(f) presents variation in the areal density of Ru as a function of the ZrO₂ ALD cycle. The open circles indicate O₃ injection without precursor injection, carried out under the equivalent conditions as shown in Fig. 4(e). The solid circles represent the conventional ZrO₂ ALD cycle with sequential injection of the precursor and O₃ to evaluate the protection capability of each precursor from O₃-induced degradation of the Ru electrode. Under the open circle condition, the MePrCp–Zr precursor with a shorter O₃ injection time (3 s) than the Cp–Zr precursor (5 s) led to a smaller decrease in Ru areal density, indicating less degradation of the bottom electrode. This interpretation is consistent with the oxidation behavior of the TiN electrodes shown in Fig. 4(c) and (d). Under the solid circle condition, both processes showed a smaller decrease than the open circle condition, and the MePrCp–Zr ALD showed a smaller decrease in Ru areal density compared to Cp–Zr ALD. The MePrCp–Zr precursor exhibited lower degradation, which is attributed to the effect of shorter O₃ injection time and higher reactivity to oxygens, leading to better preservation of the Ru layer. After exceeding 200 cycles of precursor injection, both precursors appeared to fully cover the Ru surface, leading to saturation of the etching behavior. However, since the physical thickness of practically applied dielectric films is limited to about 5 nm in DRAM capacitors, degradation of electrode properties in the initial growth is considered to be significantly different depending on the precursor. As shown in Fig. S4, SI, this difference was also consistently observed during the initial growth on Ru substrates. The MePrCp–Zr precursor exhibited a faster ZrO_x nucleation rate than the Cp–Zr precursor, and, in correlation with the results presented in Fig. 4(f), both faster ZrO_x nucleation and genuine suppression of RuO₄ formation were identified. These two effects collectively contributed to the effective prevention of Ru surface degradation.

Fig. 4(g) and (h) provide schematic diagrams of oxidation and etching trends and O₃-induced degradation on TiN and Ru electrodes, respectively. Fig. 4(g) shows that the formation of TiO_x or TiO_xN_y interfacial layers is suppressed at an O₃ exposure time of 3 s compared to 5 s, as discussed in Fig. 4(e). Fig. 4(h) represents the difference in etching behavior on the Ru electrode. Similar to the trend shown in Fig. 4(g), more Ru is etched under the 5 s condition due to a longer O₃ exposure time relative to the 3 s condition, without precursor injection. (Upper panel) for the deposition of ZrO₂ on the Ru electrode, the MePrCp–Zr ALD exhibited superior substrate preservation through decreased O₃ injection and enhanced reactivity compared to the Cp–Zr ALD, as indicated in Fig. 1. (Lower panel) based on the results, the MePrCp–Zr precursor is believed to offer potential applicability to both the TiN electrode used in mass production and the Ru electrode gaining attention as a promising next-generation electrode material.^18,36

Improved electrical performance of ZrO₂/TiN capacitors enabled by superior interfacial properties by MePrCp–Zr

Fig. 5 presents the electrical properties of Pt/ZrO₂/TiN-based capacitors fabricated using Cp–Zr and MePrCp–Zr precursors, respectively. Fig. 5(a) shows the variation in t_ox with frequency for Pt/6 nm ZrO₂/TiN capacitors fabricated using Cp–Zr and MePrCp–Zr precursors. The frequency-dependent change in t_ox clearly reveals the difference in dielectric response between the two precursors. In the Cp–Zr precursor, t_ox increased sharply as the frequency approached 1 MHz, indicating significant polarization loss in the high-frequency region. In contrast, the MePrCp–Zr precursor exhibited minimal t_ox variation across the entire frequency range, suggesting that MePrCp–Zr forms a more stable interfacial structure with a lower concentration of oxygen vacancies compared with Cp–Zr.^37,38 Consequently, the MePrCp–Zr precursor shows markedly decreased frequency dependence and more stable electrical properties under high-frequency operating conditions than the Cp–Zr precursor. The voltage-dependent capacitance density for each precursor measured at frequencies ranging from 1 kHz to 1 MHz is presented in Fig. S5, SI.

