Formic and Levulinic Acid from Cellulose Via Heterogeneous Catalysis

Formic and Levulinic Acid from Cellulose via Heterogeneous Catalysis

Johan Ahlkvist

Copyright © Johan Ahlkvist 2014 This work is protected by the Swedish Copyright Legislation (Act 1960:729) ISBN: 978-91-7459-798-1 Digital version available at http://umu.diva-portal.org/ Printed by: VMC-KBC Umeå Umeå, Sweden 2014 Abstract

The chemical industry of today is under increased pressure to develop novel green materials, bio-fuels as well as sustainable chemicals for the chemical industry. Indeed, the endeavour is to move towards more eco-friendly cost efficient production processes and technologies and chemical transformation of renewables has a central role considering the future sustainable supply of chemicals and energy needed for societies. In the Nordic countries, the importance of pulping and paper industry has been particularly pronounced and the declining European demand on these products as a result of our digitalizing world has forced the industry to look at alternative sources of revenue and profitability. In this thesis, the production of levulinic and formic acid from biomass and macromolecules has been studied. Further, the optimum reaction conditions as well as the influence of the catalyst and biomass type were also discussed.

Nordic sulphite and sulphate (Kraft) cellulose originating from two Nordic pulp mills were used as raw materials in the catalytic synthesis of green platform chemicals, levulinic and formic acids, respectively. The catalyst of choice used in this study was a macro-porous, cationic ion-exchange resin, Amberlyst 70, for which the optimal reaction conditions leading to best yields were determined. Cellulose from Nordic pulp mills were used as raw materials in the catalytic one-pot synthesis of ‘green’ levulinic and formic acid. The kinetic experiments were performed in a temperature range of 150–200 °C and an initial substrate concentration regime ranging from 0.7 to 6.0 wt %. It was concluded that the most important parameters in the one- pot hydrolysis of biomass were the reaction temperature, initial reactant concentration, acid type as well as the raw material applied. The reaction route includes dehydration of glucose to hydroxymethylfurfural as well as its further rehydration to formic and levulinic acids. The theoretical maximum yield can hardly be obtained due to formation of humins. For this system, maximum yields of 59 mol % and 68 mol % were obtained for formic and levulinic acid, respectively. The maximum yields were separately obtained in a straight-forward conversion system only containing cellulose, water and the heterogeneous catalyst. These yields were achieved at a reaction temperature of 180 °C and an initial cellulose intake of 0.7 wt % and belong to the upper range for solid catalysts so far presented in the literature.

The reaction network of the various chemical species involved was investigated and a simple mechanistic approach involving first order reaction kinetics was developed. The concept introduces a one-pot procedure providing a feasible route to green platform chemicals obtained via

i conversion of coniferous soft wood pulp to levulinic and formic acids, respectively. The model was able to describe the behaviour of the system in a satisfactory manner (degree of explanation 97.8 %). Since the solid catalyst proved to exhibit good mechanical strength under the experimental conditions applied here and a one-pot procedure providing a route to green platform chemicals was developed. A simplified reaction network of the various chemical species involved was investigated and a mechanistic approach involving first order reaction kinetics was developed.

Table of Contents

Abstract i Table of Contents iii List of Papers v Abbreviations vi Introduction 1 Levulinic Acid 3 Reaction Network upon Production of Levulinic Acid from Biomass 3 The Influence of Reaction Parameters to the Yield of Levulinic Acid 4 The Influence of Raw Material Varieties 4 The Influence of Reaction Temperature 5 The Influence of Substrate Concentration 6 The Influence of Acid Type 7 The Influence of Acid 8 The Influence of Reaction Time 10 The Influence of Pre-treatment 11 The influence of Additives 13 The Influence of Agitation Speed 15 Materials and methods 17 Chemicals 17 Raw materials 17 Acidic polymer catalyst 18 Reactor system 18 Substrate and catalyst characterization 19 Nitrogen physisorption 20 Scanning electron microscopy 20 Acid-base titration 21 Product analysis 21 Liquid chromatography 22 Gas chromatography 23 Mass spectrometry 23 Derivatization for GC/MS 24 Definitions for the yield calculations of levulinic acid (LA) and formic acid (FA) 24 Modelling software and techniques 25 Results and Discussion 26 Catalyst Characterization 26 Substrate characterization 27 Auto-catalysis 29 Reusability of the catalyst, catalyst deactivation 29 Reaction network 30 The influence of reaction conditions and raw material on product yields 32

iii

Comparison between different wood pulps 32 Reaction with D-glucose 33 The influence of reaction temperature and catalyst to substrate dosage 33 The influence of reaction gas and pressure 36 Selectivity 37 Kinetic modelling 38 Fit of the model to experimental data 41 Conclusions 46 Acknowledgements 48 References 49

List of Papers

I Johan Ahlkvist, Samikannu Ajaikumar, William Larsson, Jyri- Pekka Mikkola (2013). One-pot catalytic conversion of Nordic pulp media into green platform chemicals. Applied Catalysis A, 454: 21- 29.

II Johan Ahlkvist, Samikannu Ajaikumar, William Larsson, Johan Wärnå, Tapio Salmi, Jyri-Pekka Mikkola (2013). Reaction Network upon One-pot Catalytic Conversion of Pulp. Chemical Engineering Transactions, 32: 649-654.

III Johan Ahlkvist, Päivi Mäki-Arvela, Jyri-Pekka Mikkola. Macro- molecules as a Source of Levulinic Acid (2013). The International Review of Chemical Engineering. (submitted)

IV Johan Ahlkvist, Johan Wärnå, Tapio Salmi, Jyri-Pekka Mikkola (2013). Experimental and Kinetic Modelling Studies upon Conversion of Nordic Pulp into Levulinic Acid. Industrial & Engineering Chemistry Research. (manuscript)

Authors’ contributions

I Planning, kinetic experiments, data analysis, most of the practical work except some parts of the catalyst characterization, GC and GC/MS measurements, writing article

II Planning, most of the practical work and data analysis except mathematical modelling, major writing

III Major writing

IV Planning, kinetic experiments, most of the practical work and data analysis except mathematical modelling, major writing

Abbreviations

5-HMF 5-hydroxymethylfurfural AD Air dried AHP Alkaline hydrogen peroxide pretreatment APPO Aqueous phase partial oxidation BET Stephen Brunauer, Paul Hugh Emmett and Edward Teller BJH Barrett, Joyner and Halenda CE-MS Capillary electrophoresis-mass spectrometry CPD Critical point dried D Domsjö DP Degree of polymerization FA Formic acid GC Gas chromatography GC-MS Gas chromatography-mass spectrometry GVL Gamma-valerolactone HAS High amylose corn starch HMDS Hexametyldisilazane HMF Hydroxymethylfurfural HPLC High-performance liquid chromatography IUPAC International Union of Pure and Applied Chemistry LA Levulinic acid LC-MS Liquid chromatography-mass spectrometry MCMC Markov Chain Monte Carlo method MG Mechanical grinding MilliQ Deionized water MGRS Mechanical grinding rice straw MGSERS Mechanical grinding steam explosion rice straw MS Mass spectrometry M-S Metsä-Serla MW Microwave NS Normal starch PAH Pressurized acid hydrolysis RH Rice husk RI Refractive index RID Refractive Index Detector RS Rice straw SE Steam explosion SEM Scanning electron microscopy SERS Steam explosion rice straw SG Superfine grinding SGRS Superfine grinding rice straw

