The power consumption of information and communication technology is growing exponentially and is expected to become unsustainable. Nowadays, the von Neumann architecture is the foundation for computers used in data processing. Due to the data bus shuttling between the physically separated memory and computing units, there is an intrinsic bottleneck which was possible to cope thanks to the increase of performances connected to the scaling of transistors down to 10 nm or less. However, the scaling is now reaching its economic and physical limits, claiming new routes. Among the plethora of possibilities, the electronics industry points towards the development of logic-in-memory architectures that offer the possibility to merge storage and computational capabilities in a single unit. In 2019 Intel proposed a breakthrough known as magneto-electric spin-orbit (MESO) device which is intended to provide a novel logic-in-memory unit operating in the attoJoule energy regime. The MESO concept comprises a magneto-electric (ME) module, that takes advantage of collective phenomena like ferroelectricity and ferromagnetism to enable information storage and retention, and a spin-orbit (SO) module, used for information processing and readout which exploits mechanisms that conveniently scale with the size of the device. Examples of such mechanisms are the spin-to-charge current conversion (S2CC) phenomena, like the inverse spin Hall and the inverse Rashba-Edelstein effects, which translate a spin current in a voltage/current output that can be used by other transistors. However, in order to operate in the desired 1-10 aJ regime, the MESO requires the identification of several highly efficient materials and the actual proof-of-concept devices are far from applications. In the search for an efficient and low-energy consumption readout module, this thesis first presents a thorough investigation of the MESO readout unit made by NiFe/Pt heterostructures. NiFe is a soft ferromagnet and is used to lower the energy for the magnetization reversal. Platinum, instead, provides the spin-to-charge current conversion by spin Hall effect, necessary to retrieve the information stored in the ferromagnet. This section of the thesis is devoted to the readout measurements in MESO-like structures. This demonstrates the capability to perform state-of-the-art S2CC experiments, while accessing to the different contributions of the output signal (i.e. spin, anomalous and planar Hall effects). By fitting the results with micromagnetic and analytical models, we study in deep the uncertainty in the estimation of the spin Hall angle and we infer for our Pt a value of 3.5%; moreover, we carry out a detailed energetic analysis showcasing the capacity to work in the aJ framework. This work continues with the proposal of a radically new approach to low-power logic-in-memory devices. When it comes to energy-efficient materials that could be employed in such ultralow power components, an emerging class of materials called ferroelectric Rashba semiconductors (FERSC) may represent an ideal candidate. Owing to its inherent multifunctionality, this family of materials is predicted to be suitable for in-memory computing, thanks to the combination of ferroelectricity (memory) and the Rashba effect (computing) in a single semiconductor. We started from a solid background, i.e. we recently demonstrated that in the father compound GeTe, the ferroelectric polarization can be reversed by electric pulses and used to change the sign of spin-to-charge current conversion, offering the non-volatile ferroelectric control of information processing within the same physical location. Nonetheless, the activity presented here intends to increase the efficiency of S2CC mechanism with respect to GeTe as well as to lower the voltages required to switch the ferroelectric polarization (i.e. coercive voltage below 1 V). Thus, here we provide a suitable approach to tailor material properties to achieve low switching voltage and large spin-to-charge current conversion. Specifically, we describe how the alloying of the two Rashba semiconductors GeTe and SnTe allows to tune both the ferroelectric and the Rashba properties in Ge(x)Sn(1-x)Te. Endorsed by ab-initio calculations, spin- and angular- resolved photoemission spectroscopy (SARPES) and ferroelectric characterizations, we prove the engineering of the giant Rashba effect and the intimately related ferroelectricity, which persist up to room temperature in a certain range of concentrations. Indeed, the Curie temperature in Ge(0.3)Sn(0.7)Te increases up to 500 K (for pure SnTe Curie temperature is 100 K) and a significant reduction of the coercive voltage with respect to pure GeTe is found (i.e. coercive voltage for pure GeTe is 3.5 V, while for GeSnTe is 1 V). The two aforementioned studies were finally conveyed in the realization of an innovative device that aims at going well beyond the MESO component, integrating its spin-orbit module with spin-textured ferroelectric materials as FERSCs: the so called ferroelectric spin-orbit (FESO) logic. Herein, the information is stored in the ferroelectric polarization (not in the magnetization), and both processing and readout are enabled by polarization-dependent spin-to-charge current conversion. The FESO paradigm for logic-in-memory is focused on just one active material. It could allow a non-volatile voltage-controlled switching without external magnetic fields, offering new perspectives in electronics to surpass both the currently employed CMOS transistors and the von Neumann architecture. In the third final part, thanks to the knowledge acquired both with spin-to-charge current conversion experiments in MESO-like structures and the FERSC's materials investigation, we present the complete fabrication of the FESO device and its seminal magneto-transport measurements. On the basis of these findings, this work sheds light on the feasibility of developing a scalable and energy-efficient MESO readout module capable to operate in the aJ range, as well as insights on the energy bands structure of FERSC alloys. Harnessing the spin-orbit coupling and the ferroelectricity enabled preliminary investigations with FESO devices. The FERSC materials and the avant-garde change of paradigm represented by FESO offer an auspicious strategy towards the realization of logic-in-memory spintronics devices with ultralow power consumption.
