Comparative performance of fluorite-structured materials for nanosupercapacitor applications Open Access

Immediate remedies are essential to address the challenges posed by the exponential increase in energy consumption. In particular, pivotal technologies related to the fourth industrial revolution—such as the Internet of Things and Big Data—are witnessing an exponential surge in energy consumption linked to the storage, processing, and transmission of digital information.¹ Harnessing the potential of ferroelectric (FE) and antiferroelectric (AFE) materials compatible with complementary metal–oxide–semiconductor (CMOS) technology is a compelling strategy for the creation of energy-efficient electronic devices, more specifically for energy conversion applications.² 

The term “fluorite structure” denotes a prevalent pattern observed in compounds represented by the formula MX₂. In this arrangement, the X ions are situated in the eight tetrahedral interstitial sites, while the M ions occupy the regular sites within a face-centered cubic structure. This structural configuration is commonly observed in various compounds, notably the mineral fluorite (CaF₂), which gave its name to the structure. Typical fluorite-structured ferroelectrics and antiferroelectrics are, respectively, doped with hafnium oxide (HfO₂) and zirconium oxide (ZrO₂). HfO₂ has been introduced since 2007 by Intel as a high-k (high dielectric constant) in the gate stack of MOS (metal–oxide–semiconductor) transistors,³ while ZrO₂ is also widely used to fabricate DRAM cells.⁴ In 2011, a ferroelectric phase in Si-doped HfO₂ was first reported,^5,6 followed by the discovery of ferroelectricity in a solid solution of Hf_0.5Zr_0.5O₂ (HZO) the same year,⁷ paving the way for the re-introduction of ferroelectric materials in the existing CMOS technology.

At atmospheric pressure, bulk HfO₂ and ZrO₂ have centrosymmetric, non-polar crystal structures.⁸ However, under certain conditions of TiN encapsulation, doping, and/or film stress, it is possible to stabilize a meta-stable orthorhombic phase, which gives rise to ferroelectricity (FE) in HfO₂ and HfO₂–ZrO₂ thin film solid solutions.⁹ Contrary to antiferroelectricity originating from anti-parallel dipole moments, the origin of the functional AFE properties in ZrO₂ is attributed to an electric field induced non-polar to polar phase transition.¹⁰ The field-induced structural phase transition in ZrO₂ is attributed to a reversible non-polar tetragonal phase to a polar orthorhombic phase,¹¹ ensuring a double hysteresis polarization loop (PE loop) with an applied electric field.¹² 

FE and AFE nanosupercapacitors can be used for solid-state electrostatic energy storage,¹³ where high energy storage performances have been reported in ferroelectric HfO₂ or ZrO₂ films.^14–16 The field-induced transitions observed in AFE ZrO₂ hold promise for energy storage applications. However, several physical parameters of the fluorite films can limit their energy storage performances. Ferroelectric thin films exhibit a “wake-up” (WU) effect, which corresponds to an increase in the remnant polarization with cycling. This effect depends on the amplitude and frequency of the cyclic applied voltage stress.¹⁷ Another limiting factor is the film thickness scaling. Fluorite films have limited energy storage and scaling properties due to the increase in the monoclinic phase proportion at large thicknesses.¹⁸ 

Ferroelectrics (FE) excel at achieving high polarization, leading to high energy storage density (ESD). However, they have rather low efficiency, while linear dielectrics (LD) demonstrate remarkable efficiency but low polarization.¹⁹ In the context of prior research, it becomes evident that antiferroelectrics (AFE) offer an optimal balance, offering superior characteristics by combining elevated ESD and higher efficiency. Surprisingly, the exploration of these distinct attributes within the same material, thickness, and capacitor structures has been notably limited. In this context, this work aims to address this gap, employing ZrO₂ and HZO as prototype materials for a comprehensive investigation. This study proposes a performance assessment and comparison between AFE, FE, and LD fluorites for nanosupercapacitor applications. Fluorite thin films were grown with the same chemical composition but with different deposition techniques and parameters, leading to different nanosupercapacitor electrical properties, from LD to FE and AFE. AFE fluorite thin films exhibit beyond state-of-the-art energy storage capabilities.

