The accelerating rise of data-intensive applications is pushing the conventional computing technologies towards physical and environmental limits. A necessity to rethink computing has emerged, starting from the study of alternative device physics, materials, and computational paradigms. A broad research field that is referred to as unconventional computing is emerging. Particularly interesting for this work is the so-called in-materia computing, based on the exploitation of the intrinsic physical properties and non-linear effects of materials to perform computation. The work presented here focuses on the analysis of charge and spin-dependent transport in devices based on networks of shallow donors in silicon. Two different systems have been investigated: the dopant network processing units (DNPU) and silicon-on-insulator (SOI) based devices. The DNPU are multiple electrodes nanoelectronic devices based on a network of arsenic atoms in silicon. They have already been proposed as a promising scalable platform for in material computing. In this work, we investigated the network’s physical properties more in depth, in order to understand the still unclear functional mechanisms and possibly expand the DNPU’s computational capability. Charge transport has been analysed, identifying the temperature ranges in which variable range hopping is the dominant mechanism. Hopping transport is responsible for the high non-linearity required for computation. In the hopping regime, it is also possible to obtain negative differential resistance behaviour, necessary to solve linearly inseparable classification problem. The DNPU’s response was also investigated in response to external stimuli, such as photoexcitation and a static magnetic field, showing field dependent photoconductivity. Finally, we studied the possibility to incorporate the spin physics of the dopants, to obtain novel spin dependent functionalities. Spin dependent transport has been investigated performing electrically detected magnetic resonance on the DNPU. Spin dependent recombination processes via clusters of arsenic dopants and silicon/oxide interface defects have been identified. A second approach starts instead from the investigation of the material's properties, in particular phosphorus doped SOI substrates with different device layer thicknesses. Knowing the fundamental properties of the material and the associated computational power will allow to properly design a device, with tailored functionalities. In particular, we focus on the analysis of the material’s spin properties and on the identification of spin dependent transport mechanisms. It was possible to observe spin dependent scattering responsible for a reduction in the device’s current in resonant condition. In addition, we observed spin dependent recombination between conduction electrons and silicon dangling bonds.

Le attuali tecnologie informatiche stanno raggiungendo i propri limiti fisico-tecnologici e di impatto ambientale, a causa del crescente uso intensivo di dati e dell’alta performance richiesta. Emerge la necessità di rinnovare i concetti di computazione convenzionale, ripartendo dallo studio di dispositivi e materiali alternativi. Nasce quindi un ampio campo di ricerca denominato ‘unconventional computing’, computazione non convenzionale. Di particolare interesse per questo lavoro è il cosiddetto ‘in-materia computing’, che si basa sullo sfruttamento delle proprietà intrinseche dei materiali e sul loro comportamento non lineare. Il lavoro qui presentato si concentra sull'analisi del trasporto di carica e spin dipendente in dispositivi basati su donori poco profondi in silicio. In particolare sono stati studiati due sistemi: i dopant network processing units (DNPU) e dispostivi basati sulla tecnologia silicon-on-insulator (SOI). I DNPU sono dispositivi nanoelettronici a più elettrodi basati su un network di atomi di arsenico in silicio, già utilizzati per in-materia computing. In questo lavoro di tesi, abbiamo approfondito le proprietà fisiche del network, al fine di comprendere i meccanismi funzionali ancora poco chiari e espandere la capacità computazionale del DNPU. È stato analizzato il trasporto di carica, identificando gli intervalli di temperatura in cui il meccanismo di trasporto dominante è variable range hopping, responsabile del comportamento non lineare. In regime di hopping è anche possibile ottenere una resistenza differenziale negativa, necessaria per risolvere problemi di classificazione linearmente inseparabili. La risposta del DNPU è stata studiata anche tramite stimoli esterni, come illuminazione e un campo magnetico statico, mostrando la dipendenza della foto-conduttività dal campo. Infine abbiamo studiato la possibilità di incorporare le proprietà di spin dei droganti, per ottenere nuove funzionalità spin dipendenti. Il trasporto dipendente dallo spin è stato studiato eseguendo la tecnica di risonanza magnetica rilevata elettricamente. Sono stati identificati processi di ricombinazione spin dipendenti che avvengono tramite droganti di arsenico e difetti di interfaccia silicio/ossido. Un secondo approccio parte invece dall'indagine delle proprietà del materiale, in particolare substrati SOI drogati con fosforo con diversi spessori del device layer. Conoscere la potenziale capacità di calcolo del material potrà consentire di progettare un dispositivo con funzionalità su misura. In particolare, ci siamo concentrati sull'analisi delle proprietà di spin del materiale e sull'identificazione dei meccanismi di trasporto spin dipendente. È stato possibile osservare fenomeni di scattering spin dipendente e fenomeni di ricombinazione spin dipendente tra elettroni di conduzione e dangling bonds del silicio.

