Symposium EL09—Ferroelectricity and Negative Capacitance—Fundamentals, Applications and Controversies

While power consumption density significantly increases with scaling down of MOSFETs, novel low power devices have triggered lots of interest, among which ferroelectric-based negative capacitance (NC) FET is regarded as a promising candidate due to its capability of sub60 subthreshold swing (SS) and high on current. Up to now, although domain nucleation and growth theory is reasonable to explain experimental phenomena observed such as steep SS in NCFET and V-drop in R-FE circuit, the physical mechanism of hysteresis in FeFET lacks a thorough analysis, making it difficult to unambiguously determine the contribution of domain switching to hysteresis. Besides, the specific model aimed at hysteresis is missing, while the conventional method relies on extraction from pre-existing transfer curve, which is too time-consuming to approximate and systematically offer the explicit guideline with high efficiency. In this work, an accurate efficient quantitative model of hysteresis in multi-domain FeFET is proposed, based on modeling of FE polarization switching dynamics mediated by domian nucleation and growth, including history effect, minor loop operation, voltage and time dependence. The proposed model can accurately capture the dynamic mechanism of hysteresis without time-consuming iteration process, unveiling its non-monotonic dependence on voltage sweeping rate, coupled with sweeping range, which is also briefly discussed in physics. Moreover, the FE material-related parameters impact on hysteresis in FeFET is investigated by model and physically analyzed to draw a comprehensive conclusion. Furthermore, we re-evaluated the prerequisite of NC origin and requirement of hysteresis-free, showing that there exists a fundamental challenge of FeFET for high-speed and low-voltage operation as a logic device.

Since the physical dimension of integrated circuit device (especially, memory device) has been aggressively scaled down over the past a few decades, it has been faced with a few critical/technical issues such as ever-increasing power dissipation and even incommensurate integration density because of the physical scaling limit [1]. As one of the promising solutions, ferroelectric field effect transistor (FeFET) has been emerged as a candidate for future nonvolatile memory (NVM) device because of its scalability and compatibility to CMOS (complementary metal oxide semiconductor) technology. In real, among various ferroelectric materials, HfO₂-based ferroelectric material doped with various dopants such as Zr, Al, and Si are being adopted in FeFET. Al-doped HfO₂ (HAO) shows its stable ferroelectricity in thermal process (e.g., 500 °C ~ 900 °C). Moreover, by modulating various conditions of fabrication process, the ferroelectricity of HAO can be optimized. This indicates that HAO can be widely used for future FeFETs.
In this work, the performance of HAO-based ferroelectric capacitor is presented with various fabrication conditions since the ferroelectricity of ferroelectric material requires appropriate conditions for crystallization. To explore many fabrication options for HAO-based ferroelectric capacitor, following conditions are taken into account and modulated: 1) the doping concentration of HAO (i.e., the cycle ratio of ALD deposition); 2) oxygen exposure time in ALD deposition process; 3) ALD substrate temperature; 4) annealing temperature. Then, the HAO capacitor is externally connected to the gate stack of the baseline MOSFET to electrically form a FeFET. Note that 500nm-long-channel MOSFET (metal oxide semiconductor field effect transistor) is fabricated to explicitly analyze the effect of the body biasing as well as to avoid short channel effects [2]. The modulation of memory window (MW) by body biasing in FeFETs with HAO-ferroelectric capacitor are experimentally investigated.

[1] Wang, P., Wang, Z., Shim, W., Hur, J., Datta, S., Khan, A. I., & Yu, S. (2020). Drain–erase scheme in ferroelectric field-effect transistor—Part I: Device characterization. IEEE Transactions on Electron Devices, 67(3), 955-961.
[2] Kim, H. W., & Kwon, D. (2020). Impact of body-biasing for negative capacitance field-effect transistor. Journal of Physics Communications, 4(9), 095019.

Acknowledgements
This work was supported by the National Research Foundation of Korea (NRF) through a grant funded by the Korea government (MSIP; No. 2020R1A2C1009063, 2020M3F3A2A01081672, 2020M3F3A2A01082326, and 2020M3F3A2A02082473).

This tutorial will start with a general introduction to the atomic layer deposition (ALD) technique for industrial applications, the use of super-cycle ALD for fluorite structure oxides such as hafnium oxide and their nanolaminate and alloyed variants, and ferroelectricity therein. Afterwards, the main focus will be on microscopic aspects of such nanoscale, polycrystalline ferroelectrics for memory applications. Memory write and read operations require local reorientation and sensing of polar crystallographic domains, respectively. We will discuss the impact of such structural aspects on the miniaturization of the memory devices, physical principles relevant to polar domains, and methods for characterizing nanoscale domain structure in order to obtain a local picture for interpreting electrical data measured from ferroelectric memories. Lastly, we will elucidate the role of the interface layer structure, polarization charge screening, interface states and fixed charges on the performance and reliability of ferroelectric field-effect transistor based memories.

With the discovery of ferroelectricity in HfO2 based thin films 2011 and the co-integration of ferroelectric field effect transistors (FeFET) into standard high-k metal gate (HKMG) CMOS platforms 2016/17 by GLOBALFOUNDRIES, the FeFET has emerged from a theoretical dream to an applicable reality. Maturing in the beginning as a low-cost, low power eFLASH replacement, the FeFET yet is much more than a classical stiff eNVM cell. With its great HKMG CMOS compatibility, its flexibility and its unique switching properties, it is rather to be seen as a new versatile device that promises to open up new worlds. Especially the neuromorphic design community has shifted focus towards this novel device with game-changing potential. In this talk we will discuss the actual status of GLOBALFOUNDRIES FeFET technology, look into the operation and use of this device and sketch a potential roadmap.

We will discuss the effect known as “negative capacitance”, whereby some ferroelectric materials display an anomalous dielectric response (a negative permittivity) when subject to suitable electric boundary conditions. We will describe our current theoretical understanding, from both phenomenological and atomistic perspectives, drawing a clear distinction between equilibrium and transient negative-capacitance responses. We will also review the experimental evidence for both flavors of negative capacitance, and briefly present its expected technological applications (most notably in field-effect transistors, because of the concomitant “voltage amplification”). Finally, we will give our view of the current challenges and opportunities in the field.

Ferroelectric memories based on the doped HfO2 ferroelectric, including capacitor based 1T1C FeRAM and transistor based 1T FeFET, have received significant interests recently. The excellent scalability and CMOS compatibility of ferroelectric HfO2 has allowed its integration into 2x technology node. Such scaling reveals novel physical phenomena not observed before in perovskite ferroelectric devices, such as degraded device variation with scaling, switching stochasticity, switching probability accumulation. In order to understand such phenomena and exploit them for novel energy efficient computation, accurate device models would be indispensable. In this lecture, we will present ferroelectric memory models developed so far, covering computationally efficient compact models, Monte Carlo models, and computationally intensive phase field models. By comparing with experimental data, the application range of each model will be identified and the potential for scaling/miniaturization of ferroelectric devices will be discussed.

Thermodynamics dictate that charge in an ordered system can be switched with lower energy than the Boltzmann limit known for non-interacting systems. Thinking about how to exploit it for computing led to the identification of negative capacitance in materials that show (anti)ferro-electric order [1]. In this presentation, we shall discuss this connection. Stabilization of negative capacitance needs careful consideration of electrostatic boundary condition. Especially, accounting for screening by metal electrodes is very important. In this regard, intuitive theoretical reasoning and simple schematics have often led to much confusion. We shall discuss our views on how to design experimental systems that are conducive to achieving the appropriate boundary conditions. While a single domain theory was used to elucidate the effect for simplicity, negative capacitance is not limited to single domains, as it has been pointed out early [2], and later by many authors [3,4].

Going back to the thermodynamic origin of switching more charge with less energy, the increase in capacitance has remained to be the simplest macroscopic test of the negative capacitance effect. However, modern imaging techniques allow probing the internal electric fields in a polar system with atomistic resolution. These experiments have revealed local energy maxima stabilized in the material, providing a direct imaging of stabilized negative capacitance [5]. These experiments confirm the prediction of local nature of the negative capacitance. In addition, they confirm the fact that negative capacitance originates in the regions of suppressed polarization.

One of the implications of Negative Capacitance is the possibility of subthermal (<kT/q) subthreshold swing in a MOSFET[6]. However, observation of subthermal swing has often been accompanied with Hysteresis or such swing has only been seen at very low currents. These led to questions such as: (i) Is subthermal subthreshold behavior intimately connected to Hysteresis? (ii) Does not observing this behavior mean that negative capacitance effect is fundamentally incorrect? In fact, in our view, most questions, confusions and so-called controversies can be traced back to these two questions. We will discuss these scenarios- specifically, the fact that these scenarios can be predicted from Landau’s energy expressions together with appropriate boundary conditions, as it has been discussed by many authors in literature. Therefore, observation of subthermal subthreshold swing is really a function of the properties of the channel material and he (anti)ferroelectric oxide and not observing it does not constitute a fundamental question on the negative capacitance effect itself. In fact, I will discuss other non-classical subthreshold behavior that ensues from negative capacitance, even if a subthermal subthreshold swing is not present. Such non-classical behavior is clearly observed in experiments,.

Negative capacitance could find applications in different areas such as supercapacitors for energy storage, batteries, as enhanced capacitor for Dynamic Random Access Memory (DRAM) scaling, for back end capacitors in microchips, in addition to reducing supply voltage in transistors. I shall discuss some of our recent results in this regard. Specifically, I shall discuss experimental data that show how the original thermodynamic motivation of switching more charge with less energy can now be achieved[7,8] in the most advanced transistors, which could lead to supply voltage reduction and/or enhanced scaling capability beyond the roadmap.

[1] Salahuddin and Datta, Nanoletters, 8,2,2008
[2] Cano and Jimenez, APL, 97,13,2010
[3] Zubko et al, Nature 534 (7608), 524-528
[4] Lukyanchuk et al, Comm. Physics, 2,22, 2019
[5] Yadav et al Nature, 565, 7740, 468–471, 2019
[6] Salvatore et al, IEDM, 2008
[7] Cheema et al, Nature, 580 (7804), 478-482
[8] DW Kown et al, IEEE EDL, 40,6,2019

In this talk we will present novel progress in the device functionality and energy efficiency when applying gate stacks with negative capacitance (NC) features to 2D devices. We will explore NC as technology booster for 2D/2D tunneling FETs based on WSe 2/SnSe 2 vdW heterojunctions, from low to high temperatures, including the energy losses per switching operation, together with the analysis of the advantages and challenges of using NC in doped high-k dielectrics like Si:HfO2. We will report both DC and pulsed experiments trying to have further insights in recent controversies concerning the negative capacitance regime and the dynamics of nucleation processes in ferroelectrics.

The idea of ferroelectric-based devices with negative capacitance (NC) in low-power nanoscale electronics triggered explosive activity in the field. This revived the thoughts of engineers of the early 1930-s on the possibility of negative circuit constants and the 40-years old fundamental question raised by Rolf Landauer, whether the capacitance can be negative, that is if the increase in the charge of the capacitor can decrease its voltage. Having stated that ferroelectrics can harbor the NC during the transient processes of switching, Landauer indicated that the very existence of the static NC as a part of steady-state is challenged by instability against spontaneous domain formation. For today, most of the research addressed transient NC, leaving the basic question of the existence of the steady-state NC unresolved. Here we demonstrate* that the ferroelectric nanodot capacitor hosts a stable two-domain state realizing the static reversible NC device.

* I. Luk’yanchuk, Y. Tikhonov, A. Sené, A. Razumnaya & V. M. Vinokur, Harnessing ferroelectric domains for negative capacitance, Communications Physics volume 2, Article number: 22 (2019)

Topological solitons such as magnetic skyrmions have drawn attention as stable quasi-particle-like objects. The recent discovery of polar vortices and skyrmions in ferroelectric-oxide superlattices has opened up new vistas to explore topology, emergent phenomena, and approaches for manipulating such features with electric fields. Using macroscopic dielectric measurements, coupled with direct scanning convergent-beam electron diffraction (SCBED) imaging at the atomic scale, theoretical phase-field simulations, and second-principles calculations, we demonstrate that polar skyrmions in (PbTiO₃)_n/(SrTiO₃)_n superlattices are distinguished by a sheath of negative permittivity at the periphery of each skyrmion. This enhances the effective dielectric permittivity compared to individual SrTiO₃ and PbTiO₃ layers. Moreover, the response of these topologically protected structures to electric field and temperature show a reversible phase transition from the skyrmion state to a trivial uniform ferroelectric state, accompanied by large tunability of the dielectric permittivity. Pulsed-switching measurements show a time-dependent evolution and recovery of the skyrmion state (and macroscopic dielectric response). The interrelationship between topological and dielectric properties presents an opportunity to simultaneously manipulate both of them by a single, and easily controlled, stimulus, the applied electric field.

Domains in ferroelectric materials have been at the forefront of recent scientific research, mostly due to the fascinating properties of their boundaries, domain walls. As well as enhancing the macroscopic properties of ferroelectrics, domain walls exhibit symmetry and properties different from the bulk material, offering novel functionality. One scalable method to control domain wall configurations in ferroelectrics is epitaxial strain. Epitaxial strain and the interplay between electrostatic and mechanical energy allow for the formation of novel domain structures, such as regular arrays of flux closure domains or chiral vortex centres.

Here, we show that a stable supercrystal phase comprising a 3D ordering of nanoscale domains with tailored periodicities can be engineered in PbTiO₃/SrRuO₃ ferroelectric-metal superlattices under tensile strain. A combination of x-ray diffraction, piezoresponse force microscopy, transmission electron microscopy and phase-field simulations reveals a complex hierarchical superstructure, which consists of alternate vertical and horizontal flux closure patterns, and forms in order to minimize the elastic and electrostatic energy. Large local deformations of the ferroelectric lattice are accommodated by periodic lattice modulations of the SrRuO₃ layers, presenting a paradigm for engineering correlated materials with tailored modulated structural and electronic properties.

My group is interested in the behavior of ferroelectric materials that, in particular circumstances, display a persistent voltage drop that opposess the overall applied DC bias; that is, they behave as a static negative capacitance [1]. More precisely, we investigate PbTiO3/SrTiO3 ferroelectric/dielectric superlattices as a convenient model system [2,3], using second-principles simulation methods [2,4] to monitor their behavior as a function of the available control knobs: temperature, applied electric field, epitaxial strain and layer thickness. This allows us to better understand the factors affecting the negative-capacitance response of the PbTiO3 layers, and thus identify strategies to optimize it. In this talk I will review our most recent results, focusing on the possible persistence of negative capacitance down to very low temperatures and the evolution of the differential capacitance as a function of applied electric field.

Work done in collaboration with M. Graf and H. Aramberri, postdocs at the Luxembourg Institute of Science and Technolgoy. Funded by the Luxembourg National Research Fund through Grant INTER/RCUK/18/12601980.

[1] J. Íñiguez et al., "Ferroelectric negative capacitance", Nature Reviews Materials 4, 243 (2019).

[2] P. Zubko et al., "Negative capacitance in multidomain ferroelectric superlattices", Nature 534, 524 (2016).

[3] S. Das et al., "Local negative permittivity and topological phase transition in polar skyrmions", Nature Materials (2020), https://doi.org/10.1038/s41563-020-00818-y

[4] J.C. Wojdel et al., "First-principles model potentials for lattice-dynamical studies: general methodology and example of application to ferroic perovskite oxides", J. Phys. Condens. Matt. 25, 305401 (2013).

