ISBN 91-7283-447-1 ISRN KTH/KKE-03/1-SE ISSN 0349-6465 TRITA-KKE-0301 ======

Free Radical Mediated Cellulose Degradation

Erik E Johansson

Doctoral Thesis Department of Chemistry Nuclear Chemistry Royal Institute of Technology

======Stockholm, Sweden 2003

Abstract This thesis addresses the mechanisms involved in cellulose degradation in general and Totally Chlorine Free (TCF) bleaching of pulp in particular. The thesis shows that the cellulose degradation during high consistency ozone bleaching is explained by free radical chain reactions. By simulation, it has been shown that the number, weight and viscosity average of liner polymer chain length can be used to calculate the number of random scissions in a linear polymer of any molecular weight distribution, provided that there is a calibrated Mark- Houwink equation. A model describing partial degradation of molecular weight distributions of linear polymers measured with viscometry was developed and verified experimentally. The model predicts viscometric measurement of chemical cellulose degradation by a rapidly reacting reagent to be strongly dependent on cellulose accessibility. The role of free radical reactions in cellulose degradation was studied by varying the amount of ferrous ions and ozone added to the cotton linters. The result was compared to the results obtained from cellulose of lower crystallinity (cellulose beads) by measuring average chain length. When a ferryl ion reacted with cotton linters in the presence of ozone, the very formation of one glycosidic radical was more significant to degradation than the final step of forming one oxidised glycoside. The inefficient degradation observed of the oxidation step is explainable by the amount of accessible glycosides being too small to influence viscometry. The efficient degradation observed in association with the glycosidic radical formation is explained by initiation of free radical chain reactions that are propagated as long as there is ozone in the system. As none of these phenomena were found in the less crystalline cellulose, cellulose structure appears to be important for how free radical mediated cellulose degradation develops. The theory of free radical chain reactions coupled with diffusion suggests a concentric expansion of the chain reactions outwards from the initial site of radical formation during ozonation of carbohydrates. This was confirmed by demonstrating free radical chain reactions spreading from a spot of initiation outwards during ozonation of a filter paper, using a pH- indicator to monitor acid formation. Furthermore, the interior and exterior of cellulose fibres doped with initiator were shown to be permeated by small holes after ozonation. Ethylene glycol was shown to improve the selectivity during ozone bleaching of oxygen bleached kraft pulp at pH 3. Optimal conditions were obtained at pH 3 for 25 wt% ethylene glycol. The influence of ethylene glycol on selectivity is explained by a proportion of the free radical chain reactions being carried by the ethylene glycol instead of the cellulose during ozone bleaching. The observations were summarised in the form of a model where the observed degradations for pulp, bleached pulp and cotton fibres during both ozone bleaching and ethylene glycol assisted ozone bleaching were shown to agree with each other. From γ-irradiation of ozonised aqueous solutions of alcohol, the rate constant of superoxide formation from the peroxyl radical of methanol was estimated to be 10 s-1. Rate constants of the reactions between ozone and alkylperoxyl radicals were determined to be around 104 M-1s-1. The possibility of the reaction between alkylperoxyl radicals and ozone contributing significantly to free radical chain reactions during ozonation of carbohydrates and alcohols could therefore be ruled out.

Keywords: Cellulose, degradation, free radical, ozone, selectivity, ethylene glycol, alcohol, bleaching, kraft pulp, cotton linters, delignification, fibre, fibril, crystallinity, ferryl ion, free radical chain reactions, TCF, viscometry, molecular weight distributions, random scissions. Abstrakt

Denna avhandling handlar om mekanismerna som är inblandade i cellulosanedbrytning, framförallt den som sker vid Helt KlorFri (TCF) blekning av pappersmassa. Avhandlingen visar att cellulosanedbrytningen som sker under ozonblekning (hög konsistens) förklaras av fria radikalers kedjereaktioner. Genom simulering har det visats att nummer-, vikts- och viskositetsgenomsnitt av linjära polymerkedjors längd kan användas till att beräkna antalet slumpmässiga kedjebrott som skett under nedbrytning oavsett molekylviktsfördelning, förutsatt att det finns en kalibrerad Mark- Houwinkekvation. En modell som beskriver partiell nedbrytning av linjära polymers molekylviktsfördelningar mätt med viskometri har utvecklats och verifierats experimentellt. Modellen förutsäger att viskometrisk mätning av kemisk cellulosanedbrytning från snabbt reagerande ämnen kommer att vara under stark påverkan av cellulosatillgänglighet. Fria radikalers roll under cellulosanedbrytning studerades genom att variera mängden ferrojoner och ozon som tillsatts bomulls-linters . Resultatet jämfördes med resultat erhållna från cellulosa av lägre kristallinitet (cellulosakulor) genom att mäta genomsnittlig kedjelängd. När ferryljoner reagerade med bomulls-linters i närvaro av ozon var själva bildningen av glykosidradikaler mer signifikant för nedbrytningsresultatet än det slutliga reaktionssteget, bildandet av en oxiderad glykosid. Den oeffektiva nedbrytningen, oxidationsteget, kan förklaras av att mängden tillgänglig cellulosa i provet var för låg för att påverka viskometrin. Den effektiva nedbrytningen, initiationsteget, förklaras av radikala kedjereaktioners verkan, som propageras så länge det finns ozon i systemet. Eftersom inga av dessa fenomen observerades i den mindre kristallina cellulosan, verkar cellulosastruktur vara viktigt för hur kedjereaktioner utvecklar sig. Teori för fria radikala kedjereaktioner kopplad till diffusion för tanken till en koncentrisk expansion av kedjereaktionerna utåt från den ursprungliga radikalbildningsplatsen, under ozoneringen av kolhydrater. Detta bekräftades genom att visa hur fria radikal kedjereaktioner sprider sig från en initieringsplats utåt vid ozonering av ett filterpapper genom att använda pH-indikator för att följa syrabildningen från nedbrytningen. Dessutom uppvisade det yttre och inre av cellulosafibrer som var dopade med initiator, sig vara genomträngda av små hål efter ozonering. Etylenglykol visades förbättra selektiviteten under ozonblekning av syrgasblekt sulfatmassa vid pH 3. Optimala betingelser vid pH 3 rådde vid 25 vikt% etylenglykol. Etylenglykols inverkan på selektiviteten kan förklaras av att en andel av kedjereaktioner har etylenglykol som substrat i stället för cellulosa under blekning Observationerna sammanfattades i form av en modell där de observerade nedbrytningarna för oblekt massa, blekt massa och bomulls-linters under ozonblekning med eller utan etylenglykol visades överenstämma med varandra. Genom γ-bestrålning av ozoniserade vattenlösningar med alkohol, kunde hastighetskonstanten för superoxid-bildning ur metanols peroxylradikal att bestämmas till 10 s-1. Hastighetskonstanter runt 104 M-1s-1 bestämdes för reaktioner mellan ozon och peroxylradikaler. Möjligheten att reaktioner mellan alkylperoxylradikaler och ozon skulle signifikant bidra till kedjereaktioner vid ozonering av kolhydrater och alkoholer kunde därför uteslutas. List of publications

This thesis is based on the following publications, referred to in the text by their Roman numerals:

Paper I. Johansson, E. E. and Lind, J. (2003). "The general link between random scissions, changes in average chain length and monomer reactions."

Paper II. Johansson, E. E. and Lind, J. (2003). "Viscometry of heterogeneously degraded cellulose samples "

Paper III. Johansson, E. E. and Lind, J. (2003). "Spreading of free radical reactions in cellulosic materials."

Paper IV. Johansson, E. E. and Lind, J. (2003). “Free radical mediated cellulose degradation: Cotton linters during HC pulp ozonation conditions”. J Wood Chem Tech. Submitted.

Paper V. Johansson, E. E., Lind, J. and Ljunggren, S. (2000). "Aspects of the chemistry of cellulose degradation and the effect of ethylene glycol during ozone delignification of kraft pulps." Journal of Pulp and Paper Science, 26(7): 239-244.

Paper VI. Lind, J., G. Merenyi, Johansson, E. and Brinck, T. (2002). "The reaction of peroxyl radicals with ozone in water." Journal of Physical Chemistry A, 107: 676-681

Some general texts on bleaching and theory in this thesis are more or less been verbatim citations. This concerns especially Ed: Dence, C. W. and Reeve, D., W., “Pulp bleaching. Principles and Practice”, TAPPI Press, 1998 and Gedde U., W., “Polymer Physics”, Chapman & Hall, 1995. This kind of citations is annotated after the section dot, ”.34”, instead of the normal ” 34.” within sentences.

Previously published material Johansson, E. E. (2000). Free Radical Mediated Cellulose Degradation During High- Consistency Ozone Bleaching Conditions. Licentiate thesis. Department of Chemistry - Nuclear Chemistry. Stockholm, Royal Institute of Technology.

Johansson, E. and Lind, J. (1999). Free radical degradation of fibres during HC pulp ozonation conditions. 10th International Symposium on Wood and Pulping Chemistry, June 7-10 1999, Yokohama, Japan.

Johansson, E. E., Lind, J. and Ljunggren, S. (1998). Influence of ethylene glycol on ozone bleaching of kraft pulps. 1998 International Pulp Bleaching Conference.

Contents

1 INTRODUCTION 1

2 BACKGROUND 1

2.1 Bleaching 1

2.2 Ozone pulp bleaching 2 2.2.1 Ozone 2 2.2.2 Wood 2 2.2.2.1 Cellulose 2 2.2.2.2 Lignin 2 2.2.2.3 Hemicellulose 3 2.2.2.4 Fibre morphology 3 2.2.3 Treatment 4 2.2.3.1 Kraft pulping 4 2.2.3.2 Oxygen bleaching 5 2.2.3.3 Ozone bleaching at high pulp consistency (HC) 5

2.3 Free radical chemistry in the cellulosic environment 6 2.3.1 Ozone bleaching chemistry 9 2.3.1.1 Formation of free radicals - initiation reactions 9 2.3.1.1.1 Initiation from phenol reactions 9 2.3.1.1.2 Initiation from transition metal reactions 10 2.3.1.1.3 Initiation from carbohydrate reactions 10 2.3.2 The ozone – Fe(II) - carbohydrate system 10 2.3.3 Alcohol chemistry 11 2.3.4 Accessibility: Cellulose structure models 11

3 EXPERIMENTALS 13

3.1 Cellulosic substrates 13

3.2 Pulp ozonation 13

3.3 Measuring lignin: Kappa number 14

3.4 Measuring average cellulose chain length: viscosity 14

3.5 Molecular weight averages 15

4 RESULTS 17

4.1 Scissions, random 17 4.1.1 Investigation 18 4.1.1.1 Random scissions in chains 18 4.1.1.2 The Mark-Houwink equation. 20 4.1.1.3 Summary 21

4.2 Scissions, partial 22 4.2.1 Investigation 22 4.2.1.1 Summary 24

4.3 Free radical mediated cellulose degradation 25 4.3.1 Spreading of free radical degradation in cellulosic substrates 26 4.3.1.1 Investigation: Development of free radical degradation in carbohydrates 26 4.3.1.1.1 Summary 28 4.3.1.2 Investigation: Cellulose structure effects on free radical cellulose degradation 28 4.3.1.2.1 Sub-study: ds-als 31 4.3.1.2.2 Summary 34

4.4 Cellulose degradation in pulp 34 4.4.1 Investigation: Influence of lignin and alcohols upon ozone degradation of pulp 35 4.4.1.1 Discussion 39 4.4.1.1.1 Lignin influence on final viscosity 39 4.4.1.1.2 The influence of ethylene glycol on final viscosity 40 4.4.1.1.2.1 Ethylene glycol and ds 41 4.4.1.2 Summary 41 4.4.2 Investigation: Free radical chain reactions in ozonised aqueous solutions of alcohol 42 4.4.2.1 Summary 43

5 DISCUSSION 44

5.1 Scissions 44 5.1.1 Charged groups 44 5.1.2 Alternative interpretations of mechanisms behind scissions 44 5.1.2.1 Chemistry 44 5.1.2.2 Solubility 45 5.1.2.3 Amorphous zones 45

5.2 Chemistry 46 5.2.1 Free radical formation from direct ozone attack on cellulose 46 5.2.1.1 Transition metals and the relationship to background degradation 46 5.2.1.2 Phenolate anions 47 5.2.1.3 Observed degradation of pulp 47

5.3 Fibres and free radical mediated cellulose degradation 47

5.4 Pulps: Coherency of the results 48 5.4.1 Quantification of the relative influence of the free radical chain reaction parameters in cotton linters 48 5.4.1.1 Calculation 48 5.4.2 Extrapolating from cotton linters to oxygen bleached kraft and fully bleached pulp 50 5.4.2.1 Calculation 50 5.4.2.2 Results 52 5.4.3 Summary 52

6 CONCLUSIONS 54

6.1 Conclusions 54 6.1.1 Rationale of free radical chain reactions 54 6.1.2 Cellulose degradation 54 6.1.2.1 Free radical chain reactions, ds-als 54 6.1.3 Ethylene glycol and methanol 55 6.1.4 Pulp 55 6.1.5 Spreading of chain reactions 55 6.1.6 Scissions 55

ACKNOWLEDGEMENTS 57

REFERENCES 58 1 Introduction Cellulose comprises about 33% of all vegetable matter (90% of cotton and 50% of wood is cellulose). Cellulose is of great economic importance to the world and the most important products are papers and fibres1. Chemistry is fundamental to the manufacture of cellulosic products and has engaged a great number of scientists over the years. In the 1980s, the environmental issue of chlorinated organics in the effluents of paper mills spawned research for ways to decrease or replace chlorine in bleaching processes. One of many results was the TCF-bleaching technology (Totally Chlorine Free). TCF-bleaching normally refers to the use of oxygen, peroxide and ozone as bleaching agents. A common property of these agents is their disposition for free radical chemistry in aqueous dispersions of carbohydrates. Free radical chemistry has long been suggested to be involved with the observed by-reaction of cellulose degradation during TCF-bleaching. This cellulose degradation is unwanted. When a chemist sets out to limit cellulose degradation from oxygen based free radical chemistry during TCF-bleaching, he is confronted by physics and mathematics as well as chemistry. The mechanisms of free radical chemistry of oxyradicals in water are not all clear. The specific nature of free radical chemistry of carbohydrate solids has only recently been realised and much is yet to discover. There are open issues in the understanding of the macrostructure of cellulose, likewise in the relationship between polymer chain length and degradation. In order to limit cellulose degradation in a rational manner, the chemist has to deal with these questions as well, e. g. the mechanisms involved during free radical mediated cellulose degradation. This is the subject of the thesis.

2 Background

2.1Bleaching Bleaching is a chemical process applied to materials to increase their brightness. Brightness is the reflectance of visible light. The absorbance of visible light by wood pulp fibres is caused mainly by the presence of lignin. Lignin in native wood is slightly coloured and residual lignin remaining from some chemical processes (e.g. alkaline processes) is highly coloured. Bleaching processes increase brightness by lignin removal or lignin decolourisation. In the manufacture of pulps, most of the lignin is removed; when these pulps are later bleached, lignin removal is continued2. Ozone was suggested as a means of bleaching pulp on the basis of its specific and effective reaction with double bonds. Nevertheless, ozone delignification has been proven to be associated with considerable degradation of other pulp constituents, most importantly cellulose. How to increase ozone bleaching selectivity, i.e. maximum delignification to minimum cost in terms of cellulose degradation, is still a central question. Characterisation of pulp and bleaching operations is necessary in order to improve selectivity. In this thesis, traditional measures are used, i.e. the kappa number for the lignin content and viscosity for the average cellulose chain length, Sections 3.3 and 3.4. Both methods have been subject to debate but are nevertheless established for communication of research results and plant mill experiences. The chemical constituents are organised into higher order structures in wood and pulp (supermolecular structure, fibre morphology). Cellulose can, for example, be discussed with reference to glycosidic units, cellulose chains, fibrils, cell wall layers, fibres and fibre populations (pulp) or woods. Ideas on the nature of cellulose degradation during pulp

1 bleaching are strongly linked to how cellulose structure is conceived. Cellulose structure models and fibre morphology are discussed in Sections 2.2.2.4 and 2.3.4.

