Hydrogenated amorphous silicon

(1)

,ǇĚƌŽŐĞŶĂƚĞĚ

ĂŵŽƌƉŚŽƵƐƐŝůŝĐŽŶ͗

ŶĂŶŽƐƚƌƵĐƚƵƌĞ

(2)

(3)

Propositions belonging to the thesis:

Hydrogenated amorphous silicon: nanostructure and defects

by Jimmy Melskens

1. The disordered network with vacancies is a more accurate way to describe the nanostructure of hydrogenated amorphous silicon than the continuous random network.

This thesis, chapters 2 – 7 2. The light-induced creation of defects in hydrogenated amorphous silicon over time

cannot be described by a single time dependence according to a power law.

This thesis, chapters 3 and 6 3. None of the currently existing models of the Staebler-Wronski effect in hydrogenated amorphous silicon that are known from literature can completely and correctly describe the nanoscopic changes induced by light soaking and annealing.

This thesis, chapter 3, 6, and 7 4. Defects that are associated with vacancies and voids are needed in a model that

accurately describes the light-induced degradation of hydrogenated amorphous silicon. This thesis, chapter 7 5. It is easier to climb up the mountains of science than it is to climb up the mountains of

bureaucracy.

6. When the variety of food choices available at a university campus is extremely limited, both employees and students are encouraged to limit their lunch and dinner moments to a minimum amount of time, which promotes an undesirable choice between good health and meeting your colleagues.

7. A government that reduces funding for education on the national budget thereby reduces the right to take pride in the quality of the educational system.

8. The subsidies which the Dutch railways (Nederlandse Spoorwegen) are receiving from the government to provide every Dutch student with free public transport would be more wisely spent on the quality of higher education in the Netherlands.

9. A patent application is a waste of time and money when the idea behind it is not to commercialize it.

10. Comparing different photovoltaic module technologies solely based on the costs per Watt peak is insufficient to be able to make a fair comparison and assess their economic value in the context of electricity generation.

These propositions are regarded as opposable and defendable, and have been approved as such by the promotors prof. dr. M. Zeman and prof. dr. E. H. Brück.

(4)

Stellingen behorende bij het proefschrift:

Hydrogenated amorphous silicon: nanostructure and defects

door Jimmy Melskens

1. Het wanordelijke netwerk met vacatures is een meer accurate manier om de nanostructuur van gehydrogeneerd amorf silicium te beschrijven dan het continu willekeurige netwerk.

Dit proefschrift, hoofdstukken 2 – 7 2. De lichtgeïnduceerde creatie van defecten in gehydrogeneerd amorf silicium in de tijd kan niet worden beschreven door een enkele tijdsafhankelijkheid volgens een machtswet.

Dit proefschrift, hoofdstukken 3 en 6 3. Geen van de momenteel bestaande modellen van het Staebler-Wronski effect in

gehydrogeneerd amorf silicium die bekend zijn uit de literatuur kunnen compleet en correct de nanoscopische veranderingen geïnduceerd door “light soaking” en temperen beschrijven.

Dit proefschrift, hoofdstukken 3, 6 en 7 4. Defecten die gerelateerd zijn aan vacatures en holtes zijn benodigd in een model dat de

lichtgeïnduceerde degradatie van gehydrogeneerd amorf silicium accuraat beschrijft. Dit proefschrift, hoofdstuk 7 5. Het is eenvoudiger om de bergen van de wetenschap te beklimmen dan het beklimmen

van de bergen van de bureaucratie is.

6. Wanneer de voedingskeuzevariëteit op een universiteitscampus extreem beperkt is, worden zowel medewerkers als studenten aangemoedigd om hun lunch- en dinermomenten tot een minimale hoeveelheid tijd te beperken, wat een ongewenste keuze tussen een goede gezondheid en het ontmoeten van je collega’s bevordert.

7. Een overheid die de financiering van onderwijs op de nationale begroting vermindert, vermindert daarmee het recht om trots te zijn op de kwaliteit van het onderwijssysteem.

8. De subsidies die de Nederlandse Spoorwegen ontvangen van de overheid om elke Nederlandse student van gratis openbaar vervoer te voorzien zouden verstandiger besteed kunnen worden aan de kwaliteit van het hoger onderwijs in Nederland.

9. Een patentaanvraag is een verspilling van tijd en geld als het idee erachter niet is om het te commercialiseren.

10. Verschillende fotovoltaïsche moduletechnologieën vergelijken uitsluitend op basis van de kosten per Wattpiek is onvoldoende om in staat te zijn om een eerlijke vergelijking te maken en hun economische waarde te beoordelen in de context van elektriciteits-opwekking.

(5)

Hydrogenatedamorphoussilicon:

nanostructureanddefects

JimmyMelskens

(6)

(7)

Hydrogenatedamorphoussilicon:

nanostructureanddefects

Proefschrift

terverkrijgingvandegraadvandoctor aandeTechnischeUniversiteitDelft, opgezagvandeRectorMagnificusprof.ir.K.C.A.M.Luyben, voorzittervanhetCollegevoorPromoties, inhetopenbaarteverdedigenop25februari2015om12:30uur door JimmyMELSKENS ingenieur,TechnischeUniversiteitDelft geborenteRotterdam,Nederland

(8)

This dissertation has been approved by the promotors: Prof. dr. M. Zeman

Prof. dr. E.H. Brück and

Copromotor:

Dr. ir. A.H.M. Smets

Composition of the doctoral committee: Rector Magnificus, chairperson

Prof. dr. M. Zeman, Technische Universiteit Delft, promotor Prof. dr. E.H. Brück, Technische Universiteit Delft, promotor Dr. ir. A.H.M. Smets, Technische Universiteit Delft, copromotor Prof. dr. I.M. Richardson, Technische Universiteit Delft

Prof. dr. ir. W.M.M. Kessels, Technische Universiteit Eindhoven

Prof. dr. K. Lips, Helmholtz-Zentrum Berlin für Materialen und Energie GmbH, Duitsland

Prof. dr. C.R. Wronski, Pennsylvania State University, Verenigde Staten van Amerika

Financial support for this research from ADEM, A green Deal in Energy Materials, of the Ministry of Economic Affairs of The Netherlands is acknowledged. (www.adem-innovationlab.nl)

Front cover photo by J. Melskens Dome of the Sheikh Lotfollah Mosque, Isfahan, Iran Back cover photo by R.A.C.M.M. van Swaaij The author

Cover design by S.M. Klingens

No part of this material may be reproduced, stored in a retrieval system, nor transmitted in any form or by any means without the prior written permission of the copyright owner.

ISBN 978-94-6203-749-6

(9)

“What we don’t understand” is far greater than “what we know” – Mitsuru Ezaki

(10)

(11)

Summary

Since the first report on the fabrication of hydrogenated amorphous silicon (a-Si:H) in 1965, this material has found many applications, for instance in the fabrication of solar cells, sensors and transistors. An especially notable example in the context of this thesis is thin-film silicon solar cells, which have attracted a lot of attention. This is largely due to the fact that this type of photovoltaic technology can be implemented on flexible substrates and is a potentially cheap and lightweight product. However, a-Si:H-based solar cells suffer from light-induced degradation (LID) which can only partially be recovered by annealing. This issue, which has become known as the Staebler-Wronski effect (SWE), is obviously an undesirable material property in the context of photovoltaic applications. Although the solar cell conversion efficiency of thin-film silicon solar cells has seen significant gains over the years, research efforts are still ongoing to achieve further improve-ments. A significant portion of these improvements has come from the development of new materials or improvements in existing materials. This type of fundamental research is inherently linked to the understanding of the nanostructure and the defects therein to enable further material quality improvements.

