Research Article www.acsami.org
Engineering of Ferroelectric HfO2−ZrO2 Nanolaminates Stephen L. Weeks,* Ashish Pal, Vijay K. Narasimhan, Karl A. Littau, and Tony Chiang Intermolecular, Inc., 3011 N. 1st Street, San Jose 95134, California, United States S Supporting Information *
ABSTRACT: In this work, the ferroelectric properties of nanolaminates made of HfO2 and ZrO2 were studied as a function of the deposition temperature and the individual HfO2/ZrO2 layer thickness before and after electrical field cycling. The ferroelectric response was found to depend on the structure of the nanolaminates before any postdeposition annealing treatment. After annealing with a TiN cap, an “antiferroelectric-like” response was obtained from nanolaminates deposited in an amorphous state at a lower temperature, whereas a ferroelectric response was obtained from nanolaminates deposited at a higher temperature, where crystallites were detected in thick films before annealing. As the individual layer thicknesses were decreased, an increased lattice distortion and a concurrent increase in remanent polarization were observed from the nanolaminates deposited at high temperatures. After field cycling, nanolaminates deposited at lower temperatures exhibited an antiferroelectric-like to ferroelectric transition, whereas those deposited at higher temperatures exhibited a larger remanent polarization. Finally, we demonstrate that by leveraging the proper choice of process conditions and layer thickness, remanent polarizations exceeding those of the HfZrO4 solid solution can be obtained. KEYWORDS: hafnium oxide, nanolaminates, ferroelectrics, nonvolatile memory, wake up
■
INTRODUCTION The demand for high-performance personal electronic devices has driven a need for a new generation of high-density memory technologies that operate with minimal power consumption. Ferroelectric materials demonstrate a hysteresis in electric-fieldinduced polarization, which has spurred their proposed use in next-generation memory applications.1,2 The development of ferroelectric random access memory (FeRAM) has traditionally focused on perovskite materials, which suffer from a small band gap, leading to a low barrier height. Therefore, thick films are typically required to avoid large leakage currents in perovskitebased FeRAM, 3 which is a major obstacle for the implementation of these materials at advanced technology nodes. The observation of ferroelectricity in HfO2-based materials and HfZrO4 solid solutions4−8 has led to a large research effort aimed at developing these materials into a viable alternative to perovskite-based FeRAM. HfO2-based ferroelectrics offer a number of advantages over traditional perovskite-based ferroelectrics. While perovskites typically have a band gap (Eg) of 3.2−4.3 eV,3 both HfO2 and ZrO2 have larger Eg values between 5 and 5.8 eV,9 allowing thinner films to be utilized in memory capacitor applications. HfO2-based ferroelectrics resist polarization degradation during forming gas annealing more effectively than conventional perovskite-based ferroelectrics, an observation that has been attributed to the strong HfO and ZrO bonds.10 Furthermore, these materials are already widely utilized as high-k materials in advanced technology nodes,11 which has led to the development of robust atomic © 2017 American Chemical Society
layer deposition (ALD) processes required to deposit these materials in three-dimensional structures,12−16 making this material system compatible with modern semiconductor integration schemes. Control of the polymorphism present in HfO2 and ZrO2 became a focus of research to encourage the high-k phase for their use as a dielectric in complementary-metal−oxide− semiconductor applications. In both bulk HfO2 and ZrO2, at ambient pressure and temperature, the monoclinic (P21/c) phase is the most stable, whereas the tetragonal (P42/nmc) and cubic (Fm3̅m) phases become stable only at higher temperatures.17,18 Decreasing the film thickness increases the stability of the tetragonal and cubic phases at room temperature,19,20 a phenomenon attributed to increasing surface energy effects.20 The ferroelectric response in HfO2-based films has been attributed to the orthorhombic Pca21 phase, and to enable the ferroelectric response in HfO2-based materials it is necessary to choose process conditions that stabilize this phase.6,7,21,22 Since the first reports of ferroelectric behavior in HfO2-based films, researchers have found that the orthorhombic phase is stabilized by oxygen vacancies,23,24 strain,25 and surface energy.22 Considering the HfO2/ZrO2 material system, the HfZrO4 solid solution, the HfO2/ZrO2 bilayer,26 and nanolaminates27 have been shown to exhibit ferroelectricity. Thus far, no report Received: January 16, 2017 Accepted: March 24, 2017 Published: March 24, 2017 13440