2.2 Ozone pulp bleaching

2.2.1 Ozone Ozone is a strong oxidant, 2.07 eV (2e, H+). The specific double bond chemistry has inspired numerous analytical and technical ozone-based approaches. Water treatment and pulp bleaching are examples of important industrial applications. Van Mauten first reported the existence of ozone in 1785. Schlobein was the first to isolate it, in 1840. A study on ozonation of dry linen was published in 18683. Modern, systematic research of ozone chemistry and cellulose refers to a series of articles by Dorée starting in 19124. A first article on ozone for modern industrial pulp bleaching was published in 19635. Technical impediments are degradative side reactions, low water solubility (0.98 g/L) and high investment and energy costs. Environmental concerns against chlorine-based bleaching methods arose in the 1980s and have created an economic incentive for TCF (totally chlorine free) bleaching methods. The first two commercial ozone-bleaching plants were introduced in 1992. There are 19 ozone pulp-bleaching plants in the world, of which 3 are situated in Sweden6.

2.2.2 Wood Trees are classified botanically as softwoods or evergreens, which are gymnosperms, and hardwood, which are angiosperms. Softwood is the dominant source in Sweden and is the pulp source used for the experiments presented in this thesis. The three main components in softwood are cellulose (41-46 wt%), lignin (26-32 wt%) and hemicellulose (14-17 wt%). Their relative quantities vary with species, growth history and the part of the plant from which the sample is taken2.

2.2.2.1 Cellulose Cellulose (β-1,4-glucan) is a long, linear polymer of D-glucopyranose monomer units in β-1,4 linkages (Figure 1). Cellulose is ordered in plants (wood) and plant derivatives (pulp). The next, higher level is fibrils where the dominant state of cellulose is crystalline or para- crystalline.

6 OH CH2OH

O

O OH 1 R 4 1 R OH β O 4 β O 2 O 3

OH CH2OH

Figure 1. Cellulose

2.2.2.2 Lignin Native lignin in softwood is a heterogeneously branched and cross-linked polymer in which phenylpropane units (C9) are linked by carbon-carbon and carbon-oxygen bonds (Figure 2). Coniferyl alcohol is the dominant precursor to lignin in softwood. Secondary reactions in the lignification process lead to a variety of linkages and functional groups (Figure 3)2.

2 CH2OH

CH

HC

OCH3

OH

Figure 2. Coniferyl alcohol (lignin precursor). Figure 3. Prominent structures in softwood lignin7.

2.2.2.3 Hemicellulose For Pinus Sylvestris (pine), typical contents of dominant hemicelluloses are: galactoglucomannan (“mannans”), 16 wt%, and arabinoglucuronoxylan (“xylans”), 9wt%, in oven dried pulp (odp). They are described as amorphous aggregates with a linear backbone of typically 100 hexose units with a certain frequency of monohexose branches8. Some hexoses of the mannans have acetylated C2 and C3 hydroxyl groups.

2.2.2.4 Fibre morphology The chemical constituents are organised into higher order structures in wood and pulp (supermolecular structure, fibre morphology). Figure 4 to Figure 6 show common morphological models. The principal construction of a wood fibre is depicted in Figure 6. The cellulose in wood and cotton fibres is organised in fibrils and the fibrils are arranged in concentric layers in the fibre wall. The layers of wood and cotton fibres differ in additional chemical content and fibrilar orientation. The average length of wood fibre varies from species to species, ranging from 1.5 to 4.5 mm for softwood fibres. The average length of cotton fibre is much longer and ranges from 10- 100 mm. The wall thickness of fibres and, more importantly, the average cellulose chain length depend on origin. The average cellulose chain length in wood is shorter than in cotton. In trees, wood fibres are bound together by the middle lamella that encloses the cells and, in the mature state, consists mainly of lignin and is the chief lignin source in pulp. Pulping isolates the wood fibres. The pulp and cotton fibres have the shape of long and hollow tubes.

3 Figure 4. Schematic section of softwood. BP, bordered Figure 5. Cell-wall model. 9 pit; Ew, earlywood; Lw, latewood; TR, tracheid; VRD, vertical resin duct; and WR, wood ray 2.

Figure 6. Models of the cellwall structure: (A) cotton fibres according to Rollins and Tripp, (B) softwood tracheids and hardwood libriform fibres according to Fengel and G. Wegener. C, surface of cuticle (rich in pectin and waxes); L, lumen; ML, middle lamella (mainly lignin); P, primary wall (approx. 1µm thick); R, reversal of the fibral spiral; S1, transition lamella or “winding layer” of secondary wall (approx. 1 µm thick); S2, main body of secondary wall (approx. 4µm thick); T, tertiary wall; W, wart layer. 10

2.2.3 Treatment As pulping and bleaching define pulp, industrial kraft pulping and oxygen bleaching are described in the following. The description of the industrial high consistency ozone bleaching procedure served as a template for the laboratory procedure.

2.2.3.1 Kraft pulping The dominant chemical pulping process is the kraft process ("kraft" = strength in German). The alkaline pulping liquor used contains sodium hydroxide and sodium sulphide. The kraft process is sometimes called the “sulphate process”, a term deriving from the use of sodium sulphate as the make-up chemical in the chemical recovery system. Wood chips are impregnated with the pulping liquor. The chips are heated to 150 –180 °C for 1-2 h in batch or continuous systems. After pulping, the chips are soft and may be fiberised with little mechanical action2. Unbleached kraft pulp has the brown colour of grocery bags or corrugated containers. The source of the brown kraft pulp colour is not known with certainty. Degradation products of carbohydrates, lignin and extractives have all been proposed2.

4 None of the commercial pulping processes can completely delignify without adversely affecting the strength properties of pulp. Traditionally, about 4-5% lignin is left in softwood pulp after pulping and the final delignification is performed under much milder conditions with bleaching2.

2.2.3.2 Oxygen bleaching Oxygen bleaching is performed in an alkaline medium under oxygen pressure and at temperatures between 85-115 °C. Oxygen attacks electron-rich sites such as phenolate and enolate groups. A 35-50% decrease in lignin content is typical for oxygen bleaching2.

2.2.3.3 Ozone bleaching at high pulp consistency (HC) In ozone bleaching, the main process variables that influence the ozone pulp reactions, pulp properties and commercial viability are: pulp consistency, ozone charge, pH, time, temperature, chemical additives, carry-over of residual dissolved organic matter to the ozone stage and pulp treatment before the ozone stage. High consistency (HC) was introduced to overcome the low ozone transfer to pulp in the bleaching stage, caused by low water solubility of ozone and large gas volumes. HC ozone bleaching occurs in a gas-phase reactor where pulp fibres are dispersed in an oxygen/ozone gas phase. Pulp preparation consists of thorough washing to reduce carry-over of organic material into the ozone stage, acidification of the pulp to pH 2-3 at low consistency, thickening to the high consistency range by pressing, shredding the pulp “mat” obtained and fluffing. Shredding and fluffing are carried out to reduce the size of pulp flocs, thereby increasing the gas-solid interface. Two reactor designs are proposed to provide good contact between ozone and pulp fibres: a vertical static bed-type reactor and a horizontal dynamic reactor, which acted as a model for the experimental set-up used here. The latter design improves the ozone pulp fibre contact and narrows the distribution of pulp residence time, which improves bleaching uniformity2.

5 Table I. The amounts and distribution of various components in oxygen bleached kraft pulp A model based on part measured values and part literature values (kappa 10, viscosity 957 ml/g, pH 3, 10 g) Organised as in Figure 5 Component Amount m (g) Distribution Reference (mmoles) over fibre wall cross-section Lignin, C9 units 1 0.2 even kappa number*0.15 = wt% lignin 11 Lignin, 0.1-0.6 12 carboxylic groups Lignin, phenol 0.1-0.2 12 groups Mannans ≤5 ≤0.9 even 13 kraft pulp 14 Xylans ≤5 ≤0.9 even 13 kraft pulp 14 Hexenuronic ≤ (0.1-0.24) ≤ (0.2-0.4) even 15 acids (hemicellulose) Metasaccharinic ≤ (0.6-1.4) ≤ (0.1-0.3) even kraft pulp 2 acid end groups (hemicellulose) Cellulose 50 8 fibril, possibly counted as fibril cluster glycosidic units • Some hexoses of the mannans have acetylated C2 and C3 hydroxyl groups in wood. Acetyl groups are easily removed with alkali and they decrease both during kraft pulping and oxygen bleaching. They are considered removed in this model. • Fully bleached pulp is treated as oxygen bleached kraft pulp without lignin, e.g. consisting of compounds of cellulose and hemicellulose as in Table I. • Bridge-like structures between fibrils have recently been suggested to be present within cell wall layers and to consist of hemicellulose 16,17. Their fate during kraft pulping, oxygen bleaching and ozone bleaching is unknown, likewise their impact on DP-estimation methods • Covalent bonds between lignin and carbohydrates, lignin-carbohydrate complexes (LCC), have long been postulated and are gaining increasing evidence14. If a fraction of lignin is bound to carbohydrates, this fraction may prove more difficult to delignify.

2.3 Free radical chemistry in the cellulosic environment The free radical chain reactions start with hydrogen being abstracted from a carbon by a hydroxyl group, R13 (Table II, Figure 7). The high reactivity of the hydroxyl radical limits the distance it is able to migrate between turns of the free radical chain cycle, R16. The preceding step in the reaction sequence is the reaction of the superoxide anion with ozone to produce the hydroxyl radical, R15. At pH 3, the reaction is slowed down, the superoxide anion being protonated to form the hydroperoxyl radical, R5b (Table IV) and this step is - therefore not as fast as R16. The superoxide ion (O2 /HO2) is hence able to migrate a longer distance in a fibre before reacting than the hydroxyl radical.

6 Table II. Principal free radical chain reactions in cellulose exposed to O3 I = initiator, (OH)RH = carbohydrate or alcohol Reactants Product #

Initiation (I) O3 I I• OH• R11 ↓ I• or OH• (OH)RH (OH) R• R12 Propagation (P) (OH)R• O2 (OH)R-OO• R13 - (OH)R-OO• R=O O2 R14 → scissions - ↓ O2 O3 OH• R15 ↓ (OH)RH OH• (OH)R• R16 i Termination (T) R-OO• R-OO• R-OO-R O2 R17

Chain reactions 2. Initiation reactions 2+ 1. - Fe O3 O2

2+ FeO OH•

WATER

FIBRIL SURFACE

ROH •ROH •O2ROH (cellulose)

O2

Figure 7. Principal reactions

i Unclear. H-OO• is likely important to termination.

7 Table III. Selected rate constants in water Reactants Rate constant Ref -6 -1 18 O3 O3 < 6e*10 s (293 K) - -1 -1 19 O3 OH 70 M s (293 K) 20 O3 EG Estimate: 0.4 M-1s-1 -1 -1 21 O3 CH3OH 0.020 M s -1 -1 21 O3 Glucose 0.45 M s OH⋅ EG 2.4*109 M-1s-1 22 8 -1 -1 23 OH⋅ CH3OH 9.7*10 M s OH⋅ Glucose 1.5*109 M-1s-1 22 2 -1 24 HO2⋅ ROO⋅of: Glucose 10 s formation ROO⋅of: EG 4.2*102 s-1 24 -1 25 ROO⋅of: CH3OH 10 s 9 -1 -1 26 O2 ⋅CH2OH 4.9*10 M s - 9 -1 -1 27 O3 O2 1.5*10 M s - + 28 HO2 O2 +H pKa 4.8 2+ 3 29 O3 Fe (8.3±0.3)*10 M-1s-1 Fe2+ FeO2+ (1.4±0.2)*105 29 M-1s-1 2+ 3 30 FeO C2H3OH (5.5±0.5)*10 M-1s-1 3 18 O3 Phenol (1.3±0.2) 10 M-1s-1 9 18 O3 Phenolate anion (1.4±0.4) 10 M-1s-1 O3 Ethene 1.8*105 M-1s-1 31

8 Table IV. Principal reactions. Free radical chain reactions in a system of ozone, carbohydrates, initiator and water, pH 3 Transition metals and phenols are included as example of initiation mechanism. Reactants Product #

Initiation O3 R-RCH-OH HO3→ R-RC⋅-OH R1 HO⋅+O2 - + O3 φ-O +H HO3→ φ-O⋅ R2 ↔ φ-OH HO⋅+O2 n+ n+ O3 M O2 M O⋅ R3a Mn+O⋅ R-RCH-OH Mn+1 R-RC⋅-OH R3b - + - O3 OH +H HO⋅+O2 HO⋅ R4

Propagation R-RC⋅-OH O2 R-RC(O2⋅)-OH R5a R-RC(O2⋅)-OH - R-RC(O) HO2↔ R5b - + O2 +H - + O2 +H O3 O2 O2+HO⋅ R5c HO⋅ R-RCH-OH H2O R-RC⋅-OH R6 Termination Recombination reactions. The influence of the solid state is unknown but undoubtedly important. Reactions are therefore left unspecified Initiators: “φ-“ = Phenyl group, “Mn+” = Transition metal

2.3.1 Ozone bleaching chemistry Ozone reacts with olefinic and activated aromatic bonds (Crigée-reaction, Figure 8). At pH 3 phenol, phenyl and ethylene structures react with ozone with rate constants in the range 103- 105 M-1-s-1 18,31. O O O O O O 3

R R R R R O R

O O H 2O + H 2O 2 R R Figure 8. The Crigee-mechanism 32. In general, ozone reacts with organic substrates by ozonolysis, i.e. ionic cyclo-addition to, and cleavage of, olefinic and activated aromatic bonds. In aqueous media the reactions become more complex.

2.3.1.1 Formation of free radicals - initiation reactions Free radical reactions have been suggested to be central to the degradation processes in all TCF-bleaching methods33. Suggested sources of free radical formation during ozonation are ozone self-decomposition and ozone reacting with hydroxide ions, transition metals, carbohydrates and phenols (lignin). The first and second reactions are too slow around pH 3 to be responsible for free radical formation during conventional ozone bleaching, see Table III

2.3.1.1.1 Initiation from phenol reactions It has been shown that addition of phenols to pulp enhances cellulose degradation during ozonation. A mechanism for free radical formation was demonstrated in 199534 and has served since as the main hypothesis for free radical chemistry during ozonation of pulp35,36. It is based on the acid-base properties of phenol groups in lignin. The pKas of these phenols are

9 in the range 9±2 depending on substituents. Even at pH<protonation of phenolates, which decreases the rate of hydroxyl radical formation.

2.3.1.1.2 Initiation from transition metal reactions Numerous studies have been performed on the influence of transition metals in native pulps or pulps where selected transition metals have been added or removed by complexation. It has been shown that addition of transition metals (Fe(II) and Cu(II)) leads to lower final viscosity averages of ozonised pulp. In kinetic model experiments, Fe(II) was shown to initiate free radical chemistry under such conditions38-40,30,41. Hydrogen peroxide has been suggested to form during ozonation. It is sometimes suggested that transition metals give rise to free radical chemistry by catalysing the decomposition of hydrogen peroxide, as in hydrogen peroxide bleaching and oxygen bleaching42,43,11. Experiments designed to determine the importance of transition metals as hydrogen peroxide decomposition catalysts have as yet shown little correlation between degradation and contents of the transition metal common in pulps or process waters44. Of particular interest are the recent data on the distribution of the transition metals in wood45. Ferrous ions, among other transition metals, have been shown to be concentrated in “spots”, i.e. locally enhanced concentrations of transition metals.

2.3.1.1.3 Initiation from carbohydrate reactions The direct attack of ozone on carbohydrates is frequently suggested to be degrading. Several aspects of these reactions are debated. Position C1 of the cellulosic glycosides appears to be the most favoured attack position. Nevertheless, C1 attack has been questioned and instead the C2 and C3 positions have been proposed as the primary target46,38. Whether the mechanism of attack on C1 is ionic or free radical is debated 40. Two electron (2e) reaction mechanisms, i.e. not free radical, are frequently considered satisfactory to explain degradation of cellulose during pulp ozonation47,48,38. Free radical formation during ozonation of α-methyl glycoside was an early observation which since then has been given little attention in the context of cellulose degradation49. It cannot be ruled out that ozone reacts with, at least, position C6 by an 1e-reaction mechanism and in this way triggers free radical chemistry50.

