Cellulose-Based Electrical Insulation Materials Dielectric and Mechanical Properties REBECCA HOLLERTZ

Home , Dissolving pulp, Paper chemicals, Paper density, Wood fibre

Cellulose-based electrical insulation materials Dielectric and mechanical properties REBECCA HOLLERTZ

Doctoral thesis KTH Royal Institute of Technology School of Chemical Science and Engineering Department of Fibre and Polymer Technology

ISBN 978-91-7729-327-9

TRITA – CHE-report 2017:21

ISSN 1654-1081

Tryck: US-AB, Stockholm, 2017

The following papers are reprinted with permission from: Paper I: Springer Papers II, III, VII and VIII: IEEE Transactions on Dielectric and Electrical Insulation

AKADEMISK AVHANDLING

Som med tillstånd av Kungliga Tekniska högskolan i Stockholm framläggs till offentlig granskning för avläggande av teknologie doktorsexamen fredagen den 12 maj 2017, kl. 10.00 i F3, Lindstedtsvägen 26, KTH, Stockholm. Avhandlingen försvaras på engelska. Fakultetsopponent: Professor Markus Biesalski, TU Darmstadt, Tyskland

ABSTRACT

The reliability of the generation and distribution of electricity is highly dependent on electrical insulation and is essential for the prosperity of our society and a ubiquitous part of our everyday life. The present study shows how some important material properties affect the electrical properties of cellulose-based electrical insulation systems which are used together with mineral oil in high-voltage transformers. Among other things, the effects of paper density and of the lignin content of the fibres on the dielectric response and charge transport of the papers have been studied. The underlying mechanisms of the inception and propagation of streamers, responsible for the most costly failures in transformers, at the oil-solid interface have been investigated and the important role of paper morphology on streamer propagation has been demonstrated. With papers, in contrast to films of synthetic polymers and microfibrillated cellulose, the branching of streamers increased and the length of slow negative streamers decreased. It was also shown that for polymers with permittivities close to that of the oil, the inception voltage was higher than with polymers with higher permittivities. This was explained by a different distribution of the electric field at the interface, and it means that with a solid restricting the initiation volume and with a permittivity of the solid matching the permittivity of the liquid, there is a lower risk for streamer inception, in negative polarity. Fibres were also modified prior to paper sheet preparation in attempts to improve the mechanical and dielectric properties. The properties of papers containing cellulosic micro- and nanofibrils and SiO2 and ZnO nanoparticles indicate that these additives can indeed be used to improve both the mechanical and dielectric properties. For example, the tensile strength index of papers with periodate-oxidised carboxymethylated CNFs increased by 56 % and by 89 % with a 2 wt% and 15 wt% addition, respectively, of the fibrils, when used together with different retention aids. A three-layered structure with two papers laminated together with a thin layer of microfibrillated cellulose also showed an increased DC breakdown strength by 47 % compared to a single-layer paper with a similar thickness.

SAMMANFATTNING

En god elektrisk isolering, i olika utrustningar i distributionskedjan, är i hög grad väsentligt för en tillförlitlig överföring och distribution av elektricitet. Detta är i sin tur en förutsättning för dagens välstånd och är av stor betydelse i vår vardag. Resultaten från doktorandprojektet visar hur flera materialegenskaper påverkar dielektriska egenskaper hos papper som används i kombination med mineralolja som elektriskt isoleringsmaterial i högspänningstransformatorer. Bland annat har inverkan av såväl papperets densitet som fibrernas ligninhalt på både laddningstransport och elektriska förluster studerats och kvantifierats. Dessutom har mekanismerna för uppkomsten och utbredning av streamers, som är den mest kostsamma orsaken till transformatorhaveri, i gränsytan mellan olja och fast fas undersökts. Resultaten visar att det finns ett klart samband mellan papperets morfologi och utbredningen av streamers. För långsamma negativa streamers på papper, jämfört med på filmer av syntetiska polymerer och mikrofibrillerad cellulosa, ses en ökad förgrening som vidare leder till kortare streamers. Genom att jämföra uppkomsten och utbredningen av streamers på olika polymerer studerades också inverkan av isoleringsmaterialets permittivitet. Med polymerer med en permittivitet som matchar oljepermittiviteten krävdes en högre elektrisk initieringsspänning för streamers. Detta innebär att risken för uppkomst av streamers i negativ polaritet minskas i gränsytan när permittiviteten hos det fasta materialet överensstämmer med oljans. Baserat på dessa resultat genomfördes olika fibermodifieringar för att förbättra de mekaniska och dielektriska egenskaperna. Resultat för papper som innehåller fibrer som modifierats med mikro- och nanofibriller och speciellt utvalda nanopartiklar visar att dessa kan användas som tillsatsmedel för att förbättra båda dessa egenskaper. Dragstyrkeindex ökade med 56 % och 89 % med 2 vikt% respektive 15 vikt% nanofibriller (karboxymetylerade fibriller som perjodat oxiderats) tillsatta i kombination med lämpliga retentionsmedel. Dessutom ökade DC genomslagsstyrkan med 47 % med en tre-skikts konstruktion med två omgivande papper som laminerats med ett skikt av mikrofibrillerad cellulosa jämfört med ett homogent papper av samma tjocklek.

LIST OF PUBLICATIONS

I. Dielectric properties of lignin and glucomannan as determined by spectroscopic ellipsometry and Lifshitz estimates of the non-retarded Hamaker constants Rebecca Hollertz, Hans Arwin, Bertrand Faure, Yujia Zhang, Lennart Bergström, Lars Wågberg, Cellulose, 20, p. 639-1648, August 2013

II. Effect of Composition and Morphology on the Dielectric Response of Cellulose-based Electrical Insulation Rebecca Hollertz, Claire Pitois, Lars Wågberg, IEEE Transactions on Dielectrics and Electrical Insulation, 22, p 2239-2248, August 2015

III. First mode negative streamers at mineral oil-solid interfaces David Ariza, Rebecca Hollertz, Marley Beccera, Claire Pitois, Lars Wågberg, Accepted for publication in IEEE Transactions on Dielectrics and Electrical Insulation

IV. Inception of first mode negative streamers at mineral oil-solid interfaces David Ariza, Marley Beccera, Rebecca Hollertz, Ralf Methling, Lars Wågberg, Sergey Gortschakow Manuscript

V. Influence of paper properties on streamers creeping in mineral oil David Ariza, Marley Beccera, Rebecca Hollertz, Lars Wågberg, Ralf Methling, Sergey Gortschakow, Accepted for publication in the Proceedings of the 2017 International Conference on Dielectric Liquids

VI. Chemically modified cellulose micro- and nanofibrils as paper- strength additives Rebecca Hollertz, Verónica López Durán, Per Larsson, Lars Wågberg Submitted

VII. Dielectric Response of Kraft Papers from Fibres Modified by Silica Nanoparticles Rebecca Hollertz, David Ariza, Claire Pitois, Lars Wågberg 2015 Annual Report Conference on Electrical Insulation and Dielectric Phenomena, p 459-462

VIII. Comparison of Oil-impregnated Papers with SiO2 and ZnO Nanoparticles or High Lignin Content, for the Effect of Superimposed Impulse Voltage on AC Surface PD Roya Nikjoo, Hans Edin, Nathaniel Taylor, Rebecca Hollertz, Martin Wåhlander, Lars Wågberg, Accepted for publication in IEEE Transactions on Dielectrics and Electrical Insulation

CONTRIBUTIONS TO PUBLICATIONS

I. Main author*. Hollertz prepared the purified lignin and spin-coated the surfaces of lignin and hemicellulose. The purified hemicellulose was prepared by Dr Yujia Zhang. The spectroscopic ellipsometry measurements were performed with the instruction and supervision of Professor Hans Arwin. The calculations of the non-retarded Hamaker constants were made by Dr Bertrand Faure. All the co- authors were engaged in writing the manuscript. II. Main author*. Hollertz prepared all the samples and performed all the measurements. Dr Uno Gäfvert assisted in the interpretation of the data. III. Co-author. Hollertz took part in planning the experiments and gave input into the sample preparation. Hollertz prepared the paper samples and characterized the papers and polymers. The streamer tests were performed and the data analysed by David Ariza. Hollertz wrote the descriptions of the materials and took an active part in the full manuscripts, with analysis and writing. IV. See III. V. See III. VI. Main author*. The fibrils were prepared by PhD-Student Veronica Lopez Duran. Hollertz prepared all the paper sheets, while the characterization was performed together with Veronica Lopez Duran. Although Hollertz had the main responsibility for the analysis of the data and the paper writing, it was done together with the other co-authors. VII. Main author*. Prepared all the paper samples and characterized them. VIII. Co-author. Hollertz prepared the paper samples and characterized the material properties, besides their dielectric response and behaviour under discharges, and took part in writing the manuscript. The ZnO nanoparticles were surface-modified and characterized by Martin Wåhlander.

*Main author embodies the work of formulating hypotheses, planning the experimental work and having the main responsibility for the manuscript. At the beginning of the PhD project, these tasks were undertaken together with Professor Lars Wågberg, but later more independently under his supervision.

Conference papers (peer-reviewed) not appended to this thesis:

Novel cellulose nanomaterials – towards use in electrical insulation Rebecca Hollertz, Claire Pitois, Lars Wågberg International Conference on Dielectric Liquids (ICDL, Bled 2014)

Measurements of the charge and velocity of streamers propagating along transformer oil-pressboard interfaces David Ariza, Marley Becerra, Rebecca Hollertz, Claire Pitois International Conference on Dielectric Liquids (ICDL, Bled 2014)

Kraft-Pulp Based Material for Electrical Insulation Rebecca Hollertz, Claire Pitois, Lars Wågberg Nordic Insulation Symposium on Materials Components and Diagnostics (NORD- IS, Copenhagen 2015)

On the initiation of negative streamers at mineral oil-solid interfaces David Ariza, Marley Beccera, Rebecca Hollertz, Claire Pitois IEEE Conference on Electrical Insulation and Dielectric Phenomena (CEIDP, Ann Arbor 2015), no. 149, session 6B-3

Propagation of streamers along mineral oil-solid interfaces David Ariza, Marley Beccera, Rebecca Hollertz, Claire Pitois IEEE Conference on Electrical Insulation and Dielectric Phenomena (CEIDP, Ann Arbor 2015) no. 362, session 4C-1

Abstracts were also presented at American Chemical Society (ACS) Meeting in New Orleans in 2013 and in San Diego in 2016.

PREFACE

I would like especially and sincerely to thank my supervisor Professor Lars Wågberg. Besides being the brain behind the many ideas and hypotheses in the project, he has also created a welcoming environment built on trust and responsibility. The level of independence, encouraged and allowed, has also been of uttermost importance for me, enabling me to pursue my research with a growing family. The initiative for this project was taken by Dr Claire Pitois, at the time at ABB Corporate Research. Pitois contributed greatly and followed the project as co- supervisor until 2016. At ABB Figeholm, the initiative was led by Dr Torbjörn Brattberg. The contacts with ABB in Västerås, Figeholm and in Ludvika have been inspiring and extremely helpful. I would therefore like to acknowledge ABB AB for their crucial role in the establishment and funding of this project and also for their professionalism and attentiveness throughout the project: Dr Claire Pitois, Dr Torbjörn Brattberg, Dr Lars Schmidt, Adjunct Professor Henrik Hillborg, Adjunct Professor Tord Bengtsson, Dr Uno Gäfvert, Assistant Professor Mikael Unge, Dr Santanu Singh, Jan Hajek and Anders Ericsson for their contribution to the project and also for their support over these years. Together with ABB AB, the Swedish Energy Agency is gratefully acknowledged for funding the project through the ELEKTRA and ELFORSK programs. Early we discovered two crucial parameters in the project that were important to address. Firstly, the dielectric properties stipulated by the type of wood pulp and the paper quality were not clear to us. Secondly, to characterize the streamer propagation and to create an understanding of the mechanisms behind streamer inception a completely new set-up had to be designed and built. To address the first question two initial projects were initiated to investigate the influence of pulp and paper composition and processing on the dielectric response. The first paper was written in collaboration with Professor Lennart Bergström at Stockholm University and Professor Hans Arwin at Linköping University to whom I express my gratitude for their expertise, time and effort. In the second paper, we wanted to characterize the dielectric response of different papers. In connection with this work I would like to acknowledge my collaborators at the Electromagnetic Engineering department for their valuable input and for welcoming me to their laboratory. For the interpretation of the dielectric response data, I had the privilege on several occasions to meet Dr Uno Gäfvert who gave me a highly valuable input and introduced me to many new concepts. To be able to characterize the streamers, a collaboration with Associate Professor Marley Beccera, working both at KTH Electromagnetic Engineering Department and ABB Corporate Research, was initiated and resulted in a PhD

project with the PhD-student David Ariza. For the great collaboration and all excellent and thorough work I would like to thank David Ariza. The characterization of streamers in positive polarity was also strengthened by collaboration with the Leibniz Institute for Plasma Science and Technology in Greifswald, Germany. For DC breakdown tests performed at ABB, I would also like to acknowledge the work done by Dr Zargham Jabri and Dr Anneli Jedenmalm. The addition of chemically modified nanofibrils resulted in a significant improvement and I would like to acknowledge my co-authors Veronica López Durán and Dr Per Larsson. The results would not have been obtained nor published without you! Considering previous work on dielectrics, we found it attractive to use the Layer-by-Layer (LbL) technique, well-known in the Fibre Technology group, to introduce nanoparticles into insulating papers. I would like to acknowledge Eka chemicals (especially Dr Michael Persson) for suppling silica nanoparticles, Martin Wåhlander for modifying the zinc oxide nanoparticles and Oruç Köklükaya for being most generous with both time and knowledge. The partial discharge measurements were performed by Dr Roya Nikjoo. Unfortunately, the samples characterized were the last she managed before her dissertation, and additional complementary measurements were not possible. I would also like to thank co- authors Dr Nathaniel Taylor, Professors Hans Edin and Eva Malmström. I would like to direct my thanks to the whole Fibre Technology group; for the warm and helpful atmosphere. I especially wish to thank my roommates and also Johan for the pleasant discussions, many times scientific, during early breakfast and for all the help in the laboratory. I feel very grateful for the possibility to work with Professor Lars Ödberg who has been involved as co-supervisor and has given valuable input and support. I would also like to thank Mia Hjertén for all her support. In summary: This has been a lot of fun! Finally, I would like to thank my parents and Mikael, Tage and Klara for being there, and also Elisabeth and Bengt for all your support.