SGSERS Superfine grinding steam explosion rice straw TMCS Chloromethylsilane (Trimethylchlorosilane) UV-vis Ultraviolet-visible spectroscopy

vii

Introduction

One of the main reasons for developing biorefinery processes for our bioresources is to reduce the emission of carbon dioxide and other greenhouse gases. Also, fossil fuels are a limited source of energy and will be depleted in the future. However, today the cost of processing renewables to chemicals and fuels is often too high, partly due to the fact that the traditional synthesis routes developed and optimized for hydrocarbons are not well adaptable for renewables [1]. Development of new processes for woody biomass or other lignocellulosic materials has attracted increasing attention for some time now. In fact, commercial operations and pilot units are either integrated as biorefineries combined with adjacent pulp and paper mills (utilizing e.g. black liquor and other secondary streams as sources of new products) or built as stand-alone processing units. In order to combine and re-invent the material streams of the old industry, together with other processes and practices of chemical industry, a lot of new R&D efforts are required thus utilizing the existing and new technology solutions to produce a range of fossil-free alternatives of future.

Lignocellulosic biomass such as wood is a superb raw material for converting renewables into fuels or fuel additives and chemicals, in view of the fact that the utilization of this kind of biomass does not compete with feed or food production. Chemical products produced from food-crops (e.g. corn) are questionable in a world where there is a lack of adequate nourishment for all its inhabitants. In fact, sugar crops like e.g. sugarcane is often grown for non- food purposes (primarily transportation fuel blends) on farmland that competes with human related food production. After the initial boom of first generation biochemicals and fuels made from edible biomass, an ever increasing number of the so-called second generation biofuels produced from non-edible biomass (e.g. agricultural residues, lignocellulosics and other waste materials) have been introduced.

Wood, the most common source of biomass in the world, is typically composed of 40-50% cellulose, 23-33% lignin and 25-35% of hemicellulose [2]. Cellulose is considered as a promising alternative to non-renewable natural resources for the sustainable supply of fuels and chemicals in the future. Currently, extensive research is being carried out worldwide to identify and study chemical or biological transformation pathways to convert cellulose into biofuels and feedstock chemicals [3, 4]. Furthermore, the hydrolysis process of cellulose has been cited as the main gateway to a biorefinery scheme based on the transformation of carbohydrates [5]. The Scholler process, which was developed in Germany in the 1920’s, was the first hydrolysis process for acidic conversion of lignocellulosic biomass [6].

To date, several studies have reported the conversion of biomass into levulinic acid (LA) using a dilute mineral acid, such as HCl and H2SO4 as catalysts [7-10]. Although these hydrolysis reactions proceed smoothly, the use of the mineral acids causes serious pollution and promotes equipment corrosion. Moreover, any homogeneous catalysts are difficult to recover from the reaction products for recycling [11]. Very limited attention has been given to the use of solid acid catalysis for the depolymerization of cellulose in water under realistic conditions for the possible exploitation, i.e. temperatures below about 200 °C (to avoid secondary transformation of the produced sugars). However, Rackemann et al. [12] summarizes the studies conducted on the synthesis of levulinic acid using solid acid catalysts from various feedstocks. The catalyst used in this experimental study is a macro-porous, cationic ion-exchange resin, Amberlyst 70 [13]. Today Amberlyst 70 is reasonably well known and has been used with good results as a catalyst in many types of reactions. Rinaldi et al. [14], Agirrezabal-Telleria et al. [15] as well as Hu and Li [16] have recently used Amberlyst 70 in similar type of reactions as our case.

Several kinetic studies using a various range of cellulosic materials can be found in the literature. Saeman [17] was one of the pioneers when he in 1945 presented a modelling study based on hydrolysis reaction of Douglas fir into glucose performed in batch mode. Also, several later studies are based on Saemans model. In papers by Fagan et al. [18], Thompson et al. [19], and Franzidis et al. [20], a plug-flow reactor was used to study the kinetics of the hydrolysis reaction of Kraft paper slurries, Solka Floc as well as filter paper. Also, Malester and co-workers [21, 22] were using Saemans model when the kinetics of dilute hydrolysis of cellulose originating from municipal solid wastes was investigated in a batch reactor. Nevertheless, only a few kinetic reports [7, 23, 24] are available for the acid-catalyzed hydrolysis of cellulose to levulinic acid before Girisuta et al. [9] who presented a more detailed reaction mechanism of levulinic acid (LA) formation from cellulose through glucose and 5-hydroxymethylfurfural (HMF), including the humin formation reactions. Jing and Lu [25] also published a model describing LA production based on a similar reaction mechanism performed in a batch autoclave using glucose as the reactant. Further, Shen and Wyman [26] presented a sophisticated model describing LA production in tubular batch reactors from microcrystalline cellulose and their model accounted for the major compounds as well as the side-reactions leading to by-products.

Hereupon we describe an experimental and modelling study of the acid- catalyzed hydrolysis of soft wood dissolving pulp into levulinic acid through glucose and HMF as well as xylose, furfural and formic acid from the transformation of xylans in the pulp.

Our aim was to develop a procedure whereupon a reasonable cellulose-to- catalyst ratio was applied. Also, avoiding the use of energy-intensive pretreatments methods for cellulose, such as long-time ball-milling typically used in most of the literature examples was the criteria set by ourselves [27]. Herein we report the direct catalytic conversion of wood pulp to levulinic acid and formic acid by sequential hydrolysis, dehydration–hydration reactions under relatively mild conditions.

Levulinic Acid

The first synthesis of LA by heating sucrose with mineral acids at high temperature was reported in the 1840’s by the Dutch professor G. J. Mulder [28]. This versatile renewable platform molecule has also been identified by the US Department of Energy as one of the `12 top value-added biochemicals [29].

Levulinic acid is a commodity chemical that finds applications for several purposes, such as source of polymer resins, animal feed, food as well as components of flavoring and fragrance industry, textile dyes, additives, extenders for fuels, antifreeze products, antimicrobial agents, herbicides and also plasticizers [12, 23]. Traditionally, the most convenient industrial route to levulinic acid involves production from hexose sugars [23]. More recently, the utilization of biomass and agricultural wastes for production of levulinic acid has been subject to increased interest as a research topic, since it would render the process cheaper due to the high availability and low cost of raw materials. In fact, even ‘negative’ cost of the raw materials can be achieved when wastes such as paper mill sludge are applied [30]. Importantly, a novel process based on the use of waste products would not compete with food supply. One-pot biomass hydrolysis to sugars followed by their dehydration can be tuned to produce many different derivatives, including 5- hydroxymethylfurfural and furfural. The former one can then consequently be rehydrated to levulinic and formic acids. There are a few reviews available that concern levulinic acid [12, 30, 31]. In a paper by Timokhin [31], mainly the properties of levulinic acid and its utilization in organic synthesis were reviewed, whereas transformation of lignocellulosics to levulinic acid is emphasized in ref. [12].