2.3.2 The ozone – Fe(II) - carbohydrate system Fe2+ reacts with ozone producing a ferryl ion, FeO2+, R7. The ferryl ion initiates the free radical chain reactions by hydrogen abstraction. The ferryl ion reacts with primary alcohols, R8. The logarithms of the rate constants for hydrogen abstraction have been shown to display a linear dependence on the C-H bond dissociation energy (BDE) for a variety of primary alcohols30. A similar linear relation is found for the hydroxyl radical. The hydroxyl radical rate constants for primary alcohols are roughly a factor of 105 higher than those for the ferryl ion30. Ferryl ions are able to abstract hydrogen from: Possibly all carbons (C1-C6) and at least one carbon (C1) of the glycosidic unit. The C6 rate constant should be of the same order of magnitude as for ethanol, 5*103 M- 1s-1. Being secondary carbons, hydrogen removal from C1-C5 should have higher rate constants than ethanol,. The ferryl ion reacts with ferrous ions, R9, (Table V) with a rate constant of (1.4±0.2)*105 M-1s-1.

10 The competition between R9 and R8 is unclear. The ferrous ion is likely to be adsorbed onto glycosidic units of accessible fibril surfaces51, which may favour the initiation reactions. Oxygen addition to glycosidic radicals, R5a (chain reactions) follow after R8. Table V. Additional principal reactions, when ferrous ions are in the system (Table IV) Reactants Product # 2+ 2+ Initiation O3 Fe O2 FeO R7 FeO2+ R-RCH-OH R-RC⋅-OH R8 Propagation Termination FeO2+ Fe2+ Fe3+ Fe3+ R9 HO⋅ Fe2+ Fe3+ R10 HO2⋅ - O2

2.3.3 Alcohol chemistry Alcohols (ethylene glycol (EG), methanol) have many reaction types in common with the cellulose and hemicellulose towards ozone. Direct ozone reactions are slow. All are able to maintain free radical chain reactions. One interesting feature of EG is the formation of formylmethyl radical (Figure 9). The kinetics are consistent with an acid-base catalysed equilibrium, pKa 0.74, where the acidic form expels water with rate constant of 8.6*106 s-1. The base catalysis is second order, 2k = 9*108 M-1s-1 52,53. The formylmethyl radical is an oxidative radical that is able to oxidise various phenols and terminate free radical chain reactions.

OH OH OH OH OH O2 - O2 - O O H -H+ HO HO HO O H+/OH- - H2 O catalysis O O O2 O O Figure 9. Ethylene glycol free radical chemistry 52,53. Methanol differs from EG by the relative stability of the peroxymethanol intermediate in the - O2 expulsion step, R5, which slows down the free radical chain reactions and also in that “formylethyl radical”–like mechanisms described in Figure 9 are excluded.

2.3.4 Accessibility: Cellulose structure models One principle is fundamental to the interpretation of this thesis: cellulose crystallinity determines the access of reactants and solvents to cellulose. Suggested average values for cellulose crystallinity or accessibility vary broadly with species, chemical treatment, measurement method and models and have been intensively discussed 54,55. With the recent advent of 13C-NMR, estimates of accessibility values have tended to decrease56-58. We use the specifics of a model developed from 13C-NMR studies of various celluloses subjected to acid hydrolysis59. The model distinguishes between cellulose at accessible fibril surfaces from inaccessible fibril surfaces and hemicellulose (Figure 10). Inaccessible fibril surfaces would occur within aggregates of coaxially aligned fibrils with close fibril-to-fibril contact (“fibril cluster”). Accessible surface cellulose has been estimated to be 6% for cotton linters, 8% for bleached birch kraft pulp and 13% for kraft spruce pulp, which are low values compared to some of the older estimates57,58.

11 Results in this thesis are discussed from a fibril point of view, not cellulose chains as such. This state of cellulose limits the fate of cellulose chains in all system levels (clusters, wall layers, fibre, pulp) but is not necessarily the level that governs degradation. Additional features in conjunction with the model are collected from selected sources with a similar approach55,60,10,61: The cellulose chains within fibrils are treated here as coaxially aligned in a solid para-crystalline or crystalline state57. Crystal lattices on fibril surfaces are at least partially distorted. Local crystalline disorder due to molecular ends inside the fibrils and mechanical bending may also occur61. In the literature, fibril cross-sections are given in shapes of square to round61. In this thesis, the cross-sections are treated as square and considered built from the cross-sections of 25-100 cellulose chains. The accessible cellulose in fibrils would then consist of 64%-36% of the cellulose chain content. The width of a fibril is commonly estimated to be in the range 1.5-35 nm55,60,61ii. The sizes of fibrils and hence clusters of fibrils vary with the plant species and processing.

Accessible surface

Inaccessible surface

Crystalline

Para-crystalline Figure 10. Schematic picture of aggregated cellulose fibrils,62.

ii The assumed cross-section range and the fibril dimensions are independently taken from literature and may not be consistent with each other.

12 3 Experimentals

3.1 Cellulosic substrates Oxygen bleached kraft pulp, kappa no. 10 and viscosity 975 ml/g, was used. In comparison to oxygen bleached kraft pulp, a pulp free of hemicellulose and lignin, cotton linters, average(viscosity 1500 ml/g), was used. Aqueous dispersions of cellulose beads were used to study the response of decreased cellulose 2+ 63 crystallinity towards O3- and Fe - controlled free radical degradation . Cellulose beads are synthetically produced gel “pearls” containing less ordered cellulose (≤ 5% crystallinity) and water. Cellulose beads were chosen to approximate the theoretical case of cellulose chains dissolved directly in water. Fully-bleached (viscosity 889 ml/g) pulp was used to compare a pulp containing hemicellulose but no lignin to an oxygen bleached kraft pulp.

3.2 Pulp ozonation Figure 11 depicts the pulp ozonation procedure and is based on the operations described in Section 2.2.3.3. residual O3 Consistency adjustment Rotovapour Ozone consumption: setup Thiosulphate ii H2O H2SO4 Air

O3 Characterisation: Pulp Storage Kappa determination Viscometry H2O Air Washing H2SO4

Additives

pH-adjustment

A fluff Z

Figure 11. Experimental set-up. High consistency pulp ozonation. Pulps are initially subject to acid washing (A). This serves to remove some of the transition metal content and to adjust pH to, default, 3. Ozone was produced in oxygen gas (about 1 mole%) at a constant flow rate by passing oxygen through an ozone generator. The outflow was directed into a laboratory reactor consisting of a rotating glass evaporator. The gas inlet was at the bottom of the vessel. The vessel was under constant rotation throughout the ozonation. Varying the flow time of ozone gas through the reactor controlled the ozone charge.

The outgoing flow passed through an ozone absorption flask containing an alkaline KI- solution. The KI-solution was titrated with sodium thiosulphate to determine the ozone content of the gas flow.

13 To inhibit the influence of keto-groups in cellulose on viscometry, the pulp was subject to chemical reduction using sodium borohydride ( R).

An alkaline heat treatment, (E), was used for extraction of lignin. The same procedure was used for degradation of cellulose when keto-groups were present in the cellulose. Pulp at 10% consistency, pH adjusted to 12.0, was subject to 60°C (water bath) for 1 h. The pulp was then washed until neutral pH was reached.

Additives (Fe2+, EG, methanol) were introduced in the A-stage, before the pH adjustment.

3.3 Measuring lignin: Kappa number The most commonly used measure of the residual lignin in pulps is the kappa number, which is based on oxidation of residual lignin with acidified permanganate2. Kappa numbers were determined according to SCAN- C 1:77. The kappa number is defined as the volume in ml of a 20 mM potassium permanganate solution that is consumed by 1 g oven dried pulp64. It is important to note that it is a measure of the total amount of substrates it is possible to oxidise rather than the actual lignin content. It has been shown that hexenuronic acids (originating from hemicellulose) contribute to the kappa number. Between 13 and 26% of the kappa number has been shown to be of non-lignin origin in industrial and laboratory-made pine kraft pulps15.

3.4 Measuring average cellulose chain length: viscosity Pulp viscosity was determined according to SCAN-CM 15:88. Viscometry is a method to determine the value of the intrinsic viscosity. The procedure for cellulose in pulp consists of dissolving the sample in copper ethylene diamine to a dilute solution. The viscosities are obtained from flow-through times (t and t0) of a defined volume of solution in a capillary viscometer: η t ≈ η0 t0 Equation 1 where t is the flow-through time and η is the viscosity. The pure solvent is indexed 0, while the sample solution carries no index. The intrinsic viscosity is given by extrapolating the concentration of pulp to infinite dilution to approach the state of zero entanglement of cellulose chains:  η−η  []η = lim 0  c→0 cη0  Equation 2 where c is the concentration of the polymer in the solution.

The intrinsic viscosity [η] is converted to the viscosity average molar mass according to the Mark-Houwink viscosity equation, Equation 3, where K and a are the empirical Mark- Houwink parameters 65. a []η = K * DP v Equation 3

14 In the present work, the degree of polymerisation (DP) of cellulose in pulps was calculated using the following variant of the Mark-Houwink equation: 1  η  0.76 DP =   for DP>950  2.28  1  η DP =  for DP<950  0.42 Equation 4 where η is the viscosity (ml/g) of the pulp66.

3.5 Molecular weight averages To describe the molecular weight distribution of linear polymers such as cellulose, several statistical measures are employed. Molecular size can be described by using the degree of polymerisation (DP): MW DP = MWrep Equation 5 where MWrep is the molecular weight and the index “rep” denotes the constitutional repeating unit (here: glycosidic unit).

The number average of chain length is defined in Equation 6, where nP is the number of molecules of degree of polymerisation DP of category i.

∑ n P (i) * DP(i) i DPn = ∑ n P (i) i Equation 6 The mass or weight average is given by Equation 7 where m(i) is the mass of chains of category i and nM the number of monomers in category i. ∑ m(i) * DP(i) i DP w = ∑ m(i) i 2 ∑n P (i) * DP (i) ∑ n M (i) * DP(i) = i = i ∑ n P (i) * DP(i)i ∑ n M (i) i i Equation 7 The viscosity average obtained by viscometry is defined in Equation 8 where a is a constant that takes values between 0.5 and 0.8 for different combinations of polymer and solvent 65.

15 1/ a  ∑ m(i) * DPa (i)   i  DPv =    ∑ m(i)   i 

1+a 1/ a  ∑ n M (i) * DP (i)   i  =    ∑ n M (i) * DP(i)   i 

Equation 8 Some related concepts: The averages are equal only for a perfectly monodisperse polymer. In all other cases the averages are different: MW n < MW v < MW w The viscosity average is normally quite close to the weight average 65.

16 4 Results

4.1 Scissions, random In chemistry, the amount of product (yield) in response to a reactant and a substrate is fundamental. Cellulose being a linear polymer and the process being degradation, the product is scissions along the chains, which is reflected by a change in polymer chain length. Equation 9 and Equation 10 are common ways of calculating a measure of the amount of scissions, s or k*t depending on tradition, using average polymer chain length and assuming random scission generation. Random scission generation means that all bonds in the polymer sample are equally probable to react.

DP0 1 1 s = −1 k * t = − DP1 DP1 DP0 Equation 9 Equation 10 Staudinger developed the hydrodynamic volume theory of Einstein into a relationship between specific viscosity and polymer chain length. The relation was semi-empirically improved by Kuhn, Houwink, Sakurada and Mark, who turned it into a relationship between intrinsic viscosity and polymer chain length, called by different names but herein called the Mark-Houwink equation, Equation 367. Numerous attempts have been made to explain the Mark-Houwink equation. This equation is often used to obtain the degree of polymerisation from the intrinsic viscosity of a polymer sample. It should be noted that the viscosity average defined in Equation 8 for an exponent equal to 1 is the same as the mass average and –1 the same as number average, Equation 7. Furthermore, the value of the Mark-Houwink equation must be an average value of polymer chain length, although this is not always explicit in literature. Concerning the meaning of exponent, the explanation based on the “rectitude” of chains in the measured state is sometimes suggested for the viscosity average. An exponent 1 is then related to the fact that the chain is straight and 0.5 that polymer chains curl into a random coil 68.

From the tradition of radiation degradation, we know the frequency of reacted monomers is calculable from the number average and the mass average by assuming the most probable size distribution (Shulz-Zimm)69,70. In the field of pulp and paper bleaching chemistry, cellulose degradation kinetics are analysed using the same formula, but derived differently. In this type of derivation, the number average is used and monodispersity is assumed71. If the formula is explained in articles, it is derived rather than referenced. The derivation seems to originate from the work of Ekenstam in the 1930s, but the discussion has been as lively as in the case of the Mark-Houwink equation72.

Facing the application of scission formulae in work on the chemistry of free radical mediated cellulose degradation during ozone bleaching, several problems were realised: • Experimental average values are predominantly viscosity averages, and not number averages. • Random scission generation presupposes that all cellulose chains are accessible to degradation. This is not in line with the effect of rapidly reacting chemical degradants and modern views of cellulose accessibility. • Free radical degradants do not spread through cellulose samples as other degradants, and the produced scissions should therefore not be directly compared.

17 • A cellulose molecular weight distribution is not monodisperse. • Without more information than averages as is often the case, no molecular weight distribution can be assumed with any precedence over others in terms of correctness. The use of derivations based on specific molecular weight distributions is hence limited • Conclusions need to be drawn, and these conclusions are limited by a vague control of mechanisms behind the change of chain length. Therefore, the subject of scission yield can scarcely be ignored by a chemist in pulp and paper science.

It is noteworthy that the scission formulae produce nice linear plots between what is predicted and what is observed, in spite of the fact that molecular weight distributions and the types of averages do not necessarily agree with the conditions for which the scission formulae were derived. The easiest thing to do in this situation is to study the effect of deviations from the conditions stipulated by a derivation, in terms of cellulose molecular weight distribution and type of average inserted in the formula. This is investigated in Paper I.

4.1.1 Investigation

4.1.1.1 Random scissions in chains Modelling the scissions within a chain, we know that 1 scission causes 2 chains, 2 scissions cause 3 chains, 3 scissions cause 4 chains and z scissions cause z+1 chains. By induction we

conclude that the number of scissions in a monodisperse sample is z = DP0/DP1-1 which

transforms into











DP1 = DP0/( z +1).

Pseudo-ring










Chain Degraded

chain

Monomer

Scission, z

Gap, g

Bond, b

Figure 12. The pseudo-ring is the hypothetical case of all possible bonds utilised (DP = 8, b = 8, z = 0, mM/MW, nP = 0, g = 0). In a chain, the ring has lost 1 bond (DP = 8, b = 7, z = 1, mM/MW = 8, nP,0 = 1, g = 1), which is the initial state. A degraded chain is a ring that has lost the first bond and some additional bonds, scissions (here: DP = 2, b = 4, z = 3, mM/MW = 8, nP,1 = 4, g = 1). The logic of the formula stems from the fact that if all bi-functional monomers in the sample had two bonds, no chain ends would exist. Only a chain shaped like a ring can satisfy this condition (Figure 12). The number “1” in DP0/(z+1) relates to the fact that a linear polymer is a ring that has received 1 scission. Degradation adds scissions to this first one. A similar explanation has been found in the literature69. The length of a degraded chain in a monodisperse sample can then be described by:

18 DP0 DP0 DP0 DP1 = = = n rM n rM f rM * DP0 + 1 * DP0 + 1 + 1 n G,0 n P,0

where n rM = f rM * n G,0

Equation 11

The amount of reacted monomers (or scissions) is nrM, the global frequency of reacted monomers frM and the total number of monomers nM,0 which is assumed constant. Applying the formula to each chain length of a polydisperse molecular weight distribution, a final molecular weight distribution is produced.

In order to do the reverse, the calculation of the frequency of reacted monomers, a way to reduce the information on the molecular weight distribution (a vector) to a single value (scalar) is needed. This is done by assuming the sample to be monodisperse (norm), the polydispersity of the true sample being a deviation from the norm. Equation 12 is obtained from combining the assumption with Equation 7. Equation 13 is the general form of Equation 12 and is correct for the number, mass and viscosity averages (d-values: –1 or 0.5-1):

DPw,0 − DPw,1 DPn,0 − DPn,1 DPv,0 − DPv,1 = = = frM DPw,1 * DPw,0 DPn,1 * DPn,0 DPv,1 * DPv,0 Equation 12

1 1  1  d  1  d f =   −   rM,  d   d   DP w,1   DP w,0 

Equation 13 Equation 13 was used to simulate the apparent number of reacted monomers for varied molecular weight distributions: monodisperse, even, diagonal, antidiagonal and step-function (Figure 13 to Figure 15). Equation 11 was used when the number of scissions per chain was larger than or equal to 1. If the number of scissions per chain was less than one, the number of reacted monomers equalled the number of polymer chains receiving one scission, which is the basis of Equation 14.

f rM * DP0 (i) < 1 →

n rM (i) = x * n P,0 (i)  DP1(i) = (1 − x) * DP0 (i) + x * DP1 (i) d d   f rM * DP0 (i)  DP1 (i) =  DP0 (i) * 1 −    2  Equation 14

19 2.5 1

2

0.1

1.5 nM0(i) rM* f M

n nM1(i)

1 0.01

0.5

0.001 0 0.001 0.01 0.1 1

0 100 200 300 400 500 frM DP

Figure 13. Simulation of random scissions Figure 14. Calculation of apparent random generation, frM =0.05, in a monodisperse sample of scissions generation, frM* from simulated initial DP 246, d=1 random scissions frM, in a monodisperse sample of initial DP 246.