(2023). Charge and spin-dependent transport in devices for unconventional computing. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2023).

Charge and spin-dependent transport in devices for unconventional computing

TAGLIETTI, FABIANA
2023

Abstract

The accelerating rise of data-intensive applications is pushing the conventional computing technologies towards physical and environmental limits. A necessity to rethink computing has emerged, starting from the study of alternative device physics, materials, and computational paradigms. A broad research field that is referred to as unconventional computing is emerging. Particularly interesting for this work is the so-called in-materia computing, based on the exploitation of the intrinsic physical properties and non-linear effects of materials to perform computation. The work presented here focuses on the analysis of charge and spin-dependent transport in devices based on networks of shallow donors in silicon. Two different systems have been investigated: the dopant network processing units (DNPU) and silicon-on-insulator (SOI) based devices. The DNPU are multiple electrodes nanoelectronic devices based on a network of arsenic atoms in silicon. They have already been proposed as a promising scalable platform for in material computing. In this work, we investigated the network’s physical properties more in depth, in order to understand the still unclear functional mechanisms and possibly expand the DNPU’s computational capability. Charge transport has been analysed, identifying the temperature ranges in which variable range hopping is the dominant mechanism. Hopping transport is responsible for the high non-linearity required for computation. In the hopping regime, it is also possible to obtain negative differential resistance behaviour, necessary to solve linearly inseparable classification problem. The DNPU’s response was also investigated in response to external stimuli, such as photoexcitation and a static magnetic field, showing field dependent photoconductivity. Finally, we studied the possibility to incorporate the spin physics of the dopants, to obtain novel spin dependent functionalities. Spin dependent transport has been investigated performing electrically detected magnetic resonance on the DNPU. Spin dependent recombination processes via clusters of arsenic dopants and silicon/oxide interface defects have been identified. A second approach starts instead from the investigation of the material's properties, in particular phosphorus doped SOI substrates with different device layer thicknesses. Knowing the fundamental properties of the material and the associated computational power will allow to properly design a device, with tailored functionalities. In particular, we focus on the analysis of the material’s spin properties and on the identification of spin dependent transport mechanisms. It was possible to observe spin dependent scattering responsible for a reduction in the device’s current in resonant condition. In addition, we observed spin dependent recombination between conduction electrons and silicon dangling bonds.
FANCIULLI, MARCO
in-materia computing; nanoelectronics; shallow donors; spin transport; SOI
in-materia computing; nanoelectronics; shallow donors; spin transport; SOI
FIS/03 - FISICA DELLA MATERIA
English
16-mag-2023
35
2021/2022
open
(2023). Charge and spin-dependent transport in devices for unconventional computing. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2023).
File in questo prodotto:
File Dimensione Formato  
phd_unimib_824315.pdf

accesso aperto

Descrizione: Charge and spin-dependent transport in devices for unconventional computing
Tipologia di allegato: Doctoral thesis
Dimensione 32.62 MB
Formato Adobe PDF
32.62 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10281/415720
Citazioni
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
Social impact