In this thesis it is aimed to fundamentally improve the under-standing of the a-Si:H nanostructure and the defects in this material to finally enable a reduction or even elimination of the SWE. Unfortu-nately, the SWE has proven to be a notoriously difficult problem due to the complexity of the a-Si:H nanostructure, which is the cause of the lacking consensus on the nature of the defects in this material. Because of this complication, there is a two-stage research approach in this thesis and it is not directly aimed to fabricate more stable a-Si:H. The first objective is to improve the fundamental understanding of the nano-structure and the defects in a-Si:H. Only secondly and using this newly gained knowledge, the SWE and the nature of metastable defects are studied to pave the way towards a reduction of the SWE.

A wide range of material characterization techniques is used throughout this thesis to conduct extensive, systematic defect studies on both a-Si:H films and solar cells. From the studies conducted on a-Si:H films with widely varying nanostructures it has become clear that

(16)

Doppler broadening positron annihilation spectroscopy (DB-PAS) and Fourier transform infrared (FTIR) spectroscopy are powerful complementary tools when trying to determine to dominant type of open volume deficiency in the material. For high quality a-Si:H it is concluded that divacancies are the dominant open volume deficiencies, while the dominant type of open volume deficiency can get as large as nanosized voids when the deposition rate is increased sufficiently.

When investigating the electrical properties of defects in a-Si:H, Fourier transform photocurrent spectroscopy (FTPS) is used to quantify the sub gap absorption, which contains information about defects in the a-Si:H bandgap. This method can be applied both to a-Si:H films and solar cells. By using voltage biasing, in situ light soaking, or in situ annealing on an a-Si:H solar cell it is possible to manipulate the occu-pation and density of states in the bandgap of the absorber layer and monitor any induced changes. It is this accuracy that has enabled the first ever reported observation of four sub gap contributions in the absorber layer bandgap of an a-Si:H solar cell. Considering the apparently important role of open volume deficiencies in the a-Si:H nanostructure and the fact that there are more than two distributions in the a-Si:H bandgap, it is unlikely that the commonly assumed contin-uous random network (CRN), in which isolated dangling bonds (dbs) are the dominant defects, is an accurate description of the nanostructure. Instead, the disordered network with hydrogenated vacancies (DNHV) is proposed as a nanostructural description which is more likely to be correct.

When using annealing as a probing tool to study the nanostructure of the as-deposited state, it is shown that a-Si:H films that are deposited from hydrogen-diluted silane gas, which are known to have an enhanced light soaking stability, have a smaller type of dominant open volume deficiency than a-Si:H deposited from pure silane gas, which typically exhibits a substantial LID. Additionally, the hydrogen passivation degree of small open volume deficiencies increases with increasing hydrogen dilution. Irrespective of the used hydrogen dilution, three different processes are observed by means of DB-PAS during annealing: vacancy agglomeration, hydrogen effusion, and crystallization. Since these processes take place at higher temper-atures when the a-Si:H film is deposited at increasing hydrogen dilution, a reduced mobility of open volume deficiencies is linked to an enhanced

(17)

light soaking stability. Finally, FTIR spectroscopy results indicate that the hydrogen effusion is a two-stage process, in which hydrogen initially mostly effuses from the small open volume deficiencies and only secondly predominantly from larger open volume deficiencies. These findings further underline the importance of including open volume deficiencies in an accurate description of the a-Si:H nanostructure and serve as further evidence against the CRN and in favor of the DNHV.

Using all these new fundamental insights into the a-Si:H nanostructure, the nature and kinetics of LID in a-Si:H films and solar cells are investigated. Through a combination of DB-PAS, FTIR spectroscopy, and current density-voltage (JV) characterization, it is concluded that the most stable type of a-Si:H is deposited at a low deposition rate, which is associated with the smallest type of dominant open volume deficiency, i.e. a well hydrogen-passivated divacancy. By means of FTPS it is concluded that the four observed sub gap contributions have different light-induced defect creation rates during light soaking and do not follow a single time dependence ~tE. More specifically, it does not generally hold that E = 1/3 or E = 1/2, as has been repeatedly reported in literature. Furthermore, the generation and recombination profiles are spatially correlated, causing the recombi-nation in the top and bulk parts of the absorber layer to be different.

Since these findings call for a more advanced SWE model, which involves defects related to open volume deficiencies in addition to the dbs present in the CRN, a new nanoscopic description of the LID is given based on a combined characterization approach involving DB-PAS, continuous wave electron paramagnetic resonance (cw-EPR) spectroscopy and pulsed electron paramagnetic resonance (p-EPR) spectroscopy. It appears that there is a long term process in the LID which does not depend on the a-Si:H nanostructure. More importantly, it is found that there are fast and slowly relaxing defect types, which are predominantly linked to defects created in small and large open volume deficiencies. This newly found link between LID and the a-Si:H nano-structure is an important step on the road towards fully understanding and further reducing the SWE and it once again shows that the DNHV is a more appropriate description of the nanostructure than the CRN.

(18)

(19)

Samenvatting

Sinds de eerste melding van de fabricatie van gehydrogeneerd amorf silicium (a-Si:H) in 1965 wordt dit materiaal op vele manieren toegepast, bijvoorbeeld in de fabricage van zonnecellen, sensoren en transistors. Een bijzonder noemenswaardig voorbeeld in de context van dit proefschrift is dunne-film silicium (TF Si) zonnecellen die veel aandacht getrokken hebben. Dit is grotendeels te danken aan het feit dat dit type fotovoltaïsche technologie kan worden geïmplementeerd op flexibele substraten en een potentieel goedkoop en lichtgewicht product is. Echter, zonnecellen gebaseerd op a-Si:H lijden onder licht-geïnduceerde degradatie (LID) die slechts gedeeltelijk kan worden hersteld door temperen. Dit probleem, dat bekend is geworden als het Staebler-Wronski effect (SWE), is vanzelfsprekend een ongewenste materiaaleigenschap in de context van fotovoltaïsche toepassingen. Hoewel het zonnecelconversierendement van dunne-film silicium zonnecellen significante toenames heeft laten zien in de loop der jaren, is het onderzoek nog steeds gaande om verdere verbeteringen te bewerk-stelligen. Een significant deel van deze verbeteringen is gekomen uit de ontwikkeling van nieuwe materialen of verbeteringen van bestaande materialen. Dit type fundamenteel onderzoek is inherent gekoppeld aan het begrip van de nanostructuur en de defecten daarin om verdere verbeteringen in materiaalkwaliteit mogelijk te maken.

In dit proefschrift wordt gepoogd om het begrip van de a-Si:H nanostructuur en de defecten in dit materiaal fundamenteel te verbe-teren om uiteindelijk een reductie of zelfs een eliminatie van het SWE te realiseren. Helaas heeft het SWE bewezen een berucht lastig probleem te zijn door de complexiteit van de a-Si:H nanostructuur die de oorzaak is van de ontbrekende algemene instemming over de aard van de defecten in dit materiaal. Door deze complicatie is er een tweeledige onderzoeks-benadering in dit proefschrift en wordt er niet direct geprobeerd om meer stabiel a-Si:H te fabriceren. Het eerste doel is om het fundamentele begrip van de nanostructuur en de defecten in a-Si:H te verbeteren. Pas ten tweede en met gebruikmaking van de nieuw opgedane kennis, worden het SWE en de aard van metastabiele defecten bestudeerd teneinde een route te vinden naar een reductie van het SWE.

(20)

Een uitgebreide selectie van materiaalkarakterisatietechnieken wordt in dit proefschrift gebruikt om uitvoerige, systematische studies op zowel a-Si:H films als zonnecellen uit te voeren. Uit de studies die zijn uitgevoerd op a-Si:H films met breed variërende nanostructuren is duidelijk geworden dat Dopplerverbreding-positronannihilatie-spectroscopie (DB-PAS) en Fouriertransformatie-infraroodDopplerverbreding-positronannihilatie-spectroscopie (FTIR) krachtige complementaire gereedschappen zijn bij het vaststellen van het dominante type open volumedefect in het materiaal. Voor hoge kwaliteit a-Si:H wordt geconcludeerd dat di-vacatures de dominante open volumedefecten zijn, terwijl het dominante type open volume-defect zo groot kan worden als een nanoholte wanneer de depositie-snelheid voldoende wordt verhoogd.