0.014 diagonal

0.012 antidiagonal

step-function 0.01 monodisperse

0.008

nM,0(i) 0.006

0.004

0.002

0 0 100 200 300 400 500 DP

Figure 15. Examples of initial molecular weight distribution types. For all distributions and any number, mass or viscosity averages, the calculated value using Equation 13 and the simulated value resulted in nice linear plots. Linear regression through zero between the calculated and simulated frequency of reacted monomer displayed slopes of 1.0 and regression coefficients of 1.0.

4.1.1.2 The Mark-Houwink equation. The scission calculus is based on the assumption of two samples formed from different levels of random scissions in one common ring-shaped polymer. The degradation of ring-shaped polymers would allow all reactions to lead to scissions and the formation of monodisperse samples of chain lengths inversely proportional to the frequency of reacted monomers. In Figure 16, weight averages simulated for different levels of degradation, frM, are compared to the corresponding chain lengths in the sample if chain ends had not been present in the sample.

20 Figure 16 shows an increasing deviation from the ideal with decreasing degradation of linear chain samples. This demonstrates that chain-ends of the initial chain sample are equivalent to a decrease in the relative amount of reactive bonds. Principally, the deviation is expected to increase with decreasing average chain length. Figure 16 also shows that the deviation can be described by curves obtained by data treated with regression to power-formulae. The Mark-Houwink equation in the present state is recognised from Staudinger’s original version by being a power formula, Equation 3. The reason for the power law has been long debated67. What we have in Figure 16 is a possible explanation of the adjustment: The Mark-Houwink parameters not only cover for effects like chain-curl in a solution during viscometry but also allow for the error that stems from low levels of degradation in a sample containing an initial amount of chain ends.

120 2 1 2 0.75 2 0.5 2 0.25 2 -1 0 1 0 0.75 0 0.5 0 0.25 0 -1 -1 1 -1 0.75 -1 0.5 -1 0.25 -1 -1 1 1 100 1 0.75 1 0.5 1 0.25 1 -1 Ideal 1/d ) d 80

y = 1.1482x0.9016 60 R2 = 0.9965

40 (Weight average DP y = 1.4376x0.6884 20 R2 = 0.975

0 0 50 100

DPMD

Figure 16. The simulated weight average chain length of a linear chain sample compared to the (monodisperse) chain length from one ring-shaped chain. Both samples were subject to the same degradation-value, frM. The first digit of the series name: 0 = even, -1 = anti-diagonal, 1 = diagonal, 2 = monodisperse. The second digit of the series name is the exponent d used for average calculus in the simulation: 1 = weight average, 0.75-0.25 = viscosity average, -1 = number average. Using a power- equation calibrated for a sample of linear polymers subject to random scission degradation, the frequency of a reacted monomer of any type of molecular weight distribution and average is: 1 1  − −  d d*c d d*c  DP w,1   DP w,0   f     rM* =   −     k1   k1     Equation 15 k1 and c are constants to be determined by measurement.

4.1.1.3 Summary General support has been given to a frequently challenged type of calculus - any molecular weight distribution subject to random scission generation is analysable using any of number,

21 mass or viscosity average. Scission calculus depends on the Mark-Houwink equation and the values are no more exact than the calibration of the Mark-Houwink equation, specific to initial molecular weight distribution and random scission degradation. The need for calibration applies to the s- or k*t-formulae and the use of number, mass and viscosity averages and the like. Throughout this thesis, scission values are calculated using the s-formula, Equation 9. The s- value is proportional to the true number of scissions (the frequency of reacted monomers) and gives directly the average number of scissions per chain.

4.2 Scissions, partial A second problem with the chemical cellulose degradation analysis is found when inspecting common cellulose structure models. On one hand the formulae used, Equation 9 and Equation 10, are derived under the assumption of random scission generation. On the other hand, the low cellulose accessibility suggested by the models indicates that the induced scissions will not be located randomly but in accessible cellulose. What cellulose structure models do suggest is that the accessible surfaces of cellulose are more severely fragmented during chemical cellulose degradation than bulk cellulose when exposed to rapidly reacting agents. In terms of the work we presented for random scissions, Section 4.1 and Paper I, the difference is that the scissions or the frequency of reacted monomers in the case of cellulosic materials and chemical agents is not a global but a local parameter. However, we do need the scission perspective in order to ascribe a chemical meaning to changes in average chain length. Paper II sets out to investigate the consequences of local degradation upon viscosity averages and the scission formulae (using the s-form) and formulate a model of how to interpret the data.

4.2.1 Investigation The influence of cellulose accessibility on viscometry of cellulose degradation was studied experimentally and theoretically. The systems were based on mixtures of cotton linter samples. Native cellulose was mixed with fragmented cellulose (gamma radiation, 60Co) and the result was observed using viscometry. First, the sample of solely degraded cellulose (homogeneous) was studied, Figure 17. The linearity of the scissions formula was confirmed. In addition, the Mark-Houwink exponent was determined to be 0.69-0.76 and the radiation chemical yield (G-value) was determined to G(scissions) 0.6*10-6 moles J-1 (5 changes per 100 eV), in accordance with the literature.

22 1600 4

2 1400 R = 0.9995 3.5

1200 3

1000 2.5

800 2 s

600 1.5 viscosity (ml/g)

400 1

200 viscosity-series 0.5 scission-series 0 0 0 5 10 15 t(h)

Figure 17. Viscosity and scissions calculated from the viscosities of cellulose sample (x=1, homogeneous sample) plotted versus irradiation time tirr. The observed degradation from the mixture should be described by Equation 16, where x is the mass fraction. Indices 1 and 0 refer to the native and degraded sample respectively while measured values are indexed using “*”. Intrinsic viscosity and scission values are then expected to follow Equation 17 and Equation 18. a a a DP w* = DP w,1 * x + DP w,0 * (1 − x) []η * = []η 1 * x + []η 0 * (1 − x)

Equation 16 Equation 17

1 − −a a s* = ()x * ()s1 + 1 + (1 − x) −1

Equation 18 The experimental results, Figure 18 and Figure 19, are in overall agreement with the values calculated using Equation 17 and Equation 18. However, deviations arose at low mass fractions. A mass fraction of ≤ 0.25 of degraded cellulose managed to change viscosity and scissions marginally while ≤ 0.1 had no effect. Such behaviour was not predicted from theory.

23 5 1600 4.5 1 0.75 1400 4 0.5 3.5 0.25 1200 3 0.1 2.5 1000 2

Observed scissions 1.5 x = 0, initial viscosity 800 x = 1 1

x = 0.75 Observed viscosity (ml/g) x = 0.5 600 0.5 x = 0.25 x = 0.1 0 400 012345 1600 1400 1200 1000 800 600 400 scissions, degraded fraction Viscosity (ml/g), degraded fraction

Figure 18. Observed viscosities (ml/g) from samples Figure 19. Observed scissions from samples with with a mass fraction x of degraded cotton linters mass fraction x degraded cotton linters mixed with mixed with undegraded cotton linters of viscosity undegraded cotton linters. 1500 ml/g. Equation 17. Chemical degradation of partly accessible cellulose should thus be evaluated according to Equation 19 for any initial molecular weight distribution, where na is the amount of degrading agents that have been consumed and nM,accessible the amount of accessible monomers. Less than 25wt% accessible cellulose should result in an underestimate of scissions and/or a decrease in average chain length of the degraded fraction. 1 −a − n = n     a a rM   na * DPv,0   → s* = x *  + 1 + (1 − x) −1 n = n * x   n * x   M,accessible M,tot    M,tot   Equation 19

4.2.1.1 Summary Surface attack is the natural hypothesis for fast reagents reacting with cellulose in fibres. The reaction occurs in proportion to substrate accessibility and proximity. By bypassing the difficulties of varying surface to bulk ratio in cellulose of cotton linters by mixing degraded and undegraded cotton linters, the degradation of cotton linter rapid agents has been shown. The detection limit of viscometry is strongly dependent on cellulose accessibility. Reported cellulose accessibility or crystallinity values, Section 2.3.4, vary broadly. A conclusion supported by both the higher and lower accessibility estimates of cellulose in literature is that the degradation from the surface attack of a fast reactant should be difficult to observe using viscometry. This is both in terms of the interval where observable degradation does respond to an increased reactant charge, as well as the level of maximal degradation possible. Based on the lowest accessibility values, the conclusion is that the surface degradation is not only limited but cannot be observed by viscometry at all. Hence, the changes in viscosity during pulping processes might not be attributable to rapid agents (however degrading). This concerns especially TCF bleaching, where a general reference to “free radicals” is used to explain degradation.

24 4.3 Free radical mediated cellulose degradation Upon reasoning around the nature of free radical chains in a partly crystalline carbohydrate, one realises that the migration of the free radical intermediaries during free radical chain reactions determines the rate of spreading of the attack from the initial radical site in the fibre. Water is confined to amorphous carbohydrates, i.e. hemicellulose, accessible fibril surfaces and fibril segments with distorted crystal order and other fibre constituents except crystalline cellulose. We assume that the reach of water soluble reactants is to the same extent as water itself, deviating from water in reactivity and size54,61,73. Thus degradants reach amorphous carbohydrates and not the para-crystalline/crystalline bulk of cellulose.

There are at least two ways in which migration of intermediaries of free radical chain reactions can govern cellulose degradation in the fibre, Figure 20: If the migration distance is short, the chemical attack is concentrated to smaller zones over the fibre wall cross-section (“hot spots”). The zone size allows inter-fibrilar movement of the radical site and does not exclude zones extending radially over several fibrils. If the migration distance is long, free radical intermediaries may penetrate into crevices, for example those “amorphous regions in cellulose” that are sometimes used to explain degradation.

The mobile intermediaries in the system are the hydroxyl radical and the hydroperoxyl radical or superoxide anion, Figure 7. An estimate in one dimension for a homogeneous mix of carbohydrate and water of the migration of these two radicals can be obtained from: 4 ⋅D r = k ⋅c Equation 20 where r is the distance (m), D is diffusivity in m2s-1, k the first order rate constant M-1s-1 and c the coreactant concentration (M). D is in the range 10-9-10-10 m2s-1 for most solutes in water. In a system containing 10 g glucose, 40 wt%, r < 0.1 nm is obtained for a hydroxyl radical. An intermittent migration of 60 nm is estimated for the superoxide ion at pH 3, using half the ozone solubility in water as a basis. From this we conclude that spatial spreading of free radical chemistry is governed by the half-life of the hydroperoxyl radical or superoxide ion.

25 1

3a

Long 0

3b

2

Figure 20. Different scission distributions. 0: Structured cellulose, 1: Surface attack, 2: Initiation of free radical chain reactions, 3a: Fast propagation of chain reactions, 3b: Slow propagation of chain reactions. From our estimate, migration distance should be short. We predict that the oxidation products (carboxylic acids, scissions, keto-groups) of degradation should exhibit a concentric type of development. “Amorphous regions in cellulose” should not be important.

4.3.1 Spreading of free radical degradation in cellulosic substrates

4.3.1.1 Investigation: Development of free radical degradation in carbohydrates The technique exploits the fact that the oxidation products of cellulose are partly carboxylic acids 74,75. After ozonation, the distribution of carboxylic acids can easily be detected, marking the spreading of oxidation. An iron(II)-solution was added to a central spot of a filter paper, after which the paper was ozonated. By adding pH-indicator, it was clearly distinguishable that acids were formed in a zone growing outwards from the iron spot (Figure 21). The possibility of the zone originating from direct ozone oxidation could be ruled out by comparing the effect of ozonation on an iron(II)-doped filter paper, on which the ferrous ions were found to remain on the introduction spot. This rules out migration of ferrous ions as the origin. The observation of zone growth fits the expectation of concentric spreading of degradation reactions.

26 3 1

2

Figure 21. Ozonation of iron(II)-doped filter paper, treated with pH-indicator solution. Iron spot (1). Oxidation zone in the middle, red-yellow color (2). Non-oxidised area, blue-green color (3). The increase of the zone size with time is explained by an induction period where ozone predominantly reacts with ferrous ions until a steady state is reached. The steady state allows chain reactions to be propagated by ozone. Equation 20 predicts that at a constant ozone concentration, the zone radius will increase in proportion to ozonation time. This is observed in Equation 20. The absence of oxidised “islands” on the filter paper is interesting76. If the transport of HO2 in particular had occurred through the gas phase, small “islands” of oxidation would have been observed.

0.40 y = 0.0019x + 0.0316 2 0.35 R = 0.9928

0.30

0.25 0.5

0.20

m(g paper) 0.15

Initial zone size 0.10

0.05

0.00 0 50 100 150 200 Ozonation time(min)

Figure 22. Area of the oxidation zone area with degree of ozonation Fe(II)-solution was introduced to a cotton linter pulp which was then ozonated during high consistency pulp bleaching conditions, pH 3. Using ESEM (Environmental Scanning Electron Microscopy), degraded fibres were observed to exhibit a porous structure formed in response to ozonation (Figure 23 and Figure 24). Regrettably, however, we do not have pictures of ozonated fibres without iron for comparison. We can conclude that fibre agrees with degradation in zones, no matter what the chemical mechanism.

27 Figure 23. Undegraded cotton linter fibre, Figure 24. Cotton linter fibre degraded in a scale of 15 µm. ozone-Fe(II), scale of 10 µm.

4.3.1.1.1 Summary We demonstrated that free radical reactions develop differently in cellulosic materials than electron-pair chemistry, like for example in hydrolysis, and we found macroscopic evidence of concentric, stepwise spreading in agreement with predictions. Therefore, the conclusion is that data from free radical mediated degradation should be evaluated using different principles than what is common. Furthermore, some of the necessary principles have been uncovered.

4.3.1.2 Investigation: Cellulose structure effects on free radical cellulose degradation

In order to initiate and control free radical chemistry, an initiator, the ferrous ion (Fe2+), was

added. Figure 25 shows the chemical



























































system employed.

Initiation
Chain

reactions
reactions HO2

1.
2.

H+

- ds

Fe2+ O3
O2

Cellulose chain

FeO2+ cleavage





OH•

Fe2+ WATER

3+

Fe Ox.

products

FIBRIL

H-R(OH)
•R(OH)

•OO-R(OH) Keto-group

(cellulose)

formation

O2

als Figure 25. The chemical system Two extreme reaction conditions are of special interest in this work: Charging Fe2+ in excess 2+. to O3 and charging O3 in excess to Fe . When O3 is charged in excess, most of the ferrous ions are converted to ferryl ions. The ferryl ions abstract hydrogens from glycosidic units at accessible fibril surfaces. We thus have free radical chain reactions proceeding in proportion to the amount of O3 remaining when the

28 ferrous ions have been consumed. With Fe2+ charged in excess, all ozone is consumed and there is a small time window for free radical chain reactions. The degradation that could occur is from ferryl ions abstracting hydrogens from glycosidic units at accessible fibril surfaces.

2.0 4.0 ds experimental series als comparison ds comparison 1.6 3.0 1.2 2.0 ds

dsor als 0.8

1.0 0.4

0.0 0.0 1E-06 1E-05 1E-04 1E-03 1E-02 1E-6 1E-4 1E-2 moles Fe (II) added moles ozone added

Figure 26. “Fe(II) constant” series. Degradation of Figure 27. “Ozone constant” series. Degradation of cotton linters at high consistency, pH 3, Fe(II) charge cotton linters at high consistency, pH 3, constant O3 constant = 1 mmole, varying the ozone charge. charge = 8.3*10-4 mole, varying the Fe(II) charge. -4 Comparison: 8.3*10 mole O3, no Fe(II) added. Comparison: 8.3*10-4 mole O3, no Fe(II) added.