Reaction Network upon Production of Levulinic Acid from Biomass All in all, one-pot synthesis of levulinic acid from biomass is a demanding reaction since the whole process most conveniently starts with acid

hydrolysis of biomass thus giving rise to various issues in terms of waste produced as well as the occurrence of multiple side reactions. The desired raw material, lignocellulosic biomass, is mainly composed of cellulose, hemicelluloses and lignin of which the first two can yield the desired hexoses, the key intermediates in the production of levulinic acid. Thereafter, sugars need to be dehydrated and rearranged to 5- hydroxymethylfurfural and furfural depending on whether hexoses or pentoses, respectively, are to be dehydrated. Consequently, hexoses can then further rehydrate to formic and levulinic acids also requiring two equivalents (moles) of water. Since two different organic carboxylic acids are formed, the theoretical yield of levulinic acid from hexoses and cellulose has been reported to reach 64.5 wt % and 71.5 wt % only, respectively [32]. It should, however, be stated that the real yields of levulinic acid are typically lower than the theoretical yields due to the harsh reaction conditions required for the hydrolysis of macromolecules. Thus, the formation of humins is hardly to be avoided.

The Influence of Reaction Parameters to the Yield of Levulinic Acid

The influence of various reaction parameters during transformation of biomass to levulinic acid, such as biomass type, reaction temperature, initial concentration of feedstock, liquid-to-solid ratio, agitation speed in the reactor, potential adjacent pretreatments of biomass as well as the influence of additives is elucidated in this Section. The main emphasis was directed towards selection of optimal reaction conditions to the yield of levulinic acid and how it actually can be maximized.

The Influence of Raw Material Varieties

The influence of raw material upon production of levulinic acid is important since, for example, the presence of lignin results in changes in the optimum reaction conditions and usually more drastic conditions are needed for lignocellulosic raw material compared to cellulose. It should be pointed out here that it is difficult to directly compare the results on the basis of levulinic acid yield, since in some cases the yield has been given on the basis of initial raw material, cellulose content (as well as C6 content) in the raw material (mass) or based on the theoretical yield of levulinic acid. The highest theoretical yield of levulinic acid reported amounted to 92 % corresponding to 65.9 % yield, based on the cellulose content in the raw material (paper sludge) and achieved by a two-stage process at a temperature of 200–215 °C [33]. Comparative theoretical yields of levulinic acid were achieved from rice husk (82.9 %) [34], tabacco chops (82.5 %) [30], paddy straw (79.6 %) [10],

bagasse (78.8 %) [10] and wheat straw (77.4 %) [30]. All produced with a strong acid, hydrochloric acid (HCl) and from a waste material. When comparing different raw materials and the yield of levulinic acid, it was observed that - as also stated by Kang et al. [35] - marine biomass can be grown rapidly and is easily cultivated without the need for expensive equipment, compared to land biomass. In addition, annual CO2 absorption by marine biomass can be five to seven times higher than that of wood- biomass and the carbohydrate content is higher and can easily be converted to chemicals through proper chemical processes.

The Influence of Reaction Temperature

The influence of reaction temperature on the levulinic acid production has been intensively investigated using different feedstock’s, such as cellulose [9, 36-38] wheat straw [8, 30], kernel grain sorghum [23], water hyacinth [39], poplar sawdust [30], paper sludge [30], olive tree pruning [30] bagasse and paddy straw [10]. The effect of reaction temperature on the cellulose conversion into LA was investigated by Peng et al. [38] and the experiments were conducted at 180–220 °C, respectively. The temperature has a significant effect on LA yield. When the reaction temperature was increased above 200 °C, the yield of LA began to fall, and intermediates such as glucose and HMF couldn’t be detected any more. Higher temperatures thus could accelerate the rate of chemical reactions, but undesired side reactions like formation of humins also appeared at the same time. Therefore, higher temperatures are unfavorable for the extent of reaction. Analogously to the study of Peng et al. [38], an optimum temperature giving rise to high yield of levulinic acid has been reported for the transformation of wheat straw at about 210 °C [40]. Similar results with an optimum temperature was also obtained by Szabolcs et al. (cellobiose) [41] and Shen et al. (cellulose) [26]. However, by contrast to cellulose transformation by Girisuta et al. [42], the yield of levulinic acid was greater in the study by Fang et al. [23] at higher temperatures (200 °C) than that obtained at lower ones (160 °C) using kernel grain sorghum as the raw material. Very high yields of levulinic acid were also obtained from bagasse and paddy straw at 220 °C [10]. These results indicate that the presence of lignin requires a higher temperature in order to maximize the yield of levulinic acid compared to the case when cellulose is the raw material of choice although it is known that the humins originate from glucose and HMF [36]. Furthermore, it has been confirmed that levulinic acid is stable at 180 °C, even after 24 h [36]. However, it was also observed that at very high temperatures, above 230 °C, the amount of levulinic acid decreased in water solution due to dehydration to α- and β- angelica lactone [10]. Generally, around 170 °C is considered the optimal reaction temperature above which the humins are readily forming due to

higher equilibrium concentration of protonated pyranose forms of the sugars [41].

The Influence of Substrate Concentration

Theoretically, a higher liquid-to-solid ratio should result in a higher yield of levulinic acid. However, it has further been found that high carbohydrate loading had an adverse effect on levulinic acid yield in the reaction mixture. Therefore, one method to maximize the yield is to slowly add the carbohydrate material to the reaction mixture in order to reduce the concentration of the carbohydrate in the mixture at any given time and increase the yield of levulinic acid. In this manner the final concentration of levulinic acid will be gradually built up, resulting in a decreased cost of materials and handling [43]. Generally, lower concentration of cellulose resulted in higher yields of levulinic acid. For each levulinic acid molecule generated, two molecules of water are needed; otherwise 5- hydroxymethylfurfural (HMF) may be converted into humins [42]. These observations are in line with other studies on the acid-catalyzed decompositions of HMF and glucose, where the low substrate concentrations favored high LA yields. The highest levulinic acid yield (60 mol %), at full cellulose conversion, could be obtained at a temperature of 150 °C, whereupon an initial cellulose addition of 1.7 wt % and a sulfuric acid concentration of 1 M were applied [9]. The same phenomena have also been noted in transformation of sorghum flour [23], corn starch [44], cotton straw [45], and cellulose [26, 46].