DOI: 10.1021/acsami.7b00776 ACS Appl. Mater. Interfaces 2017, 9, 13440−13447
Research Article
ACS Applied Materials & Interfaces
annealing,6,7 and in this work, GIXRD scans were collected on films with a blanket 5 nm TiN capping layer. The film thickness was determined using X-ray fluorescence collected on a PANalytical 2830 wafer analyzer. Cross-metrology thickness confirmation of the nanolaminate stacks was accomplished using cross-sectional high-resolution transmission electron microscopy (HRTEM) data collected at Nanolab Technologies Inc. (Milpitas, CA). The film composition was determined using X-ray photoelectron spectroscopy on a Thermo Fisher Kα system. The ferroelectric characterization was accomplished using polarization−voltage (P−V) measurements collected on a Radiance Precision II ferroelectric tester connected to a Cascade probe station. All P−V scans were collected using a triangular waveform at a frequency of 0.25 kHz. The P−V scans were collected both before and after applying electrical cycling for 1 s at 1 kHz with an applied voltage of ±3.5 V to enable characterization of the as-prepared stack as well as to determine the performance of the devices after the electrical “wake up” commonly observed in HfO2-based ferroelectrics.7,23,28−30
exists detailing the effects of process conditions on this material system as it progresses from a bilayer through a nanolaminate and ultimately to fully blended HfZrO4. Herein, we report on the ferroelectric response of HfO2/ZrO2 nanolaminates as a function of the deposition temperature and demonstrate that at 285 °C of individual layers induces an increase in remanent polarization beyond that observed for fully blended HfZrO4 processed under identical conditions. Moreover, we show that the final ferroelectric response of the deposited film is sensitive to the deposition temperature and structure of the nanolaminate before annealing.
■
EXPERIMENTAL SECTION
To assess the performance of the ferroelectric thin film, metal− insulator−metal (MIM) capacitors were fabricated using a 5 nm TiN bottom electrode deposited using physical vapor deposition (PVD) on a 300 mm 1000 Å SiO2/Si substrate. The HfO2/ZrO2 nanolaminates were deposited by ALD at 260 and 285 °C using Hf[N(CH3)2]4 (TDMA-HF), ZyALD ((tris(dimethylamino)cylcopentadienyl)zirconium, Air Liquide), and O3 as the Hf and Zr precursors and oxidant, respectively. The nanolaminates were deposited using a 300 mm deposition chamber (Intermolecular, A-30) with precursors delivered by bubbling Ar carrier gas through the liquid precursor sources into the main chamber volume via a showerhead. The ALD process used a 15 s precursor exposure for both the Zr and Hf sources and 40 s of 4 wt % O3 exposure. These conditions yielded a growth of 1.35 and 0.9 Å/cycle for the pure HfO2 and ZrO2 films, respectively, at 260 °C, and at 285 °C, the HfO2 growth rate increased to 1.39 Å/ cycle, whereas the ZrO2 growth rate remained constant. The number of HfO2 and ZrO2 cycles in a given supercycle required to deposit the (1 nm HfO2/1 nm ZrO2) × 4, (2 nm HfO2/2 nm ZrO2) × 2, and (4 nm HfO2/4 nm ZrO2) nanolaminates were determined from these growth rates. Using a HfO2−ZrO2 cycle ratio of 1:1, alternating single cycles of HfO2 and ZrO2, yielded a HfxZr1−xO2 solid solution. Nanolaminates were obtained by first depositing HfO2 and then ZrO2, each with the desired layer thickness by applying the appropriate number of cycles and then repeating the layer deposition to obtain a constant total stack thickness. Lee et al. reported the formation of the monoclinic phase in the nanolaminates formed with ZrO2 as the bottom layer;27 to avoid the nonferroelectric monoclinic phase in this study, all nanolaminates used HfO2 as the starting layer. At 285 °C, the total film thickness was 8 nm, whereas the individual layer thicknesses were varied between 4, 2, and 1 nm. At 260 °C, the growth rate of HfO2 was slightly lower, whereas the number of ALD cycles was kept constant, which led to a slightly lower individual layer