HfO2‐Based Ferroelectrics Applications in Nanoelectronics

This article is dedicated to HfO2‐based ferroelectrics applications in nanoelectronics, especially to topics not well developed up to now, such as microwaves, energy harvesting, and neuromorphic devices working as artificial neurons and synapses. Other well‐covered topics in the literature, such as memories or negative‐capacitance ferroelectric field‐effect transistors, will be only briefly mentioned. The main impact of HfO2‐based ferroelectrics is the possibility of using them for fabricating at the wafer‐level complementary metal oxide semiconductor (CMOS) compatible high‐frequency devices, such as phase‐shifters, antenna arrays, or filters with a high degree of tunability and miniaturization, as well as energy harvesting devices and neuromorphic key components. In addition, the recent transfer of 2D materials on HfO2 ferroelectrics has demonstrated new physical effects, such as opening a 0.2 eV bandgap in graphene monolayers, and allows the manufacture of very high‐mobility field‐effect transistors (FETs) based on graphene/HfZrO.


Introduction
Hafnium oxide (HfO2) is the dielectric commonly used in field effect transistors (FETs) fabricated in CMOS technologies and, as such, retrieved in any very large-scale integrated (VLSI) circuits.
Examples of such circuits include microprocessors or complex analog integrated circuits found in daily applications, for instance in computers, smartphones, laptops, or incorporated in cars, airplanes or medical equipments.Since the discovery of HfO2 as a dielectric for modern integrated circuits, increasing its permittivity was a leading research issue in itself because higher permittivity values allow decreasing the effective oxide thickness (EOT), and so the leakage current; this is especially important when billions of Si FETs are integrated on a single chip.The solution found to boost the permittivity of HfO2 is represented by structural phase transformations [1].HfO2 has its lowest electrical permittivity (εr ≈ 18) in its stable phase, which is the monoclinic phase.Higher permittivity values, of εr ≈ 27 in the cubic phase and εr ≈ 70 in the tetragonal phase, are achieved, nonetheless, in phases that are stable only at very high temperatures, exceeding 1700 o C. The decrease of these temperatures is possible, however, by HfO2 doping.Thus, the cubic and the tetragonal phases were first obtained at lower temperatures via Y doping [1] and, respectively, Si doping [2], while a stable HfO2 tetragonal phase at nearly room temperature was observed by growing HfO2 via atomic layer deposition (ALD) and doping it with ZrO2 [3], [4].As described in the recent review [5], the stabilization of these phases is determined by the concentration of dopants, the annealing temperature, and the growing methods.
These initial investigations focused on increasing the permittivity of HfO2 lead eventually to the discovery of the orthorhombic phase of HfO2, which was associated with HfO2 ferroelectricity [6], [7].This phase is obtained under different growth conditions in the presence of dopants, including those mentioned above, and requires strict control of dopant concentration, temperature, and mechanical strain.The most used orthorhombic HfO2-based ferroelectric, named further as HfZrO, is doped with Zr and is usually grown at the wafer level, up to 300 mm, using ALD.It thus