Chang et al. [8] observed that the yield of levulinic acid in fact deteriorated after a certain optimum ratio was reached as noted by Carlson [47]. Also, Peng et al. [38] observed that the optimal concentration of substrate is very important in terms of the efficient use of cellulose and the final levulinic acid (LA) concentration reached. Higher LA yield was achieved at lower substrate concentrations, but a higher LA concentration in the product is extremely favorable since it results in lower energy consumption upon purification of LA and also reduces the amount of waste water produced. When catalyst dosage was fixed, increasing substrate concentration resulted in more cellulose available for conversion which, in turn, meant that higher LA concentration was achieved in the reaction product. However, the yield of LA was found to decrease significantly by increasing substrate concentration from 10 wt % to 15 wt %. The plausible reason for this is the product feedback inhibition, reactivity diminution or insufficient catalyst dosage for the additional cellulose [38]. In accordance to Yan et al. [10] also too much or too little water inhibits the hydrolysis process. The primary hydrolysis of the reaction system is significantly depressed if the water-to-solid ratio is too

low, at least in the case of bagasse. The reaction with paddy straw shows a much less sharp change in terms of the yield of levulinic acid reached. The highest practical yield was achieved when the water-to-solid wt-ratio was 8.5 for both raw materials.

The Influence of Acid Type

In terms of the mineral acid used, it seems that the use of hydrochloric acid results in better selectivities than that of sulphuric acid. This was observed e.g. by Galetti et al. [30] when levulinic acid was produced from paper sludge. However, the highest theoretical yield ever, 92 %, is produced from paper sludge by sulfuric acid [33]. When water insoluble biomass is processed, homogeneous acid catalysts generally appear more suitable in comparison to heterogeneous ones, considering their ability to deliver the active species into the solid or swelled biomass [48]. So far H2SO4 has been the most commonly used mineral acid, even if HCl, HBr, and H3PO4 also have been frequently reported as applicable homogeneous catalysts [49].

However, HBr and H2SO4 have been attempted, albeit with poor performance compared to HCl at low temperatures [50]. In all cases, diluted

HCl always demonstrated superior behavior in comparison to H2SO4 as the catalyst. Also, significant deposition of humin solid by-products on the reactor walls was observed and when the raw biomass had a significant content of calcium salts, like paper-mill cellulose wastes or tobacco chops, significant deposition of calcium sulphate was observed, even at lower H2SO4 concentrations. When higher H2SO4 concentrations were adopted, reactor clogging took place. On the contrary, when HCl was the catalyst of choice under optimized reaction conditions and at limited exposure times, no precipitation of salts and humins deposition was observed. Consequently,

HCl always afforded higher LA yields with respect to H2SO4. However, when Chang et al. [8] reported the optimization of levulinic acid production from wheat straw using H2SO4 under optimal conditions the yield reached, based on the weight of raw material, 19.86 %, which is comparable with the 19.3 % value obtained upon use of HCl. The use of this volatile acid catalyst enables furthermore the recovery of the produced LA by atmospheric/vacuum distillation and steam stripping whereupon about 95 % of the acid catalyst and water can be recycled [51]. These results are important in terms of the economics of the whole process and considering the possible exploitation of the solid residue, mainly due to lignin component. In fact, the residual lignin has great potential for many applications such as a substrate for soil conditioning, as a free radical scavenger for antioxidation, as a low-density filler taking place of inorganic fillers, as a component for biodegradable formulations or to be incinerated to energy. Again, if water-soluble carbohydrates are employed as substrates, heterogeneous acid catalysts

appear attractive. A key property of heterogeneous catalysts is their very easy separation from the reaction products. Furthermore, heterogeneous catalysts are easily recycled. Also, they are environmentally friendly with respect to safety and corrosiveness. Up until now, very few heterogeneous acid catalysts have been tested for the conversion of carbohydrates to yield LA. Examples thereof include ion-exchange resins [52], zeolites [53], pillared clays [54], 2- - - nafion [55], and solid super acid S2O8 /ZrO2 SiO2 Sm2O3 [20]. NbP was observed to be active in the conversion of inulin to yield LA under traditional heating. Yields in the range of 20 wt % were achieved in 1 hour without significant formation of solid by-products. LA was the predominant monomer product, while a soluble monomer by-product identified in very low concentrations was the intermediate 5-HMF. Interestingly, the use of microwave (MW) irradiation allowed for significant time savings and, at the same time, enhanced the yield from 20 up to about 30 wt %, corresponding to 40 % of the theoretical yield [30]. This investigation, carried out in a slurry batch reactor, appears very promising. Many heterogeneous catalysts up to now investigated in the synthesis of LA generally required very long reaction times (15-25 h) to reach comparable yields as e.g. in refs. [52-55]. The results of the study presented above can well be compared with those obtained with homogeneous HCl under similar thermal treatment conditions. Nevertheless, the heterogeneous NbP has one significant drawback which is that NbP remains encapsulated in the solid lignin. The encapsulated catalyst can be recovered and recycled only after suitable thermal treatments [30]. In terms of Lewis or Brönsted acids, the former types are assumed more efficient concerning levulinic acid production [56]. As a general observation, hydrochloric acid appears to be slightly better catalyst than sulfuric acid although the potential release of chlorine into the biosphere speaks against the use of HCl [41].

The Influence of Acid

Fang and Hanna [23] observed that higher sulfuric acid concentration also significantly increased the levulinic acid yield. Yield of levulinic acid was positively correlated with mineral acid concentration, which greatly boosted the yield of levulinic acid by the higher rate of hydrolysis of carbohydrates to organic acids. As observed by many authors, also Peng et al. [38] noticed that the yield of LA increased rapidly with increasing catalyst dosage until an optimum level was reached. After this only a minor increase of the LA yield was achieved. Consequently, the addition of excess catalyst has little or even negative effect on the LA yield. Accordingly, when the requirements of the reaction are satisfied, no more should be added. Similar observation was noted upon LA production catalyzed by other catalysts such as H2SO4 [39] and HCl [8]. When working with CrCl3 (0.02 M), less catalyst was needed

than in the case of H2SO4 (0.1–1 M) [8, 9, 39] and HCl (~1.2 M) [10]. In this case, the most beneficial concentration of CrCl3 was found to be 0.02 M when considering the economic aspects and the efficiency. Too high concentration and long reaction time were in favor of causing side-reactions that may negatively affect the rate of hydrolysis of the cellulose to levulinic acid [37]. The plausible reason for this is the existence of parallel reactions upon transformations occurring [45]. A higher catalyst concentration was not only unnecessary but undesirable in that it was found to complicate the purification of the final product [57]. If higher acid concentrations are used, difficulties are encountered in terms of the operation of the process and, at the same time, no beneficial yield gains could be observed [44]. Chang et al. [8] also observed that mineral acid with higher concentration also caused severe corrosion in the reactor equipment – a well known fact. Mineral acid concentration in combination with increasing reaction temperature, determines how corrosive the reactants are to the metal parts of the reactor. At a given temperature there is a maximum allowable concentration of a mineral acid for a specific metal or alloy [23]. The disposal of the catalysts will not only waste resources, but also cause environmental pollution, so it is extremely important that particularly metal catalysts like chromium are recycled [38].