Stockholm, 2017 Rebecca Hollertz

TABLE OF CONTENTS 1. INTRODUCTION ...... 1 2. BACKGROUND ...... 2 2.1 Paper and pressboard in oil-filled transformers ...... 2 2.1.1 Outlook ...... 3 2.2 A short introduction to wood fibres ...... 6 2.3 Paper physics and chemistry ...... 8 2.3.1 Cellulose micro- and nanofibrils ...... 10 2.3.2 Layer-by-Layer (LbL) adsorption ...... 11 2.4 A short introduction to dielectric properties ...... 12 2.4.1 Dielectric response ...... 12 2.4.2 Streamers in oil and at oil-solid interfaces ...... 15 3. EXPERIMENTAL...... 17 3.1 Materials ...... 17 3.1.1 Wood pulps ...... 17 3.1.2 Transformer oil ...... 17 3.1.3 Polymers...... 17

3.1.4 SiO2 and ZnO nanoparticles ...... 18 3.2 Methods –material preparation ...... 18 3.2.1 Lignin and glucomannan model materials (Paper I) ...... 18 3.2.2 Paper sheets ...... 19 3.2.3 Cellulose micro- and nanofibrils ...... 20 3.2.4 Paper sheets with micro- and nanofibrils as strength additives (Paper VI) ...... 21 3.2.5 Films from micro- and nanofibrils ...... 21 3.2.6 Three-layered sandwich structure with microfibrillated cellulose ..... 22

3.2.7 Papers from fibres modified with SiO2 and ZnO nanoparticles (Papers VII and VIII) ...... 22 3.3 Methods –material characterization ...... 23 3.3.1 Charge density measurements ...... 23 3.3.2 Thickness measurements ...... 24 3.3.3 Mechanical testing ...... 24 3.3.4 Spectroscopic ellipsometry (Paper I) ...... 24 3.3.5 Dielectric spectroscopy (Papers II, VII and VIII) ...... 25

3.3.6 DC Breakdown test ...... 26 3.3.7 Streamer measurements (Papers III-V) ...... 27 3.3.8 Partial discharge (PD) measurements (Paper VIII) ...... 27 4. RESULTS AND DISCUSSION ...... 29 4.1 The influence of paper density and morphology on dielectric response (Paper II) ...... 30 4.2 The influence of wood fibre composition on dielectric properties .... 31 4.2.1 Spectroscopic ellipsometry measurements of wood biopolymers (Paper I) ...... 32 4.2.2 Dielectric response (Papers II, VII and VIII) ...... 33 4.2.3 Surface partial discharge (PD) resistance (Paper VIII) ...... 33 4.3 The influence of permittivity, density and morphology on streamer inception and propagation (Papers III-V) ...... 34 4.4 Papers with cellulosic fibres and fibrils combined ...... 37 4.4.1 DC breakdown strength of a three-layered sandwich with microfibrillated cellulose ...... 37 4.4.2 Micro- and nanofibrils as paper strength additives (Paper VI) ...... 39

4.5 Papers from wood fibres with adsorbed SiO2 and ZnO nanoparticles (Papers VII and VIII) ...... 42 4.5.1 Adsorption ...... 42 4.5.2 Dielectric response ...... 45 4.5.3 Surface partial discharge (PD) resistance ...... 46 5. CONCLUSIONS ...... 48 6. FUTURE WORK ...... 51 7. REFERENCES ...... 52

I have strived, when writing this thesis, to make it digestible for colleagues with a background in chemistry or in electrical engineering and I hope therefore that both can acquire some new understanding or even discover an interest into, for them, a new field. A dielectric material is an electrical insulating material and a summary of some dielectric properties of materials relevant for this study is presented in Table A1 in the Appendix on page 60. A list of abbreviations is found on page 59.

INTRODUCTION

1. INTRODUCTION

The integration of renewable energy sources into the electric grid, along with higher voltage ratings and a transition to the use of DC grids are factors that contribute to a more sustainable electricity production and an efficient energy transmission and distribution. These processes are already underway, or planned, and they are necessary for a transition away from power generation based on the burning of fossil fuels. As a result, there are new demands on the components of the power transmission system; to upgrade the power grid requires reliable insulating materials that have low energy losses and which prevent failures that can lead to costly maintenance work or even blackouts. A key component in a power transmission system is the high-voltage power transformer. The insulation material in these is of paper and pressboard impregnated with and surrounded by mineral oil. Insulation failure has been identified as the leading cause of transformer failure in several studies1,2. Recently, increasing efforts to develop new and also more value-added products from wood fibres have resulted in new and more sophisticated modifications of wood fibres, which can potentially be used to improve cellulose-based dielectric materials. This project has focused on how the chemistry and the morphology of the solid insulation material is affected by, and can affect, the electrical phenomena in a high-voltage transformer. It is, however, difficult to separate the effects of individual parameters because they can seldom be independently changed. With a novel experimental set-up, studying streamer inception and propagation, and an interdisciplinary collaboration, we have created a solid foundation which has enabled us to investigate the underlying mechanisms behind this dielectric phenomenon and to begin to establish connections with the properties of the solid insulation. This is essential for identifying which and how material properties should be altered in order to improve their dielectric performance. We also present novel fibre-engineering techniques that could be explored to alter the mechanical and electrical insulating properties of the wood-fibre-based insulation. The aim was to provide new understandings that can inspire and suggest new strategies to the producers of paper and pressboard used in transformers.

BACKGROUND

2. BACKGROUND

2.1 Paper and pressboard in oil-filled transformers The combination of oil and paper has been used for more than a century as electrical insulation in power transformers (Figure 2.1) and it still today plays an important role in high-voltage transformers and cables.  The liquid insulation is used as coolant and to provide dielectric strength.  The solid insulation is used as a structural material to create barriers and thereby subdivide the highly stressed oil. A good dielectric material has low dielectric losses and a high resistance to degradation and electrical breakdown both on the surface and in the bulk. In a transformer, the insulating solid barriers should be perpendicular to the electrical field to achieve the maximum dielectric strength. The solid insulation should be impregnable by oil and able to withstand a substantial mechanical load. The pressboard and paper elements are inserted and then dried exhaustively and impregnated in the oil-filled transformer tank. The insulation remains in the transformer tank throughout the lifetime of up to 30-50 years where it is subjected to high mechanical and electrical stress at elevated temperatures. A useful introduction to transformer insulation is presented by Oommen and Prevost3,4. Paper and pressboard are made from a 3D structured network of wood fibres. High-density pressboard (up to 1200 kg/m3) for transformer insulation is made by pressing paper sheets together and drying them under heat and high pressure in a specially designed hydraulic press. To obtain structural strength, the pressboard parts have a thickness of at least 1 mm. Structural parts are subsequently made from several sheets of pressboard that are glued together.

BACKGROUND

Figure 2.1. The world’s first 1100kV HVDC converter transformer, successfully tested by ABB in 2012. It weighs 800 tons and is 32 m long. The transformer tank conceals iron cores, copper windings and an insulation system of paper, pressboard and oil.

2.1.1 Outlook Increasing voltage levels in electrical systems are a clear market development for both HVAC (high voltage alternating current) and HVDC (high voltage direct current) networks, demanding new insulating materials able to withstand higher electrical stress and displaying less losses than current designs. A predominant part of the research and knowledgebase regarding dielectrics and the design of electrical equipment is, for historical reasons, concerned with an AC-field dependence. This provides a great challenge, as DC applications are becoming increasingly important for the long-distance transmission needed with the renewable energy sources located far from populated regions, and it highlights the importance of adapted diagnostics, design and material properties of electrical insulation for different applications. A variety of cellulose-based products for electrical insulation are available, with different geometries, densities and thermal properties, from different suppliers. The wood pulps used are, according to suppliers, characterized by a high level of cleanliness. Polyaramid fibres, known under the trade name NOMEX, have been presented as a heat-resistant alternative to the cellulose-based materials. However, this synthetic material is, hitherto, far more expensive and it is therefore only used in niche applications or to solve local hot-spot problems in the windings. The

BACKGROUND chemical company Dupont, that developed NOMEX, has also patented an insulating system based on NOMEX and cyanoethylated nanocellulose5. Recent advances in giving wood fibres and materials obtained therefrom novel functionalities may provide an opportunity to obtain new, relatively cheap cellulose-based electrical insulation with tailored properties. Although a modification would certainly increase the cost, the relatively small scale of the production, the high-value application and the opportunity to decrease the total amount of material needed favours the possibility for investments. Improved dielectric performance of the solid insulation could bring several competitive advantages throughout the value-chain, for example by contributing to:  A reduction in the size, mass and hence cost of the transformer,  A reduction in energy losses and increased reliability,  Increased voltage rating. In addition to the rapid development of new materials from cellulose, progress within nanotechnology has also provided new possibilities to tailor the properties of a material from a bottom-up way of design. Components made from renewable resources sometimes have properties that can be drawbacks in applications but one way to overcome these is by learning to mimic and make use of nature’s brilliance in constructing durable and functional materials. An opposition towards new materials is naturally expected and reasonable due to the risks associated, in a conservative business, with the introduction of new and maybe more expensive alternatives. The validation of a new material, which besides dielectric and mechanical testing, includes investigations of e.g. thermal and ageing properties, also requires considerable resources and time. Hence any new material must show substantial improvements in dielectric properties, at a reasonable price, to motivate investments. The cost factor is naturally of less importance in basic research but unreasonably expensive solutions should be avoided. Dry type transformers, with e.g. polymeric composites, are generally used indoors and at lower voltages than oil-filled transformers, but epoxy

BACKGROUND resin, commonly filled with micro-sized silica, has for some applications also substituted oil-paper as insulation in outdoor applications. However, the cellulose-based insulation has advantages due to its low cost and compatibility with oil (which is needed as coolant in high voltage applications). Among the weaknesses of the oil-paper insulation system, in addition to its relatively low flashpoint, is its hygroscopic behaviour, while its resistance to the propagation of discharges and electric stress in the cellulosic bulk is considered to be a major advantage of this system. In dielectrics produced from synthetic polymers, nanoparticle reinforcement has shown many promising effects on the insulating performance6–12. Due to their intrinsic properties (e.g. bandgap or permittivity), or to their interaction with the matrix, nanoparticles can provide e.g. a higher breakdown strength and greater resistance to degradation. The interaction zone, which can affect the polarizability and function as trap sites and scattering obstacles for charge carriers, is dependent on and emphasizes the importance of good nanoparticle dispersion6,13–18. Recent work has also demonstrated similar improvements (such as improved breakdown strength and mechanical strength, decreased permittivity and a more uniform distribution of space charges) with nanoparticles in cellulosic electrical insulation and in transformer oil19–24. A substantial amount of research has been conducted on the performance of alternative or improved transformer oil. Synthetic or natural ester oils as alternatives to mineral oils are already used in some distribution transformers where the dimensions are small. Some esters have a greater fire resistance than mineral oil, and natural esters are also fully or partly biodegradable25. Additives with lower excitation energies than the base liquid have been shown in streamer tests to increase the breakdown voltage in dielectric liquids26. The lower excitation energy of the additives decreases the threshold propagation voltage of the streamer, and this facilitates both propagation and branching. The large number of streamer branches produces an electrostatic shielding of the field in front of the streamer so that when the voltage is increased, new branches are formed keeping the field in front of the streamer nearly unchanged. This

BACKGROUND mechanism is known as ‘self-stabilization’ and is responsible for delaying the appearance of the fast streamers responsible for breakdown27.