1.00 out

3 0.10 ( ) %O

experimental series comparison Figure 28. The percentage of ozone leaving the 0.01 reactor. “Ozone constant” series. Cotton linters at -4 1E-06 1E-05 1E-04 1E-03 1E-02 high consistency, pH 3, constant O3 charge = 8.3*10 mole, varying the Fe(II) charge. Comparison: 8.3*10-4 moles Fe (II) added mole O3, no Fe(II) added. In parentheses: value below detection limit. Figure 27 shows the ds (direct scissions) and als (alkali induced scissions) from cotton linters 2+ given a constant charge O3 and a varying charge of Fe . The curve Figure 27 can be divided into two parts around the degradation maximum. The first part, 10-6-10-4 moles Fe2+ added, is limited by the initiation step. Initiation decreases when moving from the maximum in the direction of decreasing Fe2+ charge (R7). Very little ozone is consumed in this interval, thus the ozone concentration can be taken as constant within the interval.

29 -4 -2 2+ The second part, 10 -10 moles Fe added, is limited with respect to the supply of O3, which decreases when moving from the maximum in the direction of increasing Fe2+ charge. This is accompanied by an increase in ozone consumption, as seen in Figure 28. The ratio of initiation to termination (R7, R9) is unknown in this region. The degradation is depressed, as there is not enough O3 for the free radical chain reactions to gain from the increased initiation. Degradation at 10-6 mole charge Fe2+ is similar to the degradation of pure sample (no Fe2+ added) and constitutes the background degradation (R11, R12). The low ozone consumption means high degradation efficiency with respect to ds or viscosity drop. This description is also valid for the background degradation. This is discussed in Section 5. The minimum is lower than the background level. This indicates that the termination reaction (R9) is more rapid than the hydrogen abstraction reaction (R7) when the amount of Fe2+ is around 10-2 moles. Maximum degradation (zero limitation of the degradation) is expected when ozone is fairly equally spent on ferryl ion production (initiation (R7, R8)) and propagation of the free radical chain reactions (R15, R16). In the curve we interpolate the maximum to occur at 3*10-4 mole Fe2+ added.

A second representation of the nature of the free radical chain reactions on cellulose degradation is given in Figure 26. No degradation (0 ds) was observed unless O3 was charged in excess to Fe2+, ≥ 1 mmole. 2+ When O3 is charged < 1 mmole in Figure 26, the reaction between FeO and carbohydrates, R8, could theoretically be quenched by the competing reaction Fe2+, R9. Yet, we observed 2+ degradation for the initial reactant charges of O3 and Fe 1:1, Figure 26 and Figure 27. As this means that at least some ferryl ions must have reacted with carbohydrates, the same must also be the case when ozone charges are decreased. The observation indicates that reaction R8 is favoured. This is possibly an effect of ferrous ions being adsorbed onto accessible hexoses 51. As free radical chain reactions are disfavoured in this interval, the observed degradation should reflect solely the effect of ferryl ions reacting with accessible carbohydrates in cotton linters. As we did not observe any degradation in this interval, we can conclude that the cellulose accessible to attack in the sample is too small a fraction to be observed viscometrically.

Figure 29 shows the corresponding results for free radical degradation of cellulose beads for a similar system as used for cotton linters, above.

30 0.5 0.2 0.4

0.3 ts ts

0.1 0.2

0.1

0.0 0.0 0.1 1.0 10.0 0.1 1.0 10.0 2+ 2+ n(O3) *n(Fe ) n(O3) / n(Fe )

Figure 29. Left: Ozone degradation of cellulose beads, pH 5 and a fixed molar sum, 0.4 mmoles reactant 2+ added per g cellulose bead, varying the molar ratio O3 to Fe . Right: Ozone degradation of cellulose 2+ beads, pH 5 and a fixed molar ratio, 1:1 O3 to Fe , varying the absolute amounts of degradants: 0.2, 0.4 and 0.8 mmoles reactant added each per g cellulose bead. The explanation used for free radical degradation of cotton linters relies strongly on the nature of crystalline cellulose. Then logically, the degradation in non-crystalline cellulose should be in proportion to the free radicals formed in the system. Figure 29 shows that degradation was maximal when reactants were charged 1:1 in a series where molar ratios of the reactants were 0.5:2, 1:1, and 2:0.5. Degradation was limited by the 2+ lowest charged reactant. The response was similar irrespective of whether O3 or Fe was in excess. These observations are in keeping with reaction R7: 2+ 2+ O3+Fe → FeO (R7) According to reaction R7, the FeO2+ yield should be optimal when reactants are charged 1:1, be limited by the lowest reactant charge and depend to the same extent on both reactants. In this context it should be noted that a degradation that is governed by free radical chain 2+ reactions would exhibit higher dependence on O3 than on Fe , while a degradation governed by direct attack by ozone on cellulose would appear invariant with the Fe2+ charge.

4.3.1.2.1 Sub-study: ds-als The ds-als (direct scissions/alkali induced scissions) concept is an approach for interpreting measured final viscosities of pulps. Oxidised cellulose is labile in an alkalinic environment. From this lability, some information can be extracted on the nature of degradation chemistry. Direct cleavage of the glycosidic bond follows oxidation of carbons C1 and C4 (Figure 30). Oxidising carbons C2, C3 and C6 leads to keto-groups at these positions. The keto-groups cause cellulose chain cleavage of the glycosidic bonds when conditions favour the β-alkoxy elimination mechanism77. Degradation occurs during viscometry as the alkaline conditions favour β−alkoxy elimination78,79.

31 Attacked positions Scission categories

Q ds direct scissions O als alkali induced scissions (ß-alkoxy elimination)

HO•

HO• O O O O HO• R R R O HO•, O R 3 O HO• HO•

Figure 30. Reactive positions and effect of glycosidic unit reactions. The ds-als is as follows: An E-stage (alkali and heat) imposes β−alkoxy elimination. The sample thus obtained reflects the overall oxidation of cellulose. The cellulosic material then contains cellulose chains subjected to both direct scissions and the additional scissions from keto-groups (C1, C2, C3, C4, C6). The R-stage reduces the keto-groups formed. The viscosity measured of the same sample from an RE-stage thus relates to direct scissions from cellulose oxidation (C1, C4). The total number of scissions (ts) is calculated from the viscosity of the E-sample. The number of direct scissions (ds) is calculated from the viscosity of the RE-sample. An estimate of the number of scissions from keto-groups is given by als = (ts-ds).

In our system, the reactants initiating free radical chain oxidation by hydrogen abstraction are the hydroxyl radical and the ferryl ion, the latter reacting several orders magnitude slower (see Section 2.3.1.1). Ozone degrades by driving chain reactions, the direct degradant being the hydroxyl radical. It should be noted that if ozone initiated free radical chemistry by some mechanism, the effect of ozone would be difficult to distinguish from that of the hydroxyl radical. Ozone is most commonly suggested to oxidise C1 by a non-radical mechanism (See 5.2.1). Free radical chemistry would oxidise all positions in the glycoside (ds and als) through the hydroxyl radical intermediary. Ozone would only contribute to ds, by the non-radical mechanism. Then the reactant solely responsible for als would be the hydroxyl radical. For the hydroxyl radical, we assume 2/6 of the attack on C1 and C4 and 4/6 on the remaining carbons and 5/6 of the attack leads to direct and indirect scissions. Thus, 2/5 of the attack should yield direct scissions while 3/5 of the attack should be convertable to scissions in the presence of alkali and heat. In Figure 31 the relative ratio of ds to ts is determined to be 0.4. This agrees very well with what has been predicted for the hydroxyl radical reactivity towards a glycosidic unit. It does not support a mechanism where ozone reacts non-radically.

32 2.5

2

1.5 ds1/ds0 1

0.5 y = 0.3786x R2 = 0.9818

0 02468 ts1/ts0

Figure 31. ds vs ts in the 106-104 mole Fe(II) interval in the “Ozone constant” series In the “Ozone constant series” (Figure 27), the samples were limited with respect to either the 2+ initiation (Fe ) or the propagation (O3). The following responses are expected from the assumption of localised evolution of free radical chain reactions: • From the point of maximum degradation and in the direction of decreasing iron charge, the number of initiation points from which free radical chains reactions extend decreases. The ratio between direct (ds) and indirect scissions (als) should change little as the chain length of free radical reactions should be almost constant. The sum (ds+als = ts) should decrease since the number of points from which free radical chains reactions extend decreases. • From the point of maximum degradation and in the direction of increasing iron charge, the free radical chain length decreases. The difference between als and ds should decrease with the decreasing free radical chain length. Figure 27 shows that the ds and als curves can be described as parallel in the 106-104 mmole Fe(II) interval, in agreement with the first point. Figure 27 also shows that the difference between ds and als approaches zero from the point of maximum degradation to 10-2 mole added Fe2+, in agreement with the second point. At 1 mmole Fe2+ added in Figure 27, the ratio als to ds appears to decrease. This could be due to a difference in reactivity of ferryl ions and hydroxyl radicals.

33 0.45

0.4 als 0.35 0 ds 0.3

/scissions 0.25 1

0.2

0.15

log scissions 0.1 y = 0.1621x 2 0.05 R = 0.9861

0 0 0.5 1 1.5 2 2.5

log nFe(II),1/nFe(II),0

Figure 32. Als and ds in the 106-104 mole Fe(II) interval in the “Ozone constant” series The slopes of the ds- and als-curve are linear (the regression coefficient 0.99 for ds) in the 106-104 mole Fe(II) interval of the “Ozone constant” series, on log-log scale (Figure 32), and ∂ log D the slope-values are in both cases 0.2 (= ), see Section 5.3 ∂ log nFe(II)

4.3.1.2.2 Summary Ferryl ion formation occurs both in the cotton linter and cellulose bead systems. In cellulose beads, cellulose degradation is a direct consequence of ferryl ion formation. In cotton linters an additional step is needed, the propagation of free radical chain reactions by O3. The importance of free radical chain reactions observed in cotton linters therefore appears to be related to crystallinity. In light of viscometry of partial degradation and cellulose accessibility, the inefficiency of the ferryl ion degradation may be explained by the amount of accessible cellulose being too low for the degradation to be observed by viscometry, Sections 2.3.4 and 4.2. As such, the observed inefficiency has general implications; surface attack is not an efficient mode of degrading cellulose (or changing intrinsic viscosity of cellulose samples). The efficiency of free radical chain reactions indicates that the mechanism involved is distinct from ferryl ion attack. This is now not only theoretically but also experimentally clear from Section 4.3.1. Little is known on how the concentric evolution pattern of free radical chain reactions influences viscometry. Given that efficient degradation necessitates the degradation to work itself into the cellulose bulk, repeated reactions in the same area, see 3a in Figure 20, is a possible mechanism. Free radicals formed in the initiation step need to be transferred to accessible cellulose (surface attack). In the propagation step, the radical reaction has to be propagated to reach inaccessible cellulose (bulk attack). The first step creates a strongly inhomogeneous scission distribution (fibril surface), while the second provides scissions partly penetrating the bulk in the proximity of the first free radical site formed.

4.4 Cellulose degradation in pulp There are several ways for free radical chemistry to be initiated during pulp ozonation (Section 2.3.1). Once free radical chemistry has started, it is central to cellulose degradation. In this context, radical scavenging is a way to control degradation. Many additives have been

34 tried, sometimes motivated by radical scavenging but often also for selectively changing solvatisation properties. Additives like methanol, ethanol and tert-butanol have been found to increase the selectivity of ozone bleaching. They may act both as scavengers and solubility regulators80-87. Some data on the effect of ethylene glycol (EG) addition have also been reported88,84,82.

The change of terms for free radical mediated cellulose degradation from cotton linters to pulp is depicted in Figure 33. Set in the context of the cotton linters experiments, it is interesting to note the impact of lignin, hemicellulose and alcohol upon cellulose degradation. The relative content of cellulose in among carbohydrates and alcohols in the system decreases. The amount of ozone decreases through consumption by lignin.

4.4.1 Investigation: Influence of lignin and alcohols upon ozone degradation of pulp

OH 1. Delignification

H+

HO2 Chain Initiation + O- reactions reactions H 3. 2. - ds O3 O2 Mn+, H-R(OH) Cellulose chain I(O)• OH• cleavage WATER I(OH) Ox. products FIBRIL

H-R(OH) •R(OH) •OO-R(OH) (cellulose, Keto-group hemicellulose, formation ethylene glycol) R ? O + H O2 •OO R als Figure 33. Reaction scheme, oxygen bleached kraft pulp+EG system. The wt% of EG was varied to find the optimal charge EG for ozone bleaching at pH 3. The selectivity was studied by varying the ozone charge at optimal EG charge. The data were compared to the corresponding data for methanol, which is known to enhance selectivity in ozone bleaching . The pH was varied in order to study the influence on the mechanisms and to determine the optimal pH for bleaching at the optimal EG charge. Fully bleached pulp was used to eliminate lignin effects when addressing cellulose degradation aspects during ozone bleaching of pulp.

35 1000

950

900

850 AZE

800 Viscosity (ml/g)

750

AZE 700 AZE, 25%w ethylene glycol in media AZE, 25%w methanol in media

650 0 5 10 15 Kappa no

Figure 34 The selectivity of EG and methanol at pH 3. 0.004, 0.006 and 0.008 g O3/g odp. Oxygen bleached kraft pulp 40% consistency.

6.5

AZE AZRE

6

5.5

Final kappa no 5

4.5

4 0.10% 1.00% 10.00% 100.00% wt% ethylene glycol in media Figure 35. The wt% EG vs final kappa at pH3. 0.004 g O3/g odp. Oxygen bleached kraft pulp 40% consistency.

36

980

AZRE

960

AZE

940

920

900

880

Viscosity (ml/g)

860

840

820

800






























































































































0.10% 1.00% 10.00% 100.00% wt% ethylene glycol in media

Figure 36. The wt% EG vs final viscosity at pH3. 0.004 g O3/g odp. Oxygen bleached kraft pulp 40% consistency.

1000 15 900 13 800 Initial viscosity 700 Final viscosity AZE, 25 wt% EG 11 Final viscosity AZE 600 Final viscosity AZRE, 25 wt% EG Final viscosity AZRE 500 9 Initial kappa Final kappa AZE, 25 wt% EG

400 Kappa no Final kappa AZE

Viscosity (ml/g) 7 Final kappa AZRE, 25 wt% EG 300 Final kappa AZRE 200 5 100 0 3 1611 pH

Figure 37. The pH vs final viscosity and kappa number, with and without ethylene glycol. 0.004 g O3/g odp. Oxygen bleached kraft pulp 40% consistency.

37 0.6 80 Fully bleached pulp, 25 wt% EG 75 0.5 Fully bleached pulp, 25wt% 70 methanol Fully bleached pulp 65 0.4 Oxygen bleached kraft pulp 60

0.3 55 EG 50 blank 0.2 45 O3 consumed (mg) Methanol 40 0.1 35 30 0 30 50 70 90 ds als tot O3 charge (mg)

Figure 38. Comparison of degradation in fully and Figure 39. Ozone consumption of pulp with 25 oxygen bleached kraft pulp, 0.004 g O3 / g odp, pH wt% EG and 25 wt% methanol at pH 3. 0.004, 3. 0.006 and 0.008 g O3/g odp. Oxygen bleached kraft pulp 40% consistency.

pH Ozone consumption (mg) 25 wt% EG blank 1 39 38 3 36 37 6 37 37 9 39 39 12 38 38

Table VI. The dependence of pH on ozone consumption with and without 25 wt% EG. Oxygen bleached kraft pulp at 40% consistency pH 3 and 0.004 g ozone / g odp. Figure 34 to Figure 39 and Table VI show the results on kappa and viscosity of varying the parameters ozone charge, pH and EG. The pulp had to be charged with ≥25 wt% EG to influence the selectivity of the ozone delignification, as seen in Figure 36. The addition of EG did not change the delignification in terms of final kappa number (Figure 35) unless very high concentrations of EG were used (≥75 wt% EG). The increased selectivity observed using EG thus originates from a decrease in the degradative processes rather than an increased delignification. The pH was varied from 1 to 12. The final viscosity showed negligible dependence on pH, whereas the delignification in the pH interval 3-6 decreased to a level of about 1/3 the initial value at pH 3 (Figure 37). An amount of 25wt% EG yielded a higher selectivity than the corresponding amount of methanol.

38 Ozone consumption appeared to be proportional to lignin content, unless very high charges of EG were used. Some ozone consumption from the reaction with EG is likely to appear at high charges. Applying similar conditions to fully bleached pulps, pH3 and 0.004 g/g odp, ozonation with or without 25 wt% EG or methanol, resulted in similar degrees of degradation. EG has been shown to decrease viscosity of polymer solutions. A 10 ml/g decrease in pulp viscosity was observed when the standard washing procedure was omitted, which is within the error limits of the method. Addition of EG to pulp did not change the pulp viscosity when the standard washing procedure was followed.