In order to verify the optimum acid concentration, Szabolcs et al. [41] performed dehydration of cellulose and cellobiose in a concentration range of 0.1–2.5 M H2SO4. The optimum conditions for cellobiose were found to be 2 M H2SO4 at 170 °C for 10 min; and for cellulose 2 M H2SO4 at 170 °C for 30 min. In fact, Yan et al. [10] also found that low concentration of hydrochloride acid resulted in too low concentration of H+ ions thus being unable to facilitate cellulose hydrolysis and, consequently, a low reaction rate was observed. In the case of raw materials containing primarily C5 sugars, low acid concentrations and high temperatures were beneficial when aiming at high furfural yields. Typical streams of this kind are e.g. pre-hydrolysis liquors or pulping/papermaking waste streams. On the other hand, as expected, low acid concentrations and short reaction times lead to low levulinic acid yields. However, using low acid concentrations is problematic, because the lignin present in the biomass consumes a fraction of the acid during the reaction. Due to the low initial concentration of acid, the amount of acid lost is significant, and the LA yield decreased from 57 % to 27 % upon third batch of corn stover being added to the reaction mixture [58]. In case of starch, higher sulfuric acid concentrations favor levulinic acid formation and a large weight ratio of 0.3 (sulfuric acid-to-starch) was found optimal [59]. In case of sulfuric acid, the yield of levulinic acid falls at higher temperatures due to oxidation of raw material by the acid demonstrated by the presence of hydrogen sulfide formed [50]. Shen and Wyman observed

that an increase in HCl acid concentration caused a slight increase in the maximum LA yield from 54.8 % to 59.7 %, respectively. This outcome is consistent with the role of hydrochloric acid as a catalyst not changing chemical equilibrium under ideal conditions even though the chemical reaction rate would increase with increasing acid concentration. As a result, the maximum LA concentration and yield changed little for the three acid concentrations applied. However, increasing acid concentration reduced the time needed to attain the maximum LA yield from about 20 min with an acid concentration of 0.309 M to only 6 min at 0.926 M. This result confirmed that acids accelerated the reactions. Moreover, LA concentrations and yields dropped if the solution was held beyond the optimum time due to by- product formation [26].

The Influence of Reaction Time

The normal trend in the most reactions carried out at moderate reaction temperatures is that increasing reaction time also results in significantly increased levulinic acid yield. However, Chang et al. [8] also observed that time had a relative significant correlation with other variables. Temperature, acid concentration and liquid-to-solid ratio influenced significantly the time required for the over-all reaction to proceed. Moreover, an optimum temperature existed in the conversion process. It was suggested that if the hydrolysis was continued for too long a time, the yield decreased, again as the result of side reactions occurring that consumed the levulinic acid already formed. Peng et al. [38] noted that the amount of LA increased very fast during the first 140 min. After this it was observed that only a slight rise occurred after 140 min of reaction time. Finally, it was found that 180 min was sufficient for completion of the hydrolysis reaction. Yan et al. [10] observed that bagasse was hydrolyzed rapidly during the first 10 min after which the hydrolysis rate slowed down and the yield rose gently during the last 20 min. whereas the rate of the hydrolysis reaction for paddy straw reached its highest point after 25 min. The reaction proceeded well for 10 min, and then the reaction progressed much more sluggishly. The reason for these differences in kinetics may lie in the different cellulose content, different cellulose type and the intrinsic composition structures. The reaction was interrupted after 45 min because side reactions started to occur when the degradation products were exposed to high temperatures for a prolonged period of time. Thomas [57] also noticed that continuing the digestion of levulose, dextrose or starch with hydrochloric acid for periods longer than one hour actually resulted in decreased yields of levulinic acid, the effect decreasing in this order. Alonso et al. [58] observed that short reaction times and low acid concentrations prevent furfural degradation reactions and leave the cellulose mostly unconverted, leading to low LA

yields (less than 11 %). On the other hand, Cha et al. [59] were using a reactive extrusion process to hydrolyze starch in the presence of 5 wt % sulfuric acid. They noticed that the levulinic acid yield increased as reaction time increased from 20 to 60 min. However, the yields in the samples containing 25 wt % extruded starch and further reacted at 40 and 60 min of reaction time were not significantly different. Wang et al. [37] also investigated the effect of the reaction temperature on the levulinic acid yield. It was concluded that the reaction rate is very sensitive to the temperature and the rate was reduced dramatically at lower temperatures. As already stated, higher temperatures favor the production of more LA, also observed by Kang et al. [35].

The Influence of Pre-treatment

First of all it should be pointed out here that it is extremely difficult to directly compare the values of levulinic acid yield reported in the literature. It is very important to note that some of the raw materials were not really ‘pure’ raw materials in its proper sense. In fact, several starting materials had already been exposed to some type of pre-treatment or purification processes. Because of these reasons the results obtained may display various ‘unfair’ advantages in comparison to studies where the starting materials were not pre-treated in any form. When Galletti et al. [30] were using poplar sawdust as a substrate, the performance could be improved not only by adopting MW irradiation, but also by carrying out traditional heating in two steps: a pre-hydrolysis at lower temperature and a successive conversion step at 200 °C. The pretreatment at a lower temperature, such as 120 °C for 2 h, dissolves part of the hemicelluloses thus yielding cellulose more easily accessible and, consequently, increasing the yield of levulinic acid during the second hydrolysis step. Upon a preliminary screening, the adoption of the pre-hydrolysis step appears as a more immediate approach in order to improve the LA yields from raw biomass in comparison to MW irradiation [30].

Chen et al. [60] found out that steam explosion (SE) combined with superfine grinding (SG) of rice straw (RS) could effectively increase levulinic acid (LA) yield due to the reduction of particle size and the increase of specific surface area which would enhancing the accessibility of cellulose [61]. It was found that LA yield reached upon a SGSERS sequence amounted to 22.8 %, accounting for 70 % of the theoretical value, which was 61 % higher than obtained when applying the sequence mechanical grinding (MG) SERS. When comparing MGRS and SGRS sequences, LA yield of MGSERS and SGSERS were increased by 95 % and 68 %, respectively. Pretreatment lowers the polymerization grade and the crystallinity of the cellulose, which

can exert a significant influence on the hydrolysis yield. Additionally, the chemical pretreatment stage of the raw material aims to partially remove the lignin bound to the hemicelluloses without degrading the cellulosic chain [62]. For this reason Bevilaqua et al. [34] conducted in their study some pretreatments focusing on the delignification of the biomass and the elimination of non-contributing substances, like extractives and metals. To establish the best precondition for pressurized acid hydrolysis (PAH), rice husk (RH) was submitted to different pretreatments. From an analysis of the pretreatment procedures, it can be seen that the strategy of applying aqueous extraction via Soxhlet extraction resulted in a higher LA yield (18.2 g L-1), probably due to the elimination of interfering metals and an improved performance PAH on RH. There was no apparent advantage in pretreatment with oxalic acid aiming at removing metals or with hydrogen peroxide upon an alkaline medium (delignification). It is evident, in this case, that a part of the glucose was released during the pretreatment. Soxhlet extraction of RH with an organic solvent, with the aim of eliminating extractives and pretreatment with chlorite (RH delignification and facilitation of cellulose hydrolysis), did not exert any significant influence on the LA yield. Also, Ghorpade and Hanna [63] patented a process for producing levulinic acid using an extrusion process, which is advantageous in that it is continuous, requires fewer steps and reduces reaction time compared with a batch reactor process. However, the extrusion process is limited due to the loss of evaporated gases containing HMF. In Cha and Hanna’s [59] study extrusion has been used as pre-treatment in a ‘two-stage’ process. After extrusion, the extrudates were reacted with additional water and sulfuric acid in a pressurized reactor. Extrusion processing increased levulinic acid yield because the extrusion processing increased the hydrolysis of starch in the presence of an acid. The yields of glucose and levulinic acid increased at the higher extrusion temperature and slower screw speed (longer residence time). High amylose corn starch (HAS) was also converted more readily to glucose and levulinic acid than normal starch (NS). Extrusion of both starch varieties resulted in even higher yields.