2.2 A short introduction to wood fibres Wood fibres consist of three main components; cellulose, hemicellulose and lignin. The material conventionally used in electrical grade paper is made from softwood fibres liberated through the kraft pulping process. In native softwood, the composition is approximately 40–45 % cellulose, 30- 35 % hemicelluloses and 25-35 % lignin28. During kraft pulping hemicellulose and lignin are removed to optimize the mechanical properties of the paper product29. The degree of delignification of the fibres is most frequently quantified by the kappa number of the pulp which is mainly a measure of the amount of residual lignin. Cellulose is a linear polymer that consists of two β-glucose molecules (Figure 2.2) as repeating unit. The degree of polymerization is often very high; in naturally occurring cellulose, the number of repeating units can be over 10 000 whereas, after isolation (depending on the method), it is lowered to an average degree of polymerization in the range of 800 to 300028.

Figure 2.2 Cellobiose, the repeating unit of cellulose.

BACKGROUND

Cellulose in wood fibres is found in both crystalline and less ordered forms. The cellulose chains are organized into square nanofibrils, with a dimension of around 5 nm which are aggregated into fibril aggregates which are approximately 20-30 nm wide30,31. The fibril aggregates are embedded in an amorphous matrix of hemicelluloses and lignin and arranged in concentric lamellae in the fibre wall, as illustrated in Figure 2.329,32. Each layer can be distinguished by the orientation of the fibrils. The three major fibre wall layers are called the primary wall (P), the secondary wall (S) and the middle lamella (M). The secondary wall is divided into three sub-layers; the outer layer (S1), the middle layer (S2) and the inner layer (S3). The middle lamella, which keeps the fibres together in wood, is mainly composed of lignin and pectin. During chemical pulping, the middle lamella is removed and the wood fibres liberated in the delignification process.

~1 nm

~5 nm

S1 P M ~mm ~20 nm ~dm

~30 µm

Figure 2.3 Schematic representation of the hierarchical structure of wood and the layered structure of the fibre wall with the primary wall (P), the secondary wall (S) and the middle lamella (M). The secondary wall is divided into three sub- layers; the outer layer (S1), the middle layer (S2) and the inner layer (S3). (Original image by Mats Rundlöf)

BACKGROUND

Hemicelluloses are combinations of different monosaccharaides. The structure is more branched than cellulose and the chains are normally much shorter. They have a linear backbone composed of linked glucose, mannose, xylose or galactose units, and they often have short side chains that may include xylose, glucose, arabinose or glucuronic acid. For example, galactoglucomannan contains galactose, mannose and glucose while xylan contains xylose units. Galactoglucomannans and arabinoglycuronoxylan are the principal hemicelluloses (ca 10-15 % and 7-10 % respectively) in softwood species such as spruce and pine29. Unlike cellulose and hemicellulose, lignin does not have a well-defined repeating unit. Lignin has a more complex 3D network structure and it contains both aromatic and aliphatic units. Lignin is an amorphous polymer and it has a lower density than cellulose, due to its supramolecular structure32.

2.3 Paper physics and chemistry The mechanical properties of wood-fibre-based products are determined by the strength and length of the fibres and the interaction between them, the density (the amount of fibres interacting in a given volume) and the formation during production32. The removal of hemicellulose and lignin during the kraft pulping process results in water-rich fibres that are more flexible and which can produce a denser network during sheet preparation with a higher number of fibre-fibre contacts and a better joined area in the fibre-fibre contact zones which in turn increases the mechanical strength of the paper. Chemical pulping is a compromise between obtaining a high degree of delignification and a high cellulose yield and low cellulose degradation. Drying of a wood pulp leads to hornification33,34, which means that the original structure of the fibre is not completely regained after re-wetting. Hence, the mechanical properties of a dried pulp are poorer than those of a similar never-dried pulp. Commercial pulps are commonly mechanically treated (beaten) to improve the conformability and thereby obtain strong, high-density products. The beating process increases the swelling of the

BACKGROUND fibres, i.e. the amount of water that can be held inside the fibre wall, the amount of fines (small fibre fragments) and fibre fibrillation which increases the strength of the final paper but has the drawback that it impedes the dewatering process in the paper machine. In addition, beating is energy consuming and, in the case of free drying, results in a higher degree of shrinkage of the paper web during drying34–37. Paper chemicals are often added in the paper-making process to e.g. aid the retention or obtain desired paper properties such as strength or printability. Fibres are negatively charged, the charge originating both from the hemicelluloses, with a major contribution from the carboxylic acid in xylan, and from oxidised lignin structures29. These charges are frequently taken advantage of when modifying fibres by the physical adsorption of charged species. Modification of the fibre wall is possible provided that the modification chemicals can penetrate into its open porous structure. Dry-strength additives may increase the paper strength by:  Increasing the joint strength,  Increasing the number of active joints in the network,  Consolidating the sheet by stabilizing the meniscus during drying,  Decreasing the build-up stress concentration during drying.

More than one of these may be working at the same time and, since many additives are cationic, their effect on retention and formation should also be considered. The fibre-fibre joint is the most interesting to address with a dry strength additive since it is most often the weak link in the dry paper. The nature of the contact zone affects the joint strength but ultimately it is the intermolecular forces and interactions that act to create the strength of the inter-fibre joints38. The relative importance of these interactions (hydrogen bonds, ionic bonds, polar interactions, van der Waals forces and possibly covalent bonds and molecular entanglements) on the paper strength is not fully understood and depend on chemical additives, the fibre furnish composition, fibre treatment and the surface tension of the water, to mention the most important factors. Latex, gums, chitosan, hemicelluloses and acrylamide-based polymers are examples of possible dry-strength additives, but cationic starches are

BACKGROUND the commercially most commonly used dry-strength additives. More recently, the adsorption of polyelectrolyte multilayers has been proposed as an alternative way of improving the mechanical properties of the paper produced from treated fibres38. Chemically modified cellulose, such as cellulose acetate, nitrocellulose and carboxymethylated cellulose is widely used in the plastics, food, textile, building and pharmaceutical industries but in these applications the cellulose has been modified to make it molecularly soluble39. As the traditional boreal forest industry is suffering increasingly due to competition from fast growing trees and a decrease in newsprint and graphic papers in general, the quest to obtain more refined and differentiated products with new functionalities from fibres has intensified the development of novel fibre modifications. Examples of recent activities that have received attention are:  The use of microfibrillated cellulose (MFC) where the fibres partially or fully have been liberated to cellulose nanofibrils (CNFs)- to obtain transparent strong lightweight structures.  Layer-by-Layer (LbL) adsorption, of e.g. electrically conducting and insulating consecutive layers for energy-storage applications.  The grafting of functionalities which e.g. make wood fibres compatible with water-insoluble polymers40. The concepts of MFC, CNFs and LbL adsorption are introduced below.

2.3.1 Cellulose micro- and nanofibrils Turbak et al. showed in the 1980s that it was possible to liberate microfibrils from wood fibres via high-pressure homogenization and to obtain fibrils with a high aspect ratio and high mechanical strength41. With pre-treatments such as TEMPO-oxidation and carboxymethylation to introduce charges, less energy is needed to liberate the fibrils and to achieve higher degree of liberation and a much improved dispersion stability42. Due to its intriguing properties, this material has received a dramatically increased interest during recent years, as a competitive alternative to synthetic polymers, from an abundant and more sustainable

BACKGROUND resource; a possible future for the forest industry as well as a less polluted planet. CNFs have been successfully used, so far mostly in laboratories, to produce strong, lightweight films, foams and aerogels, modified to obtain different functionalities43–47. The large accessible surface area is advantageous for surface modifications. In the production of paper-based products, the use of CNFs as an additive is attractive, as it can be introduced to give an improved or new functionality, as an alternative to existing paper chemicals, without modifying the entire rawmaterial. In previous work, the effect of humidity on the dielectric losses has been compared for nanocellulose from wood fibres and algea and it was concluded that their crystallinity and moisture sorption has an impact on the dielectric properties48. For example, it has been suggested that the ion mobility is higher in nanocellulose from algea (Cladophora cellulose) due to the state of the water which is less tightly associated than in CNFs from wood fibres. Both CNFs and MFC contains less ordered and crystalline cellulose parts. CNFs have widths between 5-30 nm while MFC have widths between 10-100 nm (cellulose nanocrystals on the other hand have a higher concentration of crystalline cellulose and have a width of 3-10 nm), as defined by the TAPPI WI 3021 standard.

2.3.2 Layer-by-Layer (LbL) adsorption LbL adsorption is a physical process, in contrast to chemical grafting, where charged species (e.g. polyelectrolytes) are attracted and adsorbed in multiple layers onto an oppositely charged substrate. From a thermodynamic point of view, the main driving force is the entropic gain upon the release of counter-ions associated with the adsorption49. The build-up of single- and multilayers can be tailored by an appropriate choice of adsorption conditions, such as pH and salt concentration. The control of pH and ionic strength is especially important for weak polyelectrolytes, where the charge density of the polyelectrolyte is strongly dependent on the pH. The polyelectrolyte multilayer technique was introduced by Decher in 199750 and has since achieved great interest for the surface modification of solid materials. An advantage compared to many chemical

BACKGROUND modifications is that it can be performed in water and hence is easily incorporated into the already water-based, continuous, papermaking process.

2.4 A short introduction to dielectric properties Permittivity and dielectric loss are important dielectric properties which affect the loss of energy and the build-up of electric fields in an insulator and at interfaces under normal AC operating conditions. These properties are normally measured at lower voltages (typically below line voltage), with an altering current over a chosen frequency range. While permittivity is an important parameter for the performance under AC, the electric conductivity and charge accumulation are more important for under DC conditions. The dielectric strength, used to describe the maximum electric field a material can withstand before breaking or short-circuiting, is measured at higher voltages, usually applied in steps or impulses. For insulating materials used in high voltage applications the breakdown strength is in the MV/m range. Before experiencing breakdown, the insulating material is often subjected to deteriorating discharges and streamers which can also be triggered and analysed in a laboratory. Partial discharges (PD) in dielectric materials or on the surface of a dielectric are caused by locally concentrated electrical stress and appear as pulses with a duration generally much less than 1 µs51.

2.4.1 Dielectric response A dielectric material, such as cellulose, is an electrical insulator. An ideal dielectric material has no free charges; only bound charges that are polarized by an applied electric field. Under the influence of an alternating electric field, the polarization processes and charge transportation that take place determine the dielectric response. The movement due to polarization results in a dissipation of energy to the material as heat. The dielectric constant or permittivity, ε, can be resolved into a real part, the storage part,

BACKGROUND

ε´, and an imaginary component, the loss part, ε´´. The dielectric loss tangent, tan δ = ε´´/ ε´, is a measure of the ratio of the imaginary to the real part of the dielectric function at a given frequency52. The frequency dependence of the dielectric response is known as the dielectric dispersion, shown in Figure 2.4. The origin and nature of the polarizability depend on the frequency; the higher the frequency, the shorter is the length scale of polarization. Maxima called “loss peaks” occur, at characteristic frequencies, in the imaginary part (ε´´) of the dielectric dispersion, representing different polarization mechanisms. At high frequencies (in the UV-visible region) the only polarization that is sufficiently fast to move with the electric field is the electronic polarization, while at lower frequencies, larger moieties have time to orient with the electric field. In cellulose, these movements have been shown to 53 be attributed to the displacement of the (-OH) and (-CH2OH) groups .

Figure 2.4 The dielectric dispersion; the frequency dependence of ε’ and ε’’. The origin of the polarization is written above the curves at their characteristic frequencies52 .

BACKGROUND

Even though there are no free charge that can provide conduction in insulating materials, charge transport can take place and can be studied in the low frequency part of the dielectric response. For a low frequency dispersion, which is characteristic of a dielectric such as kraft paper, no loss peak is observed but both the imaginary and real parts rise steadily towards lower frequencies54. This is a consequence of hopping charges and diffusive charge transport. Hopping charges spend most of their time in localized states but if the localized states form a connected network and the charges gain energy this can give rise to DC conduction. DC conductivity is indicated by a steep rise with decreasing frequency in the imaginary part of the dielectric response (a slope of -1 can be seen in the ε’’ curve)54. To estimate the permittivity of the oil-impregnated paper, it can be handled like a composite and the combined dielectric response can be modelled55. The electric field distribution at a liquid-solid interface will be affected by the difference in permittivity of the materials on each side52. Figure 2.5 illustrates the concentration and direction of the field lines at an interface when the material 2, in this case it could be a cellulosic pressboard, is either the high or the low permittivity material.

a) b)

Figure 2.5 Diagram of the concentration and direction of the electric field lines at an interface when material 2 is a) the high and b) the low permittivity material

BACKGROUND

The concentration of the electric field can influence the charge accumulation at the interface. In the case of paper-transformer oil, where the charges are accumulated in the oil, the mismatch in permittivity results in a concentration of the electric field at the interface, and the intensity of the field increases with increasing permittivity of the solid.