Accepted Article
benefits from the state-of-the-art CMOS technology.More recently, using Zr doping and pulsed laser deposition (PLD) another, rhombohedral phase of HfO2 was discovered to be also ferroelectric [7].The rhombohedral or orthorhombic HfZrO are ferroelectrics with similar electric performances, the origin of stable ferroelectricity in both cases being due to the induced strain between the top and/or down substrates sandwiching HfZrO [8].
Doping or strain application modifies the atomic structure of a material and can induce a phase transition, in particular can induce ferroelectricity in HfO2.Similarly, HfO2 can transform into a weak magnetic material [9], [10] due to high oxygen deficiencies or other dopants, whereas an atomically thin semiconductor, MoS2, can becoming ferroelectric by applying mechanical strain with the help of an atomic force microscope (AFM) tip.This last situation will be detailed in the next chapter.
The bottom and top electrodes encapsulating HfZrO play a crucial role in maintaining the stability of ferroelectricity in this material.A common configuration is Si/TiN/HfZrO/TiN, in which ferroelectricity is preserved by the tensile strain and thermal expansion mismatch (see [11] for details).HfZrO is ferroelectric when its thickness is, typically, 4-6 nm, its ferroelectricity being generally lost for thicknesses larger than 10 nm.However, it was demonstrated that very thick HfO2, reaching 1 m, doped with 7% Y is still ferroelectric when grown by PLD on a (111) Pt/TiOx/SiO2/(001)Si substrate [12].On the contrary, it was very recently shown that HfZrO with a thickness of only 1 nm is ferroelectric when grown by ALD directly on Si/SiO2, without the wakeup effects encountered in the Si/TiN/HfZrO/TiN structures [13].We have also adopted the growth of HfZrO directly on Si starting from 2017 [14].
A possible explanation of ferroelectricity in HfZrO grown directly on Si or Si/SiO2 is the occurrence of a novel ferroelectric paracrystalline phase due to oxygen vacancies engineering during the ALD growth directly on Si.This particular paracrystalline phase is defined as having short-and medium-range lattice ordering, but lacking crystal-like long-range ordering, and is assumed to form by the interplay of a new rhombohedral polar R3m ferroelectric phase [8] Accepted Article (evidenced typically at 30.5 o in GIXRD measurements) with a minority orthorhombic Pca21 phase.
In Fig. 1 we have represented the GIXRD pattern of the HfZrO thin film on Si(100), indicating the rhombohedral phase.Of course, more structural characterizations are necessary to validate this explanation.

(b)).
There are intensive researches regarding HfO2-based ferroelectrics, such as those previously reported in the recent reviews [15], [16], [17], [18], [19], [20] which address applications of HfO2based ferroelectric memories [17], or negative capacitance ferroelectric FETs [18], [19], [20].Therefore, we will not focus further on these issues but will concentrate in this review on the applications of HfO2-based ferroelectrics in nanoelectronics, especially in microwaves, energy harvesting, and computing.Each application will start by highlighting the relevant physical properties of HfO2 in that particular case.Nonetheless, there are some basic physical properties and Accepted Article related parameters of HfO2-based ferroelectrics, such as HfZrO, which are important for many applications, such that we list them from the beginning, in Table I.

Accepted Article
This article is protected by copyright.All rights reserved 6 Note that all these 2D ferroelectrics reported in [23] and [24] are not CMOS compatible and are not grown at the wafer-scale until now, their ferroelectricity being evidenced either in-plane or out-ofplane depending on the material.On the other hand, HfO2-based ferroelectrics are not van der Walls materials but are, however, CMOS compatible atomically-thin ferroelectrics and can be grown at wafer scale.We have shown recently that the 2D semiconductor MoS2 (see Fig. 2), grown at the wafer-scale and having a thickness of 4.6 nm (7 monolayers), is also ferroelectric due to the applied strain [25].
To demonstrate the ferroelectricity of the ultrathin MoS2, PFM spectroscopy curves were performed for various values of the AC voltage amplitude applied to the tip.Each plot of amplitude and phase was recorded by scanning the DC voltage between 10 V and + 10 V in both directions, at various points of the MoS2 surface.Starting from an AC voltage amplitude of around 0.5 V (see Fig. 3), we observe that both amplitude and phase have a hysteretic behavior with a phase shift around 200 degrees, indicating the onset of ferroelectricity via strain.Beyond 0.8 V, the phase shift becomes 360 degrees and the ferroelectricity is lost.

Accepted Article
Fig. 3 Piezo-Force amplitude and phase curves obtained for 0.5 V AC voltage applied to the tip during PFM spectroscopy [25].
Future work is needed to investigate this material, especially regarding the suitable doping route of MoS2 semiconductor and for identifying suitable capping layers in order for MoS2 to become a stable ferroelectric with atomic thickness that can be grown at the wafer scale.