Capacitors were fabricated on p-doped Si (001) substrates and followed the stack Pt/TiN/oxide/TiN/Si. Details of the deposition processes for oxides are summarized in Table I. Sputtering is performed via an AC450 magnetron sputtering chamber from Alliance Concept, while Plasma-Enhanced Atomic Layer Deposition (PEALD) is carried out by a Fiji F200 apparatus from Ultratech. The first step of ZrO₂ PEALD deposition consists of the application of an O₂ plasma to the TiN bottom electrode at 300 W before opening the valve of the TDMA-Zr precursor. Alternation of a complex sequence, notably including the TDMA-Zr valve opening and the dioxygen plasma, is then performed to grow the AFE ZrO₂ thin film. Sputtering is performed for TiN and Pt using a metallic target of Ti and Pt. Pt/TiN top electrodes are obtained after a photolithography and lift-off process. Rapid thermal annealing (RTA) was then performed for all samples. Samples were then investigated by means of physical and electrical characterization.

Glancing Incidence X-Ray Diffraction (GIXRD) was performed on a Smartlab Rigaku diffractometer using a 9 kW copper rotating anode, a parabolic multi-layer mirror for parallel beam setting, a Ni filter for CuK_α radiation selection, a 0.114° aperture parallel slit analyzer, and a 0D scintillating counter. The thickness of all thin films was measured by X-Ray Reflectivity (XRR) with the same instrument.

Electrical characterization was carried out on 50 and 20 μm diameter capacitors using a probe station in a Faraday cage and a setup composed of a Keithley 4200SCS equipped with PMU. Endurance tests were performed using a custom program interfaced with the Keithley. The cycling sequence consists of bipolar voltage square pulses (commonly called the set/reset sequence) until breakdown at 3.5–4.5 V, depending on the film thickness, with a 20 µs pulse duration. Polarization as a function of electric field (P-E) curves is established from the measurement sequence consisting of three triangle pulses (DHM: Dynamic Hysteresis Measurement) with a 60 µs rise time. The first voltage pulse poles the polarization in a given pre-set direction, while the two other pulses measure the current response. The pulse amplitude is set to obtain a 3.5 MV/cm electrical field for each sample and for all figures except Fig. 5, where a maximum electrical field of 4.0 MV/cm is applied in order to maximize the ESD of the AFE ZrO₂.

The field induced electrical properties of the fluorite thin film capacitors grown by sputtering and ALD were first examined. Figure 1 shows the electrical characteristics of the capacitors, comparing FE and AFE samples with LD ones of the same chemical composition at the same applied electric field of 3 MV/cm. In Fig. 1, the FE, AFE, and LD properties of HZO and ZrO₂ are shown after 10³ cycles. Polarization vs electrical field (solid curves) and current vs voltage (dashed curves) are systematically shown for each studied sample.

For LD samples in Figs. 1(c) and 1(d), as the dielectric capacitor is charging or discharging, the current is different from zero, leading to a non-zero electric displacement field for a non-zero electric field. Therefore, a linear relationship between polarization and the applied electric field is expected. Due to leakage currents, the P-E loop is not totally closed and will lead to an efficiency of the energy storage close but less than 100% (the theoretical value for a perfect LD).

For the FE HZO in Fig. 1(a), after 10³ cycles, two peaks can be observed: one at positive voltages and one at negative voltages. These peaks correspond to polarization switching peaks due to the FE nature of the film, which is attributed to the displacement of oxygen ions in HZO.²⁰ The remnant polarization of the FE HZO is 23 μC/cm² on the positive side and 24 μC/cm² on the negative side. The small asymmetry is attributed to the possible oxidation state of the top²¹ or bottom²² TiN electrode.

Finally, for the AFE ZrO₂ sample, a state-of-the-art curve is observed after 10³ cycles. A threshold field of about 1.0 MV/cm can be seen between an LD and FE behavior, corresponding to the field induced phase transition assumed for ZrO₂.^11,23 A very sharp linear opening is present below 1.0 MV/cm, followed by a narrow hysteresis loop above, with a saturation polarization P_s as high as 20.5 μC/cm².