2.4.2 Streamers in oil and at the oil-solid interface Streamers, conducting gaseous channels which can travel at high speed, at oil-pressboard interfaces have been identified as a significant cause of transformer failure and as the failure mode with the most dangerous consequences31. The streamers responsible for breakdown are those with a filamentary appearance and supersonic velocity. The streamer filament head, which has a tip head radius of 2-3 µm, can produce a field of about 15-20 MV/cm57. With such a field, field-assisted tunnelling and the generation of charge carriers by impact ionization can occur. Direct ionization of mineral oil contributes to the formation of the gaseous phase of the streamer and the generation of hydrocarbon gases. The interested reader is recommended to turn to the appended Papers III-V and the comprehensive and useful papers by Lesaint58 and Denant59,60 for more information regarding the mechanism responsible for streamer inception and propagation. There are positive and negative streamers. Negative streamers starts at the cathode and positive streamers at the anode. The initiation voltage for a slow negative streamer is lower than that for a positive streamer61. However, since positive streamers propagate faster they are considered more harmful62. Branching, the formation of a tree-like structure distributes the electric field and lowers the potential at the tip of the propagating streamer. At an interface between gas and solid, gas and liquid or solid and liquid the streamer may be referred to as electrical creep. The interface dielectric strength, “creep strength” is affected by the properties of both media. The mechanism involved in streamer inception and propagation along a solid- liquid interface is unfortunately not fully understood. As a consequence, it is not clear which properties of the solid insulation influence the streamer.

BACKGROUND

For pressboard it has been shown that the streamer during the fast mode (>1.4 km/s) moves in close contact with the surface63, as shown in Figure 2.6.

Figure 2.6 A positive streamer propagating at an oil/paper interface (photograph kindly provided by David Ariza).

EXPERIMENTAL

3. EXPERIMENTAL

More information concerning materials and methods can be found in the respective appended papers.

3.1 Materials

3.1.1 Wood pulps The pulp used as reference throughout the project was an unbleached dried electrical grade kraft pulp from Munksjö Aspa Bruk AB, Askersund, Sweden, with a kappa number of ca 25, corresponding to a lignin content of ca 3 %. This pulp is used to make paper products used as electrical insulation and hence needs to meet high customer requirements concerning cleanliness (evaluated e.g. by conductivity and pH of aqueous extract and dissipation factor) and mechanical properties. Laboratory-prepared never- dried unbleached softwood kraft pulps were supplied by SCA Research AB in Sundsvall, Sweden, and a sulphite softwood dissolving pulp by Aditya Birla Domsjö Fabriker AB, Domsjö, Sweden. The pulps were washed and transferred to their sodium form as described in Paper II prior to further use.

3.1.2 Transformer oil The mineral oil used in this project was Nytro 10X and 10XN, from Nynäs AB, Nynäshamn, Sweden. The oil was degassed and heated to 60 ˚C prior to use for impregnation and measurements.

3.1.3 Polymers Several different polymer films were investigated in streamer tests (Papers III-V) to evaluate the influence of permittivity, porosity and surface roughness on streamer inception and propagation. The polymer films studied were low density polyethylene (LDPE), polyethylene terephthalate

EXPERIMENTAL

(PET), polytetrafluoroethylene (PTFE) and polyvinylidene fluoride (PVDF), all purchased from Good Fellow Cambridge Ltd, England. More information can be found in Table A1 in Appendix. For the adsorption of nanoparticles (Papers VII and VIII) and cellulose micro- and nanofibrils (Paper VI), several different polyelectrolytes were used. A negatively charged, polyacrylic acid (PAA) was used to build LbLs with cationic silica nanoparticles. To construct LbLs of negatively charged particles, a positively charged polyethyleneimine (PEI) was used. For paper sheets with cellulose micro- and nanofibrils, polyvinyl amine, PVAm, and poly(diallyldimethyl-ammonium chloride), pDADMAC, were used as retention aids. PAA, PEI and PVAm are weak polyelectrolytes while pDADMAC is a strong polyelectrolyte. More details about the different polyelectrolytes can be found in the respective papers.

3.1.4 SiO2 and ZnO nanoparticles

Two types of colloidal silica (SiO2), supplied by Eka Chemicals, Sweden, were used:  Bindzil CAT220, is positively charged, with a primary particle size of approximately 12 nm. The cationic charge originates from substituted aluminium ions.  Bindzil 15/500, has a negative charge and a primary particle size of approximately 5 nm. Zink oxide (ZnO) nanoparticles (NanoGuard) with a particle size of 67 nm were purchased from Alfa Aesar.

3.2 Methods - sample preparation

3.2.1 Lignin and glucomannan model materials (Paper I) Purified lignin was prepared from Indulin AT (MeadWestvaco, Richmond, VA, USA), originating from cooking liquor, according to Norgren et. al64, while glucomannan was purified from wood chips according to Zhang et 65 al . The pure glucomannan and lignin samples, in NH4 solutions, were

EXPERIMENTAL spin-coated onto silicon substrates according to Gustafsson et. al66 to obtain surfaces for characterization by spectroscopic ellipsometry.

3.2.2 Paper sheets Paper sheets were made from pulp suspensions with a Rapid Köthen sheet former (Paper Testing Instruments, Austria). For the sheets produced for dielectric characterization, deionized water was used throughout the process. The pulp used in reference papers was beaten 4000 revolutions in PFI-mill, to achieve properties similar to those of the pulp used in pressboard. The papers were dried under a negative pressure of 95 kPa in the dryers of the sheet former at 93 ˚C. To study the effect of beating and drying pressure on paper properties, the pulp was beaten for different numbers of revolutions in the PFI-mill and the papers were dried under different pressures, giving papers with different densities. Figure 3.1 shows photographs of the PFI-mill and the Rapid Köthen sheet former used and paper sheets made from unbleached kraft pulp.

a) b) c)

Figure 3.1 Photographs of a) PFI-mill, b) Rapid Köthen sheet former and c) paper sheets made from unbleached kraft pulp.

EXPERIMENTAL

3.2.3 Cellulose micro- and nanofibrils Two types of cellulose fibril (shown in Figure 3.2) were used; one type of nanofibril produced from sulphite dissolving pulp from Domsjö, after carboxymethylation, were provided by Innventia AB (now RISE Bioeconomy). The other type, denoted kraft microfibrils or kraft MFC, was prepared from the electrical grade unbleached kraft pulp from Munksjö, without chemical pre-treatment. The kraft MFC was first prepared by beating the pulp for a total of 6000 revolutions in a PFI- mill. Both the kraft pulp and the carboxymethylated sulphite pulp were homogenized in a Microfluidizer M-110eh (Microfluidics Inc., USA) at a pressure of 1600 bar by three passes through a 400 μm and a 200 μm chamber connected in series followed by three passes through a 200 μm and a 100 μm chamber connected in series. In Paper VI, the carboxymethylated CNFs was also used with two types of chemical modification: With an aldehyde functionality, obtained through periodate oxidation29,67 and with a dopamine functionality, attached by a coupling agent (1-Ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC)) 68,69. The functionalities as well as possible further crosslinking can be found in the Appendix, Figure A1-A4. The kraft MFC in this project was produced without chemical pre- treatment: To rule out the effects of changes in chemistry on the investigated dielectric properties and thereby enable investigation of the effect of the different morphologies provided by the fibrils. From a product development point of view, the MFC produced from the well-characterized electrical grade pulp could be an attractive first approach, for the present application. Changes in chemistry due to pre-treatments could, on the other hand, influence important properties, such as polarizability and ageing, when used in an insulating context. The fibres and fibrils from the electrical grade pulp have a low content of carboxyl groups available for modifications, so the carboxymethylated CNFs were chosen for grafting with dopamine. With another grafting method (not EDC coupling) and for periodate-oxidation, untreated fibrils could be considered.

EXPERIMENTAL

a) b)

Figure. 3.2 SEM-images of a) carboxymethylated CNFs and b) kraft MFC. These samples were prepared by evaporating the water from dilute dispersions, at room temperature from materials deposited on silicon surfaces.

3.2.4 Papers with cellulose micro- and nanofibrils as strength additives (Paper VI) The fibrils were first diluted to 1g/L and dispersed by mixing in an Ultra Turrax mixer (IKA, T25 Basic, Staufen, Germany) for 15 minutes at 10 000 rpm. Papers with cellulose fibrils were produced with and without retention aid, as shown in Figure 3.3. For papers with retention aid, the polyelectrolyte (PVAm or pDADMAC) was added to a fibre suspension at pH 8 and stirred for 20 min after which excess polymer was rinsed off through filtration. The cellulose fibrils were then added to the fibre suspension and stirred for 30 min at pH 8, and paper sheet were made according to above-mentioned procedure (section 3.2.2).

3.2.5 Films from cellulose micro- and nanofibrils The CNFs and MFC suspensions were poured into the cylindrical container of the Rapid Kö then sheet former (Figure 3.1). Water was removed through vacuum filtration using a 0.45 μm Durapore membrane filter (Merck Millipore Ltd.). After the filtration, another membrane was placed on top of the wet film and the material was dried in the driers of the sheet former at 93 °C under reduced pressure for 15 minutes.

EXPERIMENTAL

Figure 3.3 a) Schematic which shows the preparation procedure of sheets with fibrils as additives and b) the additions in wt% of fibrils to the paper sheets.

3.2.6 Three-layer sandwich structure with microfibrillated cellulose The middle layer consisting of kraft MFC film was produced in the sheet former according to the method described above (section 3.2.5). After dewatering for approximately 1h, a gel-like film was left which could be peeled off from the membrane. Two paper sheets made from the reference pulp were also produced in the sheet former and the fibril film was inserted between the two sheets, all three layers being still wet. The layers were couched together and dried for 30 minutes at a temperature of 93 °C and a reduced pressure of 95 kPa in the driers of the sheet former.

3.2.7 Papers from fibres modified with SiO2 and ZnO nanoparticles (Papers VII-VIII)

Modified paper sheets with SiO2 nanoparticles were produced, after consecutive treatments of the electrical grade fibre with nanoparticles and

EXPERIMENTAL polyelectrolytes. The amounts of polyelectrolyte and nanoparticle adsorbed were regulated by changing the charge density of the adsorbed species by adjusting the pH with hydrochloric acid and sodium hydroxide. Papers with ZnO nanoparticles were produced by adsorption of the nanoparticles at pH 6.5. Prior to adsorption onto the wood fibres, the ZnO nanoparticles had been surface modified, by silanization with (3- aminopropyl) triethoxy silane (APTES, 98 %) according to a previously described procedure70. The final nanoparticle concentration was determined by thermogravimetric analysis (TGA).

3.3 Methods – sample characterization This section presents the methods used for the determination of the charge density and the mechanical and dielectric properties of the paper samples. Methods used for characterization of morphology, yield, etc. can be found in the respective papers.

3.3.1 Charge density measurements The principle of polyelectrolyte titration (PET) used for surface charge measurements on wood fibres has been described by Wågberg et al71. In brief, a strong cationic polyelectrolyte with a known charge density and a high molecular weight (to prevent penetration into the fibre-wall and access to interior charges) was adsorbed onto the fibres. The excess of cationic polymer was filtered off and the amount adsorbed was thereafter determined by titration of the filtrate with an anionic polymer in the presence of an indicator, such as Orthotoluidine Blue (OTB) which turns from blue to pink with a charge shift from positive to negative. The equilibrium concentration and the adsorption characteristics of the polyelectrolytes onto the wood fibres were studied by adsorption isotherms, obtained in similar manner71. Due to the application, where free charge carriers are undesired, the influence of ionic strength was never studied.

EXPERIMENTAL

The total charge density of fibres was determined by conductometric titration according to SCAN-CM 65:02 using a Metrohm 702SM Titrino titrator. The surface charge density and stability of the micro- and nanofibril and nanoparticle dispersions were investigated with PET by streaming potential measurements using a Stabino particle charge mapping equipment (ParticleMetrix, Germany).

3.3.2 Thickness measurements The thickness used to calculate the density of paper sheets was determined in two different ways. For the samples used in all the studies which included dielectric measurements, the thickness of a stack of 3 test pieces was measured with a disc micrometer (Absolute Digimatic Quick 227, Mitutoyo). This procedure (derived from ISO 534:2005) was used due to the configuration of the dielectric response cell, where three samples were stacked in a plate-plate configuration and the distance between the plates is important for consecutive calculations of the real and imaginary parts of the permittivity. For the study with fibrils as additives (Paper VI), the thickness of the samples was determined according to SCAN-P 88:01, which is more commonly used for paper products.