Bij het onderzoeken van de elektrische eigenschappen van defecten in a-Si:H wordt Fouriertransformatie-fotostroomspectroscopie (FTPS) gebruikt om de subgap-absorptie te kwantificeren, die informatie bevat over defecten in de a-Si:H bandgap. Deze methode kan worden toegepast op zowel a-Si:H films als zonnecellen. Door gebruik te maken van een spanningsbias, in situ lichtdegradatie of in situ temperen op een a-Si:H zonnecel is het mogelijk om de bezetting en de dichtheid van toestanden in de bandgap van de absorbeerlaag te manipuleren en enige geïnduceerde veranderingen te volgen. Het is deze nauwkeurigheid die het mogelijk heeft gemaakt om voor het eerst ooit vier subgapbijdrages in de absorbeerlaag van een a-Si:H zonnecel te kunnen observeren. De schijnbaar belangrijke rol van open volumedefecten in de a-Si:H nano-structuur in beschouwing nemende en het feit dat er meer dan twee distributies in de a-Si:H bandgap zijn, is het onwaarschijnlijk dat het algemeen aangenomen continu willekeurige netwerk (CRN), waarin geïsoleerde “dangling bonds” (dbs) de dominante defecten zijn, een accurate beschrijving van de nanostructuur is. In plaats daarvan wordt het wanordelijke netwerk met gehydrogeneerde vacatures (DNHV) voorgesteld als een nanostructurele beschrijving die meer waarschijnlijk is om correct te zijn.

Wanneer temperen wordt gebruikt als een onderzoeksmiddel om de nanostructuur in de zoals-gedeponeerde toestand te bestuderen, wordt aangetoond dat a-Si:H films die zijn gedeponeerd uit waterstofverdund silaangas, waarvan bekend is dat zij een verbeterde lichtdegradatiestabiliteit hebben, een kleiner type dominant open volumedefect hebben dan a-Si:H gedeponeerd uit puur silaangas, dat

(21)

typisch een aanzienlijke LID vertoont. Daarnaast neemt de waterstofpassivatiegraad van kleine open volumedefecten toe met toenemende waterstofverdunning. Ongeacht de gebruikte waterstof-verdunning worden er drie verschillende processen geobserveerd door middel van DB-PAS tijdens temperen: vacatureagglomeratie, waterstof-effusie en kristallisatie. Omdat deze processen plaatsvinden bij hogere temperaturen wanneer de a-Si:H film gedeponeerd is bij een hogere waterstofverdunning, wordt een gereduceerde mobiliteit van open volumedefecten gekoppeld aan een verbeterde lichtdegradatiestabiliteit. Ten slotte wijzen FTIR-spectroscopieresultaten erop dat de waterstof-effusie een tweeledig proces is, waarin waterstof initieel vooral effun-deert uit de kleine open volumedefecten en pas ten tweede vooral uit grotere open volumedefecten. Deze bevindingen onderstrepen nog-maals het belang van het opnemen van open volumedefecten in een accurate beschrijving van de a-Si:H nanostructuur en dienen als verder bewijs tegen het CRN en voor het DNHV.

Met gebruikmaking van deze nieuwe fundamentele inzichten in de a-Si:H nanostructuur worden de aard en de kinetica van LID in a-Si:H bestudeerd. Door een combinatie van DB-PAS, FTIR-spectro-scopie en stroomdichtheid-spanningkarakterisatie (JV) wordt geconclu-deerd dat het meest stabiele type a-Si:H wordt gedeponeerd bij een lage depositiesnelheid, wat geassocieerd wordt met het kleinste type domi-nante open volumedefect, nl. een goed door waterstof gepassiveerde di-vacature. Door middel van FTPS wordt geconcludeerd dat de vier geobserveerde subgapbijdrages verschillende lichtgeïnduceerde defect-creatiesnelheden tijdens lichtdegradatie hebben en niet een enkele tijdsafhankelijkheid ~tE volgen. Specifieker nog kan worden gesteld dat

E = 1/3 of E = 1/2 niet algemeen geldig zijn, zoals dat herhaaldelijk gerapporteerd is in de literatuur. Bovendien zijn het generatie- en recombinatieprofiel ruimtelijk gecorreleerd, waardoor de recombinatie in het top- en ondergedeelte van de absorbeerlaag verschillend is.

Omdat deze bevindingen roepen om een meer geavanceerd SWE-model, dat defecten gerelateerd aan open volumedefecten bevat naast de dbs die aanwezig zijn in het CRN, wordt een nieuwe nanoscopische beschrijving van de LID gegeven op basis van een gecombineerde karakterisatiebenadering die gebruik maakt van DB-PAS, FTIR-spectroscopie, continue-golf resonantie (cw-EPR) spectroscopie en gepulste

(22)

elektron-paramagnetische-resonantie (p-EPR) spectroscopie. Het blijkt dat de LID een langetermijn-proces bevat dat niet afhangt van de a-Si:H nanostructuur. Belangrijker nog is het feit dat er snel en langzaam relaxerende defecttypes gevonden zijn, die respectievelijk hoofdzakelijk gekoppeld zijn aan defecten gecreëerd in kleine en grote open volumedefecten. Deze nieuw gevonden koppeling tussen LID en de a-Si:H nanostructuur is een belangrijke stap op weg naar een volledig begrip en verder reduceren van het SWE en het laat opnieuw zien dat het DNHV een meer passende beschrijving vormt van de nanostructuur dan het CRN.

(23)

1 Generalintroduction

1.1 Brief historical overview of hydrogenated amorphous silicon:

therelevanceofdefectstudies

In the past five decades, hydrogenated amorphous silicon (a-Si:H) has made the transition from a material which was initially only investigated in an academic environment to a material which is now commonly applied in sensors and thin film transistors [1.1], but also in various types of solar cells. Especially notable examples include multi-junction thin film silicon (TF Si) solar cells and silicon heteromulti-junction (SHJ) solar cells [1.2]. The first publication on the formation of a-Si:H dates back to 1965 when Sterling and Swann managed to fabricate an a-Si:H film using radio frequency chemical vapor deposition (rf-PECVD) and silane gas (SiH4) as the silicon precursor [1.3]. A few years later, Chittick et al. demonstrated that CVD was the preferred method over evaporation and sputtering to make a-Si:H with reasonable photo-conductive properties. Although it was commonly believed at that time that a-Si:H could not be substitutionally doped due to the amorphous nature of the material, Chittick et al. were the first to suggest that doping might be possible after all [1.4]. It still took several years until Spear and LeComber demonstrated in 1975 and 1976 that a-Si:H could indeed be doped both p-type and n-type by adding either diborane gas (B2H6) or phosphine gas (PH3), respectively, to the SiH4 gas during the deposition [1.5,1.6]. This very important result sparked the interest of many around the world and effectively marked the start of a vastly increased interest in a-Si:H. Shortly thereafter, in 1976, Carlson and Wronski succeeded in fabricating the world’s first a-Si:H solar cell with a conversion efficiency of 2.4% [1.7]. In the following year, Brodsky et al. pointed out the impor-tance of hydrogen inclusions in a-Si:H [1.8], but otherwise the a-Si:H nanostructure was still rather poorly understood.