Structural characterization measurements are then conducted on each sample to elucidate the origin of the functional properties, allowing for energy storage. Figure 2 shows the XRD scans for all FE, AFE, and LD samples. The FE HZO sample exhibits a distinct orthorhombic/tetragonal (o/t) peak with (111) orientation for the o-phase and (101) orientation for the t-phase around a 2θ value of 30.5°. The non-centrosymmetric o-phase is typically considered the phase responsible for ferroelectricity. However, it also shows the presence of the monoclinic (m-) phase, which is centrosymmetric. The mixture of o/t phases is attributed to HZO, while only the (101)-oriented t-phase is attributed to ZrO₂ for the peak around 30.5°, considering the current literature explanations.^23,24

The properties of HZO thin films (∼10 nm thick), synthesized via reactive magnetron sputtering from a Hf/Zr metallic target^22,25 on a TiN layer, were explored. GIXRD measurements revealed that, depending on the deposition working pressure in the chamber—low pressure (LP) (5 × 10⁻³ mbar) or high pressure (HP) (5 × 10⁻² mbar)—the thin films were either monoclinic or amorphous after deposition. The ZrO₂ films grown by ALD⁷ are amorphous after deposition. Amorphous samples exhibit the o/t peak after Rapid Thermal Annealing (RTA), while monoclinic samples remain in their monoclinic structure after RTA.

For the HfO₂/ZrO₂ ceramic target and the Zr metallic target used in this study (respectively, non-reactive and reactive magnetron sputtering), regardless of the pressure, the HZO and ZrO₂ thin films are amorphous after deposition. After RTA, films obtained at both pressures exhibit the o/t peak, but only the HZO HP samples show some monoclinic peaks, whereas LP samples only show the o/t peak (not shown in this paper). It was demonstrated that HZO LP samples are only tetragonal,²⁶ while HP samples show a mixture of tetragonal and orthorhombic phases (in addition to the monoclinic phase for HP samples).

While all samples present the o/t peak in their GIXRD scans, their electrical behaviors are vastly different. Our observations tend to show that observing a peak around 30.5° is a necessary condition to have FE or AFE properties, but it is not a sufficient condition. For HZO samples, the FE signature is given by the presence of the (111)_o peak, which overlaps with the (101)_t peak. Their respective contributions can only be evaluated by fitting the peaks. For ZrO₂ samples, the (101)_t peak is not enough to discriminate between LD and AFE. Structural measurements alone are generally insufficient to clearly identify the electrical nature of the thin films. Therefore, electrical characterizations in Fig. 1 are needed to conclude about the electric nature of the HZO and ZrO₂ capacitors.

Endurance tests were also performed, and a wake-up effect was observed. Historically, the first observation of a wake-up effect dates back to Sim et al.²⁷ However, in the Sim et al. article, the increase in P_r seems to be attributable to the increase in leakage current (fatigue phenomena) with increasing cycling counts. There is no evidence that this effect could be similar to that observed for HfO₂. In 2011, Wu et al.²⁸ continued the work of Sim et al. They coined, for the first time, the increase in Pr with the number of cycles as the “wake-up” effect. However, here again, this so-called “wake-up” effect can be attributed to the increase in leaks and phenomena of modification of space charge. In the same article, it is also noted that if the frequency decreases, P_r increases. In 2012, finally, Mueller et al.²⁹ released the first article discussing the “wake-up” effect on Si:HfO₂. They named this effect the wake-up effect in reference to the endurance test procedure carried out by Wu et al. and then spoke of a “wake-up procedure” and not a “wake-up effect.” In 2013, Zhou et al.¹⁷ were the first to truly study the wake-up (WU) effect and to name it as such. This time, the argument of increased leaks is no longer mentioned, although this is not conclusively proven in this article. Thanks to the electrical characterization PUND technique, it was already observed in previous studies that the WU effect does not result from an increase in leakage currents.³⁰ Moreover, Zhou et al. observe that as the measurement frequency increases, Pr decreases, while it increases with increasing pulse voltage amplitudes. It will be shown later that a WU effect is also present for AFE ZrO₂, and detailed observation of the wake-up and endurance properties of the analyzed films is present in the supplementary material. This WU effect, although almost identical from the perspective of current shifts along cycling in FE HZO and AFE ZrO₂, cannot be defined as an increase in P_r.