3.3.3 Mechanical testing Tensile testing of paper sheets and CNF films was carried out at 23 ˚C and 50 % RH using an Instron 5944 equipped with a 500 N load cell. For paper sheets, strips with a width of 15 mm and a free span of 100 mm between the clamps were strained at a constant rate of 100 mm/min. For wet tensile strength, the sheets were soaked in water for 1 h prior to tensile testing according to SCAN-P 20:95. The excess water was removed before the mechanical testing.

3.3.4 Spectroscopic ellipsometry (Paper I) Measurements in the ultraviolet-visible-near IR (UV-VIS-NIR) region 245–1690 nm were made, at 23 °C and 48 % RH, using a variable-angle

EXPERIMENTAL ellipsometer with dual rotating compensators (RC2, J. A. Woollam Co. Inc.). From the measurements, the wavelength-dependence of the refractive index of films of lignin and glucomannan deposited on silicon substrates was estimated. The complex refractive index Ñ is a combination of the refractive index, n, and the extinction coefficient k as given by: 푁̃(휔) = 푛(휔) ± i푘(휔) (Equation 1) The extinction coefficient k is a measure of the absorption which has a maximum at the absorption wavelength derived from electronic or inter- band transitions in the material corresponding to an excitation of an electron from an occupied orbital to an unoccupied or partially unoccupied orbital. When k = 0 the value of the real part of the permittivity ε’ can be calculated according to: 휀′(휔) = 푛2(휔) (Equation 2)

3.3.5 Dielectric spectroscopy (Papers II, VII and VIII) Dielectric spectroscopy measurements were performed on paper sheets in vacuum and in transformer oil to obtain the complex permittivity. The diameters of the ground electrodes were 50 mm and 90 mm for the cells used in vacuum and in oil respectively. Diagrams of the test cells are shown in Figure 3.4. To reduce edge effects, a guard ring separated by a 1mm gap from the ground electrode was included in the test cells. Three paper discs with thicknesses of 100 to 150 µm each were placed between the two electrodes. The cell was then inserted into a vacuum oven for drying for 24h at 110 ˚C, and consecutively (before any measurement in oil) impregnated with oil at 60 ˚C followed by a measurement at 70 ˚C. The frequency sweep was performed from 1 mHz to 1 kHz, at 200V in vacuum and 3V in oil, and the capacitance was measured. Impregnation with oil was done in a vacuum oven without releasing the vacuum after the paper samples had been dried to ensure continued dryness. The dielectric losses and permittivity were monitored during drying and impregnation to ensure drying and stable measurement conditions. Prior to the first

EXPERIMENTAL measurements in vacuum the frequency sweep was performed at lower voltages (30 V and 50 V), without any sign of non-linearity.

Sample

Figure 3.4 Diagrams of the test cells for dielectric spectroscopy measurements in a) vacuum and b) oil.

The permittivity was calculated from the capacitance, C, using:

푑 휀 = 퐶휀 (Equation 3) 0 퐴 where ԑ0 is the vacuum permittivity, d is the distance between the electrodes and A is their area.

3.3.6 DC breakdown test The DC breakdown test was performed on a reference paper and on the sandwich structure with microfibrillated cellulose. The samples were vacuum impregnated with transformer oil for one week prior to breakdown testing. Nine specimens from each material were tested, all with a thickness of ca 150 µm. Tests were performed in oil using a ramp rate of 0.5 kV/s. The electrode set-up is shown in Figure 3.5.

EXPERIMENTAL

Figure 3.5 Sample mounted between electrodes for the DC breakdown tests.

3.3.7 Streamer measurements (Papers III-V) The solid samples, papers and films, (8 mm x 100 mm) were preheated at 105°C (LDPE at 70°) for 24 h, after which they were impregnated with the filtered and degassed transformer oil at 60°C and 5 mbar for 24 h. Prior to testing, the oil was degassed and filtered in the test chamber by pumping it through a filter with a pore size of 2 µm. After the filtration and degassing process, the chamber was cooled to room temperature and filled with dry air. All tests were performed at atmospheric pressure. The experimental set-up for studying the streamer inception and propagation in a needle-plane configuration is shown in Figure 3.6. The test chamber was half-filled with transformer oil and the solid sample was inserted at an angle of 30° with respect to the plane. A series of square high voltage pulses with a rise time of 35 ns and a width of 50 µs was applied to the plane electrode with voltage levels ranging between 11.5 and 22 kV and steps of about 0.6 kV. A single frame shadowgraph of each streamer was obtained.

3.3.8 Partial discharge (PD) measurement (Paper VIII) PD measurements were performed on three layers of oil-impregnated papers. A diagram of the set-up with a high potential needle electrode and a grounded bar on the surface of the insulation used is shown in Figure 3.7. To investigate the effect of impulses on partial discharges a 50 Hz AC source and a standard lightning impulse generator system were combined to provide an impulse voltage superimposed on the AC voltage.

EXPERIMENTAL

(1) Test chamber (11) Pump (2) Point and probe electrode (12) Particle filter (3) Plane electrode (13) Valve (4) Charge measuring system (14) Heating rod (5) High-speed camera (15) High voltage pulse source (6) Xenon lamp (16) Oscilloscope (7) Xenon lamp power source (17) Signal generator (8) Optical fibre (18) Data acquisition device (9) Photomultiplier (19) Computer (10) Valve (20) Vacuum pump

Figure 3.6 Detailed diagram of the experimental set-up used for the study of streamer propagation in oil.

Figure 3.7 Diagram of the set-up used for partial discharge measurements.

RESULTS AND DISCUSSION

4. RESULTS AND DISCUSSION

When wood fibres and papers are modified it is almost impossible to influence one property without affecting others simultaneously. An example is the beating of wood pulp, which has been referred to earlier, which is used to improve mechanical properties of paper products. Beating leads to increased flexibility of the fibres, fibrillation of the fibre surfaces, the creation of fines and to an increase in the ability the of fibre wall to swell in water. In paper, this results in a greater contact area between fibres partly due to densification. Hence, if the dielectric properties of a paper are changed when the pulp in beaten, the question that arises is whether this is an effect solely of densification or fibrillation, or both. The same question is valid for other fibre treatments. To be able to distinguish the effects of different additives or chemical or physical treatments of the fibres, and the mechanisms involved, it is hence desirable to separately investigate the effect of e.g. densification, beating and pulp composition on permittivity and other dielectric phenomena. This separation was achieved by investigating model materials, different paper grades and films of synthetic polymers and microfibrillated cellulose. The first results are presented and discussed in this context. In the second part, the possibility of using novel modification routes for wood fibres to improve the dielectric and mechanical properties of cellulose-based electrical insulation is demonstrated and discussed.

RESULTS AND DISCUSSION

4.1 The influence of paper density and morphology on the dielectric response (Paper II) Dielectric spectroscopy in vacuum was carried out on papers made from pulps beaten to different degrees or pressed at different pressures to evaluate the influence of density and mechanical refinement on the dielectric response. This means that the changes in permittivity and dielectric losses with changes in the paper/oil volume fraction of the impregnated paper could be studied. The permittivity can be estimated (as shown in Figure 4.1 for 50 Hz, which is the standard “power frequency” found in European grids) from the Wiener equation (Equation 4) if the volume fraction solid/oil or solid/vacuum and the permittivity of the pure media are known72,73.

푉 푉 1 = 푝푎푝푒푟⁄ + 표푖푙⁄ (Equation 4) ⁄휀표푖푙푖푚푝푟푒푔푛푎푡푒푑 푝푎푝푒푟 휀푝푎푝푒푟 휀표푖푙

Beating the pulp and thereby changing the fibre and paper morphology did not seem to affect the dielectric properties. Although empirical data should be collected to assess the permittivity and dielectric losses of a given paper grade in a given medium, the relations give valuable input regarding the influence of densification on the dielectric properties of paper/oil insulation. This also suggests that is relatively easy to tailor the permittivity of paper and hence the electric field distribution at the oil-solid interface by producing paper parts of different densities. The densities could be altered e.g. by changing the raw material, the degree of delignification, the mechanical refinement or the pressure during the drying step. The implications of such changes on the manufacturing process, on the overall dielectric properties and on the mechanical properties must however be compiled and addressed.

RESULTS AND DISCUSSION

Figure 4.1 The real part of the permittivity, ԑ’, (dashed lines are from calculations using the Wiener equation) and tan delta at 50 Hz: In a) and b) of papers with different densities and in c) and d) of papers with different chemical compositions. The kappa numbers (correlated to the lignin content) are listed for the SCA pulps. SCA 26 has a composition similar to that of the electrical grade pulp while SCA 75 and SCA 107 have higher contents of hemicellulose and lignin.

4.2 The influence of wood fibre composition on the dielectric properties The effect of fibre composition on the dielectric response was addressed by investigating the properties and effects of the most common biopolymers (cellulose, hemicellulose and lignin).

RESULTS AND DISCUSSION

4.2.1 Spectroscopic ellipsometry measurements of wood biopolymers (Paper I) The wavelength dependence of the refractive index (Figure 4.2) of films of lignin and glucomannan deposited on silicon substrates was estimated and compared to previous measurements on cellulose74. The results indicate that lignin has the highest electronic polarizability of the three biopolymers, cellulose, hemicellulose and lignin, considering the relationship, n2 = ԑ, between the refractive index, n, and the permittivity, ԑ, (where k ≈ 0 in the investigated wavelength region). At power frequencies, movements of larger moieties contribute to the dielectric losses, but the electronic polarizability together with the movement of dipoles influences the dielectric losses and the electric field distribution during fast events and transients (in the µs and ns range). The results were expected from the polarizabilities listed for the embodied bonds of the respective polymers75, but this is one of the few studies where all three wood components have been evaluated at the same time. In addition, the measurements of the spectral parameters and the mathematical approximations presented in Paper I, was used for calculations of non-retarded Hamaker constants and also showed that lignin has the lowest energy for the main electronic transitions compared to glucomannan and cellulose. 1.8 0.20 1.8 a 0.20

a Lignin k

n Lignin Glucomannan k n 1.6 Glucomannan 0.15 1.6 0.15

1.4 0.10 1.4 0.10

Refractive index,

Refractive index, 1.2

1.2 0.05 Extinction coefficient,

0.05 Extinction coefficient, 1.0 1.0 200 400 600 800 1000 1200 1400 1600 200 400 600 800 1000 1200 1400 1600 Wavelength (nm) Wavelength (nm) Figure 4.2 The refractive index, n, and the extinction coefficient, k, of model films of glucomannan and lignin calculated from spectroscopic ellipsometry measurements.

RESULTS AND DISCUSSION

4.2.2 Dielectric response (Papers II and VIII) Compared to a paper, made from the electrical grade pulp, with a lignin content of approximately 3 % papers made from pulps with a higher lignin content (10-15 %) have a similar permittivity but their dielectric loss at 50 Hz is slightly higher both in vacuum and in oil (Figure 4.1), and the permittivity of the paper made from pulps with a higher lignin content increases more rapidly at low frequencies, indicating a greater charge transport in these materials. In oil, where the overall charge transport is greater than that in vacuum, these trends are even clearer. The papers differ in composition (relative contents of cellulose, lignin and hemicellulose) but, as a consequence of the pulping process, they also differ in e.g. density, morphology, charge density and interaction with the transformer oil. The results presented in section 4.1 and Paper II indicate, however, that the paper morphology is insignificant and that the density differences, do not contribute to any differences in the magnitude of the charge transport. The influence of the total charge density (or the charge of the associated counter-ions) and the wetting of the papers by transformer oil are parameters that, besides the polarizability and the interaction of the wood fibre components with mobile charge carriers, could be of importance and interesting to investigate further.

4.2.3 Surface partial discharge (PD) resistance (Paper VIII) Surface PD measurements on a paper with a high lignin content (14 wt%) resulted in a high PD inception voltage (28kV compared to 23kV for the reference) and the PD pattern was not affected by impulses. Possible explanations for this could be a low density and hence low permittivity or a lower excitation energy due to the conjugated bonds in lignin76. Additives with low excitation energy have previously been shown to inhibit discharges in oil26.