The cellulose hydrolysis study of Stijn Van de Vyver et al. [64] indicates that the reaction rate is strongly dependent on the degree of crystallinity, well in agreement with previous studies. Since the cleavage of the b-1,4-glycosidic bonds is considered to be rate-determining, hydrolysis of ball-milled cellulose gave the highest conversion, carbon efficiency and selectivity among the five cellulose fractions, with a LA yield higher than 27 %. Interestingly, the observed LA yield corresponds well to the results for cellobiose, cellohexaose and even starch. The latter observation should be explained based on the low crystallinity index of 4.1 % for ball-milled cellulose when compared to the substantially higher crystallinities for

commercial cellulose substrates. Generally, a limited accessibility of the b- 1,4-glycosidic linkages of cellulose is likely responsible for slower hydrolysis rates at higher crystallinity. Alonso et al. [58] have been using corn stover which has been subjected to an alkaline hydrogen peroxide pretreatment (AHP), which partially removes the lignin [65, 66]. By using AHP corn stover, the yield of LA increases slightly to 69 % in the first batch, and remained high (overall 53 %) after the third batch of stover. Removing the lignin was not necessary for the process; however the sulfuric acid was retained in the reactor, and not consumed by the lignin. The use of the AHP corn stover allows for lower acid concentrations and shorter reaction times to be employed, resulting in furfural yields up to 96 %, with low conversion of cellulose.

The influence of Additives

Already in the 1940’s Thompson [43] noticed when converting hydrol (mother liquor from industrial, crystalline corn sugar) that the presence of sodium chloride in the hydrolysis mixture will not only increase the speed of the reaction but will also increase the yield of levulinic acid. The phenomenon was at that time not fully understood, but it was speculated that the presence of the sodium chloride in the reaction mixture increases the activity of the mineral acid. The best results were obtained when the salt- to-water wt-ratio was 6. Galletti et al. [30] also observed that the addition of a cheap electrolyte to the reaction medium, such as NaCl, increased the LA yield for some substrates such as wheat straw and tobacco chops. When water slurry of tobacco chops powder containing 25 wt % of cellulose was employed as the starting material, the actual yield of LA reached about 83 % of the theoretical yield. The reason given was the combined effect of both the addition of the electrolyte and the pre-hydrolysis step at 120 °C. This yield is one of the highest ever reported using a waste biomass in a batch process. Also, more generally, the yields reached with the different substrates are promising if compared with those that have been reported until now [51]. In essence, it is known that the addition of an electrolyte can increase the rate of the heterogeneous acid hydrolysis of cellulose to convert to glucose and then LA, while enhancing the accessibility of the rest portion of the cellulose [67]. Liu et al. [68] also reported that inorganic salts could increase the hydrolysis rate of hemicellulose components during dilute acid pretreatment. This salt-effect has been explained by applying the Donnan’s theory of membrane equilibria to the heterogeneous hydrolysis of the + cellulose in dilute acid. The difference of concentration of H3O ions in the two phases can be minimized by the addition of an inert, diffusible salt. The addition of an electrolyte should result in an increase in the concentration of H3O+ ions in the cellulose phase, thus increasing its swelling and the rate of

protonation of reaction sites [30]. The effect of added salts on Nafion as a solid acid catalyst for the hydrolysis of cellulose was also investigated by Potvin et al. [69]. The result of the investigation was that the yield of levulinic acid can be increased 5-fold from 14 % to 72 % upon incorporation of a 25 wt % NaCl solution. The use of a different salt, such as potassium chloride, was also investigated to determine if the salt effect was general. Reactions using 10 wt % KCl was less powerful than reactions with comparable concentrations of NaCl. On the contrary, 5 wt % KCl solution induced higher product yields than the use of 5 wt % NaCl solution. The yields were not significantly different between the sodium and potassium chloride trials indicating that cation size may not have a large influence on the product yield. They hypothesize that the greater concentration of NaCl allows for a greater concentration of ions in solution. These charged particles interact with the hydrogen bonding network of the cellulose structure. The dramatic increase in yield indicates that sodium chloride is effective in interrupting the hydrogen bonding network at high temperatures and pressures. Wettstein et al. [70] have also been using a solute additive for the conversion of cellulose. They have demonstrated a biphasic process to make GVL (gamma-valerolactone), in which cellulose is deconstructed to produce LA and FA using a dilute aqueous solution of hydrochloric acid (HCl) with a sufficient amount of a dissolved solute in the solution to yield a biphasic system with GVL. The top phase is an organic phase rich in GVL and LA, while the bottom phase is an aqueous phase rich in mineral acid. The maximum yield of LA achieved was 72 % after the first batch of cellulose and using 1.25 M HCl saturated with 35 wt % NaCl and a 1:1 aqueous-to-organic wt-ratio. During the reaction, the GVL continuously extracts the LA and FA from the aqueous phase, and solubilizes the humins that form. The aqueous phase can be used to carry out additional cellulose deconstructions. The organic phase is then transferred to a hydrogenation reactor, where the LA is converted into GVL (using a Ru–Sn/C catalyst) that can be used directly or further upgraded to chemicals [71-73] or transportation fuels [74, 75]. A biphasic reaction system using an aqueous phase containing a phase modifier (e.g., salt and sugars) and GVL as a solvent offers significant advantages in the production of GVL from biomass derived cellulose. This system solves the following three important issues that limit current processing approaches: eliminating the need for a filtration step after cellulose deconstruction, eliminating the need for a step to separate the product from the solvent (the product is the solvent), and using a renewable solvent. Alonso et al. [58] have taken the use of GVL a step further in the conversion of lignocellulosic biomass (corn stover). By using GVL as a solvent broadens the optimal processing conditions allowing the conversion of hemicelluloses and cellulose simultaneously in a single reactor, thus eliminating pre-treatment steps to fractionate biomass and simplifying

product separation since GVL is a product of the process. Other important advantages are elimination of reactor clogging and other solids handling problems because GVL effectively solubilizes the biomass (and humins). The GVL solvent also increases the rate of cellulose conversion and decreases the rate of furfural degradation and is completely miscible with water, allowing wet biomass to be used in the process.