thicknesses. The differences in the layer thickness due to the deposition temperature were less than the lattice parameters for tetragonal and orthorhombic HfO2 and ZrO2 and thus considered insignificant for purposes of comparing the ferroelectric properties in these films. Hereafter, the nanolaminates deposited at both 285 and 260 °C are referred to as the (4 nm HfO2/4 nm ZrO2), (2 nm HfO2/2 nm ZrO2) × 2, and (1 nm HfO2/1 nm ZrO2) × 4 nanolaminates, respectively. After deposition of the nanolaminate or the HfZrO4 solid solution, the wafer was cleaved into 44 mm × 44 mm coupons for the remainder of device processing. A 5 nm blanket TiN layer was deposited onto the coupons using PVD. The top electrode was then defined with 100 nm of additional PVD TiN deposited through a shadow mask to serve as a hard mark for subsequent wet etch. The entire stack was then annealed at 500 °C in N2 for 10 min. All of the electrical measurements performed on the devices in this study were collected after annealing with a TiN cap to induce crystallization in the films. After annealing, the devices were isolated by etching through the initial 5 nm blanket TiN layer using 30% H2O2 for 30 min, leaving behind the isolated TiN electrodes. Grazing incidence X-ray diffraction (GIXRD) scans were collected on a PANalytical X’Pert PRO MRD XL XRD system and used to identify the crystallographic phase present in the film. The crystallographic phase of HfO2-based ferroelectric films has been shown to be sensitive to the presence of a capping layer during
■
RESULTS This study focused on the ferroelectric response of the layered structures composed of HfO2 and ZrO2. The growth per cycle of both HfO2 and ZrO2 was smaller than the lattice constants reported for orthorhombic Pca21 HfZrO4.6,22 Therefore, alternating single ALD cycles of the two compounds ensured complete intermixing of the species forming a solid solution with compositions of Hf0.56Zr0.44O2 and Hf0.55Zr0.45O2 for films deposited at 260 and 285 °C, respectively, determined by XPS. For the nanolaminates in this study, it is important to emphasize that separate, distinct HfO2 and ZrO2 layers were deposited, which remained separated after the processing used to fabricate the devices. To confirm that the HfO2 and ZrO2 layers remained separated, HRTEM images (Figure 1) were collected on the nanolaminate structures deposited at 260 °C after top electrode deposition and annealing. The images show that even the thinnest individual layer case, (1 nm HfO2/1 nm ZrO2) × 4, retained its nanolaminate structure with distinct HfO2 and ZrO2 layers observable in the image. In all of the nanolaminates, lattice fringes were observed that extended throughout the entire film thickness, consistent with an interlayer epitaxial relationship previously reported for HfO2/ ZrO2 nanolaminates.31 Figures 2 and 3 show the GIXRD scans collected on the (4 nm HfO2/4 nm ZrO2), (2 nm HfO2/2 nm ZrO2) × 2, and (1 nm HfO2/1 nm ZrO2) × 4 nanolaminates and HfZrO4 solid solutions deposited at 260 and 285 °C, respectively, after annealing at 500 °C in N2 for 10 min with a 5 nm TiN cap deposited by PVD. Scattering consistent with the orthorhombic/tetragonal phases was present in all of the nanolaminate films as well as the HfZrO4 solid solution. No monoclinic content was observed in any of these films. The interplanar spacing of the o(111)/t(011) planes was observed to decrease with decreasing layer thickness in the nanolaminates for films deposited at 285 °C. This shift in the interplanar spacing of the o(111)/t(011) planes was correlated with the ferroelectric response of capacitors fabricated with these films, as discussed below. The P−V characteristics before field cycling are shown in Figure 4. At a deposition temperature of 260 °C, the (4 nm HfO2/4 nm ZrO2) nanolaminate exhibited very little hysteresis before field cycling. The hysteresis loop of the fresh (2 nm HfO2/2 nm ZrO2) × 2 nanolaminate was pinched at this temperature, indicating a mix of ferroelectric and antiferroelectric-like responses, whereas the (1 nm HfO2/1 nm ZrO2) × 13441
DOI: 10.1021/acsami.7b00776 ACS Appl. Mater. Interfaces 2017, 9, 13440−13447
Research Article
ACS Applied Materials & Interfaces
Figure 2. GIXRD scans of (a) 4 nm HfO2/4 nm ZrO2, (b) (2 nm HfO2/2 nm ZrO2) × 2, and (c) (1 nm HfO2/1 nm ZrO2) × 4 nanolaminates, and (d) HfZrO4 solid solution deposited at 260 °C and annealed at 500 °C in N2 with a 5 nm TiN cap.