It was in this early development phase of a-Si:H when Staebler and Wronski published their seminal 1977 paper. In this work, they reported a dramatic decrease of both the dark conductivity and the photoconductivity upon exposure to light which could be recovered by subsequent annealing in the dark [1.9]. These observations of metastable changes in a-Si:H, which later became known as the Staebler-Wronski

(24)

Figure 1.1: Results of in situ light soaking experiments on p-i-n a-Si:H solar cells with five different absorber layer thicknesses using (a)-(d) blue LED light and (e)-(h) red LED light for light soaking and JV measurements conducted at 25 °C. Further details are described in chapter 6.

effect (SWE), would only add to the growing interest in this new material and it was quickly recognized that defects were created in the a-Si:H bandgap through light-induced degradation (LID) of the material [1.10]. It was also understood that these light-induced defects were created in addition to the so-called native defects that were already present in the material prior to light soaking due to the disordered nature of a-Si:H. The observation of LID started the field of defect studies in a-Si:H which attempts to fundamentally understand the SWE, while ultimately trying to mitigate the reduction of the solar cell conversion efficiency caused by the LID as much as possible. An

(25)

example of how LID affects the performance of a-Si:H solar cells is depicted in Fig. 1.1, which shows the light-induced changes in the solar cell external parameters, which can be deduced from current density-voltage (JV) measurements: the open-circuit density-voltage (Voc), the short-circuit current density (Jsc), the fill factor (FF), and the solar cell conver-sion efficiency (K). The degradation behavior of a series of five a-Si:H solar cells with different absorber layer thicknesses is shown for two in situ light soaking experiments conducted using red and blue light, both at an ambient temperature of 25 °C. Further details of these light soaking experiments are discussed in chapter 6.

In the years after the first observation of the SWE, various SWE models were proposed while trying to identify the dominant defect entity in a-Si:H. Examples of dominant defect entities most famously include the dangling bond (db), an undercoordinated silicon atom, which was suggested to be formed by a light-induced breaking of weak Si–Si bonds [1.11,1.12], and the floating bond (fb), an overcoordinated silicon atom [1.13,1.14]. It should be noted that the fb theory never gained wide acceptance, mostly due to the fact that the foundation of this theory relies heavily on modelling while conclusive experimental evidence of an important role of fbs in the a-Si:H nanostructure – or their existence even – is lacking. Alternatively suggested mechanisms include the motion of hydrogen at the surfaces of voids [1.15] and the long range motion of hydrogen through the material during LID [1.16]. However, the suggested formation of dbs from broken weak Si-Si bonds became more and more accepted, as well as the fact that hydrogen should play some role in passivating dbs in a-Si:H. Furthermore, the general impor-tance of dbs in a description of the dominant defects in a-Si:H increased over time, which is illustrated by the mentioned publications that appear in SWE review papers [1.17,1.18]. In addition to fundamental reasons for researching the SWE, there was a growing industrial motivation as well, since the LID could reduce the initial conversion efficiency of a-Si:H solar cells by 10-50% until a stable conversion effi-ciency would be reached. However, near the end of the 1980s the industrial research efforts slowed down due to the absence of a correlation between the LID in a-Si:H films and solar cells, or rather due to a missing understanding of how the creation of midgap states and reduction of carrier lifetimes through LID, as was observed in a-Si:H films, affected the solar cell performance.

(26)

Without exaggeration it can be stated that an astonishing amount of time and research effort has been spent on unraveling the SWE since its first observation. Unfortunately, a complete understanding and the desired elimination of this effect have still not been obtained today, which is due to three reasons. First, the complexity of the nanostructure, which is inherently linked to the amorphous nature of the material, has made it very difficult to clearly understand experimental observations. As a direct consequence of this difficulty, it has proven to be hard to correctly extract fundamental properties of defects from simulation studies. Furthermore, early fundamental studies on the a-Si:H nano-structure were conducted on a-Si:H materials which would nowadays be classified as low quality or porous a-Si:H. Therefore, the results of those studies are not generally descriptive of the opto-electrical properties of the much higher quality a-Si:H materials that have been fabricated in the past 10-20 years. Therefore, when comparing literature results it is very important to consider the quality of the discussed materials, since a-Si:H does not generally have very well-defined material properties. Instead, the possible material properties can be located anywhere on a nanostructural quality spectrum ranging from very dense to very porous materials, where the latter of which is typically rich in nanosized voids. Secondly, since the light-induced formation of defects takes place on the level of atoms and bonds between atoms, there is no characterization tool available that can provide direct insights into the nanostructure on such a small scale. This means that only indirect methods can be used to study the SWE. Last but not least, the sheer amount of publications on this topic – more than 4000 at the time of writing [1.19] – makes it extremely challenging to interpret new experimental and theoretical findings in a meaningful way when one needs to account for all the available literature.

Despite the difficulties surrounding the a-Si:H nanostructure and the SWE, significant progress has still been made in the development of TF Si solar cells over the years. For instance, the LID of a-Si:H was reduced by diluting the SiH4 gas with hydrogen gas (H2) [1.20,1.21]. Another method to reduce the LID, which is very effective but also reduces the deposition rate to a very low level, is the use of a so-called triode reactor where a voltage biased mesh is placed between the two parallel electrodes which are also present in a conventional PECVD reactor [1.22,1.23]. Next to attempts to reduce the LID in a-Si:H, the

(27)

Figure 1.2: Schematic representation of (left) the record initial triple-junction TF Si solar cell, (middle) the record stable triple-junction TF Si solar cell, and (right) the record stable double-junction TF Si solar cell. Note that in all cases at least half of the power output is generated in the a-Si:H solar cell.

problem of the SWE also encouraged research towards materials with a better stability and solar cell structures that were less susceptible to LID. A now standardized approach is to employ a-Si:H in a so-called multi-junction configuration which was first introduced by Hamakawa in 1979 [1.24]. In such multi-junction solar cells, the current generation is shared between the separate solar cells, so the individual solar cells can be made thinner, which reduces the LID of the solar cell conversion efficiency. Especially notable in this context is the development of a tandem configuration of an a-Si:H solar cell and a nanocrystalline hydrogenated silicon (nc-Si:H) solar cell in 1994 [1.25]. This cell type was introduced as the micromorph concept, which enabled a smaller LID of the solar cell conversion efficiency because the nc-Si:H solar cell did not suffer from the SWE like the a-Si:H solar cell. The multi-junction configuration was also expanded to more than two junctions and signi-ficant contributions were made by various groups. In 1997, United Solar managed to finally achieve a stable triple-junction solar cell with a

amorphous junction 67% of power output amorphous junction 50% of power output amorphous junctions 76% of power output ~13.6 % → 12.6 % (EPFL) [1.29] 15.4 % → 13.4 % (LG) [1.28] 16.3 % → ~12 % (United Solar) [1.27]

(28)

conversion efficiency of 13.0%, which would remain the record cell for more than a decade [1.26].

Until now, the highest initial solar cell conversion efficiency of 16.3%, i.e. before LID, has been reported by United Solar in 2011 for a triple-junction solar cell where a-Si:H is included as the top cell [1.27]. Furthermore, LG has also applied a triple-junction configuration to demonstrate a record stabilized solar cell conversion efficiency of 13.4% in 2013 [1.28]. Even more recently, a new record stabilized solar cell conversion efficiency of 12.6% for a double-junction (or tandem) config-uration has also been reported [1.29]. Schematic representations of the solar cell structures corresponding to these three record efficiencies are shown in Fig. 1.2. The gap between the initial and stabilized conversion efficiency values and the fact that most of the power output in those solar cells is generated by the a-Si:H solar cell illustrate that the SWE remains a problem today and why it still deserves further investigation, even four decades after it was first reported.