RESULTS AND DISCUSSION

4.3 The influence of permittivity, density and morphology of the solid on streamer inception and propagation (Papers III-VI) Previous sections have summarized the results concerning the effect of density and fibre composition on the dielectric response of kraft paper. In order to separate the effects on streamer inception and propagation of density and morphology of the solid from that of permittivity (which is also dependent on density), polymeric films with different permittivities and a kraft MFC film were investigated with regard to streamer measurements and the results were compared to those of the kraft paper. The results obtained with streamers in negative polarity on the kraft paper show that high surface roughness and porosity are associated with a low velocity and short propagation of the streamer. The porous paper morphology provides an oil-rich composite and could possibly promote branching into the paper surface. The similarity in the behaviour to that of the oil is supported by a well-defined charge step in the early propagation of the streamer, similar to a case without solid. A low velocity may imply that the streamer during e.g. an impulse would be less likely to cause short- circuiting and failure. From shadowgraphs obtained in positive polarity, it is possible to conclude that the paper in this case also enhances branching into the oil while the propagation is more concentrated to the surfaces of the smoother polymeric films and of the film of kraft MFC (Figure 4.3). Even though the effective voltage gradient is high, the branching of a streamer may lead to a lower electrical stress in the tip of each branch, and this could limit the deterioration and evaporation of the insulation. The probability of inception is plotted against applied voltage for the investigated materials in negative polarity in Figure 4.4. The inception voltage for mineral oil without an interface is about 13 kV and the inception voltage for the kraft paper, kraft MFC film, PET and PVDF interfaces is almost the same, despite differences in their relative permittivities. On the other hand, a significantly larger inception voltage was found in the case of PTFE and LDPE (ca 15.5 kV and ca 14.7 kV

RESULTS AND DISCUSSION

respectively). The results reported for PTFE and LDPE also show that these solids, with permittivities similar to that of the mineral oil, can lower the stopping voltage condition in negative polarity. From charge and light intensity recordings, it was found that the oil- solid interfaces with PVDF, PET and kraft fibril paper, with high permittivity and low surface roughness, induce a quasi-continuous injection of charge during the early stage of the streamer propagation. This quasi-continuous injection of charge contributed to an increase in average length and velocity of the streamer on these materials. Furthermore, it was shown that the average streamer charge increases with increasing permittivity of the interface. A possible explanation for the higher inception and stopping voltages with PTFE and LDPE could thus be that solids with a lower permittivity than e.g. impregnated kraft paper may reduce the electric field strength at the interface, decreasing the electron emission and accumulation of mobile charges on the solid surface. The similarity between the inception voltage conditions of the PVDF and PET films, the kraft fibril and kraft papers and the case without a solid interface indicates that the conditions in these cases depend mainly on processes taking place in the liquid phase, including ionization of the molecules in oil. A hypothesis put forward before the measurements were made, was that the high electron affinity of fluorine (present in PTFE and PVDF) could be beneficial for limiting the propagation of streamers, but although the inception voltage was higher for PTFE, the inception voltage for PVDF and the propagation on both PTFE and PVDF were similar to that of the other polymers in negative polarity. A most surprising observation in our measurements in positive polarity (unpublished results) was, however, that the streamer was quenched only in the presence of PVDF. This could not be explained by any of the mechanisms discussed but we expect that more observations and data collection will bring clarity.

RESULTS AND DISCUSSION

Figure 4.3 Images of the propagating streamer on surfaces of a) an oil- impregnated kraft paper where the streamer branches into the oil, and b) a paper made of kraft fibrils where it is more concentrated to the surface.

Figure 4.4 The probability of streamer inception plotted against the applied voltage measured in negative polarity in oil.

RESULTS AND DISCUSSION

4.4 Papers with fibres and fibrils combined The significance of paper density for the permittivity and streamer properties as well as the potential to achieve a significant improvement in mechanical strength at a lower density directed our efforts towards producing papers containing a combination of pulp fibres and cellulosic micro- and nanofibrils.

4.4.1 DC breakdown strength of a three-layered sandwich structure with microfibrillated cellulose The properties of a sandwich structure and of the reference paper are given in Table 1. The sandwich structure is a way to tailor the paper to obtain a porous relatively low-density surface with low permittivity and a high- density interior with higher permittivity. For the complete three-layered structure, the density increased by approximately 20 % while the fibril mass was 33 % of the total mass. If it was produced as a free-standing film, the dry kraft MFC film would have had a thickness of approximately 30 µm (ca 20 % of the total thickness). The permittivity of an oil-impregnated kraft MFC film was approximately 4.577. A laminate with an internal kraft MFC film had DC breakdown strength 47 % higher than that of a reference of similar thickness produced only from pulp fibres. An increasing breakdown strength with increasing density was expected, but not to the extent found. In addition, the scatter in the DC breakdown data (Figure 4.5) was significantly smaller for the sandwich structure, as shown by the higher shape factor, 20 versus 9 in the reference paper. The microfibrils as well as the outer layers of the sandwich were produced from the fibres of the electrical grade pulp. Hence, although the total density increased by 20 %, the chemical composition of the fibrous material was basically the same.

RESULTS AND DISCUSSION

Table 1. Data and properties and of the “sandwich” and a reference sample (kraft paper).

Sample Reference Sandwich

fibres fibres Diagram fibrils fibres Thickness (µm) 148 ± 8 153 ± 4

Density (g/cm3) 0.82 0.97

Breakdown strength (kV/mm) 120 ± 15 176 ± 11

• Reference Shape-factor

9.278

▪ Sandwich

Shape-factor 20.36

Figure 4.5 Weibull plot of the DC breakdown strength.

RESULTS AND DISCUSSION

4.4.2 Micro- and nanofibrils as paper strength additives (Paper VI) The aim of the work described in Paper VI was firstly to improve the mechanical properties of papers made from the electrical grade pulp. The mechanical properties are important for the transformer application and if the tensile index were significantly increased it could be possible to the use a product with a lower density or to reduce the size and thus the cost of the component. In addition, the above-mentioned DC breakdown measurements promisingly showed improvements when a layer of microfibrillated cellulose was introduced and a recent publication reported an increase in AC and DC breakdown strength by up to ca 20 %, with micro- and nanofibrillated cellulose used as an additive (with 10 wt% addition)14. Previously reported data have shown that the addition of micro- and nanofibrils led to in an increase in density and that the improvements in the mechanical properties of the paper sheets corresponded to that obtained by beating the pulp to a similar density78. Our results have shown that it is possible with chemically modified CNFs as strength additives to produce stronger sheets without increasing the density. The tensile strength index of papers containing fibrils is shown in Figure 4.6. The addition of carboxymethylated CNFs resulted in a low retention and a small effect on the tensile properties, dewatering rate and morphology. The strengthening effect achieved with the kraft MFC appeared to be solely due to the densification of the paper, in contrast to the effect achieved with the periodate-oxidised and dopamine-grafted CNFs, where the papers showed most significant improvements. The greatest strength-enhancing effect was obtained with the periodate- oxidised CNFs with retention aid. The aldehyde groups on the periodate- oxidised CNFs are believed to crosslink (see Appendix) with the hydroxyl groups on the fibres, thereby strengthening the fibre-fibre joints79. The highest dry strength, 125 kNm/kg, was achieved with pDADMAC and ca 14 wt% of periodate-oxidised CNFs in the paper sheet, an increase in strength of 89 % compared to the reference containing only pDADMAC. With PVAm, the tensile strength index increased by 56 % with less than 1

RESULTS AND DISCUSSION

wt% CNFs in the paper sheet. It is suggested that the aldehyde groups of the periodate-oxidised CNFs also form imine bonds with the amines of the PVAm80,81, as indicated by the wet strength, which was above 10 kNm/kg with less than 1 wt% CNFs in the paper sheets and almost 30 kNm/kg with ca 14 wt%. This corresponds to approximately 10 % and 30 % of the dry strength, respectively. Dopamine-grafted CNFs gave a high retention at low additions even without a retention aid and gave a significant increase in tensile strength and a decrease in dewatering rate, which is suggested to be due to their high propensity to form films (Figure 4.7).

Figure 4.6 Tensile strength index versus sheet density of sheets with MFC and CNFs; a) with and b) without retention aid (PVAm or, if indicated, pDADMAC).

RESULTS AND DISCUSSION

b) )

Figure 4.7 SEM-images of papers a) without fibrils and after the addition of b) 2 wt% and c) 15 wt% of dopamine-modified CNFs. The film formation was most evident with the dopamine-modified CNFs but could be seen in all the samples.

RESULTS AND DISCUSSION

4.5 Papers from wood fibres with adsorbed SiO2 and ZnO nanoparticles (Papers VII and VIII)

The modification of wood fibres with ZnO and SiO2 nanoparticles was investigated due to previously reported improvements in the dielectric properties of electrical insulation materials with these nanoparticles6,8,9,82,83. In addition, physical adsorption and the Layer-by- Layer (LbL) technique were evaluated as possible routes to modify wood fibres for their intended use in electrical insulation materials.

4.5.1 Adsorption

Papers containing ZnO and SiO2 nanoparticles were obtained by physical adsorption of charged nanoparticles onto wood fibres in a wet pulp suspension before the papermaking step. Multilayers were also successfully constructed by alternating the adsorption of SiO2 nanoparticles and polyelectrolytes. The charge density of the nanoparticles is pH-dependent. Both the polyelectrolytes used, PEI and PAA, are weak polyelectrolytes with charge densities and hence adsorption characteristics dependent on the pH (Figure 4.8). With physical adsorption and the LbL method, papers with SiO2 contents up to 15.3 wt% were produced. The hypothesis was that the 5 and 12 nm diameter particles would be small enough to penetrate the fibre-wall, but this hypothesis was rejected when SEM-images of fibre cross sections revealed that the SiO2 particles were only present on the fibre surface (Figure 4.9). The stability of ZnO nanoparticles in water, as shown in Figure 4.10, is pH-sensitive, and their charge density and dissolution were therefore carefully investigated to determine the pH suitable for adsorption. The charge densities of APTES-modified ZnO nanoparticles are more than one order of magnitude lower than that of the cationic SiO2 nanoparticles. APTES-modified ZnO nanoparticles were adsorbed onto wood fibres at pH 6.5 after which paper samples with approximately 1 wt% of nanoparticles were produced. SEM-images showed that the ZnO nanoparticles were distributed on the wood fibre surface (Figure 4.11 a-c). An attempt was made to covalently bond APTES-modified ZnO particles

RESULTS AND DISCUSSION

using EDC (1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide) as coupling agent, but this led to the formation of rod-like structures of ZnO and these papers were not further studied (Figure 4.11 d). For the intended application, possible desorption and migration into the oil and its implications should definitely be investigated, taking into consideration the ageing conditions and the long operating times (several decades) of a transformer.

8 a

6.6 6 6.01 5.5

3.3 catSiO2 (1 layer)

nanoparticle content nanoparticle (wt%) 2.4

2 2

SiO 3 4 5 6 7 8 pH

16 b anSiO2, PEI pH 7 15.3 14 anSiO2, PEI pH 10 catSiO , PAA pH 3.5 12 2 10.8 10.2 catSiO2, PAA pH 6 10 10.4

8 8.2 6.01 6.1 6 5.9 4.9 4.1 4

nanoparticle content nanoparticle (wt%)

2 2 2.4 1.8

SiO 1 2 3 Number of SiO2 nanoparticle layers

Figure 4.8 The content of SiO2 nanoparticles in paper samples plotted a) versus pH for a single layer of cationic (catSiO2) particles and b) and as a function of the number of layers deposited with the aid of different polyelectrolytes at different pH’s, measured by TGA.

RESULTS AND DISCUSSION

Figure 4.9 Cross section of a wood fibre with cationic SiO2 nanoparticles (6 % of the total paper weight) covering the surface.

4 Particles are Particles are dissolving agglomerated or

2 negatively charged

0 2 4 6 8 10 12 pH (µeq/g) density Charge -2

-4

-6 Figure 4.10 Charge density of APTES-modified ZnO nanoparticles in deionized water with increasing pH. The particles dissolved below pH 6 and have their isoelectric point around pH 8.

RESULTS AND DISCUSSION

a) b)

c) d)

Figure 4.11 SEM-images of a) and b) papers containing 1 wt% APTES-modified ZnO nanoparticles showing an even distribution of small agglomerates, supported by c) the detection of backscattered electrons, and d) of a paper with rod-like ZnO structures after coupling with EDC was attempted.

4.5.2 Dielectric response

Generally, SiO2 nanoparticles lowered the permittivity of the papers, even after taking into account any change in density. The permittivity of SiO2 is lower than that of pure kraft pulp but higher than that of the paper-oil combination. ZnO has a higher permittivity (ca 8) then the kraft paper and since the density of paper samples with ZnO nanoparticles is similar to the reference sample, neither of these parameters explains the low permittivity and loss that was measured for paper sheets with ZnO nanoparticles. A low mobility in interfacial areas could have decreased the overall polarizability. The interfacial polarization present in conventionally filled materials has been demonstrated to be suppressed or even eliminated if the dimensions of the particles are reduced and the local electric fields are mitigated 15,84.