Figures 3(a) and 3(b) depict current vs voltage (I–V) curves for FE HZO and AFE ZrO₂, respectively, whereas Figs. 3(c) and 3(d) illustrate P-E loops for the same samples. Dashed lines represent the behavior of the samples in their pristine state, while solid lines indicate their behavior after 10³ endurance cycles. At the pristine stage, FE HZO exhibits four switching current peaks: one pair at positive voltages and another one at negative voltages [dashed lines in Fig. 3(a)]. Along cycling, each pair would progressively merge [solid lines Fig. 3(a)], leading to the hysteresis loop in Fig. 1(c) (solid lines). For simplicity, observed peaks at the pristine stage for the FE HZO will be referred to as “double peak” or “double peak phenomenon” from now on. For positive voltages, the left peak shifts toward higher voltage values, whereas the right peak shifts toward lower voltage values, as shown by the red arrows. In terms of FE domains, this implies that certain FE domains are switching at lower voltage levels, while others are switching at higher values. With an increasing number of cycles, low-voltage switching domains transition to higher voltage values, while high-voltage switching domains transition to lower values, eventually resulting in the merging of the two peaks. In FE HZO thin films grown by sputtering, this effect has already been well described.^30,31

Although the WU effect is rarely mentioned for AFE, we observed a similar current peak displacement for ZrO₂ than for FE HZO, as highlighted by the red arrows in Figs. 3(a) and 3(b). In contrast to FE HZO, the pristine values for AFE ZrO₂ exhibit different signs. It has to be mentioned that P_r does not apply for AFE since around 0 V AFE is showing the same behavior as LD. Nevertheless, in Fig. 3(d), the two hysteresis loops of the AFE PE curves have smaller coercive fields and a higher P_s between pristine and woken states, which will lead to an increase in the ESD, as P_s increases and loss decreases.

This intriguing similarity between the double peak phenomenon in FE and the AFE behavior has already been discussed.¹¹ The non-uniform distribution of the internal electric field is likely attributed to unevenly distributed charged defects, such as oxygen vacancies, particularly near the electrodes. This asymmetry in oxygen vacancy concentration, often induced by the reduction of the doped HfO₂ layer by metal nitride electrodes, is a potential source of the internal field in the pristine material. The non-uniform distribution of oxygen vacancies near the electrodes may create an asymmetric internal field. In the process of electric field cycling, oxygen vacancies might diffuse into the bulk regions of fluorite-based films, triggering the wake-up process and resulting in the merging of switching current peaks in the case of FE HZO. Subsequent investigations have reported a redistribution of charges associated with oxygen vacancies.^32,33 Another plausible mechanism for the wake-up effect involves field-cycling-induced phase transitions.³⁴ Lomenzo et al.¹² initially proposed that the transition from the tetragonal (t-) to the ferroelectric orthorhombic phase (o-phase) underlies the wake-up effect. They observed a decrease in dielectric permittivity and an increase in P_r with an increasing number of electric field cycles, possibly indicating a phase transition from a non-ferroelectric phase with higher permittivity to a ferroelectric phase with lower permittivity. In addition, Grimley et al. employed scanning transmission electron microscopy (STEM) and impedance spectroscopy to observe a phase transition from monoclinic (m-) to o-phase.³⁵ In essence, phase change and the redistribution of defects can induce the pinning of domains, leading to WU in both FE and AFE.^36,37

Despite the fact that some authors have considered that the double peak is not similar to AFE phenomena,¹¹ the double peak on the positive voltage side can be considered to result from interactions between some negatively charged regions and positively charged regions or screened regions that define the switching current at the pristine stage. In addition, after cycling, domains tend to homogenize, similar to the observed phenomena for both FE HZO and AFE ZrO₂. Further investigations into the microstructure and oxygen vacancy re-organization would help to clearly determine if both phenomena have the same origin or not.

These calculations for ESD, loss, and η are now standard performance indicators for fluorite-based capacitors.^13,39

Figure 4 shows the ESD and η as a function of the number of cycles for an applied field of 3.5 MV/cm on the four analyzed samples and also for ZrO₂ cycled at 4 MV/cm. Contrary to the HZO FE layer, the breakdown field of the ZrO₂ AFE layer is higher, allowing it to display the film properties at 4 MV/cm. As expected, in Fig. 4(a), one can observe the very high energy density of AFE ZrO₂ compared to the other samples. For LD HZO and ZrO₂, ESD is very low due to the low values of polarization when applying an electric field. However, the samples have better endurance properties as they do not experience a breakdown at 10⁷ cycles, contrary to FE HZO and AFE ZrO₂. For FE HZO, at the early stage of cycling, the capacitor is very similar to AFE ZrO₂ because of the double peak phenomenon. However, as the number of cycles increases, switching peaks start to merge, and the ESD is, therefore increasing, leading to a higher ESD for FE HZO than the one for AFE ZrO₂ at 10³ cycles.