Fig. 5 (a) Variations in equivalent oxide thickness (t_ox) of Pt/ZrO₂/TiN capacitors as a function of frequency using Cp–Zr and MePrCp–Zr precursors. (b) Standard deviation of leakage current density as a function of electric field for Pt/ZrO₂/TiN capacitors fabricated using Cp–Zr and MePrCp–Zr precursors. (c) The variations in equivalent oxide thickness (t_ox) of Pt/ZrO₂/TiN capacitors as a function of the physical thickness (t_phy) of the ZrO₂ dielectric layers deposited using Cp–Zr and MePrCp–Zr precursors, with t_ox intercepts of 0.50 nm and 0.12 nm, respectively. The inset shows the t_ox values measured at the top, center, bottom, left, and right positions across a 4-inch wafer, indicating a uniformity higher than 95% for both precursors. (d) The variations in leakage current density of Pt/ZrO₂/TiN capacitors at an applied voltage of 0.8 V as a function of t_ox, using Cp–Zr and MePrCp–Zr precursors. Electrical performances of previous reports are included for comparison. PMA refers to post-metallization annealing.

Fig. 5(b) shows the standard deviation of the leakage current for Pt/ZrO₂/TiN capacitors with 6 nm ZrO₂ layers deposited using Cp–Zr and MePrCp–Zr precursors. The Cp–Zr sample exhibited relatively large current variation over time, whereas the MePrCp–Zr sample showed smaller current variation, indicating higher leakage current stability in Fig. S6, SI. To quantitatively analyze this behavior, the standard deviation of the leakage current was calculated under each electric-field condition, as presented in Fig. 5(b). As the electric field increased, the MePrCp–Zr sample exhibited consistently smaller standard deviation values than the Cp–Zr sample, and showed lower variation than the Cp–Zr sample at all fields above 1.33 MV cm⁻¹ except at 1.00 MV cm⁻¹, confirming its superior current stability. The smaller deviation observed in the MePrCp–Zr sample is attributed to a lower defect density, such as TiO_x, TiO_xN_y, at the ZrO₂/TiN interface. Additionally, the capacitance–voltage endurance properties were evaluated up to 10³ cycles, and both samples maintain nearly constant t_ox values, confirming that the Cp–Zr and MePrCp–Zr film exhibited stable dielectric properties under repetitive electrical stress in Fig. S7, SI.

Fig. 5(c) shows the variation in equivalent oxide thickness (t_ox) as a function of the physical thickness (t_phy) of the ZrO₂ layer. As t_phy increased, t_ox increased linearly, indicating that the dielectric constant of ZrO₂ remained constant within the studied thickness range. The bulk dielectric constant of ZrO₂ was extracted from the inverse of the slope of the best-fitted linear plot in Fig. 5(c). The bulk dielectric constants of the ZrO₂ films were confirmed to be 31 for both the Cp–Zr and MePrCp–Zr precursors. For the MePrCp–Zr precursor, the reduction of the interfacial layer was attributed to two main factors. First, the shorter O₃ injection time compared to the Cp–Zr precursor suppressed the formation of TiO_x or TiO_xN_y. Second, the higher reactivity of the precursor enabled the reduction of the TiO_x or TiO_xN_y interfacial layer. As a result, the interfacial t_ox value (Y-axis intercept in Fig. 5(c)), representing the interfacial contribution excluding ZrO₂, was significantly reduced from 0.50 nm to 0.12 nm.