4.4.1.1 Discussion

4.4.1.1.1 Lignin influence on final viscosity The following notes from the literature refer to the progress of delignification through a fibre: After kraft cooking, the distribution of lignin in a fibre is not homogeneous. The lignin content is typically 5 times higher on the fibre surface compared to the average fibre content89,90. An oxygen stage delignifies more efficiently in the fibre bulk than on the fibre surface. Ozone reacts indiscriminately between fibre surface lignin and bulk lignin90. The lignin surface coverage is also suggested not to be uniform but patched, containing lignin- enriched surface zones intervened with lignin-poorer zones90. “Nucleic growth” and “Shrinking core” are two ways to model ozone delignification. A postulate of the latter, a well-defined reaction front traversing the fibre wall, has found support91. A consequence of these models and observations is that delignification will precede cellulose degradation, due to the high rate constant for the lignin-ozone reaction.

These points should be considered when discussing the following observations: If the phenolates are related to the observed cellulose degradation, a response to a phenol/phenolate acid-base equilibrium should have been observed when varying the pH. The final viscosity changed negligibly by varying the pH. For a given ozone charge, the ds increased when going from a pulp of a high to a low lignin content (oxygen bleached sulphate pulp kappa 10 and fully bleached pulp), Figure 38. This indicates that the lignin reactions primarily disfavour cellulose degradation, in agreement with a limiting effect of lignin on the ozone supply for free radical chain reactions.

At the time of Paper V, the discussion in support of free radical degradation of cellulose was focused on the nature of phenol initiation reactions. It was thus a surprise to find that the degradation was considerable in the absence of lignin and that the response to the variation of pH was difficult to justify from phenol initiation. The facts are that: • Lignin in oxygen delignified kraft pulps is not expected to contain high amounts of phenolic groups. • Ozone seems perfectly able to initiate free radical chemistry without the presence of phenolic groups (Paper VI). • The hypothesis of phenolic initiation is unspecific on the relationship between formation of free radicals and cellulose degradation.

However, the lack of correlation between delignification and cellulose degradation versus variation of pH is even more fundamental. It indicates that cellulose degradation occurs in lignin-free regions. Delignification in the reaction front is a feasible way to envisage ozone to cellulose direct contact in lignin-free regions. Once the reaction front has passed through half the fibre wall, the remaining ozone charge (1/2) will pass lignin-free cellulose. Our results

39 from in Section 4.3.1.1 show that free radical chain reactions will not terminate as long as there is ozone flowing into the reactor.

4.4.1.1.2 The influence of ethylene glycol on final viscosity Figure 37 shows that EG improves ozone bleaching selectivity by means of decreasing cellulose degradation rather than increasing delignification. The optimal charge is 25 wt% EG at pH 3. The ability of ethylene glycol to react directly with ozone is presumably the reason for the increase in final kappa for EG charges ≥75wt% (Figure 35).

EG and glucose have similar rate constants versus hydroxyl radicals (Table III), thus a competition for hydroxyl radicals should occur at ≈20 wt% EG (homogeneous solution rate constants). This is in agreement with the fact that the final viscosity increases in the interval 10 –25 wt% EG as seen in Figure 36. No appreciable difference in the final viscosity was observed when the pH was varied. Hence, no indication of the acid / base catalysed water expulsion (EG and glycosidic units) described in Section 2.3.3 was observed. This is most likely due to rapid oxygen addition outcompeting the water expulsion step. Thus, formylmethyl radical formation (EG) and “formylmethyl”-like mechanisms (cellulose, hemicellulose) are unimportant during ozone bleaching, pH 1-12.

The relative merit of EG over methanol cannot however be explained by hydroxyl radical scavenging alone. From rate constants, Table III, we find: 1) 25 wt% methanol should react with hydroxyl radicals to a similar extent as 25 wt% of EG; and 2), 25 wt% methanol should give a shorter free radical chain length, nO3 consumption/ninitiation, than 25 wt% EG due to inefficient superoxide expulsion (See Section 4.4.2). As a hydroxyl radical scavenger, methanol should give equal or better selectivity than EG, which is not the case as seen in Figure 37. The selectivity relationship between EG and methanol could be explained by fibre swelling but the subject will be not further elaborated upon in this thesis.

40 4.4.1.1.2.1 Ethylene glycol and ds

0,16 0,3

ds residual 0,14 0,25

0,12 als OH 0,2 0,1 0,15 0,08 0,1 scissions scissions 0,06 0,05 0,04

0 0,02 0,1 1 10 100

00,1 1 10 100 -0,05 wt% EG wt% EG

Figure 40. Als and ds with increasing wt% EG. Figure 41. Theoretical amounts of degradation attributable to free radical chain reactions, OH, and the residual. From Figure 40: OH = 5/3 * als and residual = ds – 2/3 * als

Ambiguities concerning whether or not the non-radical direct attack of ozone on glycosidic units contribute to degradation existed at the time of writing of Paper V, and it was held possible that the non-radical direct attack complemented free radical chain reactions. The non-radical direct attack appears less central to cellulose degradation from Paper VI and Section 5.1.2.1. Figure 40 displays the als and ds values with increasing wt% EG charge. The hypothesis, Section 4.3.1.2.1, predicts that only hydroxyl radicals should relate to als. It should then be possible to estimate the ds value on the basis of als values. Theoretically free radical chain reactions should produce a ratio of ds to als of 2/3. If ds values are estimated from the experimental als values, ds = 2/3*als, and these values are subtracted from the experimental ds values, the residual should relate to the degradation from other sources, e.g. direct ozone attack. Figure 41 shows that the number of scissions not accounted for by free radical chain reactions is negligible.

4.4.1.2 Summary • 25 wt% EG is the optimal charge for ozone bleaching of an oxygen bleached kraft pulp at pH 3. • The selectivity for 25 wt% EG was better than for 25 wt% methanol. • Cellulose degradation in oxygen bleached kraft pulp responded negligibly to variations in pH. Therefore, pH-dependent reaction mechanisms appear less important to the observed cellulose degradation.

With respect to free radical chain reactions: • Lignin limits free radical mediated cellulose degradation during ozonation by competing for ozone. • EG and hemicellulose limit free radical mediated cellulose degradation during ozonation by being additional substrates for free radical chain reactions.

41 4.4.2 Investigation: Free radical chain reactions in ozonised aqueous solutions of alcohol There are many technical applications as well as scientific studies founded on radical scavenging. When looking at the direct ozone substrate reaction, one important scavenging principle is to add a compound reacting with hydroxyl radicals, forming a peroxyl radical unable to expel superoxide anions or hydrogen peroxyl radicals. Typical scavengers of this type are tert-butyl alcohol and methanol. By competition, they are supposed to inhibit the chain-carrying steps of the free radical chain reactions, R5-R6. A study of the ozone consumption in water where free radicals were initiated by γ-radiation revealed that ozone was consumed in aqueous solutions of tert-butanol at a rate indicating the intermediary peroxyl radicals to be reactive. In literature, it was found that alkylperoxyl radicals in the gas phase were in fact known to react with ozone, forming alkoxy radicals (104 M-1s-1) 92,93. The possibility of this reaction pathway in aqueous media had to our knowledge never been considered and the rate constant in this type of medium was an open question. However, the immediate concern to this thesis was the validity of our model of the free radical chain reactions, R5-R6, and the impact of ethylene glycol and methanol for pulp bleaching.

Table VII. Additional possible reactions, when the ozone reacts with alkyl peroxyl radicals in water Reactants Product # . . R1R2R3COO O3 R1R2R3CO 2 O2 R18 .. R1R2HCO CR2R3OH R19 . R1R2R3CO R1 R2R3CO R20 . 2+ -. 2 R1R2R3COO Fe some O2 /HO2 other products R21

The rate constants of the reaction between ozone and alkyl peroxyl radicals in water were not very different from those in gas-phase, a few times 104 M-1s-1, and substituent effects were slight. Chain lengthsiii of ozone consumption of an aqueous solution of alcohols, being subject to γ- radiation, were determined by the rate at which superoxide anions formed reacted with ozone and the rate of their termination (disproportionation), Table VIII. For our experimental conditions, the theoretical length should be about 250. The observed chain lengths for ethylene glycol was 11.5 and for methanol about 1. The lower observed chain length for methanol is reconcilable with a rate determining superoxide expulsion step for these alcohols. The rate constant of superoxide from the peroxyl radical of methanol was estimated to be 10 s-1. This is a factor of 10 lower than for peroxyl radicals of ethylene glycol or glycosidic units.

iii n b the expression bears no connection to linear polymers

42 Table VIII. Chain lengths of ozone consumption during γ-irradiation of aqueous solutions containing initially 0.06-0.08 mM O3, 1.3 mM O2 in the presence of different substrates at mM concentrations Substrate Chain-length pH Dose rate Remarks Theory ≈250 2 0.098 Formic acid 16.7 2 0.098 Ethylene glycol 11.5 3 0.098 2-propanol 6.7 2 0.098 Ethanol 3.9 2 0.098 Ethanol 4.9 ≈6 0.098 no acid added Methanol 1.0 2 0.098 Methanol 0.9 2 0.098 CH2O yield Methanol 4.5 2 0,0061

No direct impact of the formylmethyl radical formation mechanism of ethylene glycol (Figure 9), was detected. However, the shortening of the chain reaction could be due to a competition between formylmethyl radical formation and oxygen addition. The branching ratio between formylmethyl radical formation and oxygen addition was calculated to be 0.1, yielding a chain length of 25, which is fairly close to our measurement.

4.4.2.1 Summary The concentration of ozone in water being around 10-4 M, the effective rate constant value for the reaction between the peroxyl radical and ozone will be around 1 s-1. This is a factor of 100 slower than the decay of the peroxyl radical of ethylene glycol, R14/R5b, and a factor of 10 slower than that of methanol. Hence on behalf of ozone bleaching chemistry, this reaction can be ignored in lignin-containing pulp but may have a minor impact in lignin-free pulp. As no impact of the formylmethyl radical formation mechanism of ethylene glycol (Figure 9) was detected at pH 2, it will not be found at pH 3 either. This supports the suggestion of EG simply being a carrier of chain reactions in the experiments with pulp.

43 5 Discussion

5.1 Scissions

5.1.1 Charged groups The introduction of ionic groups along cellulose chains changes the conditions that determine the random coil formation with a possible consequence for viscosity measures and related DP estimates. Examples of how ionised groups can be introduced are: • The reducing ends of cellulose, original or formed from chain cleavage, are oxidised to carboxyl groups during the course of ozonation. • Charged compounds (e.g. carboxylic groups) may form during chemical processing from constituents bound to cellulose. Bonds between cellulose and lignin or cellulose and hemicellulose could be candidates. Linearity would not prevail if the contribution from formed charged groups were substantial. As seen in Figure 37, the viscosity is invariant with pH. If charged groups were to influence viscosity, a changed apparent viscosity should have been observed.

5.1.2 Alternative interpretations of mechanisms behind scissions The non-radical direct ozone attack on cellulose is central to this discussion, as it is sometimes proposed to be the sole degradation mechanism and very frequently serves as the complement to the free radical mechanisms. The problem with discussing mechanisms is the general lack of a clear explanation of how degradants can access cellulose chains organised in fibrils. For example, proposals of non-radical direct ozone attack often rely on the concept of amorphous cellulose, e.g. accessible regions in crystalline cellulose. To evaluate degradation mechanisms, information is needed on the diffusion in crystalline or para-crystalline cellulose (fibrils) and the distribution and character of amorphous zones within a fibril and its variation over a fibre cross-section.

5.1.2.1 Chemistry There are several concepts of scissions, similar to the ds and als, in pulp and paper chemistry: What we term “ds” has been determined by others to be 0.36 of the total attack for conventionally ozonated fully-bleached hemlock kraft pulp94. Although acknowledging the possible existence of radicals, these authors neglect any contribution from hydroxyl radicals in their final evaluation of scissions both “ds” and “als”. They attribute the value to the non- ionic ozone direct attack. As a free radical chain reaction may very well occur in the system, we fail to find their conclusion convincing. The value of 0.36 does imply that all carbons in a glycosidic unit except for C5 have reacted to a similar extent, according to Section 4.3.1.2.1 and also the paper. This is most simply explained by being a result of a low-selective degradant. A radical attack fits better than a non-ionic direct attack from ozone, which is commonly understood as more selective, especially for C1. In a related article by these authors, their scission model is extended with the concept of “front” and “reacted region” of a lignin-containing pulp. The degradation by the radicals is attributed to the “front” values solely 95. We find it possible that radicals form and degrade as stated. However, these radicals will not vanish. Free radical chain reactions will ensue and prevail, provided there is ozone. The less lignin, the less competition for ozone and the faster the chain reactions will be. The proliferation of chain reactions is shown by the experiments on the spreading of chain reactions, Section 4.3.1. Hence, radicals will contribute both to “front” and “reacted zone values”.

44 In another example, free radicals have been suggested to cause “ds” but “als” are assumed to be produced from a non-radical ozone attack85. Such an assertion cannot answer the question of why the hydroxyl radical involved in “als” in cellulose would selectively choose C2, C3 and C6 and not C1 and C4 in the glycosidic unit. A different approach to motivate the idea of a non-radical ozone attack is based on the alternative, that the hydroxyl radical attack would produce fewer carbonyl groups than ozone (“als”)38. This is contrary to the observed reaction pattern towards the single monomer unit.

5.1.2.2 Solubility The solubility (in water or water/alcohol mixtures) of compounds involved in bleaching provides the basis for explanations using concepts like diffusion or selective solvation5,86. Alcohol mixed with water should, in analogy with oxygen, enhance solubility of ozone and lignin and decrease the solubility of glycosidic units. The protection of cellulose in our experiments decreased in the order 25 wt% EG>25 wt% methanol>water. This is not what is predicted if changes of ozone solubility are assumed to be responsible for cellulose degradation.

5.1.2.3 Amorphous zones A non-radical attack will produce a pattern similar to free radical chain reactions if the cellulose sample in question contains high numbers of fibril segments inwhere glycosidic units are accessible. Local fibril lattice distortion is a generally accepted possible feature in cellulose structure models that yet lack spectroscopic verification. If this indicates that these glycosides are too few to show up on a %wt total glycosidic units scale, the number of ‘distorted’ glycosidic units per cellulose chain should be in the range 0.1-1 or less, (0.1%wt total glycosides detection limit and DP 1000-10000). In order to allow for intrinsic ‘distortion’ of the glycosides to establish contact with the reaction media, the low number of glycosidic units per cellulose chain imposes the restriction that these glycosides are aligned radially in fibrils. Such an arrangement agrees with fibril ‘kinks’ of mechanical or biological origin. The concept of levelling-off DP (LODP) relates to this idea. LODP is based on an observed minimum of average cellulose chain length (viscometry) in acid hydrolysis of cellulose and is involved in the idea of amorphous regions being located in intervals over crystalline cellulose (“fringe-micellar” model ). In some articles, ozone is proposed to react with “amorphous” carbohydrates and this reaction is a way to observe LODP5. As free radical chemistry cannot be ruled out in this system, the issue of crystalline cellulose inertness and inaccessibility remains in limbo. A variant of the same approach is that the amount of amorphous fibril regions should be finite. A change of the degradation pattern, a phase-shift, should follow when the amount of easily accessible glycosidic units is consumed. Two reaction phases have been claimed to be observed during ozonation of cotton linters78. The first phase is attributed to fast ozone reactions with accessible cellulose. However, no clear indication of a phase-shift is seen when the data are plotted on the logarithmic scale as in Figure 42, nor do these data accommodate the effects of fibril distortion of the order 0.1-1 ‘distortion’ per chain.

45 2.5

2

0 1.5 /ds 1

log ds 1

0.5 y = 0.8219x + 0.0994 R2 = 0.9882 0 00.511.522.5

log nO3,1/nO3,0

Figure 42. Data of degradation (ds) and the corresponding ozone consumption78. 1 wt% odp, neutral pH

5.2 Chemistry

5.2.1 Free radical formation from direct ozone attack on cellulose The following text serves to demonstrate that free radical formation is possible from ozone and carbohydrates directly, and that alternative hypotheses, such as transition metals and phenolate ions, may prove to be less important. It is established that ozonation of α-methylglycosides or carbohydrates (glucose, EG, methanol) produces free radicals, though the exact mechanism is not known. To give one example: Ozone has been shown to attack C6 in the glycosidic unit and C6 is an α-carbon of a primary alcohol. Strong evidence has been found for a hydrogen abstraction mechanism occuring with primary alcohols. The attack on C6 is minor compared to C1 (in CH2Cl2, 18% yield and 82% yield respectively) but however minor, it cannot be neglected as the free radical chain reactions that follow have a strong impact on degradation.