Chirality, when an object cannot be superimposed on its mirror image, has been seen across a multitude of fields in materials science. From biological molecules to magnetic spin textures, chirality has played a key role in determining the functional properties of such systems. However, it is only recently that the chiral objects have appeared in ferroelectric materials systems. In PbTiO₃/SrTiO₃ (PTO/STO) thin film multilayers, boundary conditions set by epitaxial strain from the substrate and paraelectric layers of STO results the continuous rotation of electric dipoles in the PTO layer in the form of polar vortices.¹ Similar to magnetic vortices, the sense of rotation of the dipoles and the orientation of the axial polarization of the vortices can be used to classify them as left-handed or right-handed.² Although chirality and topology has been observed in a number of ferroic systems, the manipulation of the sense of chirality with electric fields has remained elusive.
Using second harmonic generation circular dichroism (SHG-CD), we show that these vortices organize together to form chiral domains over length scales of 100s of nanometers. Each chiral domain consists of a multitude of vortex tubes of the same handedness, separated by a mobile chiral boundary. Using in-situ SHG-CD, we map chiral polar vortex domains as a function of in-plane electric field. We show that these domains can be elongated along the field direction, when the electric field is applied along parallel to the vortex tube. Moreover, by applying an electric field along the direction of rotation, we deterministically switched the chirality of the film by translation of the chiral boundary. Ab-initio and phase field simulations were used to show that the chiral switching arises from a reversal in the buckling of these vortices. In addition, piezoforce microscopy studies were performed to relate how the chirality arises from polarization of domains. This work illustrates the first evidence of chiral switching in any materials system using an applied electric field, and highlights the use of lab-based mesoscopic probes of nanoscale chiral objects.

References
[1] Yadav, A. K., C. T. Nelson, S. L. Hsu, Z. Hong, J. D. Clarkson, C. M. Schlepütz, A. R. Damodaran, et al. “Observation of Polar Vortices in Oxide Superlattices.” Nature 530, no. 7589 (February 2016): 198–201.
[2] Shafer, Padraic, Pablo García-Fernández, Pablo Aguado-Puente, Anoop R. Damodaran, Ajay K. Yadav, Christopher T. Nelson, Shang-Lin Hsu, et al. “Emergent Chirality in the Electric Polarization Texture of Titanate Superlattices.” Proceedings of the National Academy of Sciences 115, no. 5 (January 30, 2018): 915–20.

The negative capacitance effect in ferroelectric (FE) based devices has not only attracted a significant research interest due to the possibility of achieving steep switching in transistors [S. Salahuddin et al., Nano let. 8, 2008], but has also intrigued a broader research community aimed at understanding the fundamental origin of such intriguing properties. In a microscopic sense, the ferroelectricity originates from the non-centrosymmetric crystal structure where the spontaneous displacement of atoms (and the corresponding electron gas and ion core) leads to a non-zero spontaneous polarization (P), which yields the minimum energy bi-stable states of the FE. However, in a heterogeneous system (e.g. FE-DE-semiconductor or MFIS stack), the value of P is determined by the minimum energy of the whole system rather than the energy of the FE itself. In such systems, the P-induced bound charge in the FE-DE interface induces a depolarization electric field in the FE layer. Such a depolarization effect tends to reduce the P magnitude of the FE layer causing an increase in its free energy. However, in this scenario, the fundamental electrostatics suggests the formation of multi-domain (MD) states that suppresses the depolarization effect without any significant increase in its free energy [A. Kopal et al., Ferroelectrics 223, 1999]. Moreover, the suppression of depolarization and free energy occurs in the cost of gradient energy (which is elastic in nature) due to the P gradient near the domain wall (DW). Therefore, the formation of the MD state appears as an interplay among different energy components for obtaining the minimum energy configuration [A. K. Saha et al., Sci. Rep. 10, 2020].

To investigate the MD state formation and its effect on the effective permittivity of FE, we have developed a phase-field simulation framework for Metal-Ferroelectric-Insulator-Metal (MFIS) heterostructure by self-consistently solving the Ginzburg-Landau equation, Poisson’s equation, and semiconductor charge equations [A. K. Saha et al., Sci. Rep. 10, 2020] . According to our analysis, the applied voltage-driven P switching in the MD state is strongly dependent on the features of its DW. This further depends on the elastic, electrostatic, and physical properties of the FE layer and the dielectric/semiconductor layers that it interacts with. We show that, in the case of low elastic coupling (low gradient energy coefficient), the types of DWs are hard and the domain density depends on material parameters and physical dimensions of the heterostructure. In the hard DW regime, P-switching is hysteretic and can take place as a combination of domain nucleation and DW motion. During such a P-switching, the effective permittivity of FE becomes negative. However, even without P-switching (with voltages < coercive voltage), the effective permittivity of the FE layer can increase (but remain positive) due to the electrostatic interactions among domains. This effect is more dominant when the domain density is large. As the domain becomes denser, the effective permittivity keeps increasing within the limit of hard-DW. On the other hand, for a sufficiently high elastic coupling (high gradient energy coefficient), the DW can become soft and the corresponding soft domain-wall displacement leads to non-hysteretic characteristics. In such a scenario, the effective permittivity of the FE layer becomes negative in a non-hysteretic fashion.

The feasibility of ferroelectric negative capacitance devices for ultra-low power electronics hinges on the availability of materials which are scalable and compatible with established semiconductor manufacturing technologies. Ferroelectrics of fluorite-structure based on the binary oxides HfO₂ and ZrO₂ fulfill both of these criteria. Recently, negative capacitance was observed in HfO₂ based ferroelectrics using pulsed electrical measurements. However, relatively large applied voltages have been necessary to observe negative capacitance in this material system so far. To move towards applications in negative capacitance devices, much lower operating voltages are needed. Furthermore, the microscopic domain dynamics underlying the negative capacitance effect in HfO₂ based ferroelectrics is still not understood, which hampers physical modeling and device design. Here, we discuss the available experimental data and outline pathways to move towards improving our understanding of negative capacitance in HfO₂ based ferroelectrics and prospective applications.

The ferroelectric (FE)-dielectric (DE) heterostructure is one of the commonly used model systems to probe negative capacitance effects.^1,2 In such a system, the change in charge due to a change in source voltage can be larger than the change in charge when the same change in source voltage is experienced only by the constituent DE capacitor. Such a differential charge boost, when achieved without a hysteresis, can lower the power supply voltage and the energy dissipation in a negative capacitance field-effect transistor (NCFET). However, there is an ongoing debate on whether the charge boost in such a heterostructure is a transient phenomenon or a steady state one.

According to a transient model³ of ferroelectric negative capacitance, the charge boost in an FE-DE capacitor is a consequence of (1) the mismatch between the time scales for polarization switching and the screen charge dynamics in the FE, and (2) a charge imbalance between the FE and the DE. The change in the DE charge, △Q in such a scenario is expected to be larger than C_DE△V resulting in the so-called charge boost. The fundamental aspect of the transient model is that as the screening charges balance out the polarization charges, the boost in both V_FE and Q diminishes to values that are expected from a series combination of two positive capacitors, thereby leading to charge boost only in transience. Hence, △Q will be larger than C_DE△V only for a short amount of time. Whereas, if the charge boost was in fact a steady state response, then the charge boost stemming from it will prevail for the entire duration of the voltage pulse. Hence, the key test for transient and steady state model of negative capacitance is as follows: for a given change in voltage △V, is the charge supplied by the voltage source to a FE-DE capacitor △Q larger than C_DE△V only for a short duration of time (transient state) or an extended period of time (steady state)?

To answer this question, we present an experiment that tests the nature of the charge boost: steady-state or transient. To that end, we study a ferroelectric-dielectric (FE-DE) heterostructure with FE Hf_0.5Zr_0.5O₂ (10nm) and DE HfO₂ (5nm). We adopt a modified form of the well-known, positive up negative down (PUND) technique where 'long' voltage pulses are applied on the FE-DE capacitor to measure its polarization vs voltage (P-V) characteristics. We find that the device under test (DUT) exhibits a differential charge boost of steady-state nature i.e., the charge boost remains intact after the initial transient effects subside and all the voltages in the system reach constant values and steady states. A SPICE-based model of the system that accounts for the multi-domain polarization switching dynamics in the FE-DE stack was also used to verify our experimental results.

In summary, the results obtained from the modified PUND measurement technique on an FE-DE structure reveals that the charge boost in a hysteretic FE-DE structure is prevalent for the entire pulse duration and hence a steady state response. The experimental results match with the predictions based on a multi-domain ferroelectric model eliciting the importance of domain dynamics in obtaining the ideal stabilized negative capacitance. Investigating NCFETs and NC Capacitors using this method will provide more insights into the charge dynamics of the systems which can be used to advance the designs and study the physics of these devices.

References:
1. S. Salahuddin and S. Datta, Nano letters 8, 405 (2008).
2. M. Hoffmann, F. P. Fengler, M. Herzig, T. Mittmann, B. Max, U. Schroeder, R. Negrea, P. Lucian, S. Slesazeck, and T. Miko-lajick, Nature 565, 464 (2019).
3. K. Ng, S. J. Hillenius, and A. Gruverman, Solid State Commu-nications 265, 12 (2017).

Prasanna Venkatesan Ravindran and Nujhat Tasneem both contributed equally to this project and will be co-presenting at the 2021 MRS Virtual Spring Meeting.

Power dissipation is one of the most critical issues for nanoelectronic circuits. For low-voltage and low-power logic applications, novel steep-slope device concepts, which can break the fundamental limitation of subthreshold swing (SS) in MOSFETs (60mV/dec at room temperature), have attracted lots of attention. By utilizing the voltage amplification effect induced by the negative capacitance (NC) in the gate stack, the ferroelectric FET (FeFET) also has the capability of sub-60 SS, which is considered as NCFET. However, the NC effect in ferroelectric film has aroused great scientific controversy due to its unclear physical picture and lacking of direct experimental evidence.
Here we experimentally observed the NC effect in HfO2-based ferroelectric film and systematically studied its fundamental physics from the perspective of material features including both relaxation polarization and transient polarization. Different from the stabilized NC, the observed NC effect of HfO2-based FE material in this work is not a steady but a dynamic behavior, which is very sensitive with sweeping voltage and can be existed only under an extremely stringent operating frequency range. Moreover, even under this required frequency range, it is found that the SS of NCFET will inherently degrade with the increased gate voltage, which is not preferred for low-voltage logic operation.
Besides, it is found that the hysteresis theoretically exists, and there is an intrinsic optimization conflict between SS and hysteresis, which is a big challenge for logic applications. In addition, the hysteresis shows a strongly non-monotonic dependence on voltage sweeping rate, coupled with sweeping range, and can be even larger than the hysteresis of standalone FE capacitor for high-frequency applications. The possibility of NCFET as a steep-slope device for high-speed and low-voltage logic operation needs to be carefully re-assessed.

The reversal of polarization in a ferroelectric material involves overcoming an energy barrier and has been previously proposed and found to yield transient negative capacitance (NC) in the intermediate states of the switching process [1-3]. Homogeneous switching was assumed to interpret the experimental results of NC; however, inhomogeneous switching is a more abundant mechanism than homogeneous switching, but its possible effect on NC is basically unknown. Here, we use first-principles-based effective Hamiltonian techniques to investigate the occurrence of NC during these two types of switching processes in the super-tetragonal phase of BiFeO₃, which can be realized by changing the magnitude of the electric field. We find NC in both cases, but with the magnitude of the inverse of the capacitance being drastically larger in the inhomogeneous (nucleation limited) switching, as compared to the homogeneous switching. We further analyze the origin of the different NC effects, and point to the different energetic trajectories between these two representative switching mechanisms [4]. In addition, we derive analytical formulas and show that the capacitance can be negative under perfect screening conditions and for very different materials and switching mechanisms [5].

References:
[1] A. I. Khan, K. Chatterjee, B. Wang, S. Drapcho, L. You, C. Serrao, S. R. Bakaul, R. Ramesh, and S. Salahuddin, Nat. Mater. 14, 182 (2015).
[2] M. Hoffmann, M. Pešić, K. Chatterjee, A. I. Khan, S. Salahuddin, S. Slesazeck, U. Schroeder, and T. Mikolajick, Adv. Funct. Mater. 26, 8643 (2016).
[3] A. K. Saha, S. Datta, and S. K. Gupta, Journal of Applied Physics 123, 105102 (2018).
[4] Bin Xu, Sergey Prosandeev, Charles Paillard, and L. Bellaiche, Phys. Rev. B 101, 180101(R) (2020).
[5] Sergey Prosandeev, Charles Paillard, B. Xu, and L. Bellaiche, Phys. Rev. B 101, 024111 (2020)

Acknowledgements:
B.X. and L.B. thank the DARPA Grant No. HR0011-15-2-0038 (MATRIX program). B.X. also acknowledges financial support from National Natural Science Foundation of China under Grant No. 12074277, the startup fund from Soochow University and the support from Priority Academic Program Development (PAPD) of Jiangsu Higher Education Institutions. S.P. thanks ONR Grant No. N00014-17-1-2818 and the support from Grant RMES No. 3.1649.2017/4.6, and also appreciates the financial support from Grant No. BAZ0110/20-3-08IF of Ministry of Science and Higher Education of the Russian Federation (State assignment in the field of scientific activity, Southern Federal University, 2020). C.P. acknowledges the support of ARO Grant No. W911NF16-1-0227. We are also thankful for the computational support from Arkansas High Performance Computer Center at the University of Arkansas.

A discovery of ferroelectricity in doped HfO₂ films opens the door of the device application using very thin ferroelectric films because stable ferroelectricity is maintained below 10 nm even in polycrystalline films. We demonstrate the fundamental understanding of HfO₂-based ferroelectric films using epitaxial films[1-8]. In addition, we demonstrate 1 mm-thick ferroelectric films that are useful for the piezoelectric applications[9-10].
Various devices including ferroelectric memories, ferroelectric tunnel junction, and piezoelectric transistors have been investigated. For these device applications, low temperature deposition of ferroelectric films is highly required to reduce the reaction between the film and the underlying layers.
In this study, we demonstrate the room temperature deposition of Y-doped (Hf, Zr)O₂ films by sputtering method[7]. Room temperature-deposited films showed almost similar ferroelectricity with the post heat-treated one above 800 ^oC. In addition, wide variety of composition show the ferroelectricity even for the films deposited at room temperature. In my presentation, we discuss the determination factors for the room temperature deposition.

[1] Shimizu et al., Appl. Phys. Lett., 107, 032910-１-５ (2015);
[2] Shimizu et al., Sci. Rep., 6, 32931-1-8 (2016)
[3] Shiraishi et al., Appl. Phys. Lett., 108, 262904-1-5 (2016)
[4] Mimura et al., Jpn. J. Appl. Phys., 59, SGGB04-1-6 (2020).
[5] Shimizu et al. Appl. Phys. Lett., 113, 212901-1-5 (2018)
[6] Mimura et al., Jpn J. Appl. Phys., 58, SBBB09-1-5 (2019).
[7] Mimura et al., Appl. Phys. Lett. 113, 102901-1-4 (2018).
[8] Mimura et al., Appl. Phys. Lett., 109, 052903-1-4 (2016)
[9] Mimura et al., Appl. Phys. Lett., 115, 032901 (2019).
[10] Shimura et al., J. Ceram. Soc. Jpn., 128, 539-543 (2020).
[11] Mimura et al., Appl. Phys. Lett., 116, 062901-1-5 (2020).

Super-elasticity of functional ferroelectric oxides offers promises for integrating ferroelectric films into flexible electronics. However, super-elastic deformation is a complex phenomenon related to possibly multiple concurrent mechanisms. Fundamentally understanding how multiple mechanisms contribute to the super-elasticity of ferroelectric oxides is crucial to realizing their potential flexible electronic applications. Here, we employ phase-field simulations to model the dynamics of ferroelectric domain patterns of freestanding BiFeO₃ membranes to understand the origin of their super-elasticity under substantial bending deformation (5% strain). It is demonstrated that both a reversible Rhombohedral-Tetragonal (R-T) phase transition and a nearly reversible domain evolution of BiFeO₃ membranes contribute to accommodating the large deformation and thus their super-elasticity. The dynamics of domain evolution also reveal the formation of an exotic ferroelectric vortex and polarization rotation before the phase transition. We constructed a diagram of phases and domain patterns as a function of the membrane thickness and bending angle, which allows one to readily predict the emergence of T phase and ferroelectric vortex in bent BFO membranes. These results not only provide fundamental understanding of mesoscale super-elastic mechanisms but also reveal exotic domain states of ferroelectric membranes.