1.2 Problemdefinitionandresearchapproach

The SWE has proven to be a very tough problem considering the huge amount of research that has been devoted to this topic over the years, as has been outlined in section 1.1. Progress has been made in diminishing the SWE both in academic and industrial research, but the fundamental reasons behind this effect have so far remained unclear and the problem of LID has not been solved. The root cause why this problem has remained unsolved lies in the complexity of the nano-structure and hence there is no consensus on how the defects in a-Si:H should be described. Traditionally, the a-Si:H nanostructure is described as a glass-like structure known as a continuous random network (CRN), in which coordination defects, i.e. isolated dbs, are the dominant defects [1.30]. Alternatively, it has been argued that a disordered network with hydrogenated vacancies (DNHV), also termed anisotropic disordered network (ADN), is a more accurate way to describe the a-Si:H nanostructure [1.31,1.32]. In this alternative description, which was formulated on the basis of extensive Fourier transform infrared (FTIR) spectroscopy studies, defects related to not fully passivated vacancies and nanosized voids are suggested to play a

(29)

Figure 1.3: Schematic representation of the nanoscopic models which are used to describe the a-Si:H nanostructure. In the CRN the dominant defects are isolated dangling bonds, while the dominant defects in the DNHV are related to not fully passivated vacancies and nanosized voids; the depicted defect configurations in divacancies are merely given as an example.

dominant role in the a-Si:H nanostructure. A schematic overview of these two nanostructural descriptions and their proposed dominant defects is shown in Fig. 1.3.

The lack of consensus on the nanostructural description of a-Si:H illustrates why the central aim of this thesis is not to directly try to make more stable a-Si:H and reduce or even eliminate the SWE. Instead, the

Continuous Random Network (CRN)

Disordered Network with Hydrogenated Vacancies (DNHV) NV Disordered silicon Dis. silicon Dis. silicon Dis. silicon Dis. silicon

Figure adapted from [1.30]

Source: www.youtube.com/watch?v=OyVUmucwhPo

Figure adapted from [1.31,1.32]

dangling bond (db) divacancies (DV) Isolated dangling bonds Not fully passivated vacancies

(30)

first objective is to improve the fundamental understanding of the a-Si:H nanostructure and to comprehend the nature of the defects therein. Only once this first aim is reached, the description of the nanostructure and the native defects can effectively be used to describe the SWE and the nature of metastable defects in a-Si:H, which is the second objective of this thesis. Such systematic defect studies of a-Si:H aimed towards achieving a thorough understanding of the a-Si:H nanostructure can prove to be of general use to those working on defects in amorphous semiconductors and specifically those that are still aiming to mitigate or eliminate the undesired LID of a-Si:H today.

The research approach described above can be condensed to the following list of research questions, which are addressed in this thesis: 1. What does the a-Si:H nanostructure look like?

2. What do the native and metastable defects in a-Si:H look like and what can be said about their spatial location in the nanostructure and their energetic position in the bandgap?

3. What is the relation between the a-Si:H nanostructure and the defects therein?

4. Using the gained information on the nanostructure and the defects, how can the kinetics of LID over time be described? 5. Using the knowledge gained in addressing the four questions

above, how can a new model for the LID be formulated which might help to mitigate the SWE?

1.3 Thesisoutline

The remaining chapters of this thesis, which form the research component of this thesis, have all been published as articles in peer-reviewed journals or have been recently submitted to a journal and are under review at the time of writing. Although the articles span the research activities of a period of 4 years, the articles are not presented in chronological order of publication date. The practical reasons for this choice are simply that some experiments took more time to conduct than others and the writing/review process was not equally speedy for all the articles. Aside these practical reasons, the more important reason is that

(31)

in this thesis the articles are shown in accordance with the two-stage research approach, as described in section 1.2.

In chapter 2, the atomic structure of a-Si:H, i.e. the nanostructure and the defects therein, is first discussed by investigating a series of different a-Si:H films with Fourier transform infrared (FTIR) spectroscopy and Doppler broadening positron annihilation spectros-copy (DB-PAS). The latter is a materials characterization technique which has only sparsely been applied in photovoltaic research, but is shown to be a powerful tool to investigate the material structure and especially the open volume deficiencies therein, in a depth-resolved manner. To illustrate the strength of DB-PAS and the need to consider the a-Si:H material properties in terms of a nanostructural quality spec-trum a set of films with widely varying deposition conditions is investi-gated. Next, in chapter 3, a description of the electrical defects in the a-Si:H bandgap is derived for both a-Si:H films and solar cells using Fourier transform photocurrent spectroscopy (FTPS), with some special attention given to the latter, since it is particularly important to under-stand how the SWE affects the solar cell performance. For the funda-mental understanding of defects in a-Si:H and modelling purposes these results are particularly interesting, since four different sub gap distri-butions have been resolved in the a-Si:H bandgap and not the commonly reported number of two distributions, as the so-called standard density of states model dictates [1.33-1.35]. Then, annealing is extensively used as a tool to precisely monitor changes in the structural and electrical properties of various a-Si:H films to better understand the nanostructure and defects in the as-deposited state. These annealing experiments were conducted simultaneously on a-Si:H films deposited from pure SiH4 and H2-diluted SiH4, where the gas flow rate ratio R = [H2] / [SiH4] was systematically varied. This variation of R is of particular interest, since H2 dilution of the SiH4 gas is a way to fabricate more stable a-Si:H, while the fundamental mechanism responsible for this difference was previously unknown. The initial results of the annealing study, employing a combination of FTIR spectroscopy, DB-PAS, and FTPS, are presented in chapter 4. This chapter covers only the R = 0 material, while further findings on the various R > 0 a-Si:H materials are described in chapter 5. Most importantly, these results underline the importance of an a-Si:H description involving open volume deficiencies together with a CRN, also known as the DNHV or

(32)

Figure 1.4: Schematic representation of a new nanoscopic model of the LID in a-Si:H. The arrows indicate light-induced changes and the dominant defect configurations in small and large open volume deficiencies after prolonged light soaking are marked in a blue rectangle. Note that A1 and A2 correspond to clustered and

non-clustered defects, respectively. Further details are described in chapter 7.

ADN, in which divacancies play a particularly important role. Note that this opposes the classical description of the a-Si:H nanostructure that only contains the CRN.

Using the fundamental insights gained into the a-Si:H nanostructure, as described in chapters 2-5, the problem of the SWE is addressed in the last chapters of this thesis. In chapter 6, the complexity of the SWE and the defect creation kinetics during LID are discussed, both for films and solar cells. Important results reported here include the finding that an enhanced H passivation degree of small open volume deficiencies such as divacancies, improves the light soaking stability and the fact that the light-induced defect creation rate, E, cannot be generally

large open volume deficiencies

H₂

A₁

small open volume deficiencies

2×A₂

H₂

dominant after prolonged light soaking

H A₂ H A₂ H₂ 2×A₁ Si Si

(33)

described by a single power law as ~tE. Next, in chapter 7, a combination of DB-PAS, continuous wave electron paramagnetic resonance (cw-EPR) spectroscopy and pulsed EPR (p-EPR) spectroscopy is employed to reveal the nature of metastable defects and their structural and electrical properties. The gained information is used to formulate a new nano-scopic model of the LID in the a-Si:H nanostructure. More specifically, in this new model it is explained that light-induced defects can be created in both small and large open volume deficiencies, which further illustrates why the DNHV is a more appropriate description of the a-Si:H nanostructure than the CRN. A schematic representation of this new nanoscopic model of LID is shown in Fig. 1.4. Finally, the main conclusions of chapters 2-7 are grouped together in chapter 8 and some recommendations for future research are given.

1.4 References

[1.1] R.A. Street (ed.), Technology and Applications of Amorphous Silicon, Springer-Verlag, Germany (2000).

[1.2] K. Jäger, O. Isabella, A.H.M. Smets, R.A.C.M.M. van Swaaij, and M. Zeman, Solar Energy – Fundamentals, Technology, and Systems (2015; monograph in preparation).

[1.3] H.F. Sterling and R.C.G. Swann, Solid State Electron. 8, 653 (1965).

[1.4] R.C. Chittick, J.H. Alexander, and H.F. Sterling, J. Electrochem. Soc. 116, 77 (1969).