Fig. 5(d) presents the variation in leakage current density under an applied voltage of +0.8 V as a function of t_ox, in order to simultaneously evaluate the capacitance and leakage properties. Solid lines were added for each sample to guide the eye. The MePrCp–Zr precursor exhibited a t_ox value that was 0.38 nm lower than that of the Cp–Zr precursor, and additionally, showed significantly improved leakage current performance, with values approximately one order of magnitude lower than those of the Cp–Zr precursor. This enhancement is attributed to the reduction of TiO_x or TiO_xN_y interfacial layer formation at the ZrO₂/TiN interface by the MePrCp–Zr precursor. Since leakage current density in MIM capacitors employing ZrO₂ as a dielectric is primarily governed by trap-assisted tunneling and Poole–Frenkel emission mechanisms, these two mechanisms are closely related to defect states within the ZrO₂ layer and interfacial regions.^39–43 The Cp–Zr precursor showed characteristics consistent with previously reported results, while the MePrCp–Zr precursor exhibited excellent electrical properties even without additional crystallization annealing or interfacial modification, except in cases involving phase transition such as HZO.^44–47 As a result, the MePrCp–Zr precursor effectively reduced the formation of interfacial layers and defect sites, leading to a significant improvement in the leakage current density. The capacitance and leakage current densities as functions of applied voltage are shown in Fig. S8, SI. Consequently, a t_ox value of 0.75 nm was obtained with a low leakage level of 2.3 × 10⁻⁸ A cm⁻² by a ZrO₂ capacitor with no post annealing process or leakage barrier layer.

Conclusions

In conclusion, the MePrCp–Zr precursor successfully suppressed the formation of low-k interfacial layers in Pt/ZrO₂/TiN capacitor structures, compared to the conventional Cp–Zr precursor. This improvement was attributed to two main factors: first, the shorter O₃ injection time employed in the MePrCp–Zr process suppressed the formation of TiO_x or TiO_xN_y interfacial layers; second, the higher reactivity of the MePrCp–Zr precursor effectively reduced the TiO_x or TiO_xN_y interfacial layers. Thus, the interfacial t_ox value was decreased from 0.50 nm (Cp–Zr) to 0.12 nm (MePrCp–Zr), and the leakage current properties were improved by approximately one order of magnitude lower than that of the Cp–Zr precursor. In addition, the MePrCp–Zr precursor exhibited wide applicability to both TiN and Ru electrodes. This study confirms that ligand modification of the ALD precursor enables precise control of interface reactions, which is critical for DRAM scaling due to the decreased dielectric thickness and increased interface proportion.

Author contributions

Hyeongjun Kim: visualization, investigation, formal analysis, data curation, conceptualization, validation, writing – original draft, and writing – review and editing. Juan Hong: validation and visualization. Sangyeon Jeong: validation and visualization. Kyunghun Lyu: validation and visualization. Seungmin Jo: validation and investigation. Seokho Cho: validation and investigation. Juhyeong Kim: validation and investigation. Byung-Kwan Kim: resource and validation. Jin-Sik Kim: resource and validation. Woongkyu Lee: conceptualization, funding acquisition, visualization, methodology, project administration, resource, supervision, and writing – review & editing.

Conflicts of interest

There are no conflicts to declare.

Data availability

All relevant data supporting this article are included within the manuscript and supplementary information (SI). Supplementary information is available. See DOI: https://doi.org/10.1039/d5mh01502b.

Acknowledgements

This work was supported by K-CHIPS (Korea Collaborative & High-tech Initiative for Prospective Semiconductor Research) (RS-2025-02220654) funded by the Ministry of Trade, Industry & Energy (MOTIE, Korea), and by the National Research Foundation (NRF) (RS-2025-02433006) funded by the Korean government (MSIT, Korea).