With the increasingly high demand for high-performance dielectric nanocapacitors in electronics and energy storage devices, huge research efforts have been placed in developing high-performance multilayer ceramic capacitors (MLCCs) with a particular focus on advanced functional nanomaterials design and development. In particular, the search for lead-free, low-loss ferroelectric oxides are considered as one of the most promising material candidates for this technology. In this study, using first-principles density-functional (perturbation) theory calculations, we investigate the influence of A-site alloying in potassium niobate, (K,A)NbO₃ on its optoelectronic, polarization, and dielectric properties, where A = Li, Na, Rb, and Cs. Using the special quasi-random structure model, we consider various alloy compositions of (K,A)NbO₃and examine their mixing thermodynamics before moving on to characterizing their polar properties, establishing specific structure-property design rules for high-performance dielectric oxides. Together with complementary experimental findings, our work underscores the importance of how cation-exchange in these lead-free ferroelectrics on neoteric MLCC materials.

During switching, the microstructure of a ferroelectric normally changes to best align internal dipoles with externally applied electric fields. Dipolar regions (domains), that are favourably oriented, grow at the expense of those in unfavourable orientations and this is manifested in a predictable field-induced motion of the walls that separate one domain from the next. In copper-chlorine boracite, however, we have discovered that head-to-head charged 90^o domain walls move in the opposite direction to that expected: the polarisation that is anti-aligned with the applied electric field increases as a result of the field-induced wall movement. Polarisation-field (P-E) hysteresis loops, associated with the movement of these walls, hence show negative gradients throughout the entire P-E cycle. Negative capacitance is therefore implied. We have measured switching currents, generated by the relative motion between domain walls and sensing electrodes, by employing charge gradient microscopy. Such measurements reveal that the signs of the switching currents are opposite to those expected conventionally. Moreover, the switching current behaviour can be analysed to allow the negative capacitance component to be extracted and probed as a function of frequency.

The control of ferroelectric domains and domain walls is the basis of a promising approach to imagine novel electronic devices. While the control of ferroelectric domains by the application of an electric field has been extensively studied, the application of a mechanical load and its consequences over the ferroelectric properties have been seldomly explored. Mechanical switching of ferroelectric domains by the application of a local stress has already been observed [1], however a complete explanation of the phenomena involved in the process may depend on the properties of the material studied. In order to propose a complete description of the mechanisms at play during switching, flexoelectricity has to be taken into account [1] in addition to piezoelectricity, which is a common property of all ferroelectrics, and along with ferroelasticity [2] and electrochemistry [3].

In this contribution, we apply a mechanical stress by pressing the surface with the tip of an atomic force microscope. We show that mechanical switching in sol-gel grown ferroelectric PbZr_xTi_1-xO₃ (x=0.2) thin films is possible under different conditions of mechanical stress up to large thicknesses (200 nm). We vary the experimental conditions that allow effective mechanical switching of layers of different thicknesses, and combine electrical and mechanical switching to modify the threshold force needed to switch the polarization. The results are supported by the cross-sectional observation of the films by scanning transmission electron microscopy (STEM), which brings information about the nanoscale structure and chemistry of the films.

Our results suggest two ways of controlling the threshold force. Firstly, the prior application of an electric field would modify the value needed to switch domains mechanically. This effect is probably due to charge trapping that decreases the magnitude of the electric field created by the applied stress. Secondly, the mechanical switching is affected by the presence of nanometer-scale cavities within the sample, either aligned in specific regions of the film or randomly distributed as evidenced by STEM images, which modify the boundary conditions of the domains created by mechanical stress and their stability. The combination of both external electrical field and cavities tends to show that it is actually possible to tune the force threshold to suit the desired purpose. We will discuss these results with the aim to provide a larger understanding of means to exploit the mechanical switching in integrated ferroelectric devices. [4]




[1[ H. Lu, et al., Science, 335(6077):59–61, (2012).
[2] D. Edwards et al, Adv. Mater. Interfaces, 3, 1500470 (2016)
[3] Y. Cao, et al., Physical Review B 96, (2017)
[4] S. Gonzalez-Casal et al. In preparation.

Ferroelectric oxide thin films and superlattices have attracted broad interest on account of their novel physical and functional properties. The strong coupling between strain and polarization in these materials allows mechanical and electrical boundary conditions to be utilized to induce novel polar phases, unusual polarization textures and surprising behaviour, as exemplified by the negative capacitance phenomenon. In such heterostructures, an epitaxial strain is usually induced through a lattice-parameter mismatch between the epitaxial layers and a monocrystalline substrate. This substrate-imposed strain is an invaluable tool, which plays a crucial role in determining the domain structures and the resulting properties; however, it also presents significant limitations to accessing a wealth of yet unexplored phase transitions in complex oxide heterostructures and hinders their applications in flexible electronics and integration in semiconductor electronic circuits.

With the development of new methods for releasing thin films from the underlying substrate, free-standing complex oxide heterostructures become excellent candidates to explore and harness the full potential of strain engineering. In this work, ferroelectric-dielectric superlattices with ordered nanoscale domains were released from the substrate, allowing the thin-film oxide components to share the strain elastically with each other. The free-standing superlattices were studied with a combination of laboratory and synchrotron x-ray diffraction, piezoresponse force microscopy, transmission electron microscopy and Raman spectroscopy. Our measurements show that the spatial distributions of strain are introduced, which lead to the three-dimensional deformation of films. The released superlattices adopt a new, complex polarisation structure with an unexpected local anisotropy, which is correlated with the local film curvature.

Relaxor ferroelectrics, such as (1-x)Pb(Mg_1/3Nb_2/3)O₃-(x)PbTiO₃ (PMN-xPT), exhibit giant piezoelectricity which arises from the presence of multiple symmetries near a Morphotropic Phase Boundary (MPB). The MPB in PMN-xPT lies around x = 35% and separates a Rhombohedral (R) phase, with spontaneous polarization along the <111> crystallographic directions, from a Tetragonal (T) phase whose polarization lies along <100>. For compositions near the MPB, the presence of bridging Monoclinic (M) and Orthorhombic (O) phases help to facilitate large lattice distortions as the spontaneous polarization rotates between the competing symmetries in order to align with the applied field.
Bulk studies using (011) oriented PMN-PT single crystals have been used to demonstrate non-volatile 71/109 degree switching between in-plane and out-of-plane R polarization directions. Such switching can result in distinct strain states and forms the basis of low-power, ultrafast, memory storage devices when coupled with a ferromagnetic (FM) material [1,2]. Achieving low power, however, necessitates the use of thin films of PMN-PT. Additionally, the films must be removed from their substrates in order to prevent substrate clamping that eliminates their giant piezoelectricity. Here we demonstrate growth of epitaxial (011) PMN-PT thin films, followed by removal of the substrate via etching of a sacrificial layer. By coupling with a FM Nickel overlayer, we achieve manipulation of in-plane magnetic anisotropy under 3V applied bias across the PMN-PT film, which is much lower than the >100V needed using bulk PMN-PT. However, we do not see evidence of the 71/109 degree switching as observed in the bulk studies. Instead we find that the piezo-driven coupling is dominated by polarization rotation from the nominally <111> oriented R state towards the [011] O state along the applied field direction. Our findings have significant implications on the design of strain-mediated devices based on relaxor-ferroelectrics such as PMN-PT.

[1] Wu, T.; Bur, A.; Zhao, P.; Mohanchandra, K. P.; Wong, K.; Wang, K. L.; Lynch, C. S.; Carman, G. P. Appl. Phys. Lett. 2011, 98 (1), 012504
[2] Hu, J.-M.; Li, Z.; Chen, L.-Q.; Nan, C.-W. Nat. Commun. 2011, 2, 553

Ferroelectric materials can host a wider range of novel functional properties as well as unusual structural features, potentially useful for nanoelectronics applications. At domain walls or in regions with high strain gradients, in particular, the complex interaction between polarisation, electrostatics, and strain can lead to localised chiral polarisation textures, electrical conductivity, and charge or chemical segregation. One promising highly strained structure is the intersection between ferroelastic twins.

In Pb(Zr0.2Ti0.8)O3 thin films, investigating such twin domain intersections with a combination of scanning probe microscopy and nonlinear optical microscopy, we find a characteristic localised piezoelectric and response and evidence of a more complex rotational or closure structure in second harmonic generation polarimetry analysis. The heart of the twin domain crossing presents an extremely high susceptibility to local application of electric bias or pressure, dominating the polarisation switching dynamics in this region. We also observe distinct mechanical properties and enhanced electrical conduction at the intersection.

Application of HfO₂-based films to ferroelectric memory and logic devices has generated considerable interest as they allow overcoming significant problems associated with poor compatibility of perovskite ferroelectrics with CMOS processing. However, detailed studies of such application-relevant properties as imprint and polarization switching dynamics with respect to the electrode material and processing condition are still sparse. Here, we use a combination of Piezoresponse Force Microscopy (PFM) and pulse switching techniques to analyze the time- and field-dependent evolution of the domain structure in HfO₂-based thin film capacitors. Switching spectroscopy-PFM (SS-PFM) maps revealed the electrode-dependent spatial variations in the local potential landscape, which strongly affect the domain switching kinetics. It is shown that stronger oxidation reduces the internal imprint bias while also leading to an increase in the remanent polarization. Development of “fluid imprint” determined by the sample switching pre-history is reported. A particularly slow polarization reversal in the absence of an external field - termed as inertial switching - due to the interface entrapment of the charge injected during pulse application is shown to have a strong impact on polarization retention.

Current research effort in transient negative capacitance phenomena is focussed on attempts to separate out the components arising from change in polarisation (dP/dt) in a Ferroelectric material from circuit components such as those expressed by the Miller model [1][2]. The phenomenon of transient negative capacitance, despite having completely disparate origins in the case of multi-domain as opposed to single-domain ferroelectric, remains closely interlinked with negative capacitance as governed by the Landau-Khalatnikov (LK) equation [3]. The single-domain regime, which is governed by the LK equation, can be expressed by a series R-C circuit in parallel with an oxide capacitor Cox, where C is non-linear and can attain negative values for a certain range of polarisation. In contrast, the multi-domain approximation is modelled by the Kolmogorov-Avrami-Ishibashi (KAI) model, which utilises a time-dependent polarisation [4] or the Miller model [5], which akin to the LK framework, can be equivalently described as a series R-C circuit. However, C in the Miller model remains strictly greater than 0. In different variations of such models at least one of the elements between R and C is always considered non-linear (with respect to polarisation) [6] to help explain the steep switching behaviour.
Here we establish that the phenomenon of transient negative capacitance can be considered to be more general than reported earlier. We demonstrate the conditions for sub-60 mV/dec switching in an RC-FET, even if the R and C were constant. The voltage dropped across an equivalent series RC circuit can be represented as Vox=R (dQ_ch/dt)+(Q_ch/C) , where V_ox and Q_ch are the voltage across gate dielectric and sheet charge density in the channel, respectively. This leads us to the necessary condition for achieving a body factor m<1 along both forward and backward sweeps. However, when combined with the semiconductor charge (dQ_ch)/(dΨ_s )=(q/(k_BT))Q_ch , where Ψs is the surface potential, it is seen that sub-60 mV/dec switching is possible only if Qch >0 (i.e. when the transistor is ON) during the backward sweep. We believe this insight contributes further understanding to the causes of hysteresis in commonly used SPICE models of FE-FETs.
References: [1] J. Gomez et al., “Hysteresis-free negative capacitance in the multi-domain scenario for logic applications,” Tech. Dig. - Int. Electron Devices Meet. IEDM, vol. 2019-Decem, no. i, pp. 138–141, 2019. [2] J. Gomez et al., “Significance of Multi and Few Domain Ferroelectric Switching Dynamics for Steep-Slope Non-Hysteretic Ferroelectric Field Effect Transistor,” in 2019 Device Research Conference (DRC), 2019, vol. 2019-June, no. 574, pp. 247–248. [3] L. D. Landau and I. M. Khalatnikov, “On the anomalous absorption of sound near a second order phase transition point,” in Dokl. Akad. Nauk SSSR, vol. 96, 1954, pp. 469–472. [4] Y. J. Kim et al., “Voltage Drop in a Ferroelectric Single Layer Capacitor by Retarded Domain Nucleation,” Nano Lett., vol. 17, no. 12, pp. 7796–7802, 2017. [5] M. N. K. Alam, P. Roussel, M. Heyns, and J. Van Houdt, “Positive non-linear capacitance: the origin of the steep subthreshold-slope in ferroelectric FETs,” Sci. Rep., vol. 9, no. 1, pp. 1–9, 2019. [6] M. A. Alam, M. Si, and P. D. Ye, “A critical review of recent progress on negative capacitance field-effect transistors,” Appl. Phys. Lett., vol. 114, no. 9, 2019.

In this work, we present the first experimental determination of nucleation time and domain wall (DW) velocity by studying switching dynamics of ferroelectric (FE) hafnium zirconium oxide (HZO) according to the nucleation limited switching (NLS) model in ferroelectricity. Experimental data and simulation results were used to quantitatively study the switching dynamics. The switch speed can be degraded in high aspect ratio devices due to the longer DW propagation time or with dielectric interfacial layer due to the required additional tunneling and trapping time by the leakage current assist switch mechanism. The work is in collaborations with Xiao Lyu, Mengwei Si from Purdue University, Pragya R. Shrestha, Kin P. Cheung from NIST, Panni Wang, Shimeng Yu from GaTech.

Since the initial proposal [1] of using ferroelectric negative capacitance to realize voltage transform for low-power transistors, substantial theoretical and experimental progress has been made in determining their fundamental properties and realizing negative capacitance transistors with improved device characteristics. Nevertheless, substantial gaps remain between observed performance and model expectations.

Most pressingly, model analyses [2,3] seem to imply substantial limits on the ability of ferroelectric negative capacitance to improve the subthreshold swing of well-designed transistors. Due to the small subthreshold gate capacitances of well-designed transistors, these models suggest that negative capacitance enhanced lowering of the subthreshold swing is only achievable with rather thick ferroelectric layers, at scales impractical for large-scale integration.

A central conclusion of these models is that the design space of negative capacitance FETs is constrained largely by the quantum capacitance of the channel, and thus that lowering the quantum capacitance of the channel allows for an expanded design space. In doing so, these models make the implicit assumption that the quantum capacitance of the channel remains positive. However, lowering the quantum capacitance exposes the channel to the prospect of negative quantum capacitance [4], itself a phenomenon of interest in enhancing low-power electronic device operation. Negative quantum capacitance can arise [5] in channels with strong electron correlation effects at a low electron density in a number of materials, including silicon.

Here, we discuss the effects of negative quantum capacitance in FET channels on the operations of ferroelectric negative capacitance FETs. In particular, we find that for small negative quantum capacitance values, the subthreshold swing begins exhibiting hysteresis. While hysteresis could be desirable in some neuromorphic applications, it puts a lower limit on the density of charge carriers in the channel for conventional low-power electronics. However, we further find that for larger magnitudes of the negative quantum capacitance, hysteresis again disappears from the subthreshold swing. This adds a new region of design space to the fundamental model of ferroelectric negative capacitance FETs. We discuss the specific regime at which large negative quantum capacitance can provide steep-slope subthreshold switching, and examine the implications for ferroelectric negative capacitance FET design.

[1] S. Salahuddin and S. Datta, Nano Lett. 8, 405 (2008)
[2] M. Kobayashi and T. Hiramoto, AIP Adv. 6, 025113 (2016)
[3] W. Cao and K. Banerjee, Nat. Comm. 11, 196 (2020)
[4] T. Kopp and J. Mannhart, J. Appl. Phys. 106, 064504 (2009)
[5] P. Tsipas et al., Adv. Elec. Mat. 2, 1500297 (2016)

Epitaxial ferroelectric thin films are currently investigated for their potential in photovoltaic (PV) applications [1], mainly because of their large open circuit voltage and switchable photovoltaic effect. Indeed, in a ferroelectric film, the sign of the photovoltaic current can be controlled by the direction of the ferroelectric polarization, allowing in theory to achieve a 100% switchability of the photocurrent with the polarization, which is of great interest for potential applications, as photoferroelectric memories. However, switchability is not always achievable in integrated ferroelectric films between electrodes, because of extrinsic parameters, such as the nature of the electrode-ferroelectric interface (Schottky barrier) [2] or the amount of non-mobile charged defects. In addition, switchable photocurrent can also be affected by the migration of charged defects, such as oxygen vacancies, under applied electric field [3]. So, disentangling the impact of defects, electrodes, and polarization on the PV properties is not an easy task.
In this work, a careful study of the switchability of the PV properties of epitaxial lead zirconate titanate (PZT) thin films has been conducted in order to investigate the role played by the ferroelectric polarization. 100 nm thick PZT films were grown using pulsed laser deposition (PLD) and integrated in capacitor geometry between SrRuO3 bottom and Pt top electrodes and the photoinduced current in the PZT devices were studied under UV illumination (above the PZT band gap), in different polarization states. A voltage pulse protocol was used to pole the device under increasing electric fields to reach different electrical states while measuring their polarization value. These results revealed the critical role of the depolarizing field as the driving force for the photocurrent and allowed us to extract to screening efficiency of electrodes. In addition, the contribution from electrode-ferroelectric interface was evidenced for particular photon energy, especially with ITO top electrodes.