[1.5] W.E. Spear and P.G. LeComber, Solid State Commun. 17, 1193 (1975).

[1.6] W.E. Spear and P.G. LeComber, Phil. Mag. 33, 935 (1976). [1.7] D.E. Carlson and C.R. Wronski, Appl. Phys. Lett. 28, 671 (1976). [1.8] M.H. Brodsky, M.A. Frisch, J.F. Ziegler, and W.A. Lanford,

Appl. Phys. Lett. 30, 561 (1977).

[1.9] D.L. Staebler and C.R. Wronski, Appl. Phys. Lett. 31, 292 (1977). [1.10] D.L. Staebler and C.R. Wronski, J. Appl. Phys. 51, 3262 (1980). [1.11] H. Dersch, J. Stuke, and J. Beichler, Appl. Phys. Lett. 38, 456

(1981).

[1.12] M. Stutzmann, W.B. Jackson, and C.C. Tsai, Phys. Rev. B 32, 23 (1985).

(34)

[1.13] S.T. Pantelides, Phys. Rev. Lett. 57, 2979 (1986). [1.14] S.T. Pantelides, Phys. Rev. B 36, 3479 (1987). [1.15] D.E. Carlson, Appl. Phys. A 41, 305 (1986). [1.16] H.M. Branz, Phys. Rev. B 59, 5498 (1999).

[1.17] H. Fritzsche, Ann. Rev. Mater. Res. 31, 47 (2001) and references therein.

[1.18] T. Shimizu, Jpn. J. Appl. Phys. 43, 3257 (2004) and references therein.

[1.19] A search on Google Scholar executed on 27 October 2014 using the keyword “Staebler-Wronski effect” yielded about 4040 articles and 270 patents.

[1.20] L. Yang and L.F. Chen, Mater. Res. Soc. Proc. 336, 669 (1994). [1.21] J. Yang, X. Xu, and S. Guha, Mater. Res. Soc. Proc. 336, 687

(1994).

[1.22] A. Matsuda, T. Kaga, H. Tanaka, and K. Tanaka, J. Non-Cryst. Solids 59-60, 687 (1983).

[1.23] T. Matsui, H. Sai, K. Saito, and M. Kondo, Prog. Photovolt. Res. Appl. 21, 1363 (2013).

[1.24] Y. Hamakawa, H. Okamoto, and Y. Nitta, Appl. Phys. Lett. 35, 187 (1979).

[1.25] J. Meier, S. Dubail, R. Flückiger, D. Fischer, H. Keppner, and A. Shah, Proc. of 1st WCPEC, 409 (1994).

[1.26] J. Yang, A. Banerjee, and S. Guha, Appl. Phys. Lett. 70, 2975 (1997).

[1.27] B. Yan, G. Yue, L. Sivec, J. Yang, S. Guha, and C.-S. Jiang, Appl. Phys. Lett. 99, 113512 (2011).

[1.28] S. Kim, J.-W. Chung, H. Lee, J. Park, Y. Heo, and H.-M. Lee, Sol. Energy Mater. Sol. Cells 119, 26 (2013).

[1.29] M. Boccard, M. Despeisse, J. Escarre, X. Niquille, G. Bugnon, S. Hänni, M. Bonnet-Eymard, F. Meillaud, and C. Ballif, IEEE J. Photovolt. 4, 1368 (2014).

[1.30] D. E. Polk, J. Non-Cryst. Solids 5, 365 (1971).

[1.31] A.H.M. Smets, W.M.M. Kessels, and M.C.M. van de Sanden, Appl. Phys. Lett. 82, 1547 (2003).

[1.32] A.H.M. Smets and M.C.M. van de Sanden, Phys. Rev. B 76, 073202 (2007).

[1.33] M.H. Cohen, H. Fritzsche, and S.R. Ovshinsky, Phys. Rev. Lett.

(35)

[1.34] N.F. Mott and E.A. Davis, Electronic processes in non-crystal-line materials, 2nd edition (International series of monographs on physics), Clarendon Press, Oxford, United Kingdom (1979). [1.35] R.A. Street, Hydrogenated amorphous silicon, Cambridge

(36)

(37)

2 TheatomicstructureofaͲSi:H

This chapter is based on the following publication:

J. Melskens, A.H.M. Smets, S.W.H. Eijt, H. Schut, E. Brück, and M. Zeman, “The nanostructural analysis of hydrogenated silicon films based on positron annihilation studies”, J. Non-Cryst. Solids 358, 2015 (2012).

Abstract

Due to the complex nature of hydrogenated amorphous and microcrystalline silicon (a-Si:H; Ǎc-Si:H) a profound understanding of the Si:H nanostructure and its relation to the Staebler-Wronski effect (SWE) is still lacking. In order to gain more insight into the nano-structure we present a detailed study on a set of Si:H samples with a wide variety of nanostructural properties, including dense up to porous films and amorphous up to highly crystalline films, using Doppler broadening positron annihilation spectroscopy (DB-PAS) and Fourier Transform infrared (FTIR) spectroscopy. The results obtained from these material characterization techniques show that they are powerful complementary methods in the analysis of the Si:H nanostructure. Both techniques indicate that the dominant type of open volume deficiency in high quality a-Si:H seems to be the divacancy, which is in line with earlier positron annihilation lifetime spectroscopy (PALS), Doppler broadening (DB) PAS and FTIR studies.

(38)

2.1 Introduction

The complex nature of hydrogenated amorphous silicon (a-Si:H) and hydrogenated microcrystalline silicon (ȝc-Si:H) is the main reason why there are still large ambiguities associated with the understanding of the nanostructure of Si:H. As a result, there is still no consensus on the mechanism behind the reversible changes in electrical properties of a-Si:H under light exposure, also known as the Staebler-Wronski effect (SWE) [2.1]. This is why an improved understanding of the Si:H nano-structure is particularly important. Therefore, in this chapter we present a detailed study on a set of Si:H samples with a wide variety of nano-structural properties including dense up to porous films and amor-phous up to highly crystalline films. To gain more insight into the Si:H nanostructure we use a combination of positron annihilation copy (PAS) [2.2-2.6] and Fourier Transform infrared (FTIR) spectros-copy [2.7-2.9].

In Doppler broadening (DB) PAS the S parameter is a measure for the size of the dominant open volume deficiency [2.4,2.5,2.10]. The nanostructure parameter K, which can be deduced from FTIR analysis, is defined as the number of hydrogen atoms in the volume of a missing Si atom and indicates the dominant type of hydrogenated vacancy [2.7,2.9]. In this chapter it is shown that the S parameter correlates strongly with the found nanostructure parameter K. The dominant type of open volume deficiency in high quality a-Si:H seems to be the divacancy, which is in line with earlier observations from positron annihilation lifetime spectroscopy (PALS) [2.4], DB-PAS [2.3,2.5] and FTIR [2.9] studies. This result further underlines the important role of hydro-genated divacancies in a-Si:H, which has recently been illustrated in detail [2.9].