Microwave devices based on HfZrO
The tunability of devices and circuits at high frequencies is a prerequisite for further developments of radars and modern communications, such as 5G, internet, and Internet-of-Things (IoT), as well as for rapid changes in the environment properties or demands for various applications.The tunability is implemented at high frequencies using specific materials, in particular ferroelectrics [26].However, in the case of the first ferroelectrics used in this frequency range, namely perovskite ferroelectrics, such as Pb(Zr,Ti)O3 or PZT, the DC voltages necessary for tuning the frequency or the phase of microwave devices is very high, of tens and even hundreds of Volts.In addition, because perovskite ferroelectrics are not CMOS compatible and are difficult to grow at large wafer scales, the ferroelectric microwave devices were almost abandoned.Therefore, since HfZrO is

Accepted Article
CMOS compatible and only a few Volts are necessary to tune a microwave device based on it, has a great potential for tunable microwave devices.
The most important property of any ferroelectric in microwave applications is the electrical permittivity dependence on the applied voltage at high frequencies.We have measured the effective permittivity and wave propagation constant of HfZrO in the frequency range 1-14 GHz using two coplanar lines (CPWs) with different lengths.We observed an increase of 27% of the effective permittivity when the applied voltage varies from 0 V to 5 V [27] and have also measured the attenuation constant, phase constant, and loss tangent of HfZrO up to 14 GHz.The fabrication of HfZrO CPWs implies the ALD growth of HfZrO with a thickness of 6 nm on a high-resistivity Si (HR Si) wafer with a thickness of 525 μm, followed by metal deposition of metallic CPWs using ebeam techniques.Various structural characterization techniques used to validate the ferroelectricity of HfZrO are presented in [27].The CPW metallization consists of a 50 nm/500 nm thick Ti/Au layer, the width of both CPWs is W = 1 mm, whereas the lengths of the two CPWs are LA = 6.671 mm and LB = 2.393 mm.The CPW lines are depicted in Fig. 4.
We have measured the S parameters of the CPWs, i.e., the return loss (S11 and S22) and transmission (S21 and S12), and we have extracted the effective permittivity using the Bianco-Parodi formula.The results are depicted in Fig. 5.The measurements are performed at room temperature