[1] C. Paillard, X. Bai, I. C. Infante, M. Guennou, G. Geneste, M. Alexe, J. Kreisel, and B. Dkhil, Photovoltaics with Ferroelectrics: Current Status and Beyond, Adv. Mater. 2016, 28, 5153–5168

[2] L. Pintilie, C. Dragoi, I. Pintilie, Interface controlled photovoltaic effect in epitaxial Pb(Zr,Ti)O3 films with tetragonal structure, J. Appl. Phys. 2011, 110, 044105.

[3] Y. Guo, B. Guo, W. Dong, H. Li, H. Liu, Evidence for oxygen vacancy or ferroelectric polarization induced switchable diode and photovoltaic effects in BiFeO3 based thin films, Nanotechnology2013,24,275201.

Heteroanionic materials have recently positioned in the center of multiple investigations. The latter is motivated by the physical, optical, and chemical properties that can emerge when properly combining different, but compatible anions, within the same compound. Oxynitride perovskite compounds have shown to offer a wide range of different tunable properties depending on the O/N rate, for instance, the absorption of light in the visible regime makes them ideal candidates for photocatalysis applications where polar compounds offer a notable advantage [1].
In this study, we theoretically explored the origin of the polar and ferroelectric responses experimentally observed in perovskite oxynitrides SrNbO₂N and SrTaO₂N [2,3]. With this aim, first-principles calculations of the structural and vibrational properties were performed. We show that such ferroelectric behavior is extremely sensitive to the anionic ordering in both Sr(Nb,Ta)O₂N systems, where cis- and trans- orderings are identified as the most probable orderings in the crystals. Furthermore, We found that both cis- and trans- type can hold ferroelectricity, although the mechanisms in how such a stable phase emerges in each ordering are quite different. Aditionally, we show that spontaneous polarization characterizing such ferroelectric phases can be considerably enhanced by applying in-plane epitaxial strain, achieved by using strontium titanate as substrate.

[1] Fuertes, A. (2015). “Metal oxynitrides as emerging materials with photocatalytic and electronic properties”. Materials Horizons, 2(5), 453-461.
[2] Kim, Y. I., Woodward, P. M., Baba-Kishi, K. Z., & Tai, C. W. (2004). “Characterization of the structural, optical, and dielectric properties of oxynitride perovskites AMO₂N (A= Ba, Sr, Ca; M= Ta, Nb)”. Chemistry of materials, 16(7), 1267-1276.
[3] Kikkawa, S., Sun, S., Masubuchi, Y., Nagamine, Y., & Shibahara, T. (2016). “Ferroelectric response induced in cis-type anion ordered SrTaO₂N oxynitride perovskite”. Chemistry of Materials, 28(5), 1312-1317.

In condensed matter, Ruddlesden-Popper (RP), considered as an intergrowth of n-layers ABX₃ perovskites with an AX layer with structure: (AO)(ABO₃) have attracted a lot of attention due to their potential applications. As such, perovskites-like materials are notably at the vanguard of technological developments thanks to their versatility offered by the partial substitution of anionic sites as in the case of oxygen replacement by nitrogen sites (i.e. oxynitrides). These substitutions possible due to the similarities in electronegativity, polarizability, ionic radius, and coordination states and bringing modifications of their physical properties such as electrical, optical, catalytic, and magnetic responses [1]. Therefore, the study and research on new RP oxynitrides phases constitute an area of huge interest.

Here we present significance outcomes due to partial substitution of nitrogen for oxygen in structures Sr₂AO_4-xN_x(A=Nb, Ta)(x=0.5, 1), that show symmetry inequivalent structures that I have analyzed, These structures are more stable thermodynamically compared with oxides and nitrides structures, and the most stable is a trans-type configuration structure. The direct bandgap is reduced to benefit the performance of photocatalysis. Likewise, by researching the relationship of vibrational behaviour of trans-type configuration, it becomes evident that these RP-phases hold a ferroelectric response.

First-principles calculation reveals that strontium oxynitrides RP-phases are strongly correlated materials due to underestimation in the bandgap. So we have used, instead DFT, DFT+U for consistent results.

[1] Inorg. Chem. 2004, 43, 25, 8010–8017

The ternary oxides of the compound family with the general formula A₂B₂O₇, where A is a trivalent cation and B is a transition metal, have been widely studied because of their interesting physical properties and their technological applications. Among the relatively less explored members of this family are the rare-earth titanates (RE₂Ti₂O₇), which have a monoclinic layered perovskite-type structure and have been of interest for their ferroelectric, piezoelectric, non-linear optical, and photocatalytic properties and their high Curie temperature. Here we present our low-temperature dielectric study of high structural quality, stoichiometric single-crystal Pr₂Ti₂O₇. This compound has a monoclinic polar crystal structure (P2₁) that consists of four layers of corner-sharing TiO₆ octahedra separated by two layers of Pr cations. We will review the crystal structure and some of the magnetic and ferroelectric properties of the compound. We will then discuss our dielectric and magnetodielectric results, in particular the field-dependent anomalies in the dielectric constant and the mutual coupling between dielectric properties and magnetization in this compound.

The work in Houston is supported by US Air Force Office of Scientific Research Grants FA9550-15-1-0236 and FA9550-20-1-0068, the T. L. L. Temple Foundation, the John J. and Rebecca Moores Endowment, and the State of Texas through the Texas Center for Superconductivity at the University of Houston.
The work in Johns Hopkins University is supported as part of the Institute for Quantum Matter, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences under Award No. DE-SC0019331.

Magnetoelectric multiferroic materials have received increasing attention by the research community since they simultaneously possess magnetization and ferroelectricity and the two order parameters are coupled by magnetoelectric effects. BiFeO₃ (BFO), one of the few room-temperature magnetoelectric multiferroic materials, is of special interest. For instance, BFO can be potentially used in the magnetoelectric spin-orbit (MESO) device¹, spin-orbit torque field-effect transistor (SOTFET)², etc. However, to date, most studies on BFO have been focused on the electric field induced switching instead of magnetic field driven behavior. Despite difficulties in directly switching ferroelectricity in BFO by an external magnetic field, the exchange interaction between BFO and an adjacent ferromagnetic (FM) material³ might enable a new way for switching by applying a magnetic stimulus via spin-orbit torque if the materials parameters fall in the right window, as proposed in the modeling work of SOTFET². In this work, a SrRuO₃(SRO)/BFO/SRO heterostructure is used to study the ferroelectric properties and magnetic-induced behavior of BFO.
The SRO/BFO/SRO (20/100/20nm) heterostructure was grown epitaxially by molecular-beam epitaxy (MBE) on (110) DyScO₃ substrates using similar growth conditions in [4]. No ion irradiation was done post growth. Ti/Pt (5/50nm) was blanket deposited by sputtering, and subsequently patterned into top metal contacts of capacitors by photolithography and ion milling. Large area Ti/Pt top contacts are effectively shorted to the bottom SRO layer thus serving as the opposite electrode of the capacitors.
First, we discovered that these MBE BFO layers are highly insulating, unlike the ones reported in [4], where post-growth MeV-ion irradiation had to be applied to curb the leakage through polarization domains. The low leakage in these MBE BFO films directly enabled us to study its ferroelectric properties by the positive-up-negative-down (PUND) measurement at both RT and 5 K, a common method to obtain polarization-electric field (P-E) loops by measuring the displacement and switching currents. Conductive AFM was also carried out: very low current level (~10 pA up to 10 V) was measured across the sample and no leakage pattern was observed. The PUND data revealed a saturation polarization (P_S) of 60-80 μC/cm² and a coercive field of 0.6-1 MV/cm, but no clear dependence on temperature.
Next, we applied out-of-plane (oop) or in-plane (ip) magnetic fields up to 6 T, respectively, during the PUND measurements, at both RT and 5 K to investigate if an external magnetic field can influence the measured P-E loop in BFO. At RT, the SRO layers serve as simple metal electrodes; while at 5 K, we expect the SRO layers will be polarized by the external magnetic fields since SRO has a ferromagnetic Curie temperature of ~150 K. In this BFO/FM heterostructure, no noticeable shift in the P-E loop is observed at 5 K under the magnetic field condition used in this study. At RT, a systematic shift in the P-E loops was observed; however, after carefully correlating the shift with the order of measurements, we believe the shift arose primarily from the ferroelectric imprint effect not the magnetic field. Overall, these observations are consistent with our modeling predictions². Due to the relatively large thickness (100 nm) and P_S (~70 μC/cm²), the switching energy barrier in BFO is too high. If the coupling strength between the two order parameters remains the same or increases, it is more likely to observe magnetic-induced behavior in thin La-doped BFO with reduced P_S.
[1] Manipatruni, et al, Nature 565, 35(2019). [2] Li, et al, APL 116, 242405(2020). [3] Heron, et al, Nature 516, 370(2014). [4] Mei, et al, APL Mater. 7, 071101(2019).
* This work was supported in part by the SRC nCORE task 2758.001 and NSF E2CDA program (ECCS 1740286). This work was performed in part at the CNF, an NNCI member supported by NSF Grant NNCI-2025233.

Magnetic refrigeration (MR) is a crucial subject as a future cooling technology based on magnetocaloric effect (MCE) that replaces the conventional vapor compression technique for energy saving, high efficiency, minimal environmental impact, and reliability. MR of magnetic materials is based on the change in entropy when the materials are subjected to the field. Therefore for more efficient, materials should have large entropy changes. Spinel ferrites are one of the essential magnetocaloric materials that consist of transition metal ions (MFe₂O₄, M = Mn, Zn, Co, Ni, Cu, and Mg). Among different spinel ferrites, the Mg-Zn spinel ferrite is considered necessary due to its remarkable magnetic and electrical properties. The MCE in ferrite is observed ~ 1.36 at 350K. The current effort is to broaden the MCE operating temperature. In view of the above, in this study, a comparative study of magnetocaloric effect (MCE) in Mg_0.5Zn_0.125A_0.375Fe₂O₄, A = Co, Cu, and Ni is reported. The samples were prepared via auto-combustion method and heat-treated in the air at 1100^oC for 12 h. The phase identification of the powders performed using x-ray diffraction shows that all the samples exhibit a high-purity cubic spinel structure. The lattice parameter is affected due to different ionic radii of substituted elements. With Co and Ni substitution, a gradual shift in the peaks to the left as compared to Cu is observed that showed a decrease in lattice parameters from 8.510 Å to 8.356 Å due to smaller ionic radii of Co²⁺ ion (0.65 Å), Ni²⁺ ion (0.69 Å) than Cu²⁺ ion (0.73 Å). Morphology and size of the Mg_0.5Zn_0.125A_0.375Fe₂O₄, A = Co, Cu, and Ni particles were investigated by scanning electron microscopy (SEM). SEM shows the particle diameter was ~ 40 – 170 nm, indicating that the nanoparticles were polycrystalline. Magnetic measurements were performed using a Quantum Design Superconducting Quantum Interference Device (SQUID) magnetometer. The magnetization as a function of the temperature for all sample, under an applied magnetic field of 100Oe, were measured to see the phase transition. When the temperature decreased, the samples exhibited a paramagnetic (PM) - ferromagnetic (FM) transition at the Curie temperature (T_C). The transition temperatures increased from 550 K to 600 K for Cu to Co. Ni substituted sample showed transition temperature in between Cu and Co. Saturation magnetization for Mg_0.5Zn_0.125A_0.375Fe₂O₄, A= Cu, Ni, and Co reached 23 emu/g, 27emu/g, and 33 emu/g at 10K. Large magnetic properties for Co is due to the superexchange interaction and magnetic coupling. To evaluate the MCE of the samples, the isothermal magnetization curves were collected at different fixed temperatures ranging from 300 to 700 K. the temperature measurements are at intervals T=10K. The magnetic entropy change ΔS_M has been calculated from the M–H curves using the Maxwell relation. The maximum entropy change |-ΔS_M^Max| as a function of temperature increases from 1.25 J.Kg^-1K^-1 to 3.16 J.Kg^-1K^-1 under an applied magnetic field of 5T Cu to Co substitution. The peak position of ΔS_M vs. T curve was independent with applied fields for the same material, whereas the peak shifted towards lower temperature as for Cu sample and higher temperature for Co sample compare to Ni sample. This magnetocaloric behavior is mainly due to superexchange interaction between Fe – Fe ions and particle size. These values are relatively high and comparable to other ferrite systems. Arrot plots showed the positive slope of the curve for all samples indicating a second-order phase transition. Cooling efficiency can be calculated from the peak position of ΔS_M called relative cooling power (RCP). The RCP was found to be 51 to 126 J/Kg for Cu, 62 to 135 J/kg for Ni, and 76 to 165 J/kg for Co sample calculated from 0.5T to 5T, respectively. Significantly higher ΔS_M^Max and curie temperature make our compounds promising materials for magnetic refrigeration technology for higher temperatures.

Highly (001) oriented 0.70Pb(Zr_0.52Ti_0.48)O₃-0.30Pb(Fe_0.5Nb_0.5)O₃(0.70PZT-0.30PFN) magnetic ferroelectric thin films were deposited on La_0.67Sr_0.33MnO₃ (LSMO) buffer layer coated on (LaAlO₃)_0.3(Sr₂AlTaO₆)_0.7 (001) substrates by following two subsequent laser ablation processes in oxygen atmosphere employing pulse laser deposition technique. The 0.70PZT-0.30PFN films were found to grow in tetragonal phase with orientation along (001) plane as inferred from x-ray diffractometry analysis. The temperature dependent dielectric measurements (80-750 K) were performed on metal-ferroelectric-metal heterostructure capacitors in the frequencies range of 10²–10⁶ Hz was observed to be diffused over a wide range of temperature 400–700 K and exhibits high dielectric constant ~2200 at room temperature. The well saturated slim loop of thin films capacitors suggesting nano-scale ferroelectric behaviors (relaxor type) of 0.70PZT-0.30PFN thin films in line with diffuse dielectric response results. An excellent high energy storage density (Ure) ~51 J/cm3 with efficiency ~62 % was estimated at applied voltage 1.5 MV/cm. High DC breakdown strength, larger dielectric constant and high restored energy density values of our 0.70PZT-0.30PFN ferroelectric thin films indicate its usage in high energy storage applications. These results will be presented in details at the meeting.

