The other model antiferroelectric: PbHfO₃ thin films from ALD precursors

Antiferroelectrics (AFE) are recognized experimentally by a phase transition between two non-polar phases accompanied by a dielectric anomaly and a characteristic double polarization hysteresis loop arising from an electric field-induced phase transition from a non-polar to a polar phase. The properties of this transition make AFEs excellent for energy storage,¹ power conditioning,² and upcoming electronic memory applications.³ The archetype AFE perovskite is considered to be PbZrO₃ (PZO), identified as early as 1951.⁴ Here, we investigate the closely related perovskite PbHfO₃ (PHO). It was recognized in 1953 that the crystal structure of PHO was similar to the archetypical AFE PZO, although only a linear dielectric behavior was observed at that time.⁵ Subsequent studies on bulk PHO confirmed the structure,⁶ but the evidence for a truly antiferroelectric behavior, i.e., the field-induced phase transition and its characteristic double loop, was only reported in 1980. Besides, PHO differs from PZO by the presence of an intermediate phase consistently observed between the paraelectric cubic phase and the orthorhombic antiferroelectric Pbam phase.⁵ This intermediate phase is observed between 160 °C and 215 °C and is often referred to as AFE₁, although there is very little evidence that this phase is really antiferroelectric itself. It was recently identified as incommensurate.⁷ Generally speaking, PHO has been much less investigated than PZO, bulk or thin film. In particular, to the best of our knowledge, there is no reported electrical characterization on thin film PHO aside from the hysteresis loop in our recent work on PHO film growth.⁸ 

The AFE-FE phase transition has been noted as an energy storage mechanism since the 1960s,⁹ and therefore, these materials are considered for applications such as pulsed power, lasers, radar, and power inverters, which require high energy storage density (ESD) electrostatic capacitors.¹⁰ 

In the bulk, PHO prepared by rolling and conventional sintering was clearly AFE, with an energy storage density of 7.6 J/cm³ for a bulk material.¹¹ Recent advances in ultra-thin, atomic-layer deposited AFE-like materials have shown high ESDs, namely, Si-doped,¹² Al-doped,¹³ and Zr-rich¹⁴ compositions and combinations thereof¹⁵ of ferroelectric hafnia. Owing to thin, conformal coating, new regimes of ESD have been reached.¹⁶ However, the characteristic properties of Pb-based ferroelectrics such as high polarization, low coercive field, and high permittivity are not available within the hafnium oxide material system.

Metal–insulator–metal (MIM) capacitors were fabricated in order to investigate the electrical properties of the PHO films grown by atomic layer deposition (ALD). The 100 nm-thick platinum bottom electrode was formed on a TiO₂/SiO₂/Si substrate stack.¹⁷ The ALD PHO films were grown using a super-cycle of HfO₂ and PbO_x constituent binary oxide processes in a Kurt Lesker ALD 150-LX platform. The binary PbO_x process was performed using 98%-purity lead bis(dimethylaminomethylpropanolate) [Pb(DMAMP)₂] procured from Strem Chemicals Inc. The Pb(DMAMP)₂ was held at 90 °C and was oxidized using a combination of H₂O and ozone. The cation composition was varied by changing the relative number of PbO_x:HfO₂ cycles. Details of the growth can be found in a previous publication by Strnad et al.⁸ PHO near the ideal PbHfO₃ stoichiometry was achieved using a 3:1 PbO_x:HfO₂ cycle-ratio at a substrate temperature of 250 °C. The growth kinetics of the ALD PHO were measured in situ using an ellipsometer, which indicated a linear growth rate of 0.07 nm/cycle. The specific cation precursors, oxidizer combinations, and dose/purge times were nearly identical to that reported for ALD PbTiO₃ in a separate work by Strnad et al.,¹⁸ except that tetrakis dimethylamino hafnium (TDMAH) was substituted for tetrakis dimethylamino titanium. Three thicknesses were grown: 20 nm, 50 nm, and 200 nm. After film growth, Pt was sputtered at room temperature and electrodes were defined using a lift-off process. Samples were then annealed at 700 °C or 800 °C for 1960s in an oxygen environment. Sample composition was measured after heat treatments using Rutherford backscattering analysis showing stoichiometric PHO within the error.