2.2 Experimentaldetails

2.2.1 Depositionoffilms

A series of Si:H films is deposited either on Corning Eagle 2000 (EA) or XG type glass substrates and all films are also deposited on crystalline silicon (c-Si) substrates for FTIR characterization. The used

(39)

# CVD type d [nm] rd [nm/min] Prf [W/cm2] p [mbar] Tsub [°C] R [-] a1 rf-PE 276 9.9 0.024 0.7 180 0 a2 rf-PE 498 12.5 0.035 1 160 0 a3 rf-PE 639 21.0 0.049 1 160 0 a4 rf-PE 564 28.2 0.069 1 160 0 a5 ETP 391 21.7 - 0.15 350 - a6 ETP 417 92.7 - 0.62 350 - Pc1 rf-PE 160 4.9 0.097 2 180 60 Pc2 rf-PE 646 21.5 0.556 9 180 63

Table 2.1: Deposition conditions and film thicknesses for samples with a wide variation in nanostructure. Samples a1-a6 correspond to amorphous samples; ȝc1-ȝc2 represent microcrystalline films.

deposition methods are radio frequency plasma-enhanced chemical vapor deposition (rf-PECVD) and expanding thermal plasma chemical vapor deposition (ETPCVD). This enables the possibility of studying the dependence of the deposition method on the nanostructure. All Si:H samples listed in Table 2.1 are intrinsic. The amorphous films are all deposited from pure silane (SiH4), whereas the microcrystalline films are deposited from hydrogen-diluted silane (R = [H2] / [SiH4]). The a-Si:H samples are deposited at different rf power densities (Prf), since this influences the nanostructure significantly without inducing a phase transition in the material. Similarly, different rf powers are used to deposit the ȝc-Si:H films. It should be clear that the samples are selected based on a wide variation in nanostructure. For example, already based on the deposition rate differences between the various amorphous films shown in Table 2.1 it is obvious that a very wide range of a-Si:H materials is discussed here and not only the high quality a-Si:H which is typically deposited at rd 10 nm/min. More specifically, the deposition rates shown in Table 2.1 indicate that the nanostructural quality spectrum is spanned between very dense and very porous materials. Note that the listed thicknesses (d) are determined from reflectance/ transmittance (RT) measurements.

2.2.2 Characterizationoffilms

DB-PAS is performed using the positron depth profiling method [2.10] to study vacancies in Si:H films deposited on Corning glass. A variable energy positron beam with an intensity of 105_{e+/s and a}

(40)

diameter of 8 mm is obtained via moderation of positrons emitted from a 22_{Na source through a tungsten foil. The annihilation radiation is} measured at room temperature by a germanium solid state detector with a resolution of 1.4 keV as the Doppler-broadened 511 keV Ȗ-ray photopeak characteristic of 2Ȗ annihilation of electron-positron pairs. A total of 5×105_{counts are collected under the Ȗ-ray photopeak for each} of the selected positron implantation energies (Epositron) in the range of 0.25-25 keV.

FTIR spectroscopy is conducted to analyze the Si:H films depos-ited on c-Si substrates using a Thermo Fisher Nicolet 5700 FTIR spec-trometer in transmittance mode.

2.3 Results

In DB-PAS two parameters are conventionally used to charac-terize the spectrum of the annihilation radiation: S and W. These two Doppler broadening parameters are calculated as the ratio of the counts in the central and wing parts of the Doppler-broadened 511 keV photopeak to the total peak count [2.10], respectively. Physically, S can be understood to represent annihilation events with low momentum electrons (small energy shift from 511 keV) which are the valence electrons of atoms in the sample under investigation. Similarly, W corresponds to annihilation events with high momentum electrons (larger energy shift from 511 keV) which are the core electrons of atoms in the sample. As an example, the positron Doppler broadening depth profiles of samples a1 and a4 (see Table 2.1) are shown in Fig. 2.1.

Generally, three regions can be identified in the depth profiles shown in Fig. 2.1a: a fairly constant S value in the low energy range (1-5 keV) that represents the Si:H film and decreasing S values for positron energies below (<1 keV) and above (>5 keV) this region due to positron annihilation at the surface and in the substrate respectively. It is clear from the measurement data in Fig. 2.1a and c that both S and W as well as the Si:H film thicknesses for samples a1 and a4 are substantially different. In order to extract the S and W values both the S(E) and W(E) depth profiles are fitted using the VEPFIT program [2.11] assuming a two-layer structure (Si:H film / Corning glass) with constant S and W values for each layer. The depth profiles in Fig. 2.1a and c are fitted by

(41)

Figure 2.1: DB-PAS depth profiles of samples with different thicknesses (a1 and a4) showing (a) S parameter, (b) result of VEPFIT analysis using S(Epositron) as input, (c) W

parameter, and (d) S-W diagram with Epositron as running parameter.

solving the second order differential equation that describes the positron diffusion process in the sample. A Makhovian implantation profile is used with a mean implantation depth ݖҧ= AEn_{/ȡ, where E is the} positron implantation energy in keV and the parameters n = 1.62 and A = 4.0 ȝg·cmí2_·keVí1.62_{[2.10]. A density of ȡ = 2.3 g/cm}3_{is assumed for} the Si:H films. The results of the fitting procedure are shown as the solid and dashed lines in Fig. 2.1a and c. This fitting procedure is conducted to finally obtain values for S and W of the Si:H film and the glass substrate; an example of the calculated S values for both the film and the substrate as well as the thickness of the film is given in Fig. 2.1b. It seems from Fig. 2.1d that both films shown here have a homogeneous composition, since the (S,W) coordinates corresponding to the surface, the bulk of the Si:H film, and the glass substrate closely approximate a straight line in the S-W diagram.

0.52 0.54 0.56 0.58 0.60 0.62 0 5 10 15 20 a1 a4 E_positron [keV] S [-] 0 200 400 600 800 1000 a1 a4 distance from surface [nm]

0 5 10 15 20 0.01 0.02 0.03 0.04 0.05 0.06 surfaces films glass substrates (d) glass substrates films (b) (a) (c) films glass substrates films W [-] E_positron [keV] glass substrates 0.52 0.54 0.56 0.58 0.60 0.62 S [-]

(42)

Figure 2.2: S-W diagram with normalized values obtained from VEPFIT. The error bars for S/Sc-Si are not visible because they do not exceed the width of the markers. The

arrows indicate an increase in deposition rate (rd). Three different types of positron traps

are distinguished: divacancies, multivacancies, and nanosized voids. The corresponding values of the nanostructure parameter K are indicated for each of these three types of dominant open volume deficiency. The second listing of sample a5 in the legend refers to the same sample after annealing (a.a.).

All samples listed in Table 2.1 are analyzed using DB-PAS and VEPFIT to obtain the (S,W) coordinate for each Si:H film. The thus obtained (S,W) values can be used to observe trends in the data, but do not enable a direct comparison with previously reported (S,W) values in the literature. This is due to the fact that the absolute (S,W) values depend on the overall instrument resolution of the Doppler broadening setup and the momentum ranges applied in the determination of S and W. In order to enable a better comparison with previously reported data, all (S,W) values are normalized to the (S,W) coordinate of c-Si, which is characterized individually by DB-PAS for this purpose (Sc-Si = 0.5744 ±

1.03 1.04 1.05 1.06 1.07 0.65 0.70 0.75 0.80 0.85 K < 1 K = 3 K = 2 a1 (rf) a2 (rf) a3 (rf) a4 (rf) a5 (ETP) a5 a.a. (ETP) a6 (ETP) Pc1 (rf) Pc2 (rf) W/ W c-Si [-] S/S_c-Si [-] divaca n cies multivacancies nanosized voids

(43)

0.0002, Wc-Si = 0.0265 ± 0.0002). The normalized (S,W) coordinates of the various studied Si:H films (see Table 2.1) are shown in Fig. 2.2.

From various positron annihilation studies on a-Si:H it has been deduced that specific S/Sc-Si values correspond to different types of positron traps, i.e. open volume deficiencies. The following S/Sc-Si values have been reported [2.3-2.5]: monovacancies 1.030, divacancies 1.034-1.038, vacancy clusters 1.047-1.061, and microvoids 1.10-1.14. Since the smallest possible stable mass deficiency in silicon at room temperature is the divacancy [2.7,2.9], monovacancies are not considered here. Further it is assumed that, depending on the a-Si:H nanostructure, a broad and continuous spectrum of S/Sc-Si values is possible, so the S/Sc-Si ranges of 1.038-1.047 and 1.061-1.10 are not excluded. Using this assumption and the vacancy terminology from Ref. [2.7], the following three types of dominant positron traps and their corresponding S/Sc-Si values are distinguished: divacancies (<1.038), multivacancies (1.038-1.061), and nanosized voids (>1.061).