Accepted Article
on-wafer using a vector network analyzer (VNA) Anritsu-37397D connected to a Karl-Suss PM5 probe station, while DC polarization is applied using the internal bias tees of the VNA.
Fig. 5 The dependence of the effective permittivity of HfZrO CPW on frequency and that of effective permittivity versus relative permittivity obtained by a fitting procedure at 10 GHz [27].
From Fig. 5 it can be seen that at 2.45 GHz the relative effective permittivity is 7.61 at 0 V and 9.64 at 5 V, so the change in the DC voltage induces a change of 27% of the effective permittivity, while at 10 GHz the effective permittivity is of 6.79 at 0 V and 7.89 at 5 V, so that the increase of 27% of this parameter is almost constant within the range 2-10 GHz.The relative permittivity values reported in the literature were obtained using other methods [11].We have found an empirical formula for the dependence of the relative permittivity of HfZrO in the microwave range depending on the microwave effective permittivity at 0 V, ) 0 ( eff  , and the substrate permittivity is the coercive field, and t is the thickness of HfZrO.
We have then calculated the loss tangent tan  at various frequencies and DC voltages using the microwave data.The results, illustrated in Fig. 6, show low values of tan  and thus moderate losses of HfZrO in the microwave domain.
The phase shifter is a key device in phased antennas arrays.Ferroelectric phase shifters are tunable via applied DC voltages, with values of few V in the case of ultrathin HfZrO.We have shown that in a 1-mm-long gold CPW interdigitated structure deposited over 6 nm HfZrO grown by ALD directly on a high-resistivity Si substrate, the phase shift in microwaves is higher than 60 o at 1 GHz and of 13 o at 10 GHz at the maximum applied DC voltages, of  3 V [14] (see Fig. 7).
Hundreds of phase shifters were fabricated on the same wafer, all showing similar characteristics.The bias is applied vertically on the CPW, such that HfZrO is polarized vertically.
The phase shift of the microwave signal in the frequency domain 1-10 GHz is represented in Fig. 8 at various voltages, in the range 3 V. We defined the phase shift as , i.e., as the difference between the phase at a certain DC voltage and that at 0 V for a given frequency value.
Figure 8 shows that beyond  3 V the phase shift does no longer increase with the bias since the field becomes higher than the coercive electric field.
The capacitance of the phase shifter is defined as the input impedance.The capacitance decreases with frequency and increases with the applied voltage, increasing 4 times in the range 1-3 GHz.This behaviour explains the large phase shift, higher than 60 0 /mm, obtained in this frequency range.Fig. 8 The phase shift of HfZrO phase shifter [14].
The next step in extending the phase shifter application is to build an antenna array based on the above results.We have designed and fabricated at the wafer scale a phased array formed from Accepted Article two gold antenna patches, phase shifters, and additional circuitry, all integrated on HfZrO/highresistivity Si 4-inch wafer (see Fig. 9) [28].We have observed that the radiation beam at 2.55 GHz is steered by 25 o when HfZrO is biased with 1 V, meaning that the phased antenna array can be biased with a simple battery.The phase shifters in Fig. 9 are the two interdigitated structures exciting the antennas.
On the same wafer with the antenna array, we have designed and fabricated another configuration including also coplanar lines and other phase shifters to verify separately the phaseshifting effect, i.e., to check if the microwave signal is shifted when a DC voltage is applied.
Although the results are not different from those reported for the HfZrO phase shifter alone, a more careful design lead to a maximum phase shift of 53.74 o at 2.55 GHz when sweeping the DC voltage between -1 V and + 1V.We have measured the radiation pattern of the configuration in Fig. 9 in the range -30 o and +30 o , with steps of 5 o , using as receiver a spectrum analyzer (Anritsu MS2668C 9 kHz-40 GHz) connected to a commercial 16 dBi gain reference antenna located on a rotating arm, while the HfZrO-based antenna array was placed at 1.2 m from the reference antenna (see Fig. 10).

Accepted Article
Fig. 10 The measurement set-up of the HfZrO antenna array [28].
We have found that the gain has a maximum value of 5.36 dBi.The steered radiated field is due to applying different DC biases on the two arms, which confer a maximum phase shift.In particular, after applying 0 V on both antennas, then -1 V to one patch and +1 V to the other one, we have observed that the radiation pattern is steered with 25 o at 2.55 GHz.
Very recently, we have shown that microwave filters based on a 6-nm-thick HfZrO layer are tunable in the range 0.  From the above figure it follows that all structures contains separate inductors and capacitors that must be tested before the measurement of the filters.The interdigitated capacitance, CIDC, and phase, ϕIDC, in the CPW configuration versus the applied DC voltage for the frequencies of 2.5 GHz, 5.5 GHz, and 10 GHz are represented in Fig. 12.At 2.5 GHz, the capacitance increases from 0.38 pF to 0.58 pF as the DC voltage is swept from 0.5 V up to 4 V, whereas during the same DC voltage sweep the phase decreases from 48.7° to 34.7°, resulting thus a phase shift of 14°.By increasing the frequency to 10 GHz and by tuning the DC voltage in the same range, the capacitance increases from 0.28 pF to 0.33 pF, while the phase decreases from 11.76° to 5.86°, i.e., the obtained phase shift is now of about 6°.This microwave measurement demonstrates that HfZrO is acting as ferroelectric at very high frequencies, including the X band.The frequency shifts (f) of the three types of HfZrO filters is shown in Fig. 13.In the case of LPF, the maximum f is 4.5 GHz, while this parameter is 1 GHz for HPF and 2.5 GHz for the BPF.The frequency shift is caused by the HfZrO ferroelectric thin film, since the change of maximum 8% of the effective permittivity, corresponding to 65% variation of the relative permittivity via formula (1) produces a phase shift of the microwave signal propagating through the filter, while the impedance is still matched to 50 Ω.The most significant results are obtained for the HPF. Figure 14 displays the frequency dependence of the transmission coefficient at 3 V, showing that the rejection with respect to the central frequency of 10 GHz, is about 8.7 dB at 6 GHz, 13.2 dB at 4 GHz, and 24 dB at 2 GHz.We note here that all the filters have much smaller dimensions than the wavelengths of working frequencies.For example, the BPFthe largest filterhas dimensions of 3.19 mm × 3.47 mm.