Notes and references

1. S. E. Kim, J. Y. Sung, J. D. Jeon, S. Y. Jang, H. M. Lee, S. M. Moon, J. G. Kang, H. J. Lim, H.-S. Jung and S. W. Lee, Adv. Mater. Technol., 2023, 8, 2200878.
2. S. K. Kim and M. Popovici, MRS Bull., 2018, 43, 334–339.
3. J. Y. Park, H. Choi, J. Lee, K. Yang, S. Y. Lee, D. I. Han, I. Jeon, C. H. Jung, H. Lim, W. Lee and M. H. Park, Appl. Surf. Sci. Adv., 2025, 27, 100733.
4. B.-E. Park, I.-K. Oh, C. Mahata, C. W. Lee, D. Thompson, H.-B.-R. Lee, W. J. Maeng and H. Kim, J. Alloys Compd., 2017, 722, 307–312.
5. M. Popovici, A. Redolfi, M. Aoulaiche, J. A. Van Den Berg, B. Douhard, J. Swerts, P. Bailey, B. Kaczer, B. Groven, J. Meersschaut, T. Conard, A. Moussa, C. Adelmann, A. Delabie, P. Fazan, S. Van Elshocht and M. Jurczak, Microelectron. Eng., 2015, 147, 108–112.
6. 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.
7. 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.
8. H. Paik, J. Lim, H. Seo, T. K. Kim, J. Shin, H. Song, D. G. Kim, W. Lee, D. S. Kwon and C. S. Hwang, Mater. Horiz., 2025, 12, 7305–7317.
9. T. Kim, J. Jeon, S. H. Ryu, H. K. Chung, M. Jang, S. Lee, Y. J. Chung and S. K. Kim, ACS Appl. Mater. Interfaces, 2024, 16, 33877–33884.
10. J. Jeon, M. Jang, S. Ye, T. Kim, S. C. Kim, S. O. Won and S. K. Kim, Small, 2025, 21, 2502409.
11. M. Pešić, S. Knebel, M. Geyer, S. Schmelzer, U. Böttger, N. Kolomiiets, V. V. Afanas’ev, K. Cho, C. Jung, J. Chang, H. Lim, T. Mikolajick and U. Schroeder, J. Appl. Phys., 2016, 119, 064101.
12. S. Lee, H. H. Seol, M. K. Nam, D. H. Han, D. Kim, H. Oh, H. Kim, Y. Park, Y. Kim and W. Jeon, ACS Appl. Electron. Mater., 2024, 6, 3362–3373.
13. H. Kim, K. Choi, H. Lee and W. Lee, Electron. Mater. Lett., 2025, 21, 559–567.
14. K. M. Rabe, Nat. Nanotechnol., 2006, 1, 171–172.
15. M. Stengel and N. A. Spaldin, Nature, 2006, 443, 679–682.
16. J. H. Lee, B. E. Park, D. Thompson, M. Choe, Z. Lee, I. K. Oh, W. H. Kim and H. Kim, Thin Solid Films, 2020, 701, 137950.
17. Y. Kim, S. Y. Kang, J. H. Choi, J. S. Lim, M. Y. Park, S. J. Chung, J. Chung, H. I. Lee, K. H. Kim, Y. K. Kyoung, S. Heo, C. Y. Yoo and H. K. Kang, J. Nanosci. Nanotechnol., 2011, 11, 8309–8312.
18. M. Jang, J. Jeon, W. C. Lim, K. H. Chae, S. H. Baek and S. K. Kim, Ceram. Int., 2024, 50, 47910–47915.
19. W. Lee, C. H. An, S. Yoo, W. Jeon, M. J. Chung, S. H. Kim and C. S. Hwang, Phys. Status Solidi RRL, 2018, 12, 1–6.
20. H. Kim, H. Yang and W. Lee, Ceram. Int., 2025, 51, 19736–19741.
21. D. K. Lee, H. B. Kim, S. H. Kwon and J. H. Ahn, Mater. Lett., 2020, 279, 128490.
22. S. Y. Lee, J. Chang, J. Choi, Y. Kim, H. J. Lim, H. Jeon and H. Seo, Curr. Appl. Phys., 2017, 17, 267–271.
23. T. Blanquart, J. Niinistö, N. Aslam, M. Banerjee, Y. Tomczak, M. Gavagnin, V. Longo, E. Puukilainen, H. D. Wanzenboeck, W. M. M. Kessels, A. Devi, S. Hoffmann-Eifert, M. Ritala and M. Leskelä, Chem. Mater., 2013, 25, 3088–3095.