Lin et al. [46] was converting cellulose over a ZrO2 catalyst by a one-pot catalytic aqueous phase partial oxidation (APPO) process. Under the APPO conditions, the molar ratio of formic acid to LA was only ~0.35 with a ZrO2 catalyst. The low formic/levulinic acid molar ratio implies that, unlike acid hydrolysis, APPO follows a pathway in which formic acid is no longer a major by-product. Nevertheless, the LA yield is sensitive to the oxygen partial pressure and the maximum yield of LA was obtained with 2.8 % O2. Redox, rather than acid-base, properties of the ZrO2 catalyst play the central role in the oxidative deconstruction of cellulose in the APPO reactions. Higher oxygen partial pressures resulted in lower LA yields as LA is further degraded to acetic acid and CO2 by strong oxidation. The results also show that there is a decrease of 5 % in LA yield when not feeding O2 compared to the case with 2.8 % O2 loading. Another unique feature of the APPO process is the low yield of water insoluble humins. It is proposed that HMF is the precursor of humins during the acid hydrolysis of cellulose [76]. However, the proposed APPO reaction mechanism indicates that gluconic acid and not HMF, is the key intermediate to produce LA. This is very likely the reason why the yield of humins was lower than in a pure acid catalyzed hydrolysis reaction.

The Influence of Agitation Speed

It is well known that agitation speed is an important factor in solid-liquid phase catalytic reaction systems. If the agitation speed is too low the reaction may suffer from low reaction rate, low conversion and low product yields etc. This is a result of high interfacial mass transfer resistances between the two phases. Increasing the agitation speed might increase the contacting area and relative velocity between the two phases, and hence diminish any external diffusion limitations [77]. The effect of agitation speed on the yields of reaction products was, in fact, investigated by Peng et al. [38]. The hydrolytic reaction was carried out by varying the agitation speed from 0– 600 rpm using cellulose as a feedstock and chromium chloride as a catalyst. As can be seen, the interfacial mass transfer resistances between the solid surface and water phase appear as negligible when the agitation speed is higher than 200 rpm considering levulinic acid yield. However, one could not observe any clear influence of the agitation speed in terms of the yields of

glucose and 5-hydroximethylfurfural (HMF) - most probably because they are properly dissolved in the liquid and thus can be well contacted with the catalyst. Nevertheless, the reader should be reminded that the design of the stirrer as well as the baffling of the reaction vessel have a significant influence in terms of the mixing performance.

Materials and methods

Chemicals

The commercially acquired reagents and chemicals used were of analytical grade and used without further purifications or other treatment procedures. D-glucose (VWR International, >99 %), 5-hydroxymethylfurfural (Aldrich, 99 %), levulinic acid (Janssen Chimica, 98+ %) and formic acid (Sigma- Aldrich, 98-100 %) were used to facilitate the calibration curves needed to follow the reaction kinetics. HPLC grade sulfuric acid (Fluka, 50 %) was used for eluent preparation. Pyridine (Fluka, ≥99 %), hexamethyldisilazane (Fluka, ≥99 %) and chloromethylsilane (Fluka ≥99 %) were used for the silylation. Phenolphthalein (Fluka) and Sodium chloride (Sigma-Aldrich, ≥99 %) were used for the acid titration of catalyst. Whenever water was used, deionized water (MilliQ) was used to prepare all the solutions.

Raw materials

Most of the focus has been given to investigate the hydrolytic conversion of Nordic, bleached, industrial pulp obtained as a gift from the Aditya Birla Domsjö pulp mill in Sweden (a mixture of softwoods Norway Spruce and Scots Pine), however Metsä-Serlas kraft pulp from Finland (hardwood birch, Betula pendula) were compared as a competing raw material. The softwood pulp utilized has a typical degree of polymerization (DP) in the range of 950 as an average whereas the DP of hardwood approaches 1900 on the basis of viscosity measurements (Scandinavian Standard SCAN-CM 15:99) [78]. Kraft (sulphate) cellulose typically contains significant amounts of residual hemicelluloses (16 wt % xylose pentose) [79] whereas sulphite pulp is lean on hemicelluloses (sum of xylose and mannose pentoses comprise 4.2 wt %) [80]. Native Birch wood (Betula pendula) has been reported to contain 23 % of xylose, bound in the form of the hemicellulose xylanes [81]. Monosaccharide’s originating from hemicelluloses of four different bleached birch pulps from Finnish paper mills has been studied earlier and it was found that the average amount of xylose found in the pulps resides in the range of 14.3 wt % (13.6–15.3 wt %) [82]. The pulps were utilized without any prior chemical treatments but both the sulphite and sulphate pulp were initially disintegrated by a centrifugal mill (ZM 200, Retch, Germany) with a cutting sieve to an average particle size of 0.35 mm in order to facilitate a realistic comparison. Since the degradation and transformations of the cellulose polymers initially proceeds via hydrolysis to glucose, also transformation of pure D-glucose was investigated in order to obtain insight into the general reaction behaviour.

Acidic polymer catalyst

The acidic polymer based catalyst, Amberlyst 70, was obtained as a gift from the Rohm and Haas Company [13]. The catalyst is a macro-porous, polymeric, cationic ion exchange resin material applicable in high- temperature applications and is designed to be used in various processes such as olefin hydration, esterification and aromatic alkylation. In the aforementioned applications, Amberlyst 70 offers good performance over conventional resins due to its high thermal stability (up to 190 °C), catalytic activity and good mechanical strength. Importantly, very low sulphonate leaching has also been confirmed. The catalyst has an appearance as dark brown and spherical particles with an effective particle size of around 0.50 mm. The average pore diameter corresponds to 220 Å (22 nm) and the nitrogen physisorption (BET area) to 36 m2 g−1 [13, 83, 84].

Reactor system

All kinetic experiments were carried out batch wise in a standard 300 ml laboratory scale, three-phase stainless-steel autoclave (Parr Inc, USA) designed for high pressures and temperatures (see Fig. 1.). The temperature was measured with a thermocouple and automatically adjusted with the inbuilt temperature regulator (Parr 4848 Reactor Controller). The autoclave was equipped with a sampling outlet fitted with a 7 μm metal sinter frit designed to prevent the catalyst particles as well as pulp pieces to leave the reaction vessel upon sampling of the liquid phase. The hydrolytic degradation of wood pulp took place in the presence of catalysts in an aqueous media and high stirring rates (1000 rpm) were applied in order to eliminate any external mass transfer limitations. The reaction parameters studied, such as the reaction time, temperature, pressure as well as the nature of the catalyst was of crucial importance in terms of optimal conversion of the biopolymers to platform chemicals. Each and every experiment was commenced by charging the reaction vessel with a slurry containing 1–9 g of pulp dispersed in 141–149 ml of deionized water along with 1–9 g of catalyst. Before heating the reaction mixture to the intended temperature (150–200 °C), the system was degassed (<0.01 atm) with a high power vacuum pump to remove any residual air in the solution. After reaching the preset temperature the agitation was engaged, gas atmosphere

(CO2, Ar, H2) was applied at set pressure (3–50 bar excluding vapour pressure from water) and this was considered as the initial start of the reaction. A blank reaction (without catalyst) has also been performed at 200

°C with a higher CO2 pressure than normal reaction conditions (90 bar instead of 50 bar) to exclude that carbonic acid catalyzes the reaction. Throughout the experimental run, samples were taken at suitable intervals

and the sampling line was flushed with small volumes of the reaction solution in between the samples to avoid any contamination between consecutive samples. For the sake of precaution, the samples were once more filtered and the composition was determined by means of high-performance liquid chromatography (HPLC).