5.2.1.1 Transition metals and the relationship to background degradation The use of the ferrous ion as an initiator is one example of how transition metals can initiate free radical degradation in ozone-based bleaching systems. Common levels of transition metal content in pulp and cotton linters are however low, typically around 10 ppm. Their oxidation states are unknown. In the case of iron, it can be said that the reduced state, +II, is not favoured, especially after an oxygen bleaching stage (oxygen bleached kraft pulp).

46 0.2

-6.4±0.1 0 s

al -0.2 or log

ds -0.4 log ds log log als -0.6 log ds (O Fe) Figure 43. Extrapolation of the charge Fe(II) added, log als(O Fe) which is equivalent to direct ozone attack “Ozone -6 -4 2+ -0.8 constant series”, 10 -10 mole Fe added, and 1 -8-6-4-2 background degradation point, similar ozone log nFe(II) charge. 10-6 mole Fe(II) was added to 10 g odp 40 wt% cotton linters prior to ozonation (Figure 43). The observed degradation was similar to that of a sample with no Fe(II) added. Addition of 10-6 mole Fe(II) corresponds to approximately 4 ppm in the reaction media (water). One can reformulate the above by saying that ozone reacts with native cotton linters as if 1*10-6 mole Fe2+ had been added. This implies that initiation of free radical chain reactions could also occur in a cotton linter sample, without addition of Fe(II). In agreement, the ratio between ds to als is similar to those of free radical chain reactions, Section 4.3.1.2.1.

5.2.1.2 Phenolate anions In this thesis it is assumed that lignin contains 0.1 mmole phenolic groups per 10 g pulp (10- 20 per phenyl propane unit, Table I). Phenolate anions may rapidly produce hydroxyl radicals in their reaction with ozone. The apparent reaction rate is slower at pH 3, where the rate constant should be around 103 M-1s-1. This reaction is in competition with the Crigée reaction with a rate constant of 104 M-1s-1. If hydroxyl radicals are formed, two points need to be noted: First, the hydroxyl radical reacts a few Angstroms away due to its high reactivity, see Section 4.3.1. Second, lignin-rich zones in a fibre are ozone-depleted. As free radical chain reactions and not the hydroxyl radical per se appear to be necessary to degrade cellulose efficiently, effective degradation will not occur until the region is delignified. Phenolate anions are one possible initiation source out of several. Phenolate mediated initiation may contribute with a delay, e.g. superoxide radicals produced during delignification react slowly until the ozone concentration is restored.

5.2.1.3 Observed degradation of pulp Degradation of cotton linters, fully-bleached pulp and oxygen bleached kraft pulp for a given ozone charge has been shown to be related to the carbohydrate content, Section 5.4.1, and this without reference to phenols, transition metals and amorphous fibril segments. The common denominator for these pulps is cellulose and free radical chain reactions. Arguing by induction, it is close at hand to propose the common initiation reaction to be ozone partly reacting by hydrogen abstraction with cellulose.

5.3 Fibres and free radical mediated cellulose degradation We note that cellulose accessibility is a relative concept. Rapidly reacting degradants like the hydroxyl radical will not migrate far into a carbohydrate matrix (without chain reaction

47 propagation). Hence, using the hydroxyl radical as the norm for cellulose accessibility would render much lower values than using acid hydrolysis. Radicals being the case, the perspective of the cross-section (the fibre distribution) of a pulp sample rather than the cross-sections of individual fibres is the informative one. In Figure 24, we observe degradation localised to zones. The zones are visible on the scale of µm, in the three-dimensional shape of holes. Regardless of the mechanism causing the holes, the correct description of the degradation of fibres on the cellulose chain level should be “random hole distribution” rather than that of random scission generation and should be treated mathematically as “discharge regions” in cellulose radiolysis literature96. One zone is equivalent to several scissions.

5.4 Pulps: Coherency of the results The aim of this thesis is to construct a model from which predictions can be made. The ideas from the work on cellulose in cotton linters should be applicable to cellulose in oxygen bleached kraft pulp. The following is an attempt to demonstrate that the results of different pulps appear coherent.

5.4.1 Quantification of the relative influence of the free radical chain reaction parameters in cotton linters A mathematical approach is offered in this section, to summarise the information on degradation of cotton linters, pulp and Table II.

5.4.1.1 Calculation The observed degradation D is described as a function of free radical chain reactions in terms of the number of free radical sites on accessible cellulose N and the length of free radical chain reactions L that extend from these sites. N is defined as a function of the initiation (I) and termination (T) reactions that occur in the system. L is defined as a function of the propagation (P) of free radical chain reactions. Initiation, propagation and termination 2+ reactions depend on the amounts of Fe and O3 added. Logarithms are used to facilitate adjustment to experimental data. D = f ()N,L N = f ()I,T L = f ()P

I = f ()n Fe()II ,n O3 T = f ()n Fe()II ,n O3

P = f ()n Fe()II ,n O3

∂ ln(D) ∂ ln(D) d ln(D) = dlnN()+ dln() L ∂ ln(N) ∂ ln(L) Equation 21 2+ where nFe(II), and no3 are the amounts of Fe and O3 charged in moles to 10 g odp cotton linters respectively, and D is ds. The observed degradation is quantified according to:  D  ln 1  ∂ ln(D) ∆ ln(D)  D  = = 0 ∂ ln(n ) ∆ ln(n )  n  i i ln i,1  ni,0

48 Equation 22 where ni is either nFe(II), or no3. The values obtained are related to Equation 21 by either of the following equations: ∂ ln(D) ∂ ln(D) ∂ ln(N) ∂ ln(D) ∂ ln(D) ∂ ln(L) = ⋅ = ⋅ ∂ ln(n i ) ∂ ln(N) ∂ ln(n i ) ∂ ln(n i ) ∂ ln(L) ∂ ln(n i )

Equation 23 In the interval 10-6-10-4 moles Fe2+ added in the “Ozone constant” series (Figure 27), there are 3 ds-data, which gives:  ∂ ln(D)    = 0.2 r = 1.0  ∂ ln(nFe(II))  O3,CH=c Fe(II)

Equation 24 where CH indicates carbohydrate content. In the interval ≥8*10-4 mole Fe2+ added in the “Fe(II) constant” series (Figure 26), there are 5 ds-data which gives:  ∂ ln(D)    = 1 r = 0.94  ∂ ln(nO3))  Fe(II),CH=c O3>Fe(II)

Equation 25 Both curves are linear within the specified intervals, which indicates that the degradation modes do not change within the intervals. In the case of the “Ozone constant” series, the formation of radical sites at accessible hexoses by ferryl ions, R8, are kinetically favoured by a factor of 2000-20 as compared to free radical disappearance, R9. This is consistent with a ∂ ln(N) constant , which can be assumed to be equal to 1. It agrees in the case of the ∂ ln(n Fe(II) ) “Fe(II) constant” series in that free radical chain reactions have uncompeted access to ozone ∂ ln(L) once the ferrous ions have been consumed and only ozone remains. The is hence ∂ ln(n O3 ) also assumed to be a constant equal to 1.

Thus:  ∂ ln(D)  ∂ ln(D)     = 0.2 = 1  ∂ ln(L)  CH= c  ∂ ln(N)  CH=c

Equation 26 Equation 27

∂  ∂ ln(D)  ∂  ∂ ln(D)  ∂  ∂ ln(D)  ∂  ∂ ln(D)    =   = 0   =   = 0 ∂ ln()L  ∂ ln(N) CH=c ∂ ln()N  ∂ ln(L) CH=c ∂ ln()L  ∂ ln(L) CH=c ∂ ln()N  ∂ ln(N) CH=c

Equation 28 Equation 29

49 ∂ ln(N) ∂ ln(L) = c1 = c 2 ∂ ln(n Fe(II) ) ∂ ln(n O3 )

c1 = c 2 = 1

Equation 30 Hence, free radical degradation of 10 g odp cotton linters can be summarised as : d ln(D) = 0.2d ln()N +1d ln ()L or D = c ⋅ N 0.2 ⋅ L

Equation 31 where c is a constant.

5.4.2Extrapolating from cotton linters to oxygen bleached kraft and fully bleached pulp In this section, quantities of pulp constituents in Table I (Section 2.2.3.3) are combined with a quantification of our results, Section 5.4.1.

5.4.2.1 Calculation We have to relate the measured free radical chain reaction efficiencies from the cotton linters experiments with Fe2+ added, Section 5.4.1, to native systems, oxygen bleached kraft pulp and fully bleached pulp, with and without EG, and adjust for differences in carbohydrate and alcohol content and ozone supply. An increased amount of bulk of carbohydrates from the introduction of hemicellulose and EG content is adjusted for according to:  ∂ ln(D)     = 1     ∂ ln(nO3) ) Fe(II),CH=c  O3>Fe(II)   ∂ ln(D)  →   = −1     ∂ ln(D)   ∂ ln(D)    ∂ ln(nCH) )    = −  O3=c       ∂ ln(nO3) ) Fe(II),CH=c  ∂ ln(nCH) )  O3>Fe(II) O3=c  Equation 32 where D means degradation in scissions (ds or als) and: −1 D  n  n 2 =  CH,2  = CH,1 D1  nCH,1  nCH,2 nCH = ncellulose(hexose) + nhemcellulose(hexose) + nEG Equation 33 where CH denotes carbohydrates, which includes cellulose, hemicellulose and EG. Cellulose and hemicellulose are counted with respect to hexose units. EG and hexoses are considered equivalent on a molar basis, given their rate constants are 0.1-1 M-1s-1.

Changed conditions for free radical chain reactions by pulp are adjusted for by:

50  ∂ln(D)    = 0.2   ∂ln(n Fe(II) ) O3, CHa = c   Fe(II) < O3  ∂ln(D)   ∂ln(D)    = −   ∂ln(n )   ∂ln(n )  O3 = c  Fe(II)  O3, CHa = c  CHa  Fe = c Fe(II) < O3 Fe(II)«O3  ∂ln(D)    = −0.2  ∂ln(n )  O3 = c  CHa  Fe = c Fe(II)«O3

Equation 34 where nCHa denotes accessible carbohydrates and alcohols. The equation is simplified: −0.2 D  n ,  2 =  CHa 2  D1  nCHa,1  nCHa = ncellulose(hexose),acc + nhemcellulose(hexose) + nEG Equation 35 The ozone charge is fixed in the pulp experiments, but pulp contains elements that will decrease the ozone amount available for free radical chain reactions. The degradation obtained from a given ozone supply is:  dlnD()()   = 1 dlnn  ()()O3  CH=c

 nO3,2  D2 =   ⋅ D1  nO3,1  Equation 36 Lignin consumes ozone in proportion to the decreased final kappa. Ozone consumption from initiation is assumed equal to the observed ozone consumption when Fe2+ is charged ≤10-6 moles, adjusted for the change of accessible surface. Then: 0.2 n ⋅ K  n  n = O3,1 final − n ⋅ CHa,2  O3,2 O3,det lim   K initial  n CHa,1  Equation 37 For a pulp (oxygen bleached kraft or fully bleached, with or without EG):

 nO3,2   nCH,1  ds* =   ⋅   ⋅ DCL  nO3,1   nCH,2  Equation 38 where DCL denotes ds observed for cotton linters and ds* the ds estimate for the specific pulp system.

51 Table IX. In-data for predictions in Table X Component Value Unit Comment

ds 0.3 Cotton linters 40 wt%, pH 3, 0.4 g O3 / g odp als 0.5 ‘’ ts 0.8 ‘’ O3 consumption 5.6E-5 mole = detection limit: 7% O3 charged, for 0.4 g O3/g odp

5.4.2.2 Results

Table X. Input for the degradation of cellulose: cotton linters, fully-bleached pulp, oxygen bleached kraft pulp and the ethylene glycol HC = hemicellulose, OKP = oxygen bleached kraft pulp, FBP = fully bleached pulp HC 16 HC 16 wt% HC 16 wt% HC 16 wt% wt% EG 25 wt% EG 25 wt% Initial kappa Initial kappa 10 10 Final kappa 5 Final kappa 5 Amount 0.01 0.06 (moles) Correction Bulk of 1.2 2.1 factors carbohydrates Surface of 1.3 1.8 carbohydrates O3 supply 0.9 0.9 0.4 0.4

Prediction ds* 0.23 0.12 0.11 0.06 Observed FBP ds 0.24 0.11 OKP ds 0.11 0.04 In spite of the rough approximation method, scaling works very well; the deviation average is 0.01.

5.4.3 Summary Equation 27 has the value 1 which is high considering that free radical chain reactions are theoretically distributed in all directions from a first radical site and that the observed degradation in response is expected to vary with direction from the initial radical relative to a fibril. Equation 26 indicates that the number of free radical sites has a significantly lower impact on degradation than the number of free radical chain cycle that extend therefrom, Equation 27. We conclude from Table X that: • the results from cotton linters appear to be consistent with the results from oxygen bleached kraft and fully bleached pulp • The impact of EG on oxygen bleached kraft and fully bleached pulp is explained on the basis of adjusting for changed amount of bulk substrate and ozone for free radical chain reactions in cellulose. • It is of significance that the model does not rely on factors such as transition metal content, phenol content and amorphous fibril segments.

52 Interestingly, direct ozone attack is the only initiation source that has a direct relationship to an increased content of accessible carbohydrates and alcohols.

53 6 Conclusions A new model for the free radical chemistry of ozone in pulps with emphasis on carbohydrate degradation is presented.

6.1 Conclusions

6.1.1 Rationale of free radical chain reactions • OH reacts fast with a wide range of compounds (unselective). As a consequence, it does not live long enough and thus cannot move far in a cellulose sample. • HO2 reacts more slowly with a more narrow range of compounds (selective). As a consequence, it has a longer lifetime and can move before reacting.

Adapting the random scissions assumption (dissolved polymer with randomly distributed degradants), we have: • In fibrils there is accessible and non-accessible cellulose. • The free radical chain reactions for degradation are confined to zones.

6.1.2 Cellulose degradation • Cellulose crystallinity is central to cellulose degradation during ozonation. • In crystalline cellulose (cotton linters), the degradation responds to the free radical chain reactions that follow from initiation by ferryl ions and not the ferryl ion reaction with cellulose directly. • Degradation for low cellulose crystallinities (cellulose beads) responds directly to the ferryl ion reaction with cellulose. • An extrapolation of the results from ozonation in the ferryl ion-cotton linters system to ozonation of oxygen bleached kraft pulp, fully bleached pulp and these pulps in combination with ethylene glycol was performed. Good agreement was found between extrapolation and experiments in all cases. • Efforts to limit degradation should focus on ozone rather than free radical formation in itself. The degradation involved in the initiation process does not seem harmful. An excess of ozone is required (unconsumed by lignin) to propagate free radical chain reactions in order to produce significant degradation. • Results show that it is not satisfactory to discuss hydroxyl radicals in the context of free radical degradation without specifying how the hydroxyl radicals were introduced to the cellulose chains (surface, initiation or free radical chain reaction). • The concept of amorphous regions is not necessary to explain the results presented here. It raises the question of whether amorphous regions have relevance to ozonation of cellulose in general. • In addition to established possible initiation reactions, it is also possible that free radicals form by a direct ozone reaction, abstracting hydrogens from carbohydrates. Arguments are also presented against cellulose degradation from a non-radical direct attack by ozone.

6.1.2.1 Free radical chain reactions, ds-als Evidence for the hydroxyl radical being the active cellulose degradant as part of a free radical chain reaction are:

54 • An observed constant relative ratio of 0.67 ds to als in initiation limited degradation is consistent with the relative reactivity of hydroxyl radicals versus hydrogens on C1 and C4 to C2, C3 and C6. • Subtracting a calculated ds=0.67*als from the observed ds value should produce a negligible residual of ds if all degradation is accounted for by hydroxyl radical activity. This has been demonstrated for oxygen bleached kraft pulp at varied wt% of added EG.

