The development of cutting-edge micro and opto-electronic devices requires increasingly high standards on the material quality. Indeed, up to nowadays, the advancement of the microelectronic industry has relied on the aggressive downscaling of the size typically featured by devices in the silicon-based technology. However, since the "standard" MOS transistor design cannot be miniaturized beyond a certain size, the microelectronic industry has pursued different solutions, ranging from different transistor architectures to the heterointegration of different semiconductors on Si substrates. The appealing aspect of the latter is the possibility to exploit the superior material qualities of a wide range of semiconductor materials while maintaining the standard manufacturability in the Si-based foundries. One of the major issues that has to be addressed while attempting an heterointegration is controlling the effect of the difference in lattice parameter between the epi-layer and the substrate. Indeed, this results in an in-plane deformation of the film lattice parameter to match the substrate one, leaving the epi-layer in a strained condition during the first mono-layers growth. The elastic energy accumulated can be released in two possible ways, elastically through the deformation of the planar morphology into three dimensional structures or plastically with the loss of its in-plane lattice coherence with that of the substrate, via nucleation of misfit dislocations. These defects are terminated by threading dislocations, which do not contribute to the relaxation but extend up to the free surface. These latter are the most detrimental defects for the applications, since they reach the active area of the final devices. Lowering the density of these defects is one of the main obstacle in view of further exploitation of heteroepitaxial systems. even the presence of single defects can severely reduce the desired performance, inhibiting any expected advantage from the superior material quality. Such requirement calls for a tight synergy between experimental and theoretical investigations, since the parameter space for a heteroepitaxial process is too large to be sampled by a trial and errors approach. In this work the various aspect of heteroepitaxy, ranging from the elastic relaxation with the formation of three-dimensional structures, to the plastic relaxation of thin films and heterostructures are investigated with the main focus on SiGe layers grown on Si substrates. The models developed are based on continuum approaches. Working at the continuum level conveniently allows to match typical experimental sizes of interest (up to several tens of micrometers) and time scales (up to several minutes). Modeling heteroepitaxy requires the proper description of several different phenomena. Purely elastic relaxation requires to describe the free surface diffusion of material responding to local gradients in the strain field and has been implemented by means of a computational code able to solve the partial differential equation for surface diffusion by means of the Finite Element Method. Modeling the plastic relaxation in heteroepitaxial systems, instead, requires the use of a code able to describe the behavior of single defects, still keeping the description of the spatial scale of interest described above. This was done by means of a Dislocation Dynamics approach. Finally, in this Thesis a very extensive comparison with experimental results was carried out. The goal was not limited to providing interpretation of available data, but also to suggest to various experimental partners better growth condition to achieve the desired results.
Lo sviluppo di dispositivi avanzati nell’ambito della micro e opto-elettronica richiede standard sempre più stringenti sulla qualità dei materiali. Fino ad oggi, l’avanzamento dell’industria microelettronica si è basato sul concetto di miniaturizzazione del design dei dispositivi, ma la riduzione delle dimensioni dei transistor “MOS” non può essere spinta oltre certi limiti fisici e quindi l’industria microelettronica sta attualmente ricercando sempre nuove alternative. Queste spaziano dall’esplorazione di diverse architetture per i transistor all’eterointegrazione di diversi materiali su substrati di silicio. Uno degli aspetti più promettenti di quest’ultima proposta è la possibilità di esplorare diversi semiconduttori con caratteristiche elettroniche superiori al silicio, pur mantenendo la compatibilità con le linee di produzione attuali. La maggiore problematica risultante dalla crescita eteroepitassiale è data dalla differenza in passo reticolare tra il film depositato e il substrato, che lascia il film epitassiale in uno stato di deformazione in quanto forzato a ricoprire conformemente il substrato. L’energia elastica associata a questa deformazione può essere rilassata in due modalità. Un rilassamento è di tipo elastico tramite la deformazione della morfologia planare del film in strutture tridimensionali (dette isole) oppure un rilassamento di tipo plastico con la perdita della coerenza tra film e substrato tramite l’inserimento di dislocazioni da misfit. Questi difetti sono terminati da “threading arms” che non contribuiscono al rilassamento ma attraversano tutto il film arrivando fino alla regione dove risiede la parte attiva dei dispositivi. Per questo motivo questi sono i difetti più dannosi nei processi eteroepitassiali e limitarne la densità è uno dei maggiori avanzamenti richiesti prima dello sfruttamento di materiali alternativi al silicio. Questo obiettivo ambizioso richiede che la ricerca in questo ambito venga effettuata tramite una stretta collaborazione tra esperimenti e modellizzazioni teoriche, le quali possono supportare e indicare i parametri fondamentali per ottimizzare il processo di crescita. In questo lavoro verranno studiati tramite approcci di