Clean and renewable solar energy is regarded as one of the most reliable and abundant sources to tackle the prevalent energy crisis. The photovoltaic effect, directly converting sunlight into electricity, is an important way to harvest solar energy, and many efforts have been devoted to the development of photovoltaic technology. Toward the next-generation photovoltaics, multiferroic materials (MFs), with the coexistence of ferroelectricity and magnetism in the same phase, show promise as candidates for harvesting energy, especially from sunlight. The spontaneous polarisation of such type of non-centrosymmetric materials is the origin of charge carrier separation. On the other hand, magnetic ordering induces a small bandgap, typically due to electronic states of the magnetic atom in the Fermi energy region. One method to find these rare materials is to employ double perovskites (DPs) with general formula A₂BB'X₆, in which one of the sublattices is magnetic, and the other one is ferroelectric. This study intends to develop and apply theoretical tools to predict and improve the photovoltaic properties in new double perovskites with multiferroic properties. In that light, the structural, magnetic, electronic, and photovoltaic properties of new MF-DPs A₂MnBO₆ (A=Sn, Bi and B=V, Ta, Ti) using ab initio simulations based on DFT method are systematically investigated. The electronic band structure and density of states demonstrate the semiconducting behaviour of these double perovskites with the values of band gaps falling within the desired range for optimal photovoltaic performance. Structural calculations, including distortion parameters and bond angles, indicate that these DPs undergo a significant octahedral tilting and distortion which may result in a considerable value of electric polarization and hence, may be suggested as potential multiferroic materials. In the MnO₆ octahedra, Jahn-Teller effect, which is related to the Mn³⁺ electron configuration, distorts the Mn-O bonds. Whereas, in Ta(Ti)O₆ octahedra, the Second Order Jahn-Teller effect originating from d⁰ electron configuration of Ta⁵⁺(Ti⁴⁺) is responsible for the off-centering displacement of Ta(Ti) atoms and the distortion of Ta(Ti)O₆ octahedra. Optical properties, including absorption coefficient (α) calculations, show that the value of α in the visible region is in the order of 10⁵ cm^-1, comparable to the absorption coefficient of some well-known photovoltaic materials. These results draw an intense theoretical research interest to explore the photovoltaic properties of new multiferroic double perovskites to characterize them as potential solar materials.

The discovery of ferroelectricity in doped hafnium oxide thin films revived the interest in ferroelectric (FE) memory concepts [1]. Compared to other dopants, zirconium doped hafnium oxide (HZO) crystallizes at low temperatures (e.g. T_cryst. < 400°C), which makes this material ideal for the implementation of ferroelectric functionalities into the Back-End-of-Line (BEoL) [2]. For this approach, metal-ferroelectric-metal (MFM) capacitors are very interesting. Integrated in the BEoL, they can be connected either to the drain- or gate-contact of a standard logic located in the Front-End-of-Line (FEoL) to realize ferroelectric random access memories (FeRAMs) or ferroelectric field effect transistors (FeFETs), respectively. It is beneficial to deposit the top electrode of the MFM stack by physical vapor deposition (PVD) processes like sputtering. It is fast, can be done at room temperature, and is a low-cost and reproducible process.
Here, we present the impact of the top TiN electrode, deposited by reactive magnetron sputtering using a metallic titanium (Ti) target and nitrogen / argon plasma, on the ferroelectric properties of HZO films embedded in MFM capacitors. By using different nitrogen (N2)- and argon (Ar) flows, TiN film with various compositions (measured by X-ray-photoelectron spectroscopy) and thicknesses (measured by ellipsometry) have been realized. The crystallinity and different phases have been determined by X-ray diffraction (GI-XRD). The influence of the mentioned parameters on the ferroelectric properties of the HZO film are discussed by taking additional effects like film stress into account. Furthermore, the low temperature PVD TiN is compared with high temperature chemical vapor deposited (CVD) TiN.

[1] T. S. Böscke et al., Appl. Phys. Lett. 2011, 99, 102903.
[2] D. Lehninger et al., pss(a) 2020, 217, 1900840

Field Effect Transistors with a ferroelectric gate stack provide a negative capacitance (NC) effect and show a sub 60mV/decade sub-threshold slope (SS) [1—3]. This improves the cell delay and total power (using supply voltage scaling) without traditional Moore’s law scaling. [4—5] study the impact of the ferroelectric properties (coercive field E_c and polarization P₀) at transistor, gate, and full-chip level. Using a wide range of non-hysteresis E_c(1, 4)MV/cm, P₀(10, 40)μC/cm² values, we introduce the idea of optimality of ferroelectrics (those within 5% of the lowest achievable power at given voltage while meeting target delay). An open-source 15nm FinFET with 0.8V supply voltage is used for target delay baselines and to generate the NCFET librarues.

Transistor Level analysis at V_DD=0.4V shows sub-threshold slope(SS) and I_on improves as the (E_c, P₀) is closer to hysteresis boundary (this is the curve beyond which NCFETs have hysteretic behavior). NC effect increases as we move closer to the boundary, and higher I_on and SS do not coincide as a strong NC effect also has a smaller range of operation in terms of V_gs. So, a steep SS caused by large NC effect saturates quickly, while a slightly smaller slope remains active for a longer range of V_gs achieving higher I_on. Gate level analysis inverter provides of power and delay metric evolution.

Gate Level: The voltage amplification effect [1] of NCFET decreases the V_th and devices go to saturation at a smaller |V_gs|. This increases the saturation mode of operation to a wider range of V_gs and the voltage transfer curves become smoother. So, short-circuit current is higher at large SS, and gate capacitance (∝ I_on) is higher when I_on is large. At low V_DD, the NC effect does not saturate for V_gs within the operation range (0, V_DD) and the effects at hysteresis boundary vary due to this. NC saturation also causes the delay to increase with V_DD in extreme cases counter-intuitive to the traditional delay-V_DD curves. The optimality region also changes with V_DD as a higher NC effect is required to meet target delay when VDD is decreased for NCFETs. At iso-delay, we achieve 92% reduction in total power using a V_DD=0.2V and a high NC effect ferroelectric. The same NCFET creates 87% reduction (best observed) in full chip power analysis as well.

Not all benefits at gate level translate to a full-chip level. Using various types of cells in full chip optimization, the NCFETs that lie out of the optimal range in gate-level analysis can be within the wider optimal range of full-chip. While a subset of the optimal range in full-chip can be estimated using inverter level analysis, different cells behave differently with NC effect, and a full chip design with a large number of cells and good variation in types of cells gives a detailed optimal NCFET ranges for a given supply voltage and technology.

[1] S. Salahuddin and S. Datta. Use of Negative Capacitance to Provide Voltage Amplification for Low Power Nanoscale Devices, 2008 Nano Lett., doi: 10.1021/nl071804g
[2] K. Li et al., "Negative-Capacitance FinFET Inverter, Ring Oscillator, SRAM Cell, and Ft," 2018 IEEE International Electron Devices Meeting (IEDM), doi: 10.1109/IEDM.2018.8614521
[3] Z. Krivokapic et al., "14nm Ferroelectric FinFET technology with steep subthreshold slope for ultra low power applications," 2017 IEEE International Electron Devices Meeting (IEDM), doi: 10.1109/IEDM.2017.8268393
[4] S. Pentapati et al., "Cross-Domain Optimization of Ferroelectric Parameters for Negative Capacitance Transistors—Part I: Constant Supply Voltage," in IEEE Transactions on Electron Devices, Jan. 2020, doi: 10.1109/TED.2019.2955018.
[5] S. Pentapati et al., "Optimal Ferroelectric Parameters for Negative Capacitance Field-Effect Transistors Based on Full-Chip Implementations—Part II: Scaling of the Supply Voltage," in IEEE Transactions on Electron Devices, Jan. 2020, doi: 10.1109/TED.2019.2955010.

We analyze the impact of drain current (I_DS) variation in 28 nm high-K metal-gate Ferroelectric FET devices on FeFET-based processing-in-memory (PIM) deep neural network accelerators. Non-Normal variation in I_DS is observed due to repeated read operation on two devices with different channel dimensions at various read frequencies. Device-circuit co-analysis using the measured current distribution shows a 1 to 3 percent accuracy degradation of an FeFET-based PIM platform when classifying the Fashion-MNIST dataset with the LeNET-5 DNN model. This accuracy drop can be fully recovered with variation-aware training methods, showing that FeFET current variation is not prohibitive to the design of DNN accelerators.

In the research field of memory devices, ferroelectric memory has been attracting much attentions because of the newly discovered ferroelectric HfO2 (FE-HfO2) material which is fully CMOS compatible [1]. Ferroelectric memory device can inherently operate at high-speed and low-power. One-transistor memory, ferroelectric FET (FeFET), has a small footprint and non-distructive read-out [2], which is promising for high-density storage memory application. Recently, 3D stacked FeFET has been proposed and demonstrated[3], which can possibly compete with NAND flash memory in terms of memory capacity and power consumption.
There are, however, several challenges to realize such 3D stacked FeFET using conventional poly-Si channel. Poly-Si channel has low mobility and forms a low-k interfacial layer which causes voltage loss and reliability degradation by charge trapping. In this work, we propose to use amorphous oxide semiconductor (AOS), IGZO [4,5], as an alternative channel material. IGZO channel provides high mobility as a deposited channel and prevent low-k interfacial layer formation. IGZO has been already in production for flat panel display, and high reliability has been already proved.
We fabricated FE-HfO2 capacitor by capping with IGZO layer. IGZO layer works as preferable capping for HfZrO2 and induces ferroelectric orthorhombic phase. Electrical characerization shows sharp hysteresis and large spontaneous polarization charge, which can be explained by process-induced strain by IGZO capping. FE- HfO2 capacitor with IGZO capping also shows wake-up free endurance characteristics, which can be due to the engineered itnerface between the FE-HfO2 and IGZO layer avoiding interface defect such as oxygen vaccancy [6]. Then we fabricated FeFET with IGZO channel, in which we introduced back gate. Back gate can help to fix the IGZO body potential so that the elecric field is effectively applied to the FE-HfO2 layer for erase operation. The field-effect mobility is as high as 10cm2/Vs, which is the same as Hall mobility and the mobility of transistor with SiO2 gate insulator. There is no mobility degradation with FE-HfO2. DC I-V sweep and pulse write operation both show >0.5V memory window with almost ideal subthreshold swing. This subthrshold characteristics is attributed to the low interface trap (<1011cm2) and bulk defect, and junctionless operation. These results are promising for high-density and low power storage memory application.
[1] J. Muller et al., Nano Lett. 12, pp. 4318−4323 (2012), [2] J. Muller et al., IEDM Tech. Dig., pp. 280-283 (2013), [3] K. Florent et al., IEDM Tech. Dig., pp. 43-46 (2018), [4] K. Nomura et al., Nature, 432, 25, pp.
488-492 (2002), [5] F. Mo et al., VLSI Tech. Symp., pp. 42-43 (2019), [6] F. Mo et al., Appl. Phys. Expr., 13,
074005 (2020).

Ferroelectric materials offer significant promise for nonvolatile memories and low-power logic transistors [1] due to stable polarization states which can be reversed under an applied electric field. With the recent discovery of ferroelectricity in HfO₂ [2], fluorite-structure binary oxides have attracted considerable interest as they are compatible advanced semiconductor processes, unlike prototypical perovskite-structure ferroelectrics. Accordingly, ultrathin HfO₂-based ferroelectrics have received significant attention in charge-based ferroelectric random-access memory (FeRAM), ferroelectric field-effect transistors (FeFETs), and even negative capacitance FETs [3]. Meanwhile, resistive-switching materials have emerged as promising candidates for novel beyond-CMOS computing paradigms [1]. In this context, ferroelectric tunnel junctions (FTJs) present a promising energy-efficient resistive switching memory as FTJs exploit the polarization-dependent tunneling barriers across thin ferroelectric barriers. However, a critical requirement for FTJs is to achieve a sufficiently high tunneling current while still exhibiting a large ON/OFF ratio (tunneling electroresistance), which requires robust ferroelectricity as the tunneling barrier is reduced to the ultrathin limit, presenting a materials challenge preventing realistic FTJ-based memories.

In this work, we demonstrate FTJs with ultrathin Zr-doped HfO₂ (Zr: HfO₂) ferroelectric barriers down to a thickness corresponding to two unit cells, grown by atomic layer deposition on silicon [4,5]. These tunnel junctions exhibit large polarization-driven electroresistance (19000 %), the largest value reported for HfO₂-based FTJs. In addition, due to the ultrathin ferroelectric barrier, these junctions provide large tunneling current (> 1 A cm^-2) at low read voltage, orders of magnitude larger than reported thicker HfO₂-based FTJs. Furthermore, the data retention (>10⁴ s) and endurance characteristics (> 10³ cycles) in these ultrathin HfO_2-based FTJs are comparable to those obtained with much thicker HfO₂-based ferroelectric layers. Scanning probe microscopy techniques – piezoresponse force microscopy (PFM) and conductive atomic force microscopy (C-AFM) – help establish polarization-driven switching is indeed responsible for the observed resistive switching. Thus, the ultrathin ferroelectric barriers not only allow robust memory operation, but also enable simultaneous occurrence of large electroresistance and large current, critical for practical applications.

[1] S Salahuddin, K Ni & S Datta. “The era of hyper-scaling in electronics.” Nat. Electron. 1, 442–450 (2018).
[2] TS Böscke et al. Ferroelectricity in hafnium oxide thin films. Appl. Phys. Lett. 99, 102903 (2011)
[3] D Kwon, S Cheema, N Shanker […] S Salahuddin. “Negative Capacitance FET with 1.8-nm-Thick Zr-Doped HfO₂ Oxide.” IEEE Electron Device Lett. 40, 993–996 (2019).
[4] S. Cheema*, N. Shanker* […] S. Salahuddin. “One nanometer HfO₂-based ferroelectric tunnel junctions on Si.” arXiv 2007.06182 (2020).
[5] S Cheema, D Kwon, N Shanker […] S Salahuddin. “Enhanced ferroelectricity in ultrathin films grown directly on silicon.” Nature 580, 478–482 (2020).

The superior electrostatic control and high scalability of nanowire (NW) and nanosheet (NS) gate-all-around field-effect-transistors (FETs) make them promising alternatives to FinFETs in advanced technology nodes. Negative capacitance (NC) effect, originating from a ferroelectric (FE) material in the gate stack, can provide an elegant solution for the much-needed voltage scaling in these aggressively scaled devices. In this work, a comparative analysis on the performance of NC-NWFETs and NC-NSFETs is presented through fully calibrated, three-dimensional TCAD simulations. For the same layout footprint (LF), both single channel NC-NSFETs and those, with vertically stacked NSs, have been considered. Single channel NC-NSFET exhibits 9% lower subthreshold swing (SS) and 35% higher ON-current (I_ON) than NC-NWFET of comparable device dimensions. In contrast to NC-NWFET, capacitance matching between the FE and the underlying metal-oxide-semiconductor (MOS) capacitance is achieved with a thinner FE layer in NC-NSFET. This is particularly significant, since a thinner FE layer enables further scaling of these NC devices. In addition, NC-NSFET, with vertically stacked NSs, can achieve 4× higher I_ON and ~50% lower SS than NC-NWFET, owing to higher effective width and better capacitance matching. A high I_ON/I_OFF ratio of 10⁶, with low operating voltage (V_DD = 0.2V), is obtained for the NC-NSFET. The significantly higher I_ON in NC-NSFET also results in faster switching, with significantly lower power consumption, at V_DD < 0.3 V. The performance of stacked NC-NSFETs can be optimized by tuning the width and thickness of the NSs. Negative differential resistance (NDR) is found to be more pronounced in NC-NSFET, enabling these devices to attain a stronger drain-induced-barrier-rising (DIBR) and steeper SS for gate lengths (L_G) as small as 10 nm. The results presented here pave the way for performance optimization of NC-NWFETs and NC-NSFETs, in ultra-scaled and high-density logic applications, for 10 nm and beyond technology node.

MoTe₂ channel-based P(VDF-TrFE) ferroelectric nonvolatile memory is fabricated, which operates at minimum switching pulse voltage and minimum drain voltage. For the minimum switching voltage of 8 V, bottom-gate architecture is employed and its advantages are investigated. By using bottom-gate structure, we could avoid a dead layer formed at the interface between thermally-deposited Al and P(VDF-TrFE) at top-gate architecture. A dead layer in top-gated ferroelectric memory transistors increases the coercive voltage so as the switching pulse voltage. And, for the minimum drain voltage, a novel method of H₂O₂ treatment is developed. By oxidizing the source/drain area of MoTe₂ surface by H₂O₂ solution, Ohmic contact between Pt and MoTe₂ is achieved even without thermal annealing which would have a destructive effect on the crystal quality of P(VDF-TrFE). To demonstrate the benefit of our memory transistor in aspect of power saving, it is integrated into an OLED operating circuit.