Il consumo energetico delle tecnologie dell'informazione e della comunicazione è in crescita esponenziale e destinato a divenire insostenibile. Attualmente, i computer impiegati per l'elaborazione dei dati si basano sull'architettura di von Neumann. Tale modello comporta la trasmissione dei dati tra la memoria e le unità di calcolo che sono fisicamente separate; questo genera un ostacolo che è stato affrontato grazie all'aumento delle prestazioni, reso possibile dal ridimensionamento dei transistor, fino a grandezze inferiori ai 10 nm. Tuttavia, questo approccio sta raggiungendo i suoi limiti, sia economici che fisici, motivo per cui nuove vie devono essere perseguite. Tra le diverse proposte, l'industria dell' elettronica sta puntando allo sviluppo di architetture logic-in-memory, le quali offrono la possibilità di unire capacità di memoria e computazionali in una singola unità. Nel 2019 Intel ha proposto un dispositivo rivoluzionario noto come magneto-electric spin-orbit (MESO), un logic-in-memory device in grado di operare nel regime degli attoJoule. Il MESO comprende un modulo magneto-elettrico (ME) che sfrutta fenomeni collettivi come la ferroelettricità e il ferromagnetismo per immagazzinare le informazioni, e un modulo spin-orbita (SO) usato per processare i dati attraverso meccanismi la cui efficienza aumenta diminuendo le dimensioni del dispositivo. Tra i più importanti fenomeni in tal senso troviamo gli effetti di conversione spin-carica, come lo spin Hall effect (SHE) e il Rashba Edelstein effect (REE), che convertono mutualmente correnti di spin e correnti di carica, la quali possono essere a loro volta impiegate per pilotare altri elementi in cascata. Tuttavia, per essere in grado di operare nel regime degli attoJoule, il MESO richiede l'identificazione di diversi materiali con prestazioni eccezionali e la sua implementazione nei dispostivi commerciali è ancora lontana. Alla ricerca di un sistema di lettura efficiente, questa tesi propone un'analisi dettagliata dell'elemento di spin-orbita del MESO, in particolare focalizzandosi su un'eterostruttura di NiFe/Pt. Il NiFe (permalloy) è un magnete dolce sfruttato per ridurre l'energia nel processo di inversione della magnetizzazione. Il platino, invece, provvede alla conversione spin-carica mediante l'effetto di spin Hall, in modo da ricavare il bit di memoria immagazzinato nel ferromagnete. Questa parte della tesi descrive gli esperimenti di spin-carica in strutture simili al MESO; in particolare, dimostriamo la capacità di realizzare misure allo stato dell'arte e di individuare i singoli contributi del segnale di lettura (effetti di spin, anomalous e planar Hall). Fittando i risultati con simulazioni micromagnetiche e modelli analitici, siamo in grado di stimare accuratamente lo spin Hall angle del nostro platino il quale si attesta a un valore di 3.5%; inoltre, proponiamo un'approfondita analisi energetica e diamo prova di poter lavorare nel regime degli attoJoule. Successivamente, il lavoro prosegue con la proposta di un nuovo approccio ai dispositivi logic-in-memory. Per farlo intendiamo sfruttare un'emergente classe di materiali, i cosiddetti ferroelectric Rashba semiconductors (FERSC). Grazie alla loro intrinseca multifunzionalità questi materiali si sono dimostrati adatti per l'in-memory