Although the matching with the earlier reported S/Sc-Si values is good, it is more important to realize that the trend in S/Sc-Si indicates a considerable variation in nanostructure between the samples. For the amorphous rf-PECVD samples the size of the dominant type of open volume deficiency in the film increases with an increase in deposition rate (due to the increased Prf). This can be understood by realizing that at a constant substrate temperature (Tsub) a low deposition rate tends to lead to dense films, whereas a high deposition rate results in more porous films [2.12]. The relatively large increase in S/Sc-Si between samples a3 and a4 may be due to the formation of a short-lived e+_-eí_pair known as positronium (Ps) which can occur if the diameter of the nano-sized voids is larger than 0.9 nm. If this is indeed the case, then the voids in film a4 are not likely to be filled with H2, as this would suppress formation of Ps and hence lower S/Sc-Si [2.3]. Further analysis of sample a4 using two-dimensional Angular Correction of Annihilation Radiation (2D-ACAR) [2.13] would be fruitful here.

Similar to samples a1-a4, it can be seen that for the ETP films a5 and a6 S/Sc-Si increases for increasing deposition rate. Again this can be understood as a change from dense to more porous material. However, the increase in S/Sc-Si with increasing deposition rate (ǻS/ǻrd) is smaller for samples a5-a6 than it is for a1-a4. This can be due to the higher substrate temperature for the ETP samples which enables the possibility

(44)

of growing dense material also at higher deposition rates. Of special interest among the ETP films is the change in W/Wc-Si for sample a5 after annealing it for 8 h in Ar at 525 °C. The annealing process removes all Si-bonded hydrogen from the film without significantly affecting the crystalline fraction; this has been verified with FTIR and Raman spec-troscopy. While the annealing cycle hardly changes S/Sc-Si (no change in the size of the dominant type of open volume deficiency), there is a notable increase in W/Wc-Si. After the effusion of hydrogen from the Si lattice the relative probability of an annihilation event with Si core electrons is significantly increased, resulting in an increased W/Wc-Si after annealing.

The interpretation of (S,W) values becomes more complicated for heterogeneous materials like ȝc-Si:H, which consists of microcrystalline grains with a certain volume of amorphous tissue between them. Since the microcrystalline grains contain far less open volume deficiencies in reference to the amorphous tissue, DB-PAS mainly probes the vacancies in the a-Si:H tissue. As a consequence, highly crystalline ȝc-Si:H that has large microscopic cracks and little a-Si:H tissue gives an S parameter closer to the value for c-Si (S/Sc-Si = 1). Because sample ȝc2 has a higher crystalline fraction than ȝc1 (79% vs. 47% according to a Raman analysis), the positrons diffusing through the film are probing more crystalline material in sample ȝc2. This may explain why sample ȝc2 has a lower S/Sc-Si value than sample ȝc2. Further studies on ȝc-Si:H would be useful to verify this presumption.

2.4 Discussion

FTIR spectroscopy is commonly used to study the configurations in which hydrogen can be found in the Si:H lattice. Particularly important is the Si-bonded hydrogen since it significantly affects the optical and electrical material properties and plays a key role in the SWE. The Si–H bonds are known to have three characteristic absorption modes in the infrared (IR) absorptance spectrum: wagging modes at Ȧ = 640 cmí1_{, a bending scissor mode at Ȧ = 840-890 cm}í1_{, and stretching} modes around Ȧ = 2000 cmí1_{and Ȧ = 2100 cm}í1_{. The latter two modes} are known as the low stretching mode (LSM) and high stretching mode (HSM). It has recently been argued that the LSM corresponds to

(45)

Figure 2.3: (a) Si–H stretching modes obtained from FTIR analysis for sample a3 showing the LSM and HSM using deconvolution of the fitted measurement data; (b) Comparison of LSM and HSM for samples a1-a4 by normalizing both stretching modes to the maximum of each LSM. The arrow indicates a decrease in the nanostructure parameter K.

divacancies while the HSM is more sensitive to larger vacancies up to nanosized voids [2.7,2.9]. Note that in this terminology a vacancy is defined as a lattice site with missing adjacent Si atoms. In a divacancy this number of missing adjacent Si atoms is two, whereas several tens to thousands could be missing in a nanosized void. Following the termi-nology in Refs. [2.7] and [2.9] the nanostructure parameter K can be

0.96 0.98 1.00 1.02 1.04 1900 1950 2000 2050 2100 2150 2200 0.00 0.25 0.50 0.75 1.00 1.25 (b) (a) a3 measured fitted background LSM HSM ab so rpta nce [a .u. ] K=3 LSM K<1 HSM normalis ed absorptance [a.u.] wavenumber [cm-1] LSM HSM a1 a2 a3 a4

(46)

defined as the number of hydrogen atoms in the volume of a missing Si atom; e.g. for a divacancy K = 3 and for nanosized voids K < 1.

Using this interpretation of the LSM and HSM more insight can be gained into the nanostructure of the films. To illustrate the effect of the nanostructure it is particularly useful to compare samples a1-a4 since they have already proven to show a large variation in S/Sc-Si. The results of this analysis can be seen in Fig. 2.3. As an example a close-up of the IR absorptance spectrum for sample a3 is shown in Fig. 2.3a. After subtraction of the background (due to the c-Si substrate) the absorptance due to hydrides can be deconvoluted in the LSM and HSM. Therefore a previously described fitting routine [2.14] is utilized which uses the non-absorbing part of the spectrum (the background spectrum) to first obtain the refractive index and thickness of the film, after which two Gaussians can be fitted to the absorptance spectrum.

Using this approach the LSM and HSM can be extracted for all samples and the resulting fitted Gaussians for samples a1-a4 are depicted in Fig. 2.3b. To allow for an easy comparison between the samples both stretching modes have been normalized to the LSM peak for each sample. Clearly there is a large increase in the HSM area for increasing rd (as is also shown in Ref. [2.12]), which illustrates the transition from dense (K = 3) to more porous (K < 1) films. The nano-structure transition is also concluded from the previously discussed DB-PAS results. The peak shift ǻȦLSM towards the higher wavenumbers can be understood as a reduced screening of the monohydride bulk configurations in the more porous films [2.7]. This leads to the important conclusion that there is a strong correlation between S/Sc-Si obtained from DB-PAS and K obtained from FTIR analysis, meaning that these methods complement each other in a nanostructural analysis of Si:H films. The correlation also shows that both DB-PAS and FTIR spectroscopy indicate that the dominant type of open volume deficiency in high quality a-Si:H seems to be the divacancy, which is in line with earlier PALS [2.4], DB-PAS [2.3,2.5] and FTIR [2.9] studies.

2.5 Conclusions

A wide variety of material characterization techniques is used on the same set of Si:H samples that have a very different nanostructure.

Hydrogenated amorphous silicon

,ǇĚƌŽŐĞŶĂƚĞĚ

ĂŵŽƌƉŚŽƵƐƐŝůŝĐŽŶ͗

ŶĂŶŽƐƚƌƵĐƚƵƌĞ

Hydrogenatedamorphoussilicon:

nanostructureanddefects

JimmyMelskens

Hydrogenatedamorphoussilicon:

nanostructureanddefects

Proefschrift

Tableofcontents

Summary

Samenvatting

1

Generalintroduction

1.1

Brief historical overview of hydrogenated amorphous silicon:

therelevanceofdefectstudies

1.2

Problemdefinitionandresearchapproach

1.3

Thesisoutline

1.4

References

2

TheatomicstructureofaͲSi:H

Abstract

2.1

Introduction

2.2

Experimentaldetails

2.2.1 Depositionoffilms

2.2.2 Characterizationoffilms

2.3

Results

2.4

Discussion

2.5

Conclusions