6.1.3 Ethylene glycol and methanol • Addition of 25 wt% EG in 40 wt% (high consistency) oxygen bleached kraft pulp (initial kappa 10.4 and initial viscosity 975 ml/g) at pH 3 is optimal for ozone bleaching. • Addition of 25 wt% EG gives better selectivity than addition of 25 wt% methanol. • EG appears to decrease degradation of cellulose by carrying a proportion of the free radical chain reactions. • No effects of acid / base catalysed water expulsion from free radicals deriving from EG and glycosidic units were observed. This is most likely due to a rapid oxygen addition outcompeting the potential water expulsion step. • Increased ozone solubility from EG appears less central to selectivity. • The rate constant of superoxide from the peroxyl radical of methanol was estimated to be 10 s-1. • Rate constants of the reactions between ozone and alkylperoxyl radicals were determined to be around 104 M-1s-1. The possibility of the reaction between alkylperoxyl radicals and ozone contributing significantly to free radical chain reactions during ozonation of carbohydrates and alcohols could therefore be ruled out.

6.1.4 Pulp • Cellulose degradation appears not to be directly correlated to delignification during ozonation. It is suggested that lignin reactions consume the ozone that would otherwise have propagated free radical chain reactions. • Modelling lignin as “ozone supply reducing,” and EG and hemicellulose as “free radical chain reaction carrying”, together with the data from cotton linters produced predictions consistent with results from oxygen bleached kraft pulp and fully bleached pulp, both in the presence and absence of EG. • Delignification was favoured by a low pH. The variation with pH of delignification could be related to swelling. If the variation is caused by the phenol/phenolate acid-base equilibrium, the equilibrium proved less central to cellulose degradation.

6.1.5 Spreading of chain reactions • Free radical chain reactions were shown to spread outwards from a spot of initiation during ozonation of a filter paper using a pH-indicator to monitor acid formation. Free radical degradation in carbohydrates hence has concentric development. • After ozonation, cellulose fibres doped with initiator displayed an exterior and interior permeated by small holes. This is in agreement with concentric development of free radical chain reactions in cellulose .

6.1.6 Scissions • Any type of average chain length can be used to calculate the number in a linear polymer of any molecular weight distribution, provided there is a calibrated Mark-Houwink equation.

55 • Rapid agent degradation (as from the hydroxyl radical in the case of TCF-bleaching), occurs in cellulose but the influence on viscometry is strongly dependent on cellulose accessibility. If the cellulose exhibits <0.1 accessibility, as stated in some literature, degradation by rapid reagents cannot be observed.

56 Acknowledgements I would like to thank my supervisor Associate Professor Johan Lind, Department of Chemistry – Nuclear Chemistry, Royal Institute of Technology, Stockholm. It has been hard but satisfying work and though it has not always been so easy to explain what I am getting at, Johan has always been listening. He is greatly appreciated for his free-mindedness and creative science, patience and loyalty. It is hard not to mention Associate Professor Gabor Merényi in this context, since he is an indistinguishable part of “Johan & Gabor”. Associate Professor Sten Ljunggren, Swedish Pulp and Paper Institute, Stockholm, shared supervision with Johan for several of my years. Sten has valuably contributed with the world of wood chemistry and bleaching - with people, conferences, laboratory work, the latest news, writing and finding blues clubs in Kyoto, Japan. Associate Professor Torbjörn Reitberger, Department of Chemistry – Nuclear Chemistry, has been a genuinely positive influence, contributing ideas and helping with writing and all sorts of scientific dilemmas. Associate Professor Mikael Lindström (STFI) has been very generous and supportive in providing laboratory space as well as the company of his research group, which has been important to me. Professor Trygve Eriksen, Department of Chemistry – Nuclear Chemistry, has been standing in the background like the guardian angel of our project. Also, he is the provider of a much- appreciated workplace, the Nuclear Chemistry group at KTH: A special thanks to the “gang” of doctorands, Dr Mats, Associate Professor Mats, Dr Mireia, (Soon Dr) Sanna, Dr Rong, Dr Daqing, Ella, David, Magnus, Mariam, Fredrik and Anders and all who have made everyday-life not only endurable but exciting.

Associate Professor Erland Johansson contributed valuable advice, articles and help and, not least, is my father. Thanks also to Viklunds and Haraldsons. Thanks to Mireia for lighting up my life and for enduring my thesis - love you, bebis! Olof was responsible for the repeated attempts to destroy my health. Part of this thesis was created in Barcelona, and the family Molera Marimon are hereby gratefully acknowledged for their help

57 References 1. E. Britannica, Cellulose, Encyclopedia Britannica Online, 2002, 12 December 2. C.W. Dence and D.W. Reeve, Pulp bleaching. Principles and Practice, TAPPI Press(1998) 3. Kolbe, Bull Soc Ind Mulhouse, 38, 94, (1868) 4. M. Cunningham and C.J. Dorée, J. Chem. Soc., 101, 497, (1912) 5. Z. Osawa and C. Schuerch, Tappi J, 46, 2, (1963) 6. R.G. Rice, 13th Ozone World Congress, Kyoto, Japan (1997) 7. E. Adler, Wood Sci. Tech., 11, 169, (1977) 8. E. Sjöström, (1993) 9. A. Kerr and D.A.L. Goring, Cellulose Chem. Technol., 9, 563, (1975) 10. H.A. Krässig, Cellulose. Structure, accessibility and reactivity,; Gordon and Breach Science Publishers 1993; Vol. 11 11. P.O. Bethge and K. Wilson, Sv. Papperstidn., 2, 28, (1963) 12. G. Gellerstedt and E.-L. Lindfors, Tappi J, 6, 119, (1987) 13. T.M. Grace, B. Leopold and E.W. Malcolm, Vol. 5: Alkaline pulping, In Pulp and paper manufacture, TAPPI: Atlanta, Georgia, US(1989) 14. O. Karlsson, Studies on the extent of bonding between lignin and polysaccharides in pulps,, Royal Institute of Technology (1997) 15. J. Li, Towards an accurate determination of lignin in chemical pulps: The meaning of kappa number as a tool for analysis of oxidizable groups, Royal Institute Technology (1999) 16. L.A. Donaldson and A.P. Singh, Holzforsch., 52, 449, (1998) 17. J. Hafrén, Ultrastructure of the wood cell wall, Royal Institute of Technology (1999) 18. J. Hoigne, H. Bader, W.R. Haag and J. Staehelin, Water Res., 19, 993, (1985) 19. J. Staehelin and J. Hoigne, Environ. Sci. Technol., 16, 676-81, (1982) 20. R. Andreotti, V. Caprio and A. Insola, Water. Res., 30, 12, 2955, (1996) 21. J. Hoigné and H. Bader, Water Res., 17, 173, (1983) 22. M.S. Matheson, A. Mamou, J. Silverman and J. Rabani, J Phys. Chem., 77, 2240, (1973) 23. G.V. Buxton, C.L. Greenstock, W.P. Helman and A.B. Ross, J Phys. Chem. Ref. Data, 17, 513, (1988) 24. Bothe, Shulte-Frohlinde and v. Sonntag, J Chem. Soc., Perkins Trans. 2, 416, (1978) 25. J. Lind, G. Merenyi, E.E. Johansson and T. Brinck, Journal of Physical Chemistry A, 107, 676-681, (2003) 26. G.E. Adams and W.R. L., Trans. Faraday Soc., 65, 2981, (1969) 27. K. Sehested, J. Holcman and E.J. Hart, Journal Phys. Chem., 87, 11, 1951-1954, (1983) 28. J.W.T. Spinks and R. Woods,Introduction to radiation chemistry, 3rd ed.; John Wiley &Sons Inc 1990 29. T. Loegager, J. Holcman, K. Sehested and T. Pedersen, J. Inorg. Chem., 31, 3523- 3529, (1992) 30. T. Loegager, J. Holcman, K. Sehested and T. Pedersen, J. Inorg. Chem., 31, 3523, (1992) 31. P. Dowideit and C. Sonntag, Environ. Sci. Techn., 32, 8, 1112, (1998) 32. R. Crigée, Rec. Chem. Progr., 18, 2, 113, (1957) 33. J. Gierer, Holzforsch., 51, 34, (1997)

58 34. T. Eriksson and T. Reitberger, 8th International Symposium on Wood and Pulping Chemistry, Helsinki, Finland (1995) 35. J. Lind, G. Merényi and K. Wegner, J. Wood Chem. Tech.,, 17, 3, 297-326, (1997) 36. M. Ragnar, Radical formation under acidic TCF-bleaching conditions, Royal Institute of Technology (1998) 37. M. Ragnar, T. Eriksson, T. Reitberger and B. P., Holzforsch., 53, 423, (1999) 38. C. Chirat and D. Lachenal, Holzforsch., 51, 147, (1997) 39. T. Jacobsen, J. Holcman and K. Sehested, Int. J. Chem. Kinet., 29, 17, (1997) 40. A. Katai and C. Schuerch, J. Pol. Sci. A, 1, 4, 2683, (1966) 41. T. Jacobsen, J. Holcman and K. Sehested, Int. J. Chem. Kinet., 30, 215, (1997) 42. T. Eriksson and J. Gierer, J. Wood Chem. Tech., 5, 1, 53, (1985) 43. B. Ericsson, Oxygen oxidation of cellulose in alkaline medium, Royal Institute of Technology, (1974) 44. Y. Ni, G.J. Kang and A.R.P. van Heiningen, J. Pulp Paper Sci., 22, 2, J53-57, (1996) 45. A. Berglund, H. Brelid, A. Rindby and P. Engström, Holzforsch., 53, 474, (1999) 46. P. Deslongchamps, P. Atlani, D. Fréhel, A. Malaval and C. Moreau, Can. J. Chem., 52, 3651, (1974) 47. J. Bouchard, H.M. Nugent and R.M. Berry, Tappi J., 78, 1, 74, (1995) 48. X.Z. Zhang, G. Kang, Ni and A. van Heiningen, International Symposium on Wood and Pulping Chemistry (1997) 49. M.N. Schuchman and C. von Sonntag, JWSRT-Aqua, 38, 311, (1989) 50. B. Plesnicar, J. Cerkovnik, T. Tekavec and J. Koller, J. Am. Chem. Soc, 120, 8005- 8006, (1998) 51. S.J. Angyal, Adv. Carbohydr. Chem. Biochem., 47, 1, (1989) 52. S. Steenken and P. Neta, J Phys. Chem., 83, 9, 1134, (1979) 53. S. Steenken, J Phys. Chem., 83, 5, 595, (1979) 54. N.R. Bertoniere and S.H. Zeronian. Chapter 15. Chemical characterisation of cellulose, The structures of cellulose. Characterisation of the solid states; Atalla, R. H. Ed.; American Chemical Society, 1987; Vol. ACS 340; pp. 255,. 55. D.H.-S. Hon, Cellulose, 1, 1, 1, (1994) 56. R.H. Atalla and D.L. VanderHart, Solid State Nuclear Magnetic Resonance, 15, 1, 1- 19, (1999) 57. P.T. Larsson, K. Wickholm and T. Iversen, Carbohydr. Res., 302, 95-129, (1997) 58. E.L. Hult, T. Liitia, S.L. Maunu, B. Hortling and T. Iversen, Carbohydrate Polymers, 49, 2, 231-234, (2002) 59. K. Wickholm, P.T. Larsson and T. Iversen, Carbohydr. Res., 312, 123-129, (1998) 60. A.O. O'Sullivan, Cellulose, 4, 173, (1997) 61. E.M. Lewin and E.M. Pearce, Handbook of Fibre Chemistry, Marcel Dekker Inc, US(1998) 62. P.T. Larsson, Personal communication, (1998) 63. B. Pettersson and K. Eriksson, E., L., Analytical Biochemistry, 285, 2, 220-224, (2000) 64. Kappatal, SCAN-C 1:77, Scandinavian pulp, paper and board testing commitee, (1977) 65. U.W. Gedde,Polymer Physics; Chapman & Hall, 1995 66. Viskositet i kopparetylendiaminlösning, SCAN-CM 15:88, Scandinavian pulp, paper and board testing commitee, (1988) 67. K. Kamide and M. Saito. Viscometric determination of molecular weight. In Determination of molecular weight; Cooper, A. R. Ed.; John Wiley & Sons, Inc., 1989; Vol. 103.

59 68. F.W. Billmeyer Jr,Textbook of polymer science; John Wiley & Sons Inc 1984 69. A. Charlesby,Atomic radiation and polymers; Pergamon Press Ltd: Oxford 1960; Vol. 1 70. C.L. Moad and D.J. Winzor, Progress in Polymer Science, 23, 5, 759-813, (1998) 71. T. Kelen,Polymer Degradation: Van Nostrand Reinhold Company Inc. 1983 72. A. Ekenstam, Ber., 69, 553, (1936) 73. K. Nakamura, T. Hatekayama and H. Hatekayama, Textile Res. J., 9, 607, (1981) 74. C. Olkkonen, H. Tylli, I. Forsskahl, A. Fuhrmann, T. Hausalo, T. Tamminen, B. Hortling and J. Janson, Holzforschung, 54, 4, 397-406, (2000) 75. Y. Wang, R.I. Hollingsworth and D.L. Kasper, Carbohydrate Research, 319, 1-4, 141- 147, (1999) 76. P. Eriksson, New Approaches to Investigate and Prevent Oxidation of Polypropylene, Royal Institute of Technology (2002) 77. O. Theander, Sv. Papperstidn., 61, 18, 581, (1958) 78. M.P. Godsay and E.M. Pearce, AIChE S. S. (1985) 79. M. Strlic, J. Kolar, M. Zigon and B. Pihlar, Journal of Chromatography A, 805, 93-99, (1998) 80. J. Lind, G. Merényi and K. Wegner, J. Wood Chem. Tech., 17, 3, 297-326, (1997) 81. N. Liebergott, B. van Lierop and S. A., Tappi J., 75, 117-124, (1992) 82. J. Xu, E. Cogo, L. Brois, S. Duprat, J. Molinier, C. Coste and P. Kalck, Pulp Paper Can., 96, 9, 49, (1995) 83. H. Kamishima, T. Fujii and I. Akamatsu, Jpn Tappi J, 31, 9, 665, (1977) 84. H. Kamishima, T. Fujii and A. I., J Jpn Wood Res Soc, 29, 7, 474-480, (1983) 85. J. Bouchard, E. Morelli and R.M. Berry, J. Pulp Paper Sci., 26, 1, 30, (2000) 86. M. Solinas and T.H. Murphy, Pulp Pap, 70, 3, 133, (1996) 87. Y. Ni, A.R.P. van Heiningen, J. Lora, L. Magdzinski and E.K. Pye, J Wood Chem Tech, 16, 4, 367, (1996) 88. G.J. Kang, Y. Ni and A.R.P. van Heiningen, 82nd CPPA Technical Section Annual Meeting, Montreal, Canada (1996) 89. U. Westermark, 10th Int. Symp. Wood Pulp Chem., Yokohama, Japan (1999) 90. J. Laine, The effect of cooking and bleaching on the surface chemistry and charge properties of kraft pulp fibres, University of Technology (1996) 91. X. Zhang, G.J. Kang, Y. Ni, A.R.P. van Heiningen, A. Mislankar, A. Darabie and D. Reeve, J Wood Chem Tech, 18, 2, 129, (1998) 92. V.V. Shereshovets, V.D. Komissarov and E.T. Denisov, Izv Akad Nauk SSSR, Ser Khim, 6, 1212, (1979) 93. G.S. Tyndall, T.J. Wallington and B.J. C, Journal of Physical Chemistry A, 102, 2457, (1998) 94. X.Z. Zhang, G.J. Kang, Y. Ni and A. van Heiningen, International Symposium on Wood and Pulping Chemistry (1997) 95. X.Z. Zhang, Y. Ni and A. Van Heiningen, Journal of Pulp and Paper Science, 26, 9, 335-340, (2000) 96. W. Schnabel. Degradation by High Energy Radiation. In Aspects of Degradation and Stabilization of Polymers; Jellinek, H. H. G. Ed.; Elsevier Scientific Publishing Company: Amsterdam, 1978.

60

Paper I

Paper II

Paper III

Paper IV

Paper V There is an error in Paper V: In all the captions for Figures 3, 4, 5, 6, 7, and 8 and also in the captions for Tables II and III, the ozone charge should read "0.004 g ozone/g odp", instead of "0.4 g ozone/g odp". In Figure 1, the ozone charge should read "0.004, 0.006 and 0.008 g ozone/g odp". The default value for ozonisation throughout this thesis was 40 mg O3 per g cellulosic sample. In the thesis, the correct values are used Paper VI