Ferroelectric HfO₂ is a promising material for new memory devices, but the microstructure of the films needs to be better controlled and some properties, mainly endurance, need to be improved. Research on ferroelectric HfO₂ has been focused mainly on polycrystalline films. In contrast, epitaxial films, of great interest to understand properties and prototyping devices, are scarcely investigated. The recently achieved stabilization of the orthorhombic ferroelectric phase in epitaxial Hf_0.5Zr_0.5O₂ films on perovskite La_0.7Sr_0.3MnO₃ electrodes has allowed to, among other results, control the crystalline polymorphs through substrate selection, achieve high polarization, endurance and retention in sub-5 nm films, and fabricate epitaxial ferroelectric tunnel junctions. Studies on epitaxial films are still in a nascent state, and little is known about the effects of dopants and interfaces, whose critical influence has been demonstrated in polycrystalline films. We will show the relevance to the stabilized phases and ferroelectric properties (polarization, endurance and retention) in epitaxial HfO₂-based capacitors of i) substrates and electrodes, and ii) chemical composition of the HfO₂-based films.

Ferroelectric field effect transistors based on hafnium oxide have become a viable option as embedded non-volatile memory solution due to its high coercive field as well as its compatibility to complementary metal-oxide-semiconductor (CMOS) processes [1]. In comparison to ferroelectric capacitors, these devices comprise an additional interface layer, which impacts the electrical field distribution in the gate stack and yields a depolarization field counteracting the polarization of the ferroelectric, affecting properties like endurance and retention [2,3]. In consequence, this is expected to affect the wake-up behaviour of the hafnium oxide layer. Additionally, this layer is already present during the crystallization of the ferroelectric at high temperature. As diffusion processes cannot be excluded at these temperatures, influences on the crystallization behaviour are expected.
Here, we present the impact of different annealing temperatures on the layer stack. Diffusion processes inside the layer stack during the crystallization anneal were investigated using time-of-flight secondary ion mass spectroscopy (ToF-SIMS). For high annealing temperatures, changes in the chemical composition of the interface layer were observed. Furthermore, electrical characterization of the ferroelectric properties revealed a strong impact of the annealing temperature, which resulted in an increased presence of the wake-up effect for lower annealing temperatures. Additional transmission Kikuchi diffraction (TKD) analysis revealed strong differences in the microstructure and crystallographic texture, which are most likely responsible for the observed behaviour.
[1] Müller, J. et al.; ECS J. Solid State Sci. Technol. 4, N30-N35 (2015).
[2] Müller, J. et al.; NVMTS, (2016).
[3] Gong, N. and Ma, T.-P.;, IEEE Electron Device Lett. 37, 1123 (2016).

With the progress in wireless networks, including the upcoming 5G and Internet of Things, there is a high concern in exploring of materials for passive or active RF devices. The main criteria are good CMOS-compatibility, low tuning voltage, low chip size and low power consumption. Ferroelectric materials, like barium strontium titanate (BST) [1] have been widely used as tunable capacitors - varactors. However they lack CMOS compatibility and require relatively high tuning voltages. Hafnium zirconium oxide (HZO) is a fully CMOS-compatible material [2], with a good back-end-of-line integration possibility due to its low thermal budget [3]. It is a good candidate for varactor applications [4], with a high miniaturization possibility [5],[6].
Here, we investigate the capacitance-voltage (CV) characteristics of Hf_xZr_1-xO₂ MFM thin film capacitors with various Zr doping, thicknesses and annealing temperatures. The impact of field cycling during the wake-up on the tunability was analyzed and an optimized bias region for the maximum tunability was determined. Additionally, the effect of antiferroelectric-like (AFE) behavior on tuning was investigated. The superposition of ferroelectric and AFE regime shows an interesting behavior, where less bias for tuning is needed, but tunability is reduced. A usefulness of such a behavior for varactor application was also examined. We have also investigated the ferroelectric and antiferroelectric-like properties at elevated temperaures. It was shown that with increase of temperature tunability deteriorates. Temperaure measurements also comply with recent studies of ferroelastic nature of AFE behavior [7].

[1] N. K. Pervez et al., Appl. Phys. Lett. 85, 19, 4451, (2004).
[2] T. S. Böscke et al., Appl. Phys. Lett. 99, 10, 102903, (2011).
[3] D. Lehninger et al., pss(a), 217(8), 1900840, (2020)
[4] M. Dragoman et al., Appl. Phys. Lett. 110, 10, 103104, (2017).
[5] S. Abdulazhanov et al., IEEE MTT-S IMWS-AMP., p.46 (2019).
[6] S. Abdulazhanov et al., IEEE MTT-S IMWS-AMP., p.175 (2019).
[7] M.Lederer et.al., Appl. Phys. Lett. 115, 22, 222902, (2019)

Introduction Despite the remarkable development in ferroelectric FETs, endurance and reliability of these devices still represent serious concerns in both memory and logic applications. In this work, we present recent results discussing the reliability of HfO₂-based FeFETs [1] employed as either non-volatile memory elements or as steep-switching NCFET logic transistors. For FeFET, we analyze the role of interface and bulk trap generation on memory window closure leading to limited endurance. For NCFET, we discuss the beneficial role of negative capacitance (NC) effect in reducing the persistent reliability issue of negative bias temperature instability (NBTI). We utilize a common theoretical framework based on the Landau Theory to describe ferroelectric energy landscape that allows capturing: i) the hysteretic I-V characteristics of FeFETs when employed for memory applications; and ii) the NC effect when aiming at steep-slope (hysteresis-free) switching for logic transistors [2].
Endurance Limits of FeFETs Endurance is the total number of repeated program/erase cycles after which the states of a memory become indistinguishable. One of the main factors limiting the endurance of FeFETs is the generation of traps at the interfacial layer inserted in the gate stack between the ferroelectric and semiconductor body. In [3], we developed an experimentally validated analytical model to describe the effects of aging on the memory window (MW). This model captures the degradation of programmed/erased states depending on the amount of generated traps that in turn is related to the amplitude and duration of the writing pulses. As such, the analytical model could serve either as an add-on to traditional techniques or as a stand-alone method to characterize generated defect densities under a variety of stress conditions. For instance, it is possible to estimate the net generated traps from the MW expression derived in [3], thus allowing to correlate MW measurements with generated traps. The estimation drawn from the simple analytical model can be helpful to develop next-generation FeFETs with improved endurance.
NBTI Suppression In NC-FETs The primary mechanism for NBTI-induced degradation in MOSFETs is the increase in interface trap density with stress time. The NBTI degradation can be fully compensated by adding a ferroelectric (FE) layer of proper thickness for stabilized NC operation. This enables steep-slope, hysteresis-free operation needed for switching. NBTI reduction comes from the extra capacitance given by the parasitic source/drain overlap regions that cause the capacitance matching (between the negative capacitance and the intrinsic MOSFET gate capacitance) to improve over stress time. Since achieving an optimum FE layer thickness needed for obtaining NBTI-free operation can be challenging, we performed simulations for various thicknesses to find that even for relatively thin FE (e.g., 7 nm) NBTI degradation is reduced [4]. In other words, even if the FE thickness needs to be determined exclusively by process integration and capacitance matching constraints, it would still lead to NBTI reduction.
Conclusions Reliability aspects in ferroelectric FETs for both memory and logic applications were investigated by means of either analytical models or self-consistent numerical simulations based on the framework based on Landau Theory. The modeling approach presented in this work allows gaining useful insights into how to improve reliability of ferroelectric FETs and accelerate the path towards widespread commercialization.
Acknowlegments Authors thank prof. L. Selmi (Univ. of Modena and Reggio Emilia) for useful discussion. This research has received partial funding from the European Commission's Horizon 2020 BeFerroSynaptic project, grant agreement n°871737 via the IU-NET consortium.
References [1] J. Muller et al., IEDM, 2013.[2] M. A. Alam et al., APL, 114, 2019. [3] N. Zagni et al., APL, 117, 2020. [4] K. Karda et al., TED, 67, 2020.

Ferroelectric tunnel junction (FTJ) memories have gained prominence among emerging non-volatile memories due to low power consumption, high switching speed, and potential for neuromorphic applications [1]. Recent development of doped hafnium oxide ferroelectric layer enables integration of FTJ memory devices with CMOS technology. In FTJ devices, the tunneling current is directly dependent on the remnant polarization of the ferroelectric layer. While the tunneling current is higher across a thin ferroelectric layer (around 1-4 nm), it is challenging to have a high remnant polarization in these layers and simultaneously achieve a high tunneling electro resistance (TER) ratio. FTJs with a composite stack of Metal-Ferroelectric-Dielectric-Metal, allow the use of a thick ferroelectric layer (~10 nm) while obtaining high TER, as demonstrated recently [2, 3]. However, the crystallization of the ferroelectric layer is carried out at high temperatures (500-600°C) which are incompatible with the CMOS back-end-of-line. In this work, we demonstrate a highly CMOS backend compatible FTJ stack with a 400°C crystallized Hf_0.5Zr_0.5O₂ (HZO) ferroelectric layer and TiN-Al₂O₃-HZO-W stack. Such a stack with a TiN bottom metal and a W top metal allow seamless integration with the CMOS back-end-of-line, particularly at contact level and has not yet been explored. The FTJ devices show on-current and TER that are comparable to the state-of-the-art. The devices also show multi-level resistance behavior necessary for neuromorphic applications. We will also discuss a detailed investigation of the switching transients, voltage, and device resistance dependence on switching cycles. As the composite stack is similar to those used in the study of negative capacitance, the polarization switching behavior of the device is of importance to a broader community. Furthermore, we will also discuss device characteristic dependence on dielectric thickness and identify parameters to improve the TER, on-state current and retention.

[1] M. Benjamin et al., “Ferroelectric Tunnel junction based on ferroelectric – dielectric Hf_0.5Zr_0.5O₂/Al₂O₃ capacitor stacks,’’ European Solid-State Device Research Conference (ESSDERC), pp. 142-145, 2018.
[2] M. Benjamin et al., “Direct correlation ferroelectric properties and memories characteristics in ferroelectric tunnel junctions,’’ IEEE Journal of Electron Device Society (J-EDS), vol. 7, pp. 1175-1181, 2019.
[3] R. Hojoon et al., “Ferroelectric tunnel junctions based on aluminium oxide/zirconium-doped hafnium oxide for neuromorphic computing,” Scientific Reports, vol. 9, pp. 1-8, 2019.

Hafnia-based films are changing the way we think of ferroelectric switching. Ferroelectricity in these materials arises from metastable phases that are easier stabilized at the smallest dimensions. In addition they present low leakage as well as CMOS compatibility [1], making them ideal candidates for memory and logic devices. In addition, their switching takes place, quite uniquely, without involving domain wall motion, which allows experimental access to negative capacitance states[2]. Multiferroic tunnel junctions (MTJs) fabricated with (La,Sr)MnO₃ electrodes and ferroelectric Hf_0.5Zr_0.5O₂ barriers show both tunneling magnetoresistance (TMR) and tunneling electroresistance effect (TER), displaying four resistance states by magnetic and electric field switching[3]. Moreover, under electric field cycling, the TER effect can reach values as large as 10⁶%. Experiments indicated that polarization switching alone cannot be responsible for those changes[4]. In this talk we will show direct evidence, by means of operando transmission electron microscopy[5], as well as synchrotron x-ray diffraction experiments with in-situ application of electric fields[6], of the mechanisms that come into play during electric field switching in hafnia-based devices, as well as their relative importance determining the properties of these devices.

[1] U. Schroeder, C. S. Hwang, and H. Funakubo, Ferroelectricity in doped hafnium oxide: materials, properties and devices,Woodhead Publishing, 2019
[2] M. Hoffmann, F.P.G. Fengler, M. Herzig, T. Mittmann, B. Max, U. Schroeder, R. Negrea, P. Lucian, S. Slesazeck & T. Mikolajick , Nature 565, 464 (2019).
[3] Y. Wei, S. Matzen, T. Maroutian, G. Agnus, M. Salverda, P. Nukala, Q. Chen, J. Ye, P. Lecoeur, and B. Noheda, “Magnetic tunnel junctions based on ferroelectric Hf_0.5Zr_0.5O₂ tunnel barriers,” Physical Review Applied 12, 031001 (2019)
[4] Y. Wei, S. Matzen, C. P. Quinteros, T. Maroutian, G. Agnus, P. Lecoeur, B. Noheda “Magneto-ionic control of spin polarization in magnetic tunnel junctions”, npj Quantum Materials 4, 62 (2019).
[5] P. Nukala, M. Ahmadi, Y. Wei, S. de Graaf, S. Matzen, H. W. Zandbergen, B. Kooi, B. Noheda, “Operando observation of reversible oxygen migration and phase transitions in ferroelectric devices”, arXiv:2010.10849 (submitted)
[6] P. Nukala, G. Carbone, E. Stylianidis, R. Hamming Green, M. Salverda, Y. Wei, A. Burema, T. Banerjee , A. Bjorling, D. Mannix, S. Matzen, B. Noheda, (in preparation)

Ferroelectric and antiferroelectric oxides have gained attraction in recent years for use as gate oxides in field effect transistors (FETs), along with potential applications toward random access memories and high-density energy storage, owing to their hysteretic charge-voltage characteristics and negative capacitance phenomena. Exploiting this negative capacitance can lead to lower operating voltages, and therefore, operation of transistors beyond their current limits. The origin of negative capacitance is thought to stem from the energy barrier of first order field-induced phase transitions. In the case of antiferroelectric materials, interesting non-linearities in their polarization-voltage characteristics (i.e. double hysteresis loops) are observed and attributed to an electric field-induced first-order, structural phase transition between a non-polar, parent phase and a polar active phase. However, direct observation and identification of such phase transitions and their relation to ferroelectric/antiferroelectric properties has yet to be unanimously determined, leaving significant gaps in our fundamental understanding of the microstructure-property relationship of such materials. Since polarization correlates with microstructure, the application of an electric field alters the microscopic features, enabling electrical properties suitable for non-volatile memory, analog synapses, and artificial neurons, but also poses significant challenges to performance such as variability, reliability, and endurance.
The first step in identifying the crystallographic pathways involved in polarization switching of antiferroelectric ZrO₂-based thin films is the direct imaging of the polarization switching at the atomic scale with applied bias via in situ transmission electron microscopy (TEM). In-situ TEM characterization can provide simultaneous biasing and imaging of atomic structure in real time. Field-induced phase transitions and microstructural evolution of antiferroelectric zirconia are investigated via high resolution, in-situ TEM biasing, coupled with DFT-calculated atomic models, to identify the origin of antiferroelectricity in these materials. Such characterization techniques highlight the ability of advanced electron microscopy to provide insight into structure-property relations of fluorite-type binary FE/AFE oxides, thereby improving our fundamental understanding of the microscopic mechanisms behind device performance for the advancement of logic and memory technologies.

Polar vortices in oxide superlattices exhibit complex polarization topologies. In this presentation, I will talk about using a combination of energy loss near-edge structure analysis, crystal field multiplet theory, and first-principles calculations to probe the electronic structure within such polar vortices in [(PbTiO₃)₁₆/(SrTiO₃)₁₆] superlattices, at the atomic scale. The peaks in Ti -edge EEL spectra shift systematically depending on the position of the Ti⁴⁺ cations within the vortices i.e., the direction and magnitude of the local dipole. First-principles computation of the local projected density of states on the Ti orbitals, together with the simulated crystal field multiplet spectra derived from first-principles are in good agreement with the experiments. This combined experimental and theoretical approach can serve as a fundamental basis to study macroscopic properties such as chirality and circular dichroism.

Negative capacitance [1] has emerged as a promising solution to overcome fundamental energy-efficiency limits in conventional electronics, in which internal ferroelectric order within the gate stack of a field-effect transistor can enable low-power operation [2,3]. Thus far, negative capacitance has been primarily demonstrated in thick perovskite- and fluorite-structure ferroelectric thin films, but integration into advanced semiconductor technology nodes will require stabilization at the ultrathin regime. Here, we present evidence of negative capacitance in atomic-scale HfO₂-ZrO₂ heterostructures [4] down to two-nanometers thickness, leveraging our recent work examining the 2D limits of fluorite-structure ferroelectricity on silicon [5-7]. HfO₂-ZrO₂-HfO₂ trilayers on Si-SiO₂ demonstrate sub-7 Å effective oxide thickness (EOT) without having to scavenge the SiO₂ interlayer, in stark contrast to conventional high-κ metal gate (HKMG) technology. Accordingly, in comparison to industrial HKMG benchmarks, these SiO₂-buffered ferroic heterostructures boast the lowest reported leakage for such aggressively-scaled EOT gate oxides, possible via negative capacitance.