Structural characterization was performed with grazing-incidence X-ray diffraction (GIXRD) and transmission electron microscopy (TEM). Diffractograms of the as-deposited material in Fig. 1(a) show the presence of crystalline peaks coinciding with the litharge phase of lead oxide (PDF Card: 00-005-0570). This supports the observations from Strnad et al. that the Pb precursor does not wet the surface but instead forms a segregated, crystalline phase.⁸ With increasing thickness, we observe the emergence of thinner Bragg peaks and a general increase in intensity that we attribute to a combination of size effect, improvement in crystallinity, and increase in volume. The annealed film diffractograms [Fig. 1(b)] show a randomly oriented PHO film (PDF card: 04-012-0549) with no significant differences in preferred orientation among the film thicknesses. If the crystallization proceeded from the bottom electrode, as it does with many deposition techniques, then the orientation would be expected to change as the film thickness is increased and stress can be relieved.¹⁹ These results support earlier observations⁸ and suggest that PHO from ALD precursors crystalizes by interdiffusion of an amorphous HfO₂ matrix with isolated crystalline PbO regions. The interdiffusion-driven growth process will make it difficult to use conventional techniques for thin film stress tuning and templating. The film stress and the ability to template growth from oriented electrodes or seeds may be affected by this unique growth process.

Transmission electron microscopy (TEM) characterization was carried out using a probe-corrected JEOL ARM 200F operated at 200 keV in both conventional and scanning modes. The specimen was prepared using a Thermofisher Helios 4 focused ion beam/scanning electron microscope. TEM characterization showed that the ∼50 nm PHO had a columnar microstructure with grains going all the way through the thickness of the film. The film thickness ranged from 54 nm to 60 nm due to roughness in both the bottom Pt electrode and the film itself. Scanning TEM-high angular annular dark field (STEM-HAADF) images, which provide mass-thickness contrast, show small, spherical pores in the interior of the film, particularly along grain boundaries, with larger, hemispherical pores along the surface of the film adjacent to the interface with the top Pt electrode (Fig. S1). One possibility is that these pores represent volumes previously occupied by crystalline PbO, which was integrated during annealing or evaporated, as it is has been observed in lead zirconate titanate (PZT).²⁰ Further work is needed to optimize the thermal processing to reduce the porosity. While both Pt layers exhibit a columnar microstructure, the top electrode also exhibited intergranular pores possibly from the room-temperature sputtering process, whereas the bottom did not. Fast Fourier transforms of high resolution TEM (HRTEM) images of the PHO layer were indexed to confirm that the PHO has an orthorhombic crystal structure (space group 55—Pbam) with lattice parameters a = 0.5856 nm, b = 1.1729 nm, and c = 0.8212 nm, which is in agreement with previous studies on bulk crystals.²¹ 

Figure 2(a) shows the dielectric hysteresis loops at 1 kHz for the 20 nm, 50 nm, and 200 nm films obtained using the Radiant Precision Premier II testing system. The electrode areas were either 1.81 or 0.492 × 10⁻³ cm². Each thickness shows a classical AFE character with AFE → FE fields decreasing with increasing thickness. The trend observed in this study aligns with previous works where thinner films have a lower polarization and an increase in E_f,²² but the precise origin is unknown. Stress is a factor affecting the phase transition field.²³ The XRD study in Figs. 1(a) and 1(b) suggests that there are no large residual stress differences among the thicknesses but thermal stress due to the thermal expansion mismatch between the film and substrate are thickness-dependent.²⁴ Another explanation is that the AFE material has a “dead layer” analogous to the dead layers for ferroelectric films,²⁵ i.e., a low-permittivity, fully linear dielectric region in series with the AFE one so that an increase in E_f and a reduction in P_max would be expected with decreasing sample thickness.²⁵ The cross-sectional TEM inspection did not reveal the presence of an interfacial layer but the samples trend in accordance with this explanation. Finally, domain mobility is known to contribute to scaling effects on the coercive fields of ferroelectrics, where E_c ∝ $d−2/3$⁠,²⁶ but this scaling becomes less accurate once grains are small enough to only contain a single domain, around 100 nm.²⁷ 

For all of the samples, the coercive field is <25% of hafnia ferroelectrics.²⁸ The lower coercive field can reduce power requirements for devices like AFE field effect transistors (AFEFeT) that take advantage of the non-polar gate dielectric to reduce gate-induced drain leakage.²⁹ Unexpectedly, the 200 nm film had the highest leakage current density and lowest breakdown voltage [Fig. 2(b)]. The porosity observed in the 50 nm sample in TEM is expected in the 200 nm sample, which acts as a field concentration center.³⁰ The local concentrations of the electric field increase leakage current and provide nucleation sites for breakdown.