2D materials/ HfZrO heterostructures
All ferroelectrics, including HfO2-based ferroelectrics grown via ALD, have a very small rootmean-square (RMS) surface roughness and are probably among the best substrates on which atomically thin materials can be transferred.In particular, HfZrO has a roughness of 0.2-0.3nm [30], while this parameter is 0.4-0.6 nm in the case of hexagonal boron-nitride [31] and is bigger, around 1 nm, for SiO2 [32].Moreover, HfZrO is not only among the best substrates to transfer on graphene monolayers grown by CVD at the wafer scale, but, as we have shown very recently theoretically [33] and experimentally [34], HfZrO is opening a bandgap in graphene, conferring switching properties to FETs based on graphene monolayers.This implies that these FETs can be in well defined on and off states, paving the way for using graphene FETs in very fast logical circuits, since graphene on HfZrO has also a very large mobility, of 7900 cm 2 /Vs at room temperature .The bandgap opening in graphene transferred on a HfZrO substrate, of around 0.2-0.4eV, is due to orbital hybridization and local deformation of the graphene lattice.In all cases described below, the graphene monolayer is transferred by Graphenea on 4-inch wafers of HfZrO/Al2O3/Si or HfZrO/SiO2/Si, where HfZrO has a thickness of 6 nm.The fabrication of these structures implies several steps: (i) graphene channel patterning by e-beam lithography and reactive ion etching (RIE) (see Fig. 15(a)), (ii) source (S) and drain (D) patterning, metallization, and liftoff (see Fig. 15(b)); the channel length of the FET is 600 nm and has a width of 1 m, (iii) gate (G) insulator deposition and patterning with 40-nm-thick HSQ (hydrogen silsesquioxane), as shown in Fig. 15(c), and (iv) metallic gate deposition (see Fig. 15(d)).More details about structural characterization and fabrication at the wafer level are found in [34].The scanning electron microscopy (SEM) image of the graphene/HfZrO/SiO2/Si FET is displayed in Fig. 16.

Accepted Article
The drain current (ID) versus drain voltage (VD) at various gate voltages (VG) for the structure in Fig. 16 is represented in Fig. 17 with corresponding error bars.When the gate voltage is increased beyond few volts, the drain current is decreasing until at VG = 10 V the transistor reaches its off state; the leakage drain current is of 1 A at a drain voltage of 3 V.

Accepted Article
room temperature is about 0.18 eV.This bandgap value is comparable with that predicted by firstprinciple studies [33].
Based on this research, we have shown that graphene/HfZrO FETs are memtransitors if the graphene channel is perforated with nanoholes having a diameter of 20 nm [35] or has as channel a graphene monolayer transferred on the 3-layers ferroelectric structure HfO2/Ge-HfO2/HfO2 [36].In structures, where the insulator is an oxide, such as TiOx, HfOx, or a 2D material [37], [38], [39].
Memtransistors, on the other hand, are active three-terminal devices based on FETs, having clear advantages over two-terminal memristors, such as: (i) no electroforming process for the wake-up,

Accepted Article
(ii) tunable memory functions, (iii) high on-off ratios, and (iv) no sneak paths in crossbar arrays.
Any memtransistor is an artificial synapse.
The first memtransistors, fabricated on flakes of MoS2, worked initially at a drain voltage of 80 V [40], which was reduced progressively up to 12 V using different methods [41], [42], but remained still too high to be used in neuromorphic applications.In contrast, the ferroelectric memtransistors based on the HfO2/Ge-HfO2/HfO2 3-layers-structure [36] are working at drain voltages in the range of -2 to +2 V and are fabricated at the wafer level.In Fig. 19 we present the cross-section of the HfO2/Ge-HfO2/HfO2 FET memtransistor and in Fig. 20  backgate-V BG Fig. 19 The cross-section of the ferroelectric HfO2/Ge-HfO2/HfO2 FET memtransistor [36].).Inset: the same dependence for both VD polarizations [36].