24. E. Choi, H. Kim, S. J. Maeng, J. Lee, D. H. Kim, K. Heo and J. Y. Yun, Mater. Today Commun., 2022, 32, 104008.
25. H. S. Lee, G. Jung, H. Yun, I. Y. Song, J. Lee, W. E. Lee, J. Y. Yun, J. H. Na and K. Heo, Mater. Today Commun., 2025, 43, 111740.
26. J. Niinistö, K. Kukli, M. Kariniemi, M. Ritala, M. Leskelä, N. Blasco, A. Pinchart, C. Lachaud, N. Laaroussi, Z. Wang and C. Dussarrat, J. Mater. Chem., 2008, 18, 5243–5247.
27. S. Park, Y. Choi, S. Park, H. Lee, K. Lee, S. Kang, J. Park and W. Jeon, Appl. Mater. Today, 2024, 40, 102354.
28. T. P. Holme and F. B. Prinz, J. Phys. Chem. A, 2007, 111, 8147–8151.
29. C. E. Nanayakkara, A. Vega, G. Liu, C. L. Dezelah, R. K. Kanjolia and Y. J. Chabal, Chem. Mater., 2016, 28, 8591–8597.
30. M. Kim, J. Y. Hwang, D. Kim, T. Song, K. H. Lee and H.-H. Park, Ceram. Int., 2023, 49, 24065–24070.
31. R. D. Purohit, S. Saha and A. K. Tyagi, Mater. Sci. Eng. B, 2006, 130, 57–60.
32. D. G. Lamas, A. M. Rosso, M. S. Anzorena, A. Fernández, M. G. Bellino, M. D. Cabezas, N. E. Walsöe de Reca and A. F. Craievich, Scr. Mater., 2006, 55, 553–556.
33. J.-C. Hsu, Y.-H. Lin and P. W. Wang, Coatings, 2020, 10, 47.
34. M. Nakahara, S. Tsunekawa, K. Watanabe, T. Arai, T. Yunogami and K. Kuroki, J. Vac. Sci. Technol., B: Microelectron. Nanometer Struct.--Process., Meas., Phenom., 2001, 19, 2133–2136.
35. J. Jeon, T. Kim, M. Jang, H. K. Chung, S. C. Kim, S. O. Won, Y. Park, B. J. Choi, Y. J. Chung and S. K. Kim, Chem. Mater., 2024, 36, 3326–3333.
36. N. G. Kang, M. J. Ha and J. H. Ahn, Chem. Mater., 2024, 36, 8496–8503.
37. R. A. Dorey, Ceramic Thick Films for MEMS and Microdevices, William Andrew, Norwich, New York, 2011.
38. A. J. Lee, S. Lee, D. H. Han, Y. Kim and W. Jeon, J. Mater. Chem. C, 2023, 11, 6894–6901.
39. A. J. Lee, S. Lee, D. H. Han, Y. Kim and W. Jeon, J. Mater. Chem. C, 2023, 11, 6894–6901.
40. D. H. Han, S. Lee, J. H. Hwang, Y. Kim, M. Bonvalot, C. Vallee, P. Gonon and W. Jeon, IEEE Trans. Electron Devices, 2021, 68, 5753–5757.
41. J. H. Kim, V. A. Ignatova, J. Heitmann and L. Oberbeck, J. Phys. D: Appl. Phys., 2008, 41, 172005.
42. Y. Seo, S. Lee, I. An, C. Song and H. Jeong, Semicond. Sci. Technol., 2009, 24, 6–11.
43. W. Jeon, Y. Kim, C. H. An, C. S. Hwang, P. Gonon and C. Vallee, IEEE Trans. Electron Devices, 2018, 65, 660–666.
44. Y. Choi, S. Lee, J. Jeong, D. Kim, H. Oh, Y. Park and W. Jeon, Appl. Surf. Sci., 2025, 712, 164149.
45. J. Nam, S. Lee, H. Oh, H. Kim, Y. Park, I.-H. Baek and W. Jeon, Mater. Sci. Semicond. Process., 2025, 192, 109485.
46. J. Y. Park, H. Choi, J. Lee, K. Yang, S. Y. Lee, D. I. Han, I. Jeon, C. H. Jung, H. Lim, W. Lee and M. H. Park, Appl. Surf. Sci. Adv., 2025, 27, 100733.
47. H. Seo, I. W. Yeu, D. S. Kwon, D. G. Kim, J. Lim, T. K. Kim, H. Paik, J.-H. Choi and C. S. Hwang, Adv. Electron. Mater., 2022, 8, 2200099.

---