computing: difatti, essi combinano la ferroelettricità (memoria) e l'effetto Rashba (processing) in un singolo semiconduttore. Il nostro lavoro poggia su solide basi scientifiche; precisamente, di recente abbiamo dimostrato che nel sistema FERSC più noto, il germanio tellurio, la polarizzazione ferroelettrica può essere controllata con impulsi elettrici e usata per modificare il segno dell'effetto di conversione spin-carica in uscita; in questo modo si ha, nello stesso dispositivo, il controllo non-volatile del processamento dell'informazione. Tuttavia, l'attività qui presentata intende migliorare l'efficienza del meccanismo di conversione spin-carica rispetto al GeTe, diminuendo allo stesso tempo le tensioni richieste per invertire la polarizzazione ferroelettrica. Nel manoscritto presentiamo un valido approccio in grado di modificare le proprietà fisiche del materiale, così da ottenere un controllo della polarizzazione a basse tensioni, migliorando anche l'efficienza di conversione del dispositivo. In particolare, mostriamo come realizzare una lega di due materiali FERSC, GeTe e SnTe, permette di modificare le proprietà Rashba e ferroelettriche nel Ge(x)Sn(1-x)Te. Supportati da calcoli DFT, misure spettroscopiche risolte in angolo e spin, e caratterizzazioni ferroelettriche, dimostriamo l'ingegnerizzazione sia dell'effetto Rashba che della ferroelettricità ad esso legata, i quali, per un certo intervallo di concentrazioni degli elementi, sopravvivono fino alla temperatura ambiente. Infatti, la temperatura di Curie per il Ge(0.3)Sn(0.7)Te raggiunge i 500 K (per SnTe da solo temperatura di Curie è 100 K) e oltretutto registriamo una significativa riduzione del campo coercitivo rispetto al GeTe (i.e. campo coercitivo del GeTe è 3.5 V rispetto a 1 V per GeSnTe). I due studi sopracitati sono infine uniti per la realizzazione di un dispositivo innovativo che punta ad andare ben oltre il MESO, integrandone il modulo spin-orbita con materiali FERSC: abbiamo così il FESO (ferroelectric spin-orbit). In questo caso l'informazione è immagazzinata nella polarizzazione ferroelettrica (non nella magnetizzazione) e sia il processing che la lettura sono abilitate dal meccanismo spin-carica che è manipolabile attraverso la polarizzazione del materiali. E' importante notare che nel FESO solo uno è il materiale attivo (FERSC). Questo dispositivo permette un controllo non-volatile attraverso la tensione applicata, ma senza l'uso di campi magnetici esterni, e offre nuove prospettive nel campo dell'elettronica, superando il CMOS e l'architettura di von Neumann. Nella terza e conclusiva parte della tesi, presentiamo la fabbricazione completa del FESO e le sue misure di trasporto preliminari. Sulla base di quanto trovato, questo lavoro dimostra la possibilità di realizzare un modulo di spin-orbita del dispositivo MESO scalabile ed energicamente efficiente; allo stesso tempo proponiamo una valida strategia per la manipolazione delle proprietà dei materiali FERSC. La regolazione dell'interazione spin-orbita e della ferroelettricità permettono infine l'analisi preliminare dei dispositivi FESO. I materiali FERSC e il cambio di paradigma rappresentato dal FESO offrono una promettente strategia verso la realizzazione di nuovi dispositivi logic-in memory a basso consumo energetico.