modellizzazione continua vari aspetti della crescita etero-epitassiale, dal rilassamento elastico con la formazione di isole, al rilassamento plastico di film sottili ed eterostrutture, con particolare attenzione alla crescita di SiGe su substrati di silicio. La scelta della modellizzazione continua è stata effettuata per poter riprodurre le tipiche scale spaziali e temporali sperimentali, in cui i sistemi investigati possono avere dimensioni di decine di micron e presentare fenomeni la cui durata può raggiungere vari minuti. La modellizzazione dei processi eteroepitassiali richiede la descrizioni di diversi fenomeni fisici. Il rilassamento elastico è descritto tramite la diffusione superficiale di materiale a seguito di gradienti nel campo di deformazione ed è stato riprodotto mediante un codice in grado di risolvere numericamente l’equazione di diffusione tramite il metodo agli elementi finiti. Il rilassamento plastico invece richiede l’utilizzo di codici in grado di descrivere il comportamento di singoli difetti pur mantenendo la possibilità di raggiungere le scale spaziali descritte in precedenza. In questo lavoro è stato scelto l’utilizzo di un approccio noto come Dinamica delle Dislocazioni, sfruttando anche un estensione di esso in grado di trattare la presenza di superfici libere. Infine, in questo lavoro è stata svolta un confronto estensivo tra i risultati ottenuti dalle simulazioni e quelli sperimentali, spesso ricorrendo ad esperimenti espressamente dedicati alla validazione dei modelli proposti. Questo ha reso possibile non solo l’interpretazione dei risultati sperimentali ma anche la proposta di parametri di crescita in grado di ottimizzare il processo eteroepitassiale.
(2020). Continuum modeling of heteroepitaxy at the mesoscale: tackling elastic and plastic relaxation. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2020).
Continuum modeling of heteroepitaxy at the mesoscale: tackling elastic and plastic relaxation
ROVARIS, FABRIZIO
2020
Abstract
The development of cutting-edge micro and opto-electronic devices requires increasingly high standards on the material quality. Indeed, up to nowadays, the advancement of the microelectronic industry has relied on the aggressive downscaling of the size typically featured by devices in the silicon-based technology. However, since the "standard" MOS transistor design cannot be miniaturized beyond a certain size, the microelectronic industry has pursued different solutions, ranging from different transistor architectures to the heterointegration of different semiconductors on Si substrates. The appealing aspect of the latter is the possibility to exploit the superior material qualities of a wide range of semiconductor materials while maintaining the standard manufacturability in the Si-based foundries. One of the major issues that has to be addressed while attempting an heterointegration is controlling the effect of the difference in lattice parameter between the epi-layer and the substrate. Indeed, this results in an in-plane deformation of the film lattice parameter to match the substrate one, leaving the epi-layer in a strained condition during the first mono-layers growth. The elastic energy accumulated can be released in two possible ways, elastically through the deformation of the planar morphology into three dimensional structures or plastically with the loss of its in-plane lattice coherence with that of the substrate, via nucleation of misfit dislocations. These defects are terminated by threading dislocations, which do not contribute to the relaxation but extend up to the free surface. These latter are the most detrimental defects for the applications, since they reach the active area of the final devices. Lowering the density of these defects is one of the main obstacle in view of further exploitation of heteroepitaxial systems. even the presence of single defects can severely reduce the desired performance, inhibiting any expected advantage from the superior material quality. Such requirement calls for a tight synergy between experimental and theoretical investigations, since the parameter space for a heteroepitaxial process is too large to be sampled by a trial and errors approach. In this work the various aspect of heteroepitaxy, ranging from the elastic relaxation with the formation of three-dimensional structures, to the plastic relaxation of thin films and heterostructures are investigated with the main focus on SiGe layers grown on Si substrates. The models developed are based on continuum approaches. Working at the continuum level conveniently allows to match typical experimental sizes of interest (up to several tens of micrometers) and time scales (up to several minutes). Modeling heteroepitaxy requires the proper description of several different phenomena. Purely elastic relaxation requires to describe the free surface diffusion of material responding to local gradients in the strain field and has been implemented by means of a computational code able to solve the partial differential equation for surface diffusion by means of the Finite Element Method. Modeling the plastic relaxation in heteroepitaxial systems, instead, requires the use of a code able to describe the behavior of single defects, still keeping the description of the spatial scale of interest described above. This was done by means of a Dislocation Dynamics approach. Finally, in this Thesis a very extensive comparison with experimental results was carried out. The goal was not limited to providing interpretation of available data, but also to suggest to various experimental partners better growth condition to achieve the desired results.File | Dimensione | Formato | |
---|---|---|---|
phd_unimib_735204.pdf
Accesso Aperto
Descrizione: tesi di dottorato
Dimensione
4.77 MB
Formato
Adobe PDF
|
4.77 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.