Stabilization of ferroelectric phase in tungsten capped Hf_0.8Zr_0.2O₂

Recently, ferroelectric HfO₂ films have gained significant interest due to their potential to achieve ultrasmall, non-volatile ferroelectric-gate memory transistors and low voltage transistors through the negative capacitance effect.^1–20 The ferroelectric properties of HfO₂ films doped with Si, Al, Y, and Zr have been comprehensively studied, and it has been shown that the lack of inversion symmetry in crystals belonging to the orthorhombic space group Pca2₁ (#29) may be responsible for the observed ferroelectricity in these films. The mechanical confinement by a metal capping layer is believed to stabilize the orthorhombic phase in doped HfO₂ films under appropriate annealing conditions. Top layers of TiN, TaN, Pt, and Ir have been shown to result in ferroelectricity in the underlying doped HfO₂ films after rapid thermal annealing (RTA). Here, we report on the observation of ferroelectricity in Hf_0.8Zr_0.2O₂ with a W capping layer (hereafter referred to as W-HZO).

The properties of W-HZO in metal-insulator-semiconductor (MIS) and metal-insulator-metal (MIM) structures have been studied and compared with the conventional TiN-HZO stack by polarization-voltage (PV) hysteresis measurement and positive up negative down (PUND) measurement. Grazing Incident X-ray Diffraction (GIXRD) has been used to study the crystallographic structure of HZO films, while the interaction of the top metal with the underlying ferroelectric film has been investigated by high resolution Transmission Electron Microscopy (TEM) and Energy Dispersive X-ray (EDX) spectroscopy. Our results show that W is a promising capping layer for the stabilization of the ferroelectric phase in HZO with high remanent polarization and negligible metal-oxide inter-mixing.

For MIS structures, a heavily (10¹⁹/cm³) P-doped Si(100) substrate is first oxidized in O₂ ambient during an RTA step at 900 °C for 20 s, forming 2 nm of thermal SiO₂ on Si. Later, 100 cycles of Hf_0.8Zr_0.2O₂ is deposited at 250 °C using an Ultratech Fiji G2 atomic layer deposition (ALD) tool, providing a 10 nm-thick amorphous HZO film. A 4:1 ratio between the HfO₂ mono-layer and the ZrO₂ mono-layer determines the stoichiometry of the deposited HZO. Next, 10 nm of W or TiN is deposited on the stack followed by a 30 s-RTA step in N₂ at 500 °C to crystallize the HZO. Finally, a stack of Ti/TiN/Al is sputtered on the sample, 60 nm each, and the top electrodes are defined by subsequent photolithography and reactive ion etching. For MIM capacitors with W or TiN electrodes, the Si substrate is first coated with 100 nm of the chosen electrode (W or TiN) followed by deposition of 10 nm HZO and 100 nm of the same electrode. Next, the stack is annealed under similar conditions mentioned above (500 °C and 30 s), and top electrodes are further defined by lithography and etching. Two methods for TiN deposition have been investigated: DC sputtering at room temperature and ALD at 400 °C. In contrast, only DC sputtering has been used for deposition of W films reported here.

The PV and PUND measurements described here are performed using a Radiant Precision Multiferroic analyzer. Figures 1(a) and 1(b) show the PUND and PV measurements on MIS structures. The highest (23 μC/cm²) and lowest (5 μC/cm²) remanent polarization (P_r) values in HZO films are observed with W and ALD TiN capping, respectively. The lower P_r for ALD TiN capping (5 μC/cm²) compared to sputtered TiN (12 μC/cm²) is most likely due to prolonged exposure (∼1 h) of HZO to a temperature of 400 °C at which atomic layer deposition of TiN takes place. This temperature is sufficient to initiate the crystallization of the HZO, and particularly during the sample loading and initial cycles of TiN deposition can modify the stabilized phases in the film.²¹ Crystallization of the HZO film with no capping is expected to result in the stabilization of a non-polar monoclinic phase, and therefore, ALD TiN-capped HZO is expected to be less ferroelectric than the sputtered TiN-capped film.^1,21

The DC current measurement on MIS capacitors, Fig. S1 in the supplementary material, shows comparable leakage current for different electrodes.

