During the last three decades, the world has changed by a rapid and capillary diffusion of mobile electronic devices like smartphones, tablets and digital cameras. This phenomenon was supported by the evolution of memory architectures able to retain the data also when the power is disconnected. These elements were intensively studied in order to decrease the occupied area, which implies a decrease in the cost per bit while increasing the speed and the storage capacity. The Flash memory is without any doubt the leader in this field due to the scalability properties, that lead to the modern 20nm technology. This progressive decrease of the device dimensions seems to approach the ultimate limit in the last years; with the scaling of the elements, physical phenomena that were before negligible are now affecting the device behavior and compromising important parameters like reliability and power efficiency. To cope with these limitations, alternative memory technologies, radically different from the Flash architecture, are studied in order to overcome modern constraints. Among the vast number of devices proposed, the Phase Change Memory (PCM) is nowadays recognized as the most mature and close to commercialization. These cells are based on chalcogenide materials able to change its electrical characteristics in a fast and reversible manner through a change of atomic structure from the crystalline phase to the amorphous one, and vice versa. The possibility to integrate the cells in the back end of the line (BEOL) drastically reducing the cost is another interesting aspect of these devices. The composition of the chalcogenide is a heavily discussed problem due to the huge influence it has on the most important properties of the technology, like data retention to high profile temperature. This last parameter is the one that drove the researchers to look for different alloy compositions with respect to the traditionally used Ge2Sb2Te5. A higher Germanium concentration proved to have better retention properties and seems to be the best material composition for future commercialization. This change in the alloy composition brought some drawbacks to the cell, like the need for an activation step, called forming, at the end of the fabrication process. The cell in the pristine state, called virgin, cannot be programmed with a standard routine but needs a much more power-consuming process to be used. The forming process was demonstrated to influence a lot of parameters such as power efficiency and durability. In this thesis project, a model for the pristine state of the cell and the activation process will be presented. Finite element method simulations were performed in order to fit the electrical behavior of these two states while replicating also the physical characteristics extracted from TEM images. All the experimental data used were provided by the R&D group led by Roberto Annunziata of the STMicroelectronics campus of Agrate Brianza. The main topics presented in this thesis are: Chapter 1 provides an introduction to memory architectures and technologies. Flash memory is briefly described from the primitive element, the floating gate transistor, to the main structures used, NAND and NOR. RRAM and PCM, two of the most promising emerging memory devices, are then presented. Chapter 2 illustrates the working principles of PCM. Conduction mechanisms of both the crystalline and the amorphous phase, as well as phase transition routines, are described. Ge-rich composition is addressed and finally, the two studies on which this work is based on are presented. Chapter 3 concentrate on the virgin state of Ge-rich PCM. The constructed finite element model to represent this condition of the cell is confronted with electrical and physical parameters from the data. Chapter 4 focuses on the expansion of the model, exposed in chapter 3, to the forming process. The assumptions done for the electrical model are verified with a thermodynamic simulation and confronted with the data.
Negli ultimi tre decenni il mondo è stato rivoluzionato da una rapida e capillare diffusione di dispositivi elettronici portatili tra cui smartphones, tablets e fotocamere digitali. Il fenomeno è stato supportato dalla rapida crescita delle unità di memoria in grado di mantenere informazioni al loro interno anche senza alimentazione collegata. Questi dispositivi sono stati intensamente studiati per diminuire l'area occupata, il che comporta una diminuzione del costo per bit, e aumentare contemporaneamente velocità e capacità di immagazzinare informazioni. La tecnologia Flash, senza alcun dubbio, è il leader del settore grazie alle proprietà di miniaturizzazione che hanno portato al nodo tecnologico dei 20 nanometri. Il trend di rimpicciolimento delle dimensioni sembra aver raggiunto il suo limite negli ultimi anni; fenomeni fisici, fino ad ora trascurabili, interferiscono con il corretto funzionamento dei dispositivi compromettendo parametri importanti come il consumo di energia o l'affidabilità. Per superare queste difficoltà architetture di memoria differenti dalla Flash sono state proposte. Tra i dispositivi emergenti le memorie a cambiamento di fase, o PCM, sono le più promettenti e studiate. Basati su materiali capaci di cambiare le proprie caratteristiche elettriche in maniera veloce e reversibile attraverso un cambio della struttura atomica dalla fase cristallina a quella amorfa e viceversa, le PCM offrono la possibilità di essere integrate durante il "Back End Of the Line" (BEOL) riducendo drasticamente l'area occupata e quindi il costo. Il materiale con cui sono composte queste celle è un argomento ampiamente discusso a causa della sua forte influenza su molte caratteristiche come la durabilità e la resistenza ad alti profili di temperatura. Quest'ultimo parametro è il motivo per cui la composizione a Germanio ricco del calcogenuro formato da Germanio Antimonio e Tellurio (GST) ha attirato molte attenzioni avendo mostrato di poter conservare i dati al suo interno anche ad alte temperature. Questa lega richiede però un processo di attivazione elettrica, chiamato forming, alla fine della fabbricazione del dispositivo. La cella, dopo la deposizione, si