Exploring negative capacitance and neuromorphic devices based on CMOS-compatible ferroelectric HfO2

Ferroelectric materials are explored for numerous applications thanks to their properties associated with electrically switchable spontaneous polarization. Perovskites are an established class of ferroelectrics used for sensors and actuators. However, they present limitations if integrated into microelectronic since they are neither CMOS compatible nor easily scalable. The game-changing solution came from the new family of ferroelectric materials based on HfO2 that emerged within the last decade, which show integrability in scaled CMOS devices.HfO2-based ferroelectrics are considered a rising candidate for different applications, among which neuromorphic devices and negative capacitance (NC) logic switches. Neuromorphic devices rely on the gradual polarization switching capability, which can reproduce synaptic plasticity. By mimicking the analog operation of the human brain, neuromorphic computing is expected to be more energy-efficient than the digital Von Neumann architecture. On the other hand, in logic switches, hysteresis is usually avoided, and a steep subthreshold swing is desired. For this purpose, negative capacitance field-effect transistors (NC-FETs) show a strong promise of overcoming the thermionic Boltzmann constraint of 60 mV/dec and continuing CMOS scaling. NC-FETs rely on exploiting the NC region of the polarization-electric field curve predicted by the Landau-Ginzburg-Devonshire (LGD) model. This thesis aims to study the NC effect and its interplay with the polarization switching mechanisms of one of the best-known members of HfO2-based ferroelectrics, the Si-doped HfO2 (Si:HfO2), and then to integrate it onto electronic devices. Temperature-dependent switching in Si:HfO2 capacitors were studied, and the results were used to calibrate the LGD model, confirming the intrinsic switching mechanism experimentally. Then Si:HfO2 was integrated into gate stack transistors to verify its functionality. In this framework, two structures were studied, one with an inner metal gate (IG) between the insulator and ferroelectric and one without it. Thanks to nanosecond-range pulse measurement, the S-shaped curve, predicted by the LGD model, was extracted, and the NC region was observed. Following the previous study, the two gate stacks were implemented in a junctionless transistor (JLFET) platform with a Si channel of 12 nm, resulting in a device with Ion/Ioff of 6 orders of magnitude. The ferroelectric-JLFET (Fe-JLFET) without IG presented a hysteresis-free transfer characteristic and an improved subthreshold swing of 35% over 1.3 decades of drain current with pulse gate voltages compared to DC. On the contrary, the Fe-JLFET with IG showed a ferroelectric hysteretic behavior, and the Si:HfO2 gradual switching is exploited to mimic the synaptic plasticity. Additionally, the back-gate voltage biasing is used to tune the synaptic weight by more than 400x.Overall, this Ph.D. work confirms the potential of Fe-JLFET for both neuromorphic synapses and NC logic devices and highlights the requirements and limits of operation by studying how HfO2-based ferroelectric properties evolve from the capacitor to integration into the gate stack.