Accepted Article
Looking at the same ID-VD dependence in the logarithmic scale, we observe 12 orders of magnitudes difference between an off state (at ) and an on state (at TG V = 1 V), so that the memtransistors can function as a logic element.It was shown recently that a transistor having an atomically thin material as a channel and both a top and a back gate can act as a logic device, whereas two FETs are needed in the case of Si technology to implement a single logic device.This simplification of the design of logic gates is possible since in the first case the carriers in the channel can be controlled independently by the two gates, which form in fact two surface channels [43].The two inputs of the single device performing logic operations are in this situation the two gate voltages and the output is the drain current, which could be in a high or low state.In Fig. 22 we have displayed the dependence

Accepted Article
This article is protected by copyright.All rights reserved on a semi-logarithmic scale [36].

Energy harvesting using HfO2-based ferroelectrics
Energy harvesting is one of the areas of research that has been developed in the last years.It aims at collecting energy using various electronic devices for self-powering of sensors arrays and many tiny devices interconnected between themthe Internet-of-Things (IoT).Therefore, all types of ambient energies such as RF energy, heat, light, or mechanical vibrations are intended to be transformed and stored in electrical energy to power IoT [44].
Electromagnetic energy harvesting in microwaves and millimeter-waves domains is based on an antenna or an antenna array terminated with a diode having the role of rectifying the electromagnetic field at zero bias.The antennas integrated with a diode are referred to as rectennas.
We will present in the following two types of rectennas.The first one is based on a MIM structure where the insulator is HfO2 while the other is based on a ferroelectric HfZrO tunneling junction (FTJ).In the case of the harvesting based on rectification, the fact that HfO2 is ferroelectric or not is less important, the most important issue is a strong nonlinear current-voltage dependence which ensure a good yield for the microwave to DC conversion without any applied bias.Moreover, there are many reports indicating the undoped HfO2 is ferroelectric.Both structures are vertical and the rectification mechanism in both cases is the tunneling of carriers in HfO2, with a thickness of 6 nm between top and bottom metallic electrodes.The first rectenna is based on the HfO2 MIM diode and is represented in Fig. 23 [45].We present integrated rectennas fabricated at the wafer level.The rectenna consists of a bow-tie antenna integrated with an Au/HfO2/Pt MIM diode for electromagnetic harvesting in the 60 GHz bandwidth allocated to IoT.The top and bottom electrodes of the HfO2 MIM diode form the bow-tie antenna arms.The area occupied by the bowtie antenna is 20.25 mm 2 and its length is of 1.54 mm.The substrate is high-resistivity silicon (HR Si) having a thickness of 525 μm, over which a 300-nm-thick film of silicon dioxide (SiO2) was deposited.
Accepted Article The rectenna output is fabricated in CPW technology for 50 Ω impedance matching.The fabrication process is explained in detail in [45].In Fig. 24 [45] Amplitude-modulated input signals having a continuous wave (CW) carrier at 61.6 GHz with 1 kHz rate are detected by this rectenna having no DC polarization (see Fig. 25).The harvested signal decreases in amplitude as the distance between the transmitter and the bow-tie is increased from 5  to 18  (see Fig. 25).We can thus harvest 250 μV at a distance of 5λ from the mm-wave transmitter, our wireless harvester having a responsivity of 5.4 V/W at an incident power of -20 dBm.Similarly, we have demonstrated that a metal-ferroelectric-semiconductor device coupled