Compared to AFE-like hafnia materials, the AFE phase of PHO is the thermodynamically stable zero field state that may provide long-term stability advantages. In Fig. 2(c), the maximum polarization is shown up to 10⁷ cycles for the 20 nm PHO sample after the AFE has been stimulated by a square wave at 1.5 V and 10 kHz. AFE-like Al:HfO₂ change to the FE phase after 10⁵ cycles,³¹ while Si:(Hf,Zr)O₂ maintains the AFE-like phase past 10⁹ cycles,¹⁵ which is in agreement with observations of related antiferroelectrics.³² 

In order to get more insight into the material properties and the stability of the antiferroelectric switching behavior, the dielectric properties were investigated as a function of temperature up to 300 °C for the 50 nm PHO sample (Fig. 3). Figure 3(a) shows polarization as a function of field and temperature, where clear double hysteresis loops are observed up to elevated temperatures. The dielectric constant and loss tangent tan(δ) measured using a TF Analyzer 2000 from aixACCT at a frequency of 5 kHz with an oscillating voltage of 100 mV as a function of temperature are reported in Fig. 3(b). We observe a steady increase in the dielectric constant that reaches a peak around T = 185 °C before decreasing again, which matches the shape expected for an AFE transition. In addition, In order to determine an E–T phase diagram, we tentatively determined the critical electric fields as a function of T by taking the maximum of current (I) curves vs field. This was done by fitting the I(E) data with an asymmetric peak (of no particular physical significance), as illustrated in Fig. 3(c). Both the switching field E_f and the backswitching field E_a decrease gradually with increasing temperature. However, they [Fig. 3(d)] do not drop to zero at T_c, as it could have been expected from past examples of E–T phase diagrams³³ or phenomenological theory.³⁴ Instead, both values of E_f and E_a seem to reach a plateau above 200 °C, suggesting that these extracted fields above T_c might not have a physical meaning. Indeed, if the material had stabilized in a ferroelectric or antiferroelectric phase, this critical field should vary with temperature. Further work is needed to clarify the interpretation for this and check whether this field corresponds to a structural transition or not. Also remarkable is the fact that the dielectric data do not show any clear evidence for the intermediate phase known in bulk PHO. The maximum of permittivity falls in between the temperatures of the two transitions observed in bulk PHO around 160 °C and 210 °C.^35,36 It is possible that the difference between the two phases is not captured by these dielectric measurements, but it could also be that the intermediate phase is suppressed in thin films due to strain or other effects. Further structural studies will be needed to clarify this issue; first principle calculations could also help to evaluate the robustness of this intermediate phase against the strain expected in thin films.

We now discuss these properties in the perspective of various electrical applications. Figure 4(a) shows the ESD and storage efficiency for each film thickness. ESD increases dramatically at the AFE-FE transition field, reaching ∼8 J/cm³ at 450 kV/cm and doubling at dielectric breakdown at 2000 kV/cm. The storage efficiency is observed by the slimness of the hysteresis loop, or the difference in E_f − E_a, which has been shown to be reducible by doping lead lanthanum zirconated titanate (PLZT).³⁷ The state of the art for hafnia-based AFE-like materials is closer to 60 J/cm³ and 80% efficiency at electric fields of 4500 kV/cm.²⁸ Although the ESD is much higher, the lower electric field values merit examinations. Compared to the PHO ESD, it requires >2000 kV/cm vs 450 kV/cm for AFE-like HFO to reach the ∼8 J/cm³ storage density. While ESD is a useful figure for comparison, the driver of pulsed power applications is the energy, which will depend on the product of film geometry and ESD. Having higher energy storage densities at lower fields could be a benefit since the voltage will increase as the film thickness increases.