RESULTS AND DISCUSSION

The presence of traps or scattering sites introduced by the incorporation of interfacial areas may also affect the charge transport mechanism (hopping charges) by decreasing the mobility of charge carriers, and this would explain the low loss and the flat slope of the permittivity at low frequencies for the nanoparticle-filled papers85,86. For nanocomposites with synthetic polymers, some investigations have shown that the greatest improvements in the dielectric and mechanical properties are achieved if the inorganic nanoparticles are surface modified in order to enhance the interactions or even create covalent bonds between filler and matrix86. In some cases, the functional groups grafted to the surfaces are also proposed to function as traps or scattering sites6,13. With wood fibres, on the other hand, the filler- matrix interactions are already favoured by the hydrophilic nature of the cellulose, and with charges on wood fibres and nanoparticles, the interactions are further enhance by electrostatics. The retained mechanical performance measured for papers containing SiO2 nanoparticle (Paper

VII), indicates that the incorporation of SiO2 nanoparticles at low concentrations does not weaken the interactions between fibres. An even distribution, a moderate nanoparticle concentration and a small particles size were probably beneficial for both the fibre-fibre interaction and for the formation of the paper sheets. In addition, a large interfacial area was created without contact between particles, which has that also previously been shown to be a criterion for improved dielectric and mechanical properties11. In some of the modified papers, remaining charge carriers are believed to have affected the permittivity and dielectric losses at low frequencies, especially in oil.

4.5.3 Surface partial discharge (PD) resistance (Paper VIII)

Papers with 2 wt% SiO2 and 6 wt% SiO2 nanoparticles had a greater the impulse resistance of AC PD than both the reference paper and a paper with ZnO nanoparticles. The paper with ZnO nanoparticles showed a lower discharge activity in respect to the amount and magnitude of charges then the reference only before being exposed to an impulse. The lower permittivty of the nanoparticle-filled papers could improve the surface

RESULTS AND DISCUSSION

discharge resistance by decreasing the mismatch with the oil permittivty. In addition, it has been suggested that the incorporation of the interfacial zone also works to increase the breakdown voltage and resistance towards discharges6,11,15, by mechanisms that have been discussed in previous section.

CONCLUSIONS

5. CONCLUSIONS

The development of new electrical insulating materials can be achieved by both conventional and novel modifications of wood-based fibres. The physical processes that occurs in the insulation system of a transformer with high applied operating voltages, impulses, elevated temperatures and high mechanical loads need to be understood and addressed if new materials are to be properly evaluated. All results are meant to contribute with some pieces to an increased understanding of the phenomena involved and of how to achieve improved mechanical and dielectric performance of cellulose-based insulation materials. The following general conclusions can be drawn:

The paper morphology contributes to the branching of streamers. This is supported by the shorter stopping length for measurements in negative polarity and shadowgraphs obtained for measurements of streamer propagation in positive polarity. Branching and shorter stopping lengths may result in less deterioration of the material and a lower probability of failure of the power transformers.

On a low permittivity polymeric film, streamers had a higher inception voltage and induced less charges than on the materials with higher permittivities. Low density polyethylene (LDPE) and polytetrafluoroethylene (PTFE) have permittivities close to that of oil and in negative polarity, higher streamer inception voltages were measured with these materials (14.7 kV with LDPE in oil and 15.5 kV with PTFE in oil, compared to ca 13 kV in oil and with kraft paper in oil). The permittivity of cellulose is an inherent material property but the permittivity of an oil-impregnated paper can be affected by the pulp composition and is highly dependent on the oil:paper ratio. Even though this makes it possible to alter the permittivity of an oil-impregnated paper,

CONCLUSIONS it is not evident that it is the combined oil-paper permittivity that is important. The construction of a low permittivity surface using e.g. a low-density cellulosic material or a coating could be beneficial for the dielectric performance and from a manufacturing viewpoint. To explore this, it would be interesting to define how thick such a surface layer should be; if and when the relative influence on streamer properties of the bulk permittivity can be superseded by that of the surface permittivity.

Fibrils can be used to increase both mechanical strength and DC breakdown strength. It was showed that the chemically modified fibrils significantly increased the tensile strength index of a paper produced from electrical grade kraft pulp. Microfibrils from electrical grade pulp (kraft MFC) were shown to increase the DC breakdown strength when incorporated as a film between two regular papers. Higher mechanical strength and higher breakdown strength could enable down-sizing of the insulating components. However, on a kraft MFC film, a propagating streamer was concentrated to the surface and its velocity and length were greater than those of a streamer propagating on a normal kraft paper. The results suggest that to improve the overall dielectric performance it is necessary to distinguish between material properties optimal for high resistance towards dielectric phenomena at the surface and in the bulk.

Nanoparticles can decrease the permittivity and the surface discharges under an applied AC voltage.

SiO2 and ZnO nanoparticles in papers were shown to decrease the permittivity and also to decrease the surface discharge. A possible mechanism suggested to be responsible for this is decreased mobility and trapping and scattering of charges in the interfacial region. Physical adsorption and the Layer-by-Layer (LbL) technique were used to obtain layers with evenly distributed nanoparticles. However, the washing to remove excess ions appeared to be more difficult than with an unmodified pulp.

CONCLUSIONS

It has been showed that fibrils and nanoparticles can be used to upgrade paper products used as electrical insulation. Hopefully, the investigation into the mechanisms of streamer inception and propagation and their relation to material properties can also be useful for future studies.

FUTURE WORK

6. FUTURE WORK

The aim of this project was to study cellulose-based electrical insulation materials and how they can be modified to improve resistance towards streamer inception and propagation. The specially designed and constructed experimental set-up for studying streamers was primarily used to investigate the relevant mechanisms and how they are related to material properties. Due to the time-consuming nature of the measurements, only a few cellulosic materials have so far been evaluated. For future work, more cellulosic samples have been and could be produced to further increase our understanding of how to control streamer inception and propagation. Paper samples of different compositions and different surface roughnesses and densities will be evaluated along with the papers made from fibres with adsorbed nanoparticles. It is also suggested that further investigations should be made into how different cellulosic micro- and nanofibrils can be used as additives or in laminates to improve dielectric properties. In this context, different chemistries as well as different fibril contents should be compared, and this should also provide valuable input to the relative influence of surface and bulk properties on streamer inception and propagation. For modified materials, the results should add a valuable contribution to the fundamental understanding of how to improve the insulating performance, and of the dielectric phenomena. Modified materials should also be evaluated with respect to their industrial relevance and overall dielectric and mechanical performance.

REFERENCES

7. REFERENCES

1. William H, Bartley P. An Analysis of International Transformer Failures, Part 1. Locomot. 2013:1-4. 2. Murugan R, Ramasamy R. Failure analysis of power transformer for effective maintenance planning in electric utilities. Eng Fail Anal. 2015;55:182-192. 3. Oommen T V., Prevost T a. Cellulose insulation in oil-filled power transformers: Part II - maintaining insulation integrity and life. IEEE Electr Insul Mag. 2006;22(2):5-14. 4. Prevost T a., Oommen T V. Cellulose insulation in oil-filled power transformers: Part I - History and development. IEEE Electr Insul Mag. 2006;22(1):28-34. 5. Harmer MA, Liauw AY, Kang BS, Scialdone MA. US 20140186576 A1: Insulating material containing nanocellulose. 2014. 6. Roy M, Nelson JK, MacCrone RK, et al. Polymer nanocomposite dielectrics - The role of the interface. IEEE Trans Dielectr Electr Insul. 2005;12(4):629-642. 7. Lewis TJ. Nanometric dielectrics. IEEE Trans Dielectr Electr Insul. 1994;1(5):812-825. 8. Tanaka T. Dielectric nanocomposites with insulating properties. IEEE Trans Dielectr Electr Insul. 2005;12(5):914-928. 9. Grabowski CA, Fillery SP, Westing NM, et al. Dielectric breakdown in silica-amorphous polymer nanocomposite films: The role of the polymer matrix. ACS Appl Mater Interfaces. 2013;5(12):5486-5492. 10. Roy M, Nelson JK, MacCrone RK, Schadler LS. Candidate mechanisms controlling the electrical characteristics of silica/XLPE nanodielectrics. J Mater Sci. 2007;42(11):3789-3799. 11. Cao Y, Irwin PC, Yousni K. The Future of Nanodielectrics in the Electrical Power Industry. IEEE Trans Dielectr Electr Insul. 2004;11(5):797-807. 12. Schadler LS, Nelson JK, Calebrese C, Travelpiece A, Schweickart D. High temperature breakdown strength and voltage endurance characterization of nanofilled polyamideimide. IEEE Trans Dielectr Electr Insul. 2012;19(6):2090-2101.

REFERENCES

13. Ma D, Hugener TA, Siegel RW, et al. Influence of nanoparticle surface modification on the electrical behaviour of polyethylene nanocomposites. Nanotechnology. 2005;16(6):724-731. 14. Huang X, Zheng Y, Jiang P, Yin Y. Influence of nanoparticle surface treatment on the electrical properties of cycloaliphatic epoxy nanocomposites. IEEE Trans Dielectr Electr Insul. 2010;17(2):635-643. 15. Nelson JK, Hu Y. Nanocomposite dielectrics—properties and implications. J Phys D Appl Phys. 2005;38(2):213-222. 16. Pourrahimi AM, Hoang TA, Liu D, et al. Highly Efficient Interfaces in Nanocomposites Based on Polyethylene and ZnO Nano/Hierarchical Particles: A Novel Approach toward Ultralow Electrical Conductivity Insulations. Adv Mater. 2016;28(39):8651-8657. 17. Lewis TJ. Interfaces are the dominant feature of dielectrics at the nanometric level. IEEE Trans Dielectr Electr Insul. 2004;11(5):739-753. 18. Kubyshkina E, Unge M, Jonsson BLG. Communication: Band bending at the interface in polyethylene-MgO nanocomposite dielectric. J Chem Phys. 2017;146(5):51101. 19. Gao F, Xiang M, Liao R, Xu Z, Wang J. The Experimental Investigation on Space Charge Distribution of Cellulose Insulation Paper Modified with Alumina Nanoparticles. 2016:773-777. 20. Liao R, Lv C, Yang L, Zhang Y, Wu W, Tang C. The insulation properties of oil-impregnated insulation paper reinforced with nano- TiO2. J Nanomater. 2013;2013. 21. Rafiq M, Lv Y, Li C. A Review on Properties, Opportunities, and Challenges of Transformer Oil-Based Nanofluids. J Nanomater. 2016;2016. 22. Jin H, Morshuis PHF, Smit JJ, Andritsch T. The effect of surface treatment of silica nanoparticles on the breakdown strength of mineral oil. Proc - IEEE Int Conf Dielectr Liq. 2014:1-4. 23. Contreras JE, Rodriguez EA, Taha-Tijerina J. Nanotechnology applications for electrical transformers—A review. Electr Power Syst Res. 2016;143:573-584. 24. Luethen V, Winkler G. US9171656 B2: Electrically insulating nanocomposite having semiconducting or nonconducting nanoparticles, use of this nanocomposite and process for producing it. 2015. 25. Bertrand Y, Hoang LC. Vegetal oils as substitute for mineral oils. Proc 7th Int Conf Prop Appl Dielectr Mater. 2003;2:491-494. doi:10.1109/ICPADM.2003.1218460.

REFERENCES

26. Unge M, Singha S, Van Dung N, Linhjell D, Ingebrigtsen S, Lundgaard LE. Enhancements in the lightning impulse breakdown characteristics of natural ester dielectric liquids. Appl Phys Lett. 2013;102(17):172905. 27. Lesaint 0., Massala. Positive Streamer Propagation in Large Oil gaps: Experimental Characterization of Propagation Modes. IEEE Trans Dielectr Electr Insul. 1998;5(3):360-370. 28. Krässig HA. Cellulose, Structure, Accessibility and Reactivity. Amsterdam: Gordon and Breach Science Publisher; 1993. 29. Sjöström E. Wood Chemistry. 2nd ed. London, UK: Academic Press; 1993. 30. Wickholm K, Larsson PT, Iversen T. Assignment of non-crystalline forms in cellulose I by CP / MAS 13 C NMR spectroscopy. 1998;312. 31. Boylston EK, Hebert JJ. Fibrillar size in native cellulose. J Appl Polym Sci. 1980;25(9):2105-2107. 32. Bristow A, Kolseth P. Paper Structure and Properties. Taylor & Francis; 1986. 33. Jayme G. Micro-Quellenmessungen an Zellstoffen. Der Pap Wochenbl Pap. 1944;6:187-194. 34. Laivins GV, Scallan AM. The influence of Drying and Beating on the Swelling of Fines. J Pulp Pap Sci. 1996;22(5):178-183. 35. Lumiainen J. Refining of chemical pulp. In: Paulapuro H, ed. Papermaking Science and Technology, Papermaking - Part1, Stock Preparation and Wet End. Book 8. Volume 8. Fapet Oy, Jyväskylä, Finland; 2000:87-121. 36. Stone JE, Scallan AM, Abrahamson B. Influence of Beating on Cell Wall Swelling and Internal Fibrillation. Sven papperstidning. 1968;(19):687-694. 37. Page D.H. The beating of chemical pulps – the action and the effect. In: 9th Fundamental Research Symposium Notes. Cambridge, UK; 1989:1. 38. Lindström T, Wågberg L, Larsson T. On the nature of joint strength in paper – a review of dry and wet strength resins used in paper manufacturing. 13th Fundam Res Symp. 2005;32(May 2005):457-562. 39. Klemm D, Heublein B, Fink HP, Bohn A. Cellulose: Fascinating biopolymer and sustainable raw material. Angew Chemie - Int Ed. 2005;44(22):3358-3393. doi:10.1002/anie.200460587. 40. Mohanty AK, Misra M, Drzal LT. Surface modifications of natural fibers and performance of the resulting biocomposites: An overview. Compos Interfaces. 2001;8(5):313-343.