In Fig. 4(b), as one could expect, the most efficient samples are the LD ones. One can observe that η is actually not totally equal to 100% because of leakage currents and has the lowest ESD of all samples. As the hysteresis of FE HZO is wide open, it has the highest losses; hence, it shows the lowest η of the four samples at ≈55%. At the same time, the high P_s of the FE sample makes it reach the highest ESD value at 95 J/cm³. Finally, the reason why AFE is considered a better option for supercapacitors compared to simple FE and LD inorganic electrostatic capacitors is that their efficiency falls in between FE and LD, with an η of 75%, while their ESD is almost as high as that of FE samples, reaching 52 J/cm³ at 3.5 MV/cm. This value can be further increased up to 84 J/cm³ at a higher electrical field (4.0 MV/cm).

The current literature for nanosupercapacitors using FE, AFE, but also relaxor-ferroelectric (RFE) hafnium- and zirconium-based fluorite materials is compared with our results in Fig. 5. One can observe that only a few papers show higher ESD and η than our present work. Moreover, FE HZO is also showing excellent properties for nanosupercapacitor applications compared to what was previously observed for similar FE materials.

A limiting factor to further improve the ESD and efficiency in thin films is the FE and AFE film thickness scaling. Fluorite films have limited energy storage and scaling properties due to the increase in the monoclinic phase proportion at large thicknesses.¹⁸ However, the ESD achieved for a thin film can be significantly enhanced by transitioning to a multilayered and three-dimensional (3D) structure,¹⁵ an aspect that can be explored in future studies. This transition holds the potential to elevate the ESD by several orders of magnitude, promising new avenues for enhanced energy storage capabilities. Investigations into multilayered and 3D architectures thus represent an exciting frontier in the quest for optimizing energy storage efficiency, offering prospects for groundbreaking advancements in the field.

We investigated the potential of ferroelectric (FE) and antiferroelectric (AFE) fluorite-structured materials, such as hafnium oxide (HfO₂) and zirconium oxide (ZrO₂), for energy-efficient applications, addressing the urgent need to curb the soaring energy demands of the digital age. By integrating these materials into existing CMOS technology, it demonstrates a forward-looking approach to enhancing electronic device efficiency through advanced energy conversion mechanisms. The research provides a comprehensive analysis of the structural and electrical properties of HZO and ZrO₂ thin films, showcasing their significant potential in solid-state electrostatic energy storage.

Moreover, the study compares the energy storage performances of FE, AFE, and LD samples, underscoring the superior energy storage density (ESD) and efficiency of AFE materials. This finding is critical, as it highlights the promise of AFE ZrO₂ in energy storage applications, offering a balanced trade-off between high ESD and efficiency. The meticulous methodology, from synthesis to characterization, provides a robust framework for assessing the capabilities of these materials and sets a benchmark for future studies in the field.

See the supplementary material for an explanation of the applied electric field in the capacitors and polarization plots in the function of the applied electric field at each measured cycle on all capacitors.

This work was carried out on the NanoLyon technology platform and implemented within the NanOx4EStor project. We would like to specifically thank Céline Chevalier, Giovanni Alaimo-Galli, and Jean-Charles Roux for their implication on the research project at the NanoLyon platform. This NanOx4EStor project has received funding under the Joint Call 2021 of the M-ERA.NET3, an ERA-NET Cofund supported by the European Union’s Horizon 2020 research and innovation program under Grant Agreement No. 958174. This work was supported by the Portuguese Foundation for Science and Technology (FCT) in the framework of M-ERA.NET NanOx4EStor Contract No. M-ERA-NET3/0003/2021; by the Executive Agency for Higher Education, Research, Development, and Innovation Funding (UEFISCDI); and by the Agence Nationale de la Recherche (ANR) under Contract No. ANR-22-MER3-0004-01.

The authors have no conflicts to disclose.

Grégoire Magagnin: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal). Jordan Bouaziz: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal). Martine Le Berre: Data curation (equal); Investigation (equal); Writing – review & editing (equal). Sara Gonzalez: Data curation (equal); Investigation (equal); Writing – review & editing (equal). Damien Deleruyelle: Project administration (equal); Resources (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). Bertrand Vilquin: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Resources (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal).

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