trova in uno stato chiamato vergine e non può essere programmata da un normale impulso di scrittura. È stato dimostrato inoltre che questa operazione di attivazione influenza molti parametri come ciclabilità e consumo di potenza. Il lavoro di tesi, qui presentato, propone un modello per lo stato vergine e per il processo di frorming per celle PCM composte da una lega di GST ricca in Germanio. Lo scenario proposto consente di replicare correttamente sia le caratteristiche elettriche, come le distribuzioni di resistenze, sia i parametri fisici estratti dalle immagini TEM. Tutti i dati sperimentali usati come riferimento sono stati forniti dal gruppo di ricerca di STMicroelectronics della sede di Agrate Brianza diretto da Roberto Annunziata. L'elaborato è strutturato come segue Il capitolo 1 fornisce un introduzione sul tema delle memorie per poi presentare alcune architetture non volatili. La tecnologia Flash viene esposta partendo dal floating gate transistor per arrivare alle strutture NOR e NAND. Successivamente due tra i dispositivi emergenti più promettenti, RRAM e PCM, vengono esposti. Il capitolo 2 illustra i principi di funzionamento delle celle PCM. Temi come la conduzione dello stato amorfo e cristallino, il processo di transizione di fase e la struttura della cella sono presentati. Le caratteristiche delle celle con alta concentrazione di Germanio sono illustrate. Infine due importanti studi su cui questo lavoro di tesi è basato sono descritti . Il capitolo 3 si concentra sullo stato vergine delle PCM. Un modello agli elementi finiti per rappresentare questo stato della cella viene introdotto e i risultati confrontati con i dati sperimentali. Il capitolo 4 espande il modello, presentato nel capitolo 3, al processo di forming. Le assunzioni fatte per il modello elettrico sono verificate con simulazioni termodinamiche e confrontate con i dati.
Modeling of virgin state and forming operation in embedded phase change memory (ePCM)
BALDO, MATTEO
2019/2020
Abstract
During the last three decades, the world has changed by a rapid and capillary diffusion of mobile electronic devices like smartphones, tablets and digital cameras. This phenomenon was supported by the evolution of memory architectures able to retain the data also when the power is disconnected. These elements were intensively studied in order to decrease the occupied area, which implies a decrease in the cost per bit while increasing the speed and the storage capacity. The Flash memory is without any doubt the leader in this field due to the scalability properties, that lead to the modern 20nm technology. This progressive decrease of the device dimensions seems to approach the ultimate limit in the last years; with the scaling of the elements, physical phenomena that were before negligible are now affecting the device behavior and compromising important parameters like reliability and power efficiency. To cope with these limitations, alternative memory technologies, radically different from the Flash architecture, are studied in order to overcome modern constraints. Among the vast number of devices proposed, the Phase Change Memory (PCM) is nowadays recognized as the most mature and close to commercialization. These cells are based on chalcogenide materials able to change its electrical characteristics in a fast and reversible manner through a change of atomic structure from the crystalline phase to the amorphous one, and vice versa. The possibility to integrate the cells in the back end of the line (BEOL) drastically reducing the cost is another interesting aspect of these devices. The composition of the chalcogenide is a heavily discussed problem due to the huge influence it has on the most important properties of the technology, like data retention to high profile temperature. This last parameter is the one that drove the researchers to look for different alloy compositions with respect to the traditionally used Ge2Sb2Te5. A higher Germanium concentration proved to have better retention properties and seems to be the best material composition for future commercialization. This change in the alloy composition brought some drawbacks to the cell, like the need for an activation step, called forming, at the end of the fabrication process. The cell in the pristine state, called virgin, cannot be programmed with a standard routine but needs a much more power-consuming process to be used. The forming process was demonstrated to influence a lot of parameters such as power efficiency and durability. In this thesis project, a model for the pristine state of the cell and the activation process will be presented. Finite element method simulations were performed in order to fit the electrical behavior of these two states while replicating also the physical characteristics extracted from TEM images. All the experimental data used were provided by the R&D group led by Roberto Annunziata of the STMicroelectronics campus of Agrate Brianza. The main topics presented in this thesis are: Chapter 1 provides an introduction to memory architectures and technologies. Flash memory is briefly described from the primitive element, the floating gate transistor, to the main structures used, NAND and NOR. RRAM and PCM, two of the most promising emerging memory devices, are then presented. Chapter 2 illustrates the working principles of PCM. Conduction mechanisms of both the crystalline and the amorphous phase, as well as phase transition routines, are described. Ge-rich composition is addressed and finally, the two studies on which this work is based on are presented. Chapter 3 concentrate on the virgin state of Ge-rich PCM. The constructed finite element model to represent this condition of the cell is confronted with electrical and physical parameters from the data. Chapter 4 focuses on the expansion of the model, exposed in chapter 3, to the forming process. The assumptions done for the electrical model are verified with a thermodynamic simulation and confronted with the data.| File | Dimensione | Formato | |
|---|---|---|---|
|
BaldoMatteo_Tesi.pdf
non accessibile
Dimensione
7.4 MB
Formato
Adobe PDF
|
7.4 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/10589/154287