Accepted Article
to an antenna array can harvest electromagnetic energy at 26 GHz with a responsivity of 63 V/W and a NEP (noise equivalent power) of 4 nW/Hz 0.5 , the harvested power being high enough to feed tiny things with a very low power consumption connected in an IoT network [46].
Fig. 25 The harvested signal from a 61.6 GHz wireless source which is AM modulated.[45] The ferroelectric tunneling junction (FTJ) configuration is depicted in Fig. 26(a) and its current-voltage dependence in Fig. 26(b).The current-voltage (I-V) characteristic is obtained in the following way: before measurement, a poling signala ramp of +10 Vwas applied, and then the I-V dependence was measured in the on state (black line).Subsequently, the polarity of the poling signal was reversed (-10 V), and after the poling signal was stopped the FTJ was measured in the off state (red line).

Accepted Article
In [48] it is noted that antiferroelectrics show a hysteresis loop and lower losses compared to ferroelectrics, and thus attaining higher ESD.
Pyroelectricity means the generation of a current i when a temperature difference T  is applied to them.The ECE is the inverse effect of pyroelectricity, i.e., it describes the appearance of a difference in temperature T  when a current/voltage is applied.
The pyroelectric coefficient, with a value of -1300 C/m 2 K in Si-doped HfO2 ferroelectrics, was termed as giant in [47] and is the consequence of the temperature dependence of the concentration of ferroelectric phase near the phase transition between ferroelectricity and paraelectricity.This result has triggered further researches for CMOS compatible pyroelectric sensors having multiple applications in many industries, such as automotive, aerospace, and security.The research of CMOS compatible pyroelectric detectors has started decades earlier, when ferroelectric polymers were discovered.Based on these materials, large arrays of pyroelectric detectors, for example having 16  16 elements, were fabricated in CMOS technology [49] with very good performances, such as a pixel voltage sensitivity of 2200 V/W.Recently, arrays of pyroelectric detectors based on Si-doped HfO2 ferroelectric were designed, fabricated and measured.Because the pyroelectric current is given by:

Accepted Article
where p is the pyroelectric coefficient, A is the device area, T is the temperature, and t is the time, the area of the sensor must be enhanced to obtain a higher current.One solution for increasing the sensor current is to use an array of pyroelectric sensors, as discussed above, or a micromachined Si array (see Fig. 27), of tens of millions of trenches [50], [51], [52].In the last case, a current of few nA is generated when changing the temperature between 29 o C and 31 o C.

Si
Top electrode Doped HfO 2 ferroelectric TiN Fig. 27 The pyroelectric detector using doped HfO2 ferroelectric based on Si micromachining.
The ECE effect was also termed as giant in the case of HfZrO, raising the hope for miniaturized solid-state refrigerators, which are key components in many industries where heat is produced.In particular, such devices are of interest for microprocessors, where for many years the dissipated heat is the main obstacle to increase significantly the performances [47], [53].In the case of HfxZr1-xO2, if x = 0.2, 0.3, the ECE coefficient is positive, producing a T  of 10 K in a broad range of temperatures, 300-448 K. On the contrary, when x = 0.5 the ECE is negative and T  is 10.8 K at 448 K.However, the design and realization of a miniaturized ECE cooler are in infancy.
Any ferroelectric is also a piezoelectric material.Piezoelectricity, which refers to the transformation in a reverse manner of the mechanical energy into an electric current, is another

Accepted Article
physical effect with many applications in energy harvesting (see the recent review [54]).In the case of HfO2 ferroelectrics, the piezoelectricity is present in both thin and thick films [55], [56], [57] and the values of piezoelectric coefficients are similar to that of AlN.For example, a 1 μm-thick Ladoped HfO2 shows a d33 coefficient of 7.7 pm/V, while d33 for AlN is 6 pm/V.This fact opens new ways to use HfZrO-doped ferroelectrics for actuation applications, bulk acoustic wave resonators (BAW), and surface acoustic wave resonators (SAW), which are widespread in daily used electronic devices, such as smartphones and laptops.