In contrast to previous reports of negative capacitance in thicker ferroelectric HfO₂-based films [8], the microscopic origin of negative capacitance in these ultrathin HfO₂-ZrO₂ multilayers arises from mixed ferroelectric-antiferroelectric order. Remarkably, this fluorite-structure family fosters the stabilization and competition of such ferroic order at the atomic-scale regime [5-7]. We first demonstrate ultrathin ferroic order – a crucial requirement for negative capacitance – down to one-nanometer in doped-HfO₂ films [5,6] and five-angstroms in conventionally-antiferroelectric ZrO₂ films [7]. To establish polarization switching in these ultrathin films, hysteretic characterization – via scanning probe, electrical tunnel junction and dielectric measurements – in various device geometries (in-plane, out-of-plane) diagnose competing antiferroelectric-ferroelectric order. In conjunction, electron microscopy and synchrotron X-ray characterization monitor structural signatures of the competing ferroic phases. These results indicate not only the absence of a ferroelectric critical thickness, but also ultrathin-enhanced polar distortion in fluorite-structure thin films [5,7], critical breakthroughs towards exploiting ferroic-based phenomena at ultra-scaled dimensions. Finally, pulsed electrical measurements and Landau phenomenology help explain the underlying origins of negative capacitance in this mixed-ferroic system. Therefore, this work not only broadens the ferroic origins of negative capacitance beyond ferroelectricity, but also allows one to design negative capacitance in atomic-scale HfO₂-ZrO₂ heterostructures on silicon [4], paving the way towards realistic ferroelectric-based computing.

[1] S Salahuddin & S Datta. “Use of Negative Capacitance to Provide Voltage Amplification for Low Power Nanoscale Devices. Nano Lett. 8, 405–410 (2008).
[2] D Kwon^*, S Cheema^* […] S Salahuddin. “Negative Capacitance FET with 1.8-nm-Thick Zr-Doped HfO₂ Oxide.” IEEE Electron Device Lett. 40, 993–996 (2019).
[3] D Kwon^*, S Cheema^* […] S Salahuddin. “Near Threshold Capacitance Matching in a Negative Capacitance FET with 1 nm Effective Oxide Thickness Gate Stack.” IEEE Electron Device Lett. 41, 179–182 (2020).
[4] S Cheema […] S Salahuddin. “Negative capacitance in atomic-scale HfO₂-ZrO₂ ferroic heterostructures for advanced transistors.” [in preparation]
[5] S Cheema^*, D Kwon^* […] S Salahuddin. “Enhanced ferroelectricity in ultrathin films grown directly on silicon.” Nature 580, 478–482 (2020).
[6] S Cheema […] S Salahuddin. “One nanometer HfO₂-based ferroelectric tunnel junctions on silicon.” arXiv 2007.06182.
[7] S Cheema […] S Salahuddin. “Emergence of sub-nanometer mixed antiferroelectric-ferroelectric order on silicon.” [in preparation]
[8] M Hoffmann et al. Nature 565, 464–467 (2019).

Ferroelectric HfO2 is being investigated for neuromorphic computing and back-end memory due to its scalability and compatibility with the current CMOS processing. However, for low power operation, low switching voltage (sub +/-1V) is desired. To achieve this, thinner FE HZO films are required along with no interfacial oxide/dead layers. In this study, Ni(top)/HZO/W/TiN MIM capacitors grown at high temperature were explored to achieve low voltage operation. Sputtered W/TiN bottom electrodes were deposited on a Si wafer. Samples were annealed in-situ prior to deposition in UHV at 350oC for 30 min in homebuilt ALD tool. HZO was deposited using TDMAH, TDMAZ and H2O ALD precursors at 350°C substrate temperature. 50 nm Ni top electrodes were deposited using thermal evaporation. Oxide thickness was measured using ellipsometry on a Si witness sample. Samples were annealed at 450°C in N2 for 2min. Electrical characterization was performed using a Keithley 4200A-SCS characterization system.

A sample with 200 cycles HZO (12 nm) was fabricated and characterized. Cross sectional TEM measurements revealed an oxidized W film which is reduced at the W/HZO interface; this was confirm by EDS (not shown). This results in abrupt W/HZO interface and low voltage switching because there is no interfacial dielectric layer which inducing a voltage drop as observed in other electrode materials such as TiN and Si. The sample shows steep switching and high 2Pr of 41μC/cm² at +/-5V. To reduce the operating voltage, the device was scale. 100 cycles of HZO were deposited; ellipsometry was used to confirm the thickness as 6 nm. Endurance test was performed at 1MHz +/-1.8V (~3 MV/cm) to study the device performance over time.

The device showed a 2Pr = 39 uC/cm2 at +/-1.8V (3MV/cm) and a maximum 2Pr = 62 uC/cm2 at +/- 3.1V along with sharp switching I-V. The voltage scaling with thickness, as shown by the comparison of two devices is consistent with the absence of interfacial oxides. The endurance performed is superior to typical TiN/HZO/TiN structures and is sufficient for many applications. Furthermore, the devices do not show any wake-up which is beneficial since this complicates operation in a practical circuit. The absence of wake-up is consistent with the absence of oxygen vacancies at the W/HZO interface because the HZO ALD was performed on oxidized W film which was reduced to W metal during high-temperature ALD; confirmed by EDS (not shown). It is well known that oxygen vacancies at the electrodes result in the pinning of FE grains which reduces Pr. By operating with subloops with less than full Pr, +/-1V operation can be achieved and this will further increase the endurance.

Relaxor ferroelectrics are characterized by diffuse dielectric and piezoelectric properties, which distinguishes them from traditional ferroelectrics. To formulate models of local polar behavior, the structural and chemical complexity in relaxor systems have largely been probed with diffuse scattering using X-rays, neutrons, and electrons. These techniques, however, provide a measure of the average global and local structure across a relatively large volume of the material. Consequently, the origin of relaxor properties continues to be debated although these materials have been studied extensively. In this talk, I will discuss how aberration corrected scanning transmission electron microscopy (STEM) can be used to directly separate nanoscale structural and chemical inhomogeneities in relaxors. Here, we apply these techniques to Pb(Mg_1/3Nb_2/3)O3-PbTiO3 (PMN-PT). Through simultaneous acquisition of images that are sensitive to chemistry (angle annular dark-field STEM) and light elements (integrated differential phase contrast STEM), we directly connect nanoscale chemical order regions, distorted oxygen octahedra, and local polarization. We find that the degree of chemical order smoothly varies within ordered domains and approaches a minimum at anti-phase boundaries, as well as regions of correlated oxygen octahedral titling are found to be anti-correlated with regions of maximal chemical order. Comparing with the projected polarization, we observe that the regions of greatest variation in polarization correspond to the regions of maximum chemical order and maximum octahedral distortion. Based on these results, we show that both structural and chemical inhomogeneities act as a barrier for polarization rotation and thus frustrate long range polar order.

The recent discovery that with certain dopants hafnium oxide can be crystallized in a form that exhibits ferroelectricity has opened up a range of device possibilities(1). We chose to work within the Hf/Zr system for our initial studies. The advantages of the Hf/Zr doping system include the similarity in precursors which can enable several running modes for the ALD process including a co-pulsing scenario, as well a cycle ratio scenario for controlling the Zr doping. In addition, the relative range of doping for which a strong remnant polarization is observed is much broader for the Hf/Zr system than for other dopants such as Al, Si, etc. which means this system should be more tolerant of small variations in composition, temperature and thickness than the other doping schemes. These factors along with a relatively low thermal budget suggest the CMOS-friendly Hf/Zr system could provide the basis for a number of ferroelectric memory devices. Other potential uses for these films include the possibility to create multi-domain ferroelectric FETs that can be used within a neuromorphic device to mimic synapse behavior. This talk will review and update our work to date on these films and their potential uses. References: 1. T. S. Böescke, J. Müller, D. Bräuhaus, U. Schröder and U. Böttger, in Technical Digest of the International Electron Devices Meeting, p. 24.5.1 (2011).

Ferroelectric materials have been explored for device structures such as ferroelectric random-access memory, ferroelectric tunnel junction, and ferroelectric field-effect transistors (FeFETs) [1]-[3]. Among the device structures, the FeFETs have been considered as one of the most versatile devices to implement the non-volatile memory and in-memory compute. It has low write energy as it is electric field driven, non-destructive read-out mechanism and relatively fast program/erase speed [3]. The discovery of ferroelectric material based on doped HfO₂ has made FeFETs more attractive. Hafnia based ferroelectric material has good compatibility with the complementary metal-oxide-semiconductor (CMOS) fabrication process. Even more, its ferroelectric properties could be controlled depending on the dopant materials and doping concentration, which enables fine-tuning of FeFETs’ performance.
Industrial development of FeFETs stays at 28nm/22nm today [4] [5], primarily because high voltage (~3V) is required to program the polarization states in the conventional ferroelectric gate MFIS stack (metal/doped HfO₂/interfacial layer/silicon). The unmatched charge between the Fe-cap and MOS-cap creates enhanced electric field in the interfacial layer, resulting in trap generation, reliability degradation and exacerbated variability.
Making FeFETs compatible with 7nm FinFET-like structure, if successful, could enable the emerged memory-logic circuit designs that take the advantages of high-performance 7nm logic transistors. Scaling down the ferroelectric layer thickness while maintaining ferroelectric properties is especially critical. In a recent research [6], 1~2 nanometers of the hafnia-zirconia-based ferroelectric layer were demonstrated. GlobalFoundries prototyped high-k/metal-gate bulk FeFETs at 28 nm [4] and fully-depleted SOI (FDSOI) FeFETs at 22 nm using silicon doped HfO₂ based material [5].
In this work, we simulated FeFETs from 22 nm node of FDSOI to 7 nm node of ferroelectric FinFET (Fe-FinFET) using a Sentaurus TCAD to assess its scalability. Simulation parameters including ferroelectric properties were fitted with experimental results of a 22nm FDSOI FeFET for accurate expectation and future suggestion. Quasi-static I-V characteristics of the FeFET were modeled using the multi-domain Preisach model built in TCAD and domain phase variations were considered. Eventually, the design guideline for 7 nm technology node of Fe-FinFET was suggested with an aim to optimize the write voltage below 1.5V. The memory window of Fe-FinFET could be enhanced using innovative gate stack design and/or employing higher-k materials for an interlayer oxide.
Reference
[1] J. Okuno et al., “SoC compatible 1T1C FeRAM memory array based on ferroelectric Hf_0.5Zr_0.5O₂,” IEEE Symposium on VLSI technology and Circuits, virtual, June 2020.
[2] M. Kobayashi, Y. Tagawa, F. Mo, T. Saraya, and T. Hiramoto, “Ferroelectric HfO₂ Tunnel Junction Memory With High TER and Multi-Level Operation Featuring Metal Replacement Process,” IEEE J. Electron Devices Soc., vol. 7, pp. 134–139, 2019.
[3] E. Yurchuk et al., “Impact of Scaling on the Performance of HfO₂ -Based Ferroelectric Field Effect Transistors,” IEEE Trans. Electron Devices, vol. 61, no. 11, pp. 3699–3706, Nov. 2014.
[4] M. Trentzsch et al., “A 28nm HKMG super low power embedded NVM technology based on ferroelectric FETs,” IEEE International Electron Devices Meeting (IEDM), San Francisco, CA, USA, Dec. 2016, p. 11.5.1-11.5.4.
[5] S. Dunkel et al., “A FeFET based super-low-power ultra-fast embedded NVM technology for 22nm FDSOI and beyond,” IEEE International Electron Devices Meeting (IEDM), San Francisco, CA, USA, Dec. 2017, p. 19.7.1-19.7.4.
[6] S. S. Cheema et al., “Enhanced ferroelectricity in ultrathin films grown directly on silicon,” Nature, vol. 580, no. 7804, pp. 478–482, Apr. 2020.

Atomic stack models in ferroelectric (FE) hafnium-zirconium oxide (HZO) based on density functional theory have been developed to elucidate the atomic and electronic structure level origin of fatigue behavior, i.e., deteriorating memory window with field cycling. Stack models to mimic metal-FE-metal (MFM) capacitor and metal-FE-semiconductor FE field effect transistor (FEFET) were developed. There were three key findings. (1) For a metal in direct contact with the FE layer, the field inside the FE is eliminated and local potential profile inside the FE is unchanged upon polarization switching. Conversely, the field inside the FE is not completely eliminated by a relatively less polarizable semiconductor electrode, and there is an asymmetric potential profile due to polarization switching. (2) When a DE interlayer is placed at the interfaces in MFM and FEFET stacks, a significant internal field is induced in both cases. Strength of the field is dependent on the thickness of the FE layer, decreasing for thicker FE layers, and for ultrathin films (<4 nm) can exceed the breakdown voltage, consistent with difficulty in synthesizing FE HZO films thinner than 5 nm. (3) Since the field within the FE layer is reversed each time the polarization switches, this will promote charge-compensating defect generation. With a significant accumulation of defect density, most likely O vacancies, the FE phase of HZO would become thermodynamically unstable, and transition to a non-polar phase losing the spontaneous polarization. This would explain the fatigue behavior of the FE devices. The suggested mechanism would be more significant for a FEFET due to the lower polarizability of semiconductor vs metal electrodes and the spontaneous formation of insulating oxide DE interlayers, for example silicon oxide on silicon, which is consistent with unsatisfactory endurance for FEFET devices compared to MFM capacitors. The atomic models have developed a fundamental understanding to enable experimental optimization strategies for desired device performance.

The multidomain nature of ferroelectric (FE) polarization switching dynamics has recently attracted a great deal of research attention due to the discovery of the doped hafnium oxides as an FE material compatible with transistor fabrication processes. To study Zr-doped hafnium oxides (HZO), we develop a physics-based circuit compatible model for phase-field simulations, where the 3-dimensional time-dependent Ginzburg–Landau (TDGL) equation and Poisson’s equation are self-consistently solved with the SPICE simulator. Systematically calibrated based on the experimental measurements of metal–ferroelectric–metal (MFM) capacitors, the model well captures transient negative capacitance (NC) in pulse switching dynamics, with domain interaction and viscosity being the key factors. The influence of pulse amplitudes on voltage transient behaviors is found to be attributed to the fact that the FE free energy profile strongly depends on how the domains interact. In addition, we extract the domain viscosity dynamics during polarization switching according to the experimental measurements. For the first time, a physics-based circuit-compatible SPICE model for multidomain phase-field simulations is established to reveal the impact of microscopic domain interaction on the NC effect. The findings of this article may have important implications for the charge boost induced by the stabilization of NC in an FE/dielectric (DE) stack at a specific operating voltage or frequency.

Conventional multi-purpose hardware based on von-Neumann architecture does not satisfy the energy efficiency requirements of large-scale implementations of deep neural networks (DNN). Hardware accelerators that reduce data movement are therefore key to improve the power efficiency of many big data applications based on deep learning, such as image classification and speech recognition. Emerging non-volatile memory devices such as resistive random-access memory, phase change memory, floating gate memory and ferroelectric field effect transistors (FEFETs) are potential candidates for these DNN accelerators due to their synaptic functionality i.e. analog conductance modulation.
FeFET analog synapses are 3-terminal devices that have been proposed to improve the classification accuracy and yield low latency in non-volatile memory-based DNN accelerators. This is due to their high conductance ratio, operation capability with sub-100 ns pulse [1] and seamless integration with CMOS process flow.
This work presents a simulation framework to evaluate novel ferroelectric and anti-ferroelectric devices, such as doped HfO₂ ferroelectric-transistors, and their impact in system-level neuromorphic applications. The study uses a new, comprehensive, physics-based compact Verilog-A model, the MIT Virtual Source Ferroelectric FET (MVSFE) model that describes transistors with ferroelectric (FE)-oxides in their gate-stack. We expanded physics-based MIT Virtual Source model benchmarked against scaled 45nm [2] MOSFET to capture the physics of ferroelectric oxide using VDST-P [3,4]model calibrated against experimental FE HZO characteristics. The model captures underlying FET incorporating ballistic transport and charges for advanced technology node FeFETs (with or without internal gate). Our model extends the scope of [1] to the highly scaled L_g required for energy efficient system-level implementation of DNN [5] We also use the framework to show that engineering the device with multiple t_FE/V_T elements along width direction engineering is effective to improve system-level performance metrics (i.e. linearity and classification accuracy) of DNN built using analog synapses. The results provides insight on improved device design, and its system-level performance impact in a multilayer perceptron neural network.