Similar to MIS structures, MIM capacitors with sputtered W exhibit higher remanent polarization (10 μC/cm²) than those with TiN (5 μC/cm²), as demonstrated in Figs. 2(a) and 2(b). The lower P_r in MIM capacitors compared to MIS has been reported before and was attributed to a weak (111) orientation of the bottom polycrystalline metal electrode which slightly suppresses the formation of the orthorhombic phase in the top HZO and results in lower P_r in MIM capacitors.¹¹ The larger coercive voltage in MIS structures is due to the presence of SiO₂ in series with the HZO film, which effectively decreases the voltage drop across the HZO. The permittivity of the HZO film as a function of voltage, derived from the C-V measurement in MIM structures, shows the typical “butterfly” profile [Fig. 2(c)], indicative of the ferroelectric switching.²² 

Following the crystallization anneal, the top electrode of MIS capacitors is chemically etched in NH₄OH:H₂O₂:H₂O in the ratio of 1:1:10 at temperatures above 40 °C, and the GIXRD on the underlying HZO is performed [Fig. 3(a)]. The ferroelectric behavior in thin HZO films has been attributed to the stabilization of the orthorhombic phase.²³ In our synthesized HZO films, a mixture of orthorhombic (at 30.4°) and tetragonal (at 30.8°) phase is observed.²³ For W-HZO, the peak is closer to the theoretical orthorhombic peak compared to TiN-HZO (both sputtered and ALD deposited TiN). In addition, the W-HZO film shows an XRD peak with smaller FWHM compared to TiN-HZO, indicating a more textured structure [along the (111) direction]. Note here that the ALD TiN capped HZO film is qualitatively different from the other two films as partial crystallization occurs during the ALD growth of TiN at 400 °C before the RTA annealing. (Sputtered) TiN-HZO and W-HZO MIS capacitors were studied in detail by TEM. Thin cross-sectional foils were prepared by mechanical polishing at a 0.5° wedge followed by Ar ion milling with a starting energy of 3 keV down to a final cleaning energy of 200 eV. High Resolution TEM (HR-TEM) and EDX were performed on an FEI-Titan instrument operating at an acceleration voltage of 200 kV. Elemental maps were acquired in the ChemSTEM mode using a Bruker EDS windowless Silicon Drift Detector (SDD) concurrently with High-Angle Annular Dark-Field (HAADF) images. Figures 3(b) and 3(c) show the cross-sectional TEM image of MIS W-HZO and TiN-HZO, respectively. A 1.9 nm thick SiO₂ layer separates the HZO layer from the Si substrate, relaxing the epitaxial strain between these layers. Considering the TEM images of Figs. 3(b) and 3(c), one can notice a clear boundary at the W-HZO interface, while the TiN-HZO boundary is smeared, indicating an intermixing of TiN and underlying HZO.

In order to investigate the diffusion of elements from the metal cap layer into the HZO, the EDX elemental analysis is performed, and the point spectrum (shown in Fig. S2 in the supplementary material) was collected from the Si substrate, the HZO layer, and the metal cap independently. All elements possible to be present in the HZO layer were considered for a final quantification, which is summarized in Table I. A summary of the elemental analysis is plotted as a 2-D false color map in Fig. 4 along with HAADF images for a better visualization of the layers.

While there is a clear trace of the TiN capping layer in the underlying HZO, the EDX measurements given in Fig. 4(a) and Table I show the absence of W in HZO. Based on these measurements, it becomes evident that annealing of TiN-HZO promotes diffusion of Ti and N atoms into the HZO film, which further confirms the smeared TiN-HZO interface. In contrast, the intact W-HZO interface indicates the negligible interaction of W with HZO during the high temperature treatments. It has been previously reported that the presence of N atoms in the HfO₂ films creates positively charged oxygen vacancies^2,24,25 and the formed vacancies are believed to help stabilize the orthorhombic phase in doped HfO₂.⁷ However, our results show that a similar (and even stronger) ferroelectric behavior can be achieved for W-HZO, where no nitrogen is present in the underlying HZO film. This shows that the presence of N and its diffusion are not critical factors to achieve ferroelectricity. Importantly, diffusion of Ti and/or N could lead to increased leakage,²⁵ especially in ultra-thin films, which is undesirable for electronic applications. In that context, the ferroelectric W-capped HZO reported in this work could be more effective.

In conclusion, we have shown that a W capping layer can be used to stabilize the ferroelectric phase in Hf_0.8Zr_0.2O₂. Compared to widely used TiN-capped Hf_0.8Zr_0.2O₂ films, W-capped HZO shows superior properties in terms of remanent polarization and metal-oxide intermixing. HR-TEM and EDX analyses show minimal W diffusion into the HZO film as opposed to significant Ti and N diffusion found in TiN capped samples. The stronger ferroelectricity in W-capped films is somewhat surprising, as the presence of N in the HZO films has often been considered as a catalyst for oxygen vacancy formation that leads to ferroelectricity in HZO. Nevertheless, our results indicate that a high quality ferroelectric film can be obtained with W capping without introducing structural defects coming from cap metal diffusion. Further studies will have to focus on understanding the formation and behavior of oxygen vacancies in such W-capped films.

See supplementary material for the data on DC leakage current of MIS capacitors and for the full EDX spectrum of W-HZO and TiN-HZO structures.

This research was funded by Berkeley Center for Negative Capacitance Transistors (BCNCT). The work at the Molecular Foundry was supported by the Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under Contract No. DE-AC0205CH11231.