Figure 1. HRTEM images of (a) 4 nm HfO2/4 nm ZrO2 bilayer, (b) (2 nm HfO2/2 nm ZrO2) × 2, and (c) (1 nm HfO2/1 nm ZrO2) × 4 nanolaminate stacks deposited at 260 °C after top electrode deposition and annealing steps.
4 nanolaminate produced an antiferroelectric-like response characteristic of tetragonal HfxZr1−xO2 systems. The HfZrO4 solid solution deposited at this temperature showed a weak loop opening with a 2Pr value of 16 μC/cm2. In contrast to the films deposited at 260 °C, those deposited at 285 °C possessed a clean ferroelectric hysteresis loop before field stressing. The 2Pr values of the (4 nm HfO2/4 nm ZrO2), (2 nm HfO2/2 nm ZrO2) × 2, and (1 nm HfO2/1 nm ZrO2) × 4 nanolaminates grown at 285 °C were observed to be 5.8, 20.6, and 36.6 μC/ cm2, respectively. Before field cycling, the 2Pr of the HfZrO4 solid solution deposited at 285 °C was 22 μC/cm2. The P−V characteristics of MIM capacitors fabricated with the nanolaminates and the HfZrO4 solid solution after field cycling at 3.5 V are shown in Figure 5. At a deposition temperature of 260 °C, it is notable that devices fabricated from all of the nanolaminates exhibited a hysteresis loop after electrical cycling. After field cycling, the 2Pr values from the (4 nm HfO2/4 nm ZrO2), (2 nm HfO2/2 nm ZrO2) × 2, and (1 nm HfO2/1 nm ZrO2) × 4 nanolaminates deposited at 260 °C were observed to be 20, 35, and 39 μC/cm2, respectively, whereas that of the HfZrO4 solid solution was found to be 39 μC/cm2. At a deposition temperature of 285 °C after field
Figure 3. GIXRD scans of (a) 4 nm HfO2/4 nm ZrO2, (b) (2 nm HfO2/2 nm ZrO2) × 2, and (c) (1 nm HfO2/1 nm ZrO2) × 4 nanolaminates, and (d) HfZrO4 solid solution deposited at 285 °C and annealed at 500 °C in N2 with a 5 nm TiN cap.
cycling, the 2Pr values from the (4 nm HfO2/4 nm ZrO2), (2 nm HfO2/2 nm ZrO2) × 2, and (1 nm HfO2/1 nm ZrO2) × 4 13442
DOI: 10.1021/acsami.7b00776 ACS Appl. Mater. Interfaces 2017, 9, 13440−13447
Research Article
ACS Applied Materials & Interfaces
Kim et al. reported that decreasing the deposition temperature of ferroelectric HfO2 to as low as 220 °C led to C incorporation in the film, which could suppress the lateral grain growth and stabilize the ferroelectric phase. In that study, they used Auger emission spectroscopy to show that decreasing the deposition temperature to 240 °C leads to an inclusion of ∼10% C, a concurrent decrease in grain size, and stabilization of the ferroelectric phase.32 Although C can contribute to the ferroelectric response in HfO2-based thin films, XPS scans of the C 1s region show no detectible carbon incorporation in our films (see Figure S1 of the Supporting Information). Beyond this, the GIXRD scans collected show no significant peak broadening in films deposited at 260 °C compared to those deposited at 285 °C (see Figures 2 and 3), suggesting minimal change in grain size. In addition to carbon, Xu et al. used secondary-ion mass spectrometry to show that the presence of trace amounts of nitrogen (