[1] M. C. Jerry, P -Y.; Zhang, J.; Sharma, P.; Ni, K.; Yu,S.; Datta, S., "Ferroelectric FET Analog Synapse for Acceleration of Deep Neural Network Training," 2017.
[2] C. Auth, A. Cappellani, J. S. Chun, A. Dalis, A. Davis, T. Ghani, et al., "45nm high-k plus metal gate strain-enhanced transistors," 2008 Symposium on Vlsi Technology, pp. 99-+, 2008.
[3] B. Jiang, P. Zurcher, R. E. Jones, S. J. Gillespie, and J. C. Lee, "Computationally efficient ferroelectric capacitor model for circuit simulation," 1997 Symposium on Vlsi Technology, pp. 141-142, 1997.
[4] J. Chow, A. Sheikholeslami, J. S. Cross, and S. Masui, "A voltage-dependent switching-time (VDST) model of ferroelectric capacitors for low-voltage FeRAM circuits," 2004 Symposium on Vlsi Circuits, Digest of Technical Papers, pp. 448-449, 2004
[5] T. Gokmen and Y. Vlasov, "Acceleration of Deep Neural Network Training with Resistive Cross-Point Devices: Design Considerations," Front Neurosci, vol. 10, p. 333, 2016.

Nanoscale ferroelectric materials hold great promises in miniaturized device applications such as ferroelectric capacitor, transducers, actuators, sensors, photovoltaics, etc. Here we present first-principles theoretical predictions and understandings of three materials classes of two-dimensional ferroelectrics, including (a) semiconducting 2D group IV monochalcogenides [1], (b) semiconducting 2D multiferroic semiconductors in monolayer transition metal phosphorus chalcogenides (TMPCs) with coexisting ferroelectricity and ferromagnetism [2], and (c) semimetallic few-layer WTe₂ [3,4]. We show that monolayer group IV monochalcogenides hold highly anisotropic and large ferroelectric polarization with visible-spectrum excitonic gap and sizable exciton binding energy and strain-tunable ferroelectric transition barrier. We will further discuss our study and understanding of domain wall, phase transition and domain switching in 2D ferroelectric group IV monochalcogenide using first-principles based machine-learning force field developed in our group. In contrast to group IV monochalcogenides, monolayer TMPCs hold coexisting ferroelectricity and ferromagnetism where Cu atoms spontaneously move away from the center atomic plane and result in out-of-plane electric dipole moment, suggesting the possibility of controlling electric polarization by external vertical electric field. Finally, we will present our study of ferroelectric polarization in semimetallic few-layer WTe₂ where small interlayer sliding induces ferroelectric transition with small kinetic barrier, enabling facile ferroelectric switching upon electric gating and ferroelectric nonlinear anomalous Hall effect demonstrated in experiment very recently [4].

References: [1] 2D Materials 4, 015042 (2017). [2] Applied Physics Letters 113, 043102 (2018). [3] npj Computational Materials 5, 119 (2019). [4] Nature Physics 16, 1028-1034 (2020).

The integrating ferroelectric gate stack into FETs with negative capacitance (NC) effect [1][2] for subthreshold swing (SS) improvement has attracted lots of attention due to hafnium (Hf)-based oxide with ferroelectricity. The prospect of ferroelectric Hf-based oxide by ALD (Atomic Layer Deposition) has been wide and intensive studied due to lots of applications. The HfO₂ with suitable dopants and annealing for ferroelectric (FE) transition would be studied, and demonstrated NC effect for steep-slope FET (SS-FET) and memory applications [3][4].
Pursuing steep-SS with accompanying hysteresis is a challenge for NC-FET development, as well as the issue of asymmetric SS of bi-directional sweep [5][6]. Furthermore, the reduced NC onset voltage is another issue to boost steep SS at low operation bias [7]. The antiferroelectric (AFE) Hf_1-xZr_xO₂ with adopting as gate would overcome partial issues and make comparison with ferroelectric Hf_1-xZr_xO₂ [8][9]. Note that the FE and AFE characteristics of HfZrO₂-based is achieved by Zr incorporation content by ALD supercycle.
Device dimension scaling down, such as FinFET and GAA, to comparable with domain size of polycrystalline ferroelectric HfZrO₂ (HZO) is evaluated for SS and drain-induced barrier lowing (DIBL) [10][11]. The multi-domain modeling of polarization randomly located in HZO and the probability with Gaussian distribution HZO are confirmed by measurement data within limitation polarization, and presented NC effect for voltage amplification with opposite direction of E-field in FE and IL. The experimental data of FinFET and planar FET are used to validate the simulation results and the direction effect of domain quantity.
The compatible CMOS process and scaling film thickness are the advantages to integrate into semiconductor industry by comparison with perovskite material. The high scalability and CMOS-compatibility of ferroelectric Hf-based oxide is a key technology enabler for energy-efficient computing to serve for smart power management in IoT applications.

Acknowledgment
The authors are grateful for the funding support from the National Science Council (MOST 109-2213-E-003-003 and 109-2622-8-002-003), process supported by Taiwan Semiconductor Research Institute (TSRI), Nano Facility Center (NFC), and computing support was provided by the National Center for High-Performance Computing (NCHC), Taiwan.

References:
[1] S. Salahuddin et al, NanoLetters, 8, 405, 2008. [2] S. Salahuddin et al, in IEDM, 2008, 693. [3] M. H. Lee et al, in IEDM, 2015, 616. [4] K.-S. Li et al, in IEDM, 2015, 620. [5] W. Chung et al, in IEDM, 2017, 365. [6] M. H. Lee et al, in IEDM, 2017, 565. [7] M. H. Lee et al, in IEDM, 2016, 306. [8] X. Lyu et al, in VLSI Symp., 2019, T44. [9] M. H. Lee et al, in IEDM, 2019, 447. [10] M. H. Lee et al, in IEDM, 2018, 735. [11] K.-T. Chen et al, SST, 35, 125011, 2020.

HfO₂-based ferroelectric thin films are promising for memory and neuromorphic applications. Compared to conventional perovskite-structure ferroelectric materials, HfO₂-based ferroelectrics have excellent CMOS compatibility, and they are better able to meet back-end-of-line (BEOL) thermal budget requirements. In this work, we demonstrate that high polarization (P_r > 20 μC/cm²) Hf_0.5Zr_0.5O₂ (HZO) thin films are achieved by a rapid thermal anneal (RTA) at a very low temperature (350°C). Orthorhombic phase formation of HZO is studied as a function of RTA time.

In the ferroelectric HZO capacitor structure, 10 nm HZO is sandwiched between 10 nm top and bottom TiN electrodes. The HZO film is deposited by a plasma-enhanced atomic layer deposition (PE-ALD) process at 250°C. The bottom and top TiN are deposited by reactive sputtering of Ti in N₂ atmosphere. After top TiN deposition, the capacitors are annealed by RTA at 350°C in N₂ atmosphere from 30 to 600 seconds. Finally, Pt contacts are deposited on the TiN top electrodes to facilitate for electrical measurements. Capacitor areas are defined by etching Pt/TiN stack in SC1 solution.

All HZO capacitors display polarization switching after annealing at 350°C, even after only 30 s anneal. The remnant polarization (P_r) increases as the RTA time increases from 30 s to 120 s, and the P_r saturates for longer RTA times (P_r > 20 μC/cm²). For all devices, there is no evidence of an anti-ferroelectric hysteresis loop observed even in the as-processed (pristine) state, and no significant polarization “wake-up” effect is observed during voltage cycling. Grazing incidence X-ray diffraction (GI-XRD) is performed to analyze the phase evolution of the HZO thin film in this low temperature anneal process. The observed O(111) peak is consistent with formation of the orthorhombic phase. HZO films annealed for longer RTA time display higher O-phase peak intensities. Also, weak peaks indexed as the monoclinic phase begin to appear as the RTA time increases. Synchrotron XRD are performed for quantitative analysis on the phase evolution structure as a function of RTA time.

The ALD temperature used in deposition of the HZO film is a key factor enabling a low RTA thermal budget. HZO deposited at 200°C required a higher temperature for formation of films exhibiting a programmable polarization (> 500°C). Also, significant wake-up behavior is observed during voltage cycling, suggesting that the tetragonal phase and orthorhombic phase co-exist in the pristine state for films deposited at this lower ALD temperature, later transforming to the orthorhombic phase during cycling. Possible structural origins of this ALD temperature-induced change in phase formation sequence will be discussed.

Since the discovery of fluorite-structure oxides with ferroelectricity, especially doped-HfO₂, the ferroelectric field effect transistor (HfO₂-FeFET) has attracted huge attention as an emerging memory device. One of the serious challenges of the HfO₂-FeFETs is their limited endurance, which has been attributed to the degradation of the interface layer before the polarization fatigue of the HfO₂-ferroelectric layer. The main mechanisms of the interface layer degradation of FeFETs represented by metal-ferroelectric-insulator-semiconductor (MFIS) are charge injection/trapping and interfacial defect generation, which is mainly induced by high electric field through the interfacial dielectric layer. In this work, we propose strategies to minimize the interfacial electric field by approaching from two viewpoints of ferroelectric material properties and device structure. In this study, we present a structural approach and ferroelectric characteristics of a device that can have a larger dielectric capacitance compared to ferroelectric capacitance to improve the memory window and reliability characteristics. When it comes to 1T-1C type FeRAM, material and process are designed to have a high spontaneous polarization (Ps) value. However, for large memory window and reliable 1T type FeFET, it requires a certain value of spontaneous polarization value. We found that MFMIS (metal-ferroelectric-metal-insulator-semiconductor) FeFET with a three-dimensional channel structure is very effective to improve the memory window and reduce the electric field through the interfacial layer. In particular, the latter contributes greatly to improving reliability characteristics.

Negative capacitance field effect transistors (NCFET) have the potential to lower the subthreshold swing below 60 mV/decade at room temperature. This can be achieved through an internal amplification for which the magnitude of ferroelectric capacitance (C_FE) must be lower than the oxide capacitance (C_OX) i.e. |C_FE| < C_OX [1]. Since the transistor architecture has evolved from traditional planar to nanowire, the geometrical considerations for C_FE and C_OX have become relevant in the choice of ferroelectric material [2]–[3]. Our analytical investigation for a long channel metal–ferroelectric–insulator–semiconductor (MFIS) NCFET has shown that the circular cross-section of the nanowire yields more negative ferroelectric capacitance along with a higher oxide capacitance than that exhibited by the planar device. Also, |C_FE| increases significantly as compared to C_OX in a cylindrical geometry as compared to planar NCFET, which translates into the requirement of a thicker ferroelectric layer (T_FE) to sustain the internal amplification and the associated sub-60 mV/decade current transition. In order to mitigate the above condition of using a relative thicker T_FE, a ferroelectric material with a higher coercive field or lower remnant polarization is more suited for nanowire architecture. For an interfacial oxide layer of 1 nm, a double gate NCFET with Al-HfO₂ [2], Y-HfO₂ [4], Gd-HfO₂ [5] and HZO [4] requires a minimum T_FE of 5 nm, 8 nm, 12 nm, and 21 nm, respectively, to sustain internal amplification. However, for a cylindrical nanowire NCFET, the minimum T_FE values increase to 7 nm, 11 nm, 23 nm, and 69 nm, respectively. This difference between the minimum values of T_FE is least (~2 nm) for Al-HfO₂ (exhibits a high coercive field) and highest (48 nm) for HZO based NCFET. This difference between the minimum ferroelectric thickness of a planar double gate and a nanowire NCFET is governed by the coercive field and remnant polarization of the ferroelectric material. Since nanowire architecture can effectively suppress short channel effects, the maximum possible nanowire diameter should be selected so as to further limit the minimum T_FE for sustaining the internal amplification. For a thinner interfacial oxide layer thickness, the minimum T_FE for cylindrical NCFET can also become comparable to that required for a planar NCFET. The above mentioned inferences indicate the suitability of a ferroelectric material with a higher coercive field (or lower remnant polarization) along with thinner interfacial oxide or larger nanowire diameter for nanowire transistor architecture.

References
[1] S. Salahuddin et al., Nano Lett., 8 (2007), 405–410.
[2] A. D. Gaidhane et al., IEEE Trans. Electron Devices, 65 (2018), 2024-2032.
[3] D. Jimenez et al., IEEE Trans. Electron Devices, 57 (2010), 2405–2409.
[4] S. Semwal et al., IEEE Trans. Electron Devices, 67 (2018), 3868 - 3875.
[5] M. Hoffmann et al., Adv. Funct. Mater. 26 (2016), 8643–8649.

Electrostatic MEMS (Micro Electro Mechanical System) actuators are widely used in various applications like RF (Radio Frequency) MEMS, Digital Micromirror Devices etc. These devices inherently consume low-power. However, they demand high operating voltages. A novel method proposed to mitigate this large voltage requirement, is to connect a ferroelectric capacitor, exhibiting negative capacitance, in series with the MEMS actuator thereby forming a hybrid MEMS actuator¹. This technique is adapted from the emerging and recently reported use of ferroelectric negative capacitance in realizing steep slope (Subthreshold Swing SS < 60 mV/decade) low-voltage transistors. The response of the electrostatic MEMS actuator depends heavily on the applied input (static and dynamic inputs). Hence, it is necessary to develop a physics-based unified framework that facilitates an easy analysis of the hybrid MEMS actuator for different inputs.

We propose an energy-based framework to systematically analyze the statics and dynamics of the hybrid MEMS actuator. The proposed method uses graphical energy-displacement and phase portrait plots to investigate static pull-in, dynamic pull-in and pull-out phenomena of the hybrid MEMS actuator. The ferroelectric capacitor is governed by the Landau-Khalatnikov equation, which relates the voltage across the ferroelectric to its charge. The MEMS actuator, however, is governed by the nonlinear force-balance differential equation, expressed in terms of the electrode displacement. Therefore, we employ a coordinate transformation from the charge to the displacement, to obtain the Hamiltonian of the hybrid MEMS actuator in terms of its displacement. A mapping function, based on the MEMS capacitor charge-voltage relationship, is used for this transformation².

Using this framework, we show that the static pull-in, dynamic pull-in and pull-out voltages of the hybrid actuator are significantly reduced due to the ferroelectric negative capacitance, as compared to the standalone MEMS actuator. The results obtained are in agreement with the numerical SPICE (Simulation Program with Integrated Circuit Emphasis) simulations of the hybrid MEMS actuator³. The framework is unified because the same Hamiltonian is used for examining the static pull-in, dynamic pull-in, and pull-out phenomena. Since the proposed framework uses only graphical plots for the analysis, it eliminates solving intricate nonlinear differential equations that govern the dynamics of the hybrid actuator. Therefore, this serves as a quick analysis tool to predict the pull-in and pull-out voltages of the hybrid MEMS actuator. The proposed energy-based framework also has the capacity to include the adhesion between the contacting surfaces. We model the adhesion using Van der Waals force and illustrate the reduction in the pull-out voltage in the hybrid MEMS actuator due to the presence of adhesion. The proposed framework could be further enhanced to include other phenomena such as fringing field capacitance. It could also be used for different MEMS structures by suitably modifying the mapping function, used for the coordinate transformation.

[1] M Masuduzzaman and M A Alam, Nano Lett., vol. 14, no.6, 2014.
[2] R. Tattamangalam Raman, A Ajoy and R Padmanabhan, IEEE Trans. Electron Devices, vol. 67, no. 10, Oct. 2020.
[3] R. Tattamangalam Raman and A Ajoy, IEEE Trans. Electron Devices, vol. 67, no. 11, Nov. 2020.