Harnessing ferroelectricity and spin-orbit coupling towards ultralow power spintronics
FAGIANI, FEDERICO
2023/2024
Abstract
The power consumption of information and communication technology is growing exponentially and is expected to become unsustainable. Nowadays, the von Neumann architecture is the foundation for computers used in data processing. Due to the data bus shuttling between the physically separated memory and computing units, there is an intrinsic bottleneck which was possible to cope thanks to the increase of performances connected to the scaling of transistors down to 10 nm or less. However, the scaling is now reaching its economic and physical limits, claiming new routes. Among the plethora of possibilities, the electronics industry points towards the development of logic-in-memory architectures that offer the possibility to merge storage and computational capabilities in a single unit. In 2019 Intel proposed a breakthrough known as magneto-electric spin-orbit (MESO) device which is intended to provide a novel logic-in-memory unit operating in the attoJoule energy regime. The MESO concept comprises a magneto-electric (ME) module, that takes advantage of collective phenomena like ferroelectricity and ferromagnetism to enable information storage and retention, and a spin-orbit (SO) module, used for information processing and readout which exploits mechanisms that conveniently scale with the size of the device. Examples of such mechanisms are the spin-to-charge current conversion (S2CC) phenomena, like the inverse spin Hall and the inverse Rashba-Edelstein effects, which translate a spin current in a voltage/current output that can be used by other transistors. However, in order to operate in the desired 1-10 aJ regime, the MESO requires the identification of several highly efficient materials and the actual proof-of-concept devices are far from applications. In the search for an efficient and low-energy consumption readout module, this thesis first presents a thorough investigation of the MESO readout unit made by NiFe/Pt heterostructures. NiFe is a soft ferromagnet and is used to lower the energy for the magnetization reversal. Platinum, instead, provides the spin-to-charge current conversion by spin Hall effect, necessary to retrieve the information stored in the ferromagnet. This section of the thesis is devoted to the readout measurements in MESO-like structures. This demonstrates the capability to perform state-of-the-art S2CC experiments, while accessing to the different contributions of the output signal (i.e. spin, anomalous and planar Hall effects). By fitting the results with micromagnetic and analytical models, we study in deep the uncertainty in the estimation of the spin Hall angle and we infer for our Pt a value of 3.5%; moreover, we carry out a detailed energetic analysis showcasing the capacity to work in the aJ framework. This work continues with the proposal of a radically new approach to low-power logic-in-memory devices. When it comes to energy-efficient materials that could be employed in such ultralow power components, an emerging class of materials called ferroelectric Rashba semiconductors (FERSC) may represent an ideal candidate. Owing to its inherent multifunctionality, this family of materials is predicted to be suitable for in-memory computing, thanks to the combination of ferroelectricity (memory) and the Rashba effect (computing) in a single semiconductor. We started from a solid background, i.e. we recently demonstrated that in the father compound GeTe, the ferroelectric polarization can be reversed by electric pulses and used to change the sign of spin-to-charge current conversion, offering the non-volatile ferroelectric control of information processing within the same physical location. Nonetheless, the activity presented here intends to increase the efficiency of S2CC mechanism with respect to GeTe as well as to lower the voltages required to switch the ferroelectric polarization (i.e. coercive voltage below 1 V). Thus, here we provide a suitable approach to tailor material properties to achieve low switching voltage and large spin-to-charge current conversion. Specifically, we describe how the alloying of the two Rashba semiconductors GeTe and SnTe allows to tune both the ferroelectric and the Rashba properties in Ge(x)Sn(1-x)Te. Endorsed by ab-initio calculations, spin- and angular- resolved photoemission spectroscopy (SARPES) and ferroelectric characterizations, we prove the engineering of the giant Rashba effect and the intimately related ferroelectricity, which persist up to room temperature in a certain range of concentrations. Indeed, the Curie temperature in Ge(0.3)Sn(0.7)Te increases up to 500 K (for pure SnTe Curie temperature is 100 K) and a significant reduction of the coercive voltage with respect to pure GeTe is found (i.e. coercive voltage for pure GeTe is 3.5 V, while for GeSnTe is 1 V). The two aforementioned studies were finally conveyed in the realization of an innovative device that aims at going well beyond the MESO component, integrating its spin-orbit module with spin-textured ferroelectric materials as FERSCs: the so called ferroelectric spin-orbit (FESO) logic. Herein, the information is stored in the ferroelectric polarization (not in the magnetization), and both processing and readout are enabled by polarization-dependent spin-to-charge current conversion. The FESO paradigm for logic-in-memory is focused on just one active material. It could allow a non-volatile voltage-controlled switching without external magnetic fields, offering new perspectives in electronics to surpass both the currently employed CMOS transistors and the von Neumann architecture. In the third final part, thanks to the knowledge acquired both with spin-to-charge current conversion experiments in MESO-like structures and the FERSC's materials investigation, we present the complete fabrication of the FESO device and its seminal magneto-transport measurements. On the basis of these findings, this work sheds light on the feasibility of developing a scalable and energy-efficient MESO readout module capable to operate in the aJ range, as well as insights on the energy bands structure of FERSC alloys. Harnessing the spin-orbit coupling and the ferroelectricity enabled preliminary investigations with FESO devices. The FERSC materials and the avant-garde change of paradigm represented by FESO offer an auspicious strategy towards the realization of logic-in-memory spintronics devices with ultralow power consumption.File | Dimensione | Formato | |
---|---|---|---|
Thesis_Fagiani_Depositata_12_3.pdf
Open Access dal 13/03/2025
Dimensione
100.71 MB
Formato
Adobe PDF
|
100.71 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/217651