The future of HfO2-based ferroelectrics
Although the ferroelectricity in HfO2 is a well-studied subject in many papers, the applications have a slower pace compared to material studies.Therefore, in the next years, nanoelectronics applications will emerge following the paths indicated above, i.e., they will concentrate on memories, low-power electronics circuits, tunable devices for wireless communications, sensors,
1-16 GHz by varying the DC voltage in the range 0-4 V [29].Three filter configurations: low-pass (LPF), high-pass (HPF), and band-pass (BPF) were designed, fabricated, and measured, showing a frequency shift of the minimum of the reflection coefficient between 90 MHz and 4.4 GHz, with a minimum insertion loss of about 6.9 dB in the high-pass configuration.The filters are illustrated in Fig. 11.All filters are made using inductors and interdigitated capacitors, which play the role of variable capacitors (varactors) due to the HfZrO ferroelectric layer underneath.The LPF and HPF have a T-network configuration: LPF is formed from two series inductances and a shunt varactor and HPF consists of two series varactors and a shunt inductor, while the BPF results from the connection of the LPF to the HPF.As a major result, the BPF has much lower dimensions (3.19 mm × 3.47 mm) than the working wavelength.

Fig. 12
Fig.12The capacitance and phase of the CPW interdigitated capacitor acting as a varactor versus

Fig. 13
Fig. 13 The frequency shifts (f) of the three filters versus applied DC voltage at 10 GHz [29].
principle, memristors are two-terminal devices, which are resistive non-volatile memories, having two distinct states, on and off, displaying different hysteresis loops of current-voltage dependence at repetitive sweeps of the voltage.Two-terminal memristors are represented mainly by valence change memristors (VCM) or filament memristors, which are metal-insulator-metal (MIM) the optical image of the HfO2/Ge-HfO2/HfO2/Si chip.

Fig. 21
Fig. 21 D I VD = 1 V.All the dependencies show hysteretic behavior in the clockwise direction.Whether this figure shows that D I is in a high or a low state depending on the back and top gate voltages, the building of a logic device based on these results is still under consideration at the present moment.

Fig. 23 The
Fig. 23 The HfO2 rectenna: (a) top-view of the bow-tie antenna and (b) MIM area; (c) cross-section

Fig. 24 (
Fig. 24 (a) SEM image of the rectenna and (b) current density versus voltage of the MIM diode.
and harvesting energy[58].We note that the next-generation communication technologies (5G: 26-95 GHz by 2025 and 6G: 95 GHz-3 THz by 2030) have stringent requirements and as yet unsatisfied needs, such as: (i) low bias voltages for beam-steering capabilities; (ii) increased power efficiency and ultrafast switching times; (iii) miniaturization of wireless nodes; (iv) reconfigurability (to implement on-chip antenna multiplexing and antenna tuning); (v) frequency scalability (for future increase in bandwidth); (vi) zero-energy devices (built-in self-powering functionalities of mobile devices, such as Internet-of-Things (IoT) and Internet-of-Services (IoS)).Presently, Si-based CMOS devices operate close to their theoretical limits in terms of power densities, operating frequencies, and temperature.Furthermore, traditional and current High-Frequency Electronics (HFE) technologies (III-V, GaN, and SiGe BiCMOS) are prohibitive for ultra-large volume manufacturing needs (50 billion mobile devices by 2030).Thus, none of the currently used materials and technologies can provide the above-mentioned requirements of low Accepted Article power, ultra-fast switching times, reduced size, reconfigurability, and scalability.Therefore, the only way to achieve this goal is to use the quantum tunneling effect to build oxide-based highfrequency emitters and receivers working up to THz frequencies while being CMOS compatible.Therefore, HfO2-based ferroelectrics could play a crucial role in the future 5G and 6G communication technologies due to CMOS compatibility and the versatility of physical properties obtained by doping engineering of materials.

Table I
Basic physical properties of HfZrO and the nanoelectronic areas where they are applied.
Atomically-thin or two-dimensional (2D) ferroelectrics are the ultimate ferroelectrics consisting of one or few monolayers.2D materials commonly originate from van der Waals materials