Capacitance density is an important metric for compact power conditioning applications. In Fig. 4(b), we show the relative permittivity of the 20 nm PHO sample measured using an HP 4192A network analyzer with an oscillating voltage and frequency of 50 mV and 10 kHz, respectively. The permittivity varies from ε = 200–600 at the phase transition and then <200 in the region of ferroelectric saturation. There are two implications derived from the permittivity data: first, the capacitance density of these films will be around ten times higher than hafnia films of an equivalent thickness and a permittivity of 20, and second, the films show a tunability of 221% within 1 V calculated from

PHO could have applications in low power microwave applications, where recently, simulations of hafnia ferroelectrics have shown promise taking advantage of the tunability over a small applied voltage range.³⁸ 

FE random access memory (FRAM), which relies on non-volatile positive and negative polarization states, has recently been reimagined by Pesic et al. for AFEs where the two states are poled and un-poled.³⁹ Both states are made accessible at 0 V by significantly imprinting the hysteresis loop such that the AFE+ → FE+ transition is at positive biases and the FE+→AFE + transition is at negative biases. This work is motivated by the generally superior fatigue performance of AFEs vs FEs. Figure 4(c) shows the polarization hysteresis for a symmetrical 20 nm Pt/PHO/Pt capacitor and the same material with an asymmetrical electrode configuration of TiN/PHO/Pt. A minor loop of +/−3 V taken with a DC bias of 1 V appears nearly identical to a ferroelectric polarization [Fig. 4(d)] but with non-polar and polar states.

In addition to potential imprint, the maximum polarization and field width of the AFE → FE hysteresis widened with the addition of a TiN electrode. It is challenging to de-convolute asymmetrical changes in polarization and coercive fields within the polarization hysteresis because the measurement is self-centering, and therefore, vertical shifts of unsaturated loops can appear as imprint. An assumption that the TiN generated an oxygen-depleted TiO_xN_y interface would increase E_f but not clearly increase polarization. PZO films of similar thicknesses grown epitaxially with pulsed-laser deposition show polarization values ∼45 µC/cm²⁴⁰ compared to PHO showing ∼30 µC/cm², so the increase in P_max associated with the TiN could be associated with a de-pinning of domains at the electrode interface. The leakage current reduced for the TiN device but gained a strong bias-dependent asymmetry, while the fatigue suffered, failing after 10⁵ cycles. Further work is needed to fully realize PHO films imprinted, centered on 0 V.

In summary, antiferroelectric PbHfO₃ films show a unique combination of properties, namely, transition fields, permittivity, and polarization compared to the similarly thick hafnia-based films. The structural and electrical analysis reveals both similarities and differences to the isomorphic PZO with an added benefit of more reliable ternary ALD growth from hafnia precursors vs zirconia.⁸ The unique crystallization route for PHO from ALD precursors may be exploited to lower the crystallization temperature. Further work should focus on process optimization to establish a minimum crystallization temperature, reduce porosity, and manipulate stress to tailor the E_f field for specific applications. The pyroelectric and electrocaloric performance of Pb-based AFEs is excellent,^41,42 so further investigation in PHO is warranted. The potential for electronic device integration, particularly through imprinting the AFE, also merits further investigation. Finally, the electric field–temperature phase diagram suggested by our study disagrees with previous observations on PHO ceramics where an intermediate phase is observed in the temperature dependence of permittivity and with x-ray diffraction, so film stress implications need to be better understood. PHO is a classic antiferroelectric identified over 60 years ago, which could realize new physics and applications as a thin film.

See the supplementary material for scanning transmission electron microscopy high-angle annular dark-field imaging of the thin film PbHfO₃.

The authors would like to acknowledge the Army Research Laboratory (ARL) Specialty Electronic Materials and Sensors Cleanroom (SEMASC) Nanofabrication Facility for their assistance with device fabrication. C.M.-B. and M.G. acknowledge financial support by the Luxembourg National Research Fund under Project No. BIAFET C16/MS/11348912/Guennou.

The data that support the findings of this study are available from the corresponding author upon reasonable request.