REFERENCES

41. Turbak AF, FW S, KR S. Microfibrillated cellulose, a new cellulose product: properties, uses, and commercial potential. In: J Appl Polym Sci: Appl Polym Symp 37. ; 1983:815-827. 42. Eichhorn SJ, Dufresne A, Aranguren M, et al. Review: Current International Research into Cellulose Nanofibres and Nanocomposites. Vol 45.; 2010. 43. Hamedi M, Karabulut E, Marais A, Herland A, Nyström G. Nanocellulose aerogels functionalized by rapid layer-by-layer assembly for high charge storage and beyond. Angew Chemie - Int Ed. 2013;52(46):12038-12042. 44. Sehaqui H, Liu A, Zhou Q, Berglund L a. Fast preparation procedure for large, flat cellulose and cellulose/inorganic nanopaper structures. Biomacromolecules. 2010;11(9):2195-2198. 45. Koga H, Saito T, Kitaoka T, Nogi M, Suganuma K, Isogai A. Transparent, conductive, and printable composites consisting of TEMPO-oxidized nanocellulose and carbon nanotube. Biomacromolecules. 2013;14(4):1160-1165. 46. Iwamoto S, Nakagaito AN, Yano H. Nano-fibrillation of pulp fibers for the processing of transparent nanocomposites. Appl Phys A Mater Sci Process. 2007;89(2):461-466. 47. Pääkkö M, Vapaavuori J, Silvennoinen R, et al. Long and entangled native cellulose I nanofibers allow flexible aerogels and hierarchically porous templates for functionalities. Soft Matter. 2008;4(12):2492-2499. 48. Le Bras D, Strömme M, Mihranyan A. Characterization of dielectric properties of nanocellulose from wood and algae for electrical insulator applications. J Phys Chem B. 2015;119(18):5911-5917. 49. Bertrand P, Jonas A, Laschewsky A, Legras R. Ultrathin polymer coatings by complexation of polyelectrolytes at interfaces: suitable materials, structure and properties. Macromol Rapid Commun. 2000;21:319-348. 50. Decher G. Fuzzy Nanoassemblies: Toward Layered Polymeric Multicomposites. Science (80- ). 1997;277(5330):1232-1237. 51. IEC Standard 60270 Partial Discharge Measurements. Geneva, Switzerland; 2000. 52. Raju GG, ed. Dielectrics in Electric Fields. New York: Macel Dekker, Inc.; 2003.

REFERENCES

53. Jafarpour G, Dantras E, Boudet A, Lacabanne C. Study of dielectric relaxations in cellulose by combined DDS and TSC. J Non Cryst Solids. 2007;353(44-46):4108-4115. 54. Jonscher AK. Dielectric Relaxation in Solids. London: Chelsea Dielectrics Press Ltd; 1983. 55. A. Sihvola. Electromagnetic Mixing Formulas and Applications. Padstow, Cornwall: The Institution of Electrical Engineers; 1999. 56. Lokhanin A K, Shneider G Y, Sokolov V V EA. Internal insulation failure mechanisms of HV equipment under service conditions. Cigre Rep 15. 2002;201:1-6. 57. Dai J, Wang Z, Jarman P. Creepage discharge on insulation barriers in aged power transformers. IEEE Trans Dielectr Electr Insul. 2010;17(4):1327-1335. 58. Lesaint O. Prebreakdown phenomena in liquids: propagation “modes” and basic properties. J Phys D Appl Phys. 2016;49. 59. Denat A. Conduction and breakdown initiation in dielectric liquids. In: IEEE International Conference on Dielectric Liquids, ICDL. ; 2011. 60. Denat A. High field conduction and prebreakdown phenomena in dielectric liquids. IEEE Trans Dielectr Electr Insul. 2006;13(3):518-525. doi:10.1109/TDEI.2006.1657963. 61. Biller P. A simple qualitative model for the different types of streamers in dielectric liquids. 1996:189-192. 62. Massala G, Lesaint O. Streamer propagation and breakdown in transformer oil under overvoltage in semi-uniform gaps. IEEE 1997 Annu Rep Conf Electr Insul Dielectr Phenom. 1997;2:0-3. 63. Lundgaard L, Linhjell D, Berg G, Sigmond S. Propagation of positive and negative streamers in oil with and without pressboard interfaces. IEEE Trans Dielectr Electr Insul. 1998;5(3):388-395. 64. Norgren M, Notley SM, Majtnerova A, Gellerstedt G. Smooth model surfaces from lignin derivatives. I. Preparation and characterization. Langmuir. 2006;22(3):1209-1214. 65. Zhang Y, Li J, Lindstroem ME, Stepan A, Gatenholm P. Spruce glucomannan: preparation, structural characteristics and basic film forming ability. Nord Pulp Pap Res J. 2013;28:323-330. 66. Gustafsson E, Johansson E, Wågberg L, Pettersson T. Direct adhesive measurements between wood biopolymer model surfaces. Biomacromolecules. 2012;13(10):3046-3053.

REFERENCES

67. Jackson EL, Hudson CS. Application of the cleavage type of oxidation by periodic acid to starch and cellulose. J Am Chem Soc. 1937;59(10):2049-2050. 68. Karabulut E, Pettersson T, Ankerfors M, Wågberg L. Adhesive layer-by- layer films of carboxymethylated cellulose nanofibril-dopamine covalent bioconjugates inspired by marine mussel threads. ACS Nano. 2012;6(6):4731-4739. 69. Nakajima N, Ikada Y. Mechanism of amide formation by carbodiimide for bioconjugation in aqueous media. Bioconjug Chem. 1995;6(1):123- 130. 70. Wåhlander M, Nilsson F, Larsson E, et al. Polymer-grafted Al2O3- nanoparticles for controlled dispersion in poly(ethylene-co-butyl acrylate) nanocomposites. Polym (United Kingdom). 2014;55(9):2125- 2138. 71. Wågberg L, Ödberg L, Glad-Nordmark G. Charge determination of porous substrates by polyelectrolyte adsorption. Nord Pulp Pap Res J. 1989;4(2):71-76. 72. Venkateswaran A. Formulas for the Dielectric Constant and Dissipation of Mixtures and Their Application to the Cellulose System. J Appl Polym Sci. 1965;9:1127-1138. 73. Morgan VT. Effects of Frequency, Temperature, Compression, and Air Pressure on the Dielectric Properties. IEEE Trans Dielectr Electr Insul. 1998;5(1):125-131. 74. Bergström L, Torbjörn Dahlfors SS, Arwin H, Ödberg L. Spectroscopic ellipsometry characterisation and estimation of the Hamaker constant of cellulose. Cellulose. 1999;6:1-13. 75. Denbigh KG. The Polarizabilities of Bonds - I. Faraday Soc. 1940;936- 948. 76. Huzayyin A, Boggs S, Ramprasad R. Quantum mechanical studies of carbonyl impurities in dielectric polyethylene. IEEE Trans Dielectr Electr Insul. 2010;17(3):920-925. 77. Hollertz R, Wågberg L, Pitois C. Novel cellulose nanomaterials. IEEE Int Conf Liq Dielectr. 2014. 78. Brodin FW, Gregersen ØW, Syverud K. Cellulose nanofibrils: Challenges and possibilities as a paper additive or coating material - A review. Nord Pulp Pap Res J. 2014;29(1):156-166.

REFERENCES

79. Larsson PA, Gimåker M, Wågberg L. The influence of periodate oxidation on the moisture sorptivity and dimensional stability of paper. Cellulose. 2008;15(6):837-847. 80. Han S, Lee M, Kim BK. Crosslinking Reaction of Oxidized Cellulose Fiber. I. Reaction Between Dialdehyde Cellulose and Multiofunctional Amines on Lyocell Fabric. J Appl Polym Sci. 2010;117:682-690. 81. Syverud K, Pettersen SR, Draget K, Chinga-Carrasco G. Controlling the elastic modulus of cellulose nanofibril hydrogels—scaffolds with potential in tissue engineering. Cellulose. 2015;22(1):473-481. 82. Huang J, Yuanxiang Z, Dong L, Zhou Z, Liu R. Enhancement of mechanical and electrical performance of insulating presspaper by introduction of nanocellulose. Compos Sci Technol. 2017;138:40-48. 83. Fabiani D, Montanari GC, Krivda A, Schmidt LE, Hollertz R. Epoxy based materials containing micro and nano sized fillers for improved electrical characteristics. In: Proceedings of the 2010 IEEE International Conference on Solid Dielectrics, ICSD 2010. ; 2010. 84. Roy M, Nelson JK, MacCrone RK, Schadler LS, Reed CW, Keefe R. Polymer nanocomposite dielectrics-the role of the interface. IEEE Trans Dielectr Electr Insul. 2005;12(4). 85. Smith RC, Hui L, Nelson JK, Schadler LS. Interfacial charge behavior in nanodielectrics. Annu Rep - Conf Electr Insul Dielectr Phenomena, CEIDP. 2009:650-653. doi:10.1109/CEIDP.2009.5377878. 86. Smith RC, Liang C, Landry M, Nelson JK, Schadler LS. The mechanisms leading to the useful electrical properties of polymer nanodielectrics. IEEE Trans Dielectr Electr Insul. 2008;15(1):187-196. 87. Larsson PA, Gimåker M, Wågberg L. The influence of periodate oxidation on the moisture sorptivity and dimensional stability of paper. Cellulose. 2008;15(6):837-847. 88. López Durán V, Larsson PA, Wågberg L. On the relationship between fibre composition and material properties following periodate oxidation and borohydride reduction of lignocellulosic fibres. Cellulose. 2016;23:3495-3510.

LIST OF ABBREVIATIONS

AC Alternating current AFM Atomic Force Microscopy APTES (3-aminopropyl) triethoxy silane CNFs Cellulose nanofibrils DC Direct current EDC 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide IR Infrared HV High voltage LbL Layer-by-Layer LDPE Low density polyethylene MFC Microfibrillated cellulose PAA Polyacrylic acid pDADMAC Polydiallyldimethylammonium chloride PEI Polyethylene imine PET Polyethylene terephthalate PD Partial discharge PTFE Polytetrafluoroethylene PVAm Polyvinyl amine PVDF Polyvinylidene fluoride SEM Scanning Electron Microscopy TGA Thermogravimetric analysis UV Ultra Violet

Symbols tan δ Dielectric loss tangent, dissipation factor ε Dielectric constant, permittivity ε0 Vacuum permittivity

APPENDIX

Table A1. Material properties Breakdown Permittivity Dielectric loss Product Material strength ԑ tan δ name (kV/mm)

LDPE1 2.2 0.0001-0.001 27 ET311201

PET1 3.0 0.002* 17 ES301400

PTFE1 2.0-2.1 0.0003-0.0007 50-170 FP301300

PVDF1 8.4 0.06 13 FV301300 Pressboard in - 0.004 45-50 Elboard oil2 Transformer oil3 2.2 0.003-0.005 >30 Nytro 10XN

1 Data from Good Fellow Cambridge Ltd, England (at 1MHz, *at 1kHz) 2 Data of high density board from Figeholm Elboard, Figeholm, Sweden 3 Data from Nynäs AB, Nynäshamn, Sweden

Diagrams showing the formation of cellulose functionalities and possible crosslinks

Figure A1. Oxidation of the C2-C3 bond of cellulose (left) into dialdehyde cellulose (right), with sodium periodate67.

Figure A2. Crosslinking can take place between aldehydes of the oxidised cellulose (Figure A1) and a) hydroxyls forming hemiacetal bonds87 and b) a primary amine (PVAm) forming imine bonds (also known as Schiff base formation)80.

Figure A3. Carboxymethylation of cellulose29.

Figure A4. Dopamine can be grafted onto the carboxyls of the carboxymethylated cellulose (Figure A3), i.e. the carboxymethylated cellulose reacts with EDC, and dopamine is then grafted onto an intermediate structure69.

Figure A5. Infrared (IR) spectra of CNF films made from the different fibrils and from periodate-oxidised carboxymethylated fibrils with retention aids. The oxidation and the reaction with dopamine was detected with FTIR (with a reduction of the carboxyl peak68 at 1593-1, the appearance of the characteristic peak of dialdehyde cellulose88 at 1740 cm-1 and the appearance of bands attributed to amide carbonyl stretching vibrations68 at 1680-1630 cm-1 for the film with dopamine fibrils). It was also shown that the periodate-oxidised fibrils have reacted further with PVAm (N-H deformation vibrations80 were detected at 1555 cm-1) but not with pDADMAC.