Gas inlet

Water in/outlet

RPMI TI Reactor Controller Sampling PI RPMIC

TIC

PI PI Pressure indicator Reactor TI Temperature indicator Heating jacket TIC Temperature indicator & Thermocouple controller PC Coil RPMI RPM indicator Sinter RPMIC RPM indicator & controller Agitator

Figure 1. Schematic picture of the experimental set-up used for the catalytic transformation of wood pulp in batch mode.

Substrate and catalyst characterization

The substrates and catalyst were characterized by means of the following methods: scanning electron microscopy, nitrogen physisorption and acid- base titration. Also the wood pulps were characterized by scanning Electron Microscopy (SEM): The morphology of the two wood pulps, kraft pulp from Metsä-Serla and the sulfite pulp from Aditya Birla (Domsjö) was investigated by SEM (Cambridge Leica Stereoscan 360). The samples were prepared with a thin layer (50 Å) of gold and carbon before they were introduced to the measurement chamber. Some of the samples were also critical-point-dried for a comparison with air-dried samples. Also the morphology of the Amberlyst 70 catalyst was investigated by SEM (Cambridge Stereoscan 360). Further nitrogen adsorption–desorption measurements were conducted for the fresh and spent catalyst using Micromeritics TriStar 3000 porosimeter. Adsorption–desorption isotherms were recorded at −196 °C after degassed the samples at 120 °C. Surface areas were calculated by B.E.T method and pore volumes were calculated from desorption isotherm. Pore size

distribution was estimated using the Barrett, Joyner and Halenda (BJH) algorithm (ASAP-2010) available as built-in software from Micromeritics. The acid site concentration of the catalyst was measured by means of a conventional titration method [85]. 1 g of the resin was added to about 50 ml of NaCl solution (200 g L-1). The cation exchange between H+ and Na+ was allowed to take place for 24 h upon stirring of the system. The mixture was then titrated with a 0.1 N NaOH solution. The measurement was carried out on fresh catalyst and on spent catalysts exposed to the reaction temperatures of 190 and 200 °C, respectively.

Nitrogen physisorption

Surface area, pore size distribution and pore volume is an attribute that is used by catalyst manufacturers and users to monitor the activity and stability of catalysts as well as a quality control tool during catalyst manufacture. There are different methods used to measure surface area and each method can yield different results. In physisorption, a monolayer formation of inert gas, mostly nitrogen, is adsorbed by weak Van der Waals attractions on the surface of a solid material or, in case of porous materials, also on the surface of pores. Most widely known is the determination of the B.E.T (Stephen Brunauer, Paul Hugh Emmett, and Edward Teller) surface area by gas adsorption. Adsorption of nitrogen at a temperature of 77 K leads to a so- called adsorption isotherm, sometimes referred to as the B.E.T. isotherm, which is the most common way to express measured porosity of materials. The principle of capillary condensation is usually applied to assess the presence of pores, pore volume and pore size distribution. Pre-treatment of the sample is performed at elevated temperature in vacuum or either by a flowing inert gas in order to remove any contaminants.

Scanning electron microscopy

A scanning electron microscope (SEM) is a type of electron microscope that produces micrographs of a solid specimen by scanning it with a focused beam of high-energy electrons. The electrons interact with atoms in the sample, producing signals that can be detected and that contain information about the surface topography, chemical composition, crystalline structure and orientation of materials making up the sample. In most applications, information is collected over a selected area of the surface of the sample, and a micrograph is generated that displays spatial variations in these properties. Generally the electron beam is scanned in a raster scan pattern, and the beam's position is combined with the detected signal. A two-dimensional micrograph with a resolution better than 1 nanometer can be produced. The most common mode of detection is by secondary electrons emitted by atoms

excited by the electron beam. The number of secondary electrons is a function of the angle between the surface and the beam. On a flat surface, the plume of secondary electrons is mostly contained by the sample. However on a tilted surface, the plume is partially exposed and more electrons are emitted. By scanning the sample and detecting the secondary electrons, a micrograph displaying the tilt of the surface is generated.

Acid-base titration

The acid-base titration process consists of the addition of a solution with known concentration, from a buret, to a measured volume of a solution of the other reactant or to a weighed sample dissolved in water, until the same number of equivalents of each substance has been used. The end point in the titration is detected by means of a suitable indicator. Phenolphthalein, which is red in basic solution and colorless in acid solution, is often used for acid- base titrations. The end point, where the permanent color change just occurs after thorough mixing, is usually very sharp. Often a drop or a fraction of a drop will bring about a color change.

Product analysis

The quantitative composition of the samples in the liquid-phase was analyzed off-line with high-performance liquid chromatography (HPLC, Agilent Technologies 1200 Series) equipped with a refractive index (RI) detector, a degasser, a double-channel binary pump and an auto sampler. The analytical column of choice (Bio-Rad Aminex HPX-87H) is a 300 x 7.8 mm, a pre-packed HPLC column suited for carbohydrate and acidic compounds analysis supplied in the proton form containing 9 μm particle size with 8 % cross linkage and is designed to operate at a pH range of 1–3. 5 3 -1 mM H2SO4 was utilized as the mobile phase with a flow rate of 0.5 cm min and temperature of 60 °C. The samples were injected into the HPLCs directly after the sample was taken without any pre-treatment other than filtering (2 µm) to prevent column damage. The analysis for a sample was completed within 1 h. The concentrations of each compound in the liquid-phase mixture were determined using calibration curves obtained by analyzing standard solutions with known concentrations. For the sake of the clarity, instead of more commonly applied UV-vis detector, refractive index (RI) detection is a must upon analysis of carbohydrates and their derivatives since UV detection is not capable of resolving most of the species. A Bio-Rad Aminex HPX-87C carbohydrate column was also tested, however the acidic main compounds, formic- and levulinic acid, peaks overlapped each other (same retention times).

Other possible by-products of the acid-catalyzed hydrolysis of cellulose are gas-phase components from thermal degradation (eventually catalyzed by the autoclave walls) reactions of reactants, intermediates and/or products. To gain insights into the extent of these reactions, a GC study has been carried out to analyze the gas phase composition during an experiment performed at 180 ◦C without any gas addition. A sample of non-condensable gases were taken with a gas syringe and analyzed with a Varian 490-GC Micro-GC. To identify the compounds in the liquid phase gas chromatography-mass spectrometry (GC-MS) analysis have been performed. The samples were derivatized to make the compounds volatile and stable, suitable for gas chromatography analysis. About three drops (150 µl) of the liquid samples were transferred to a little test tube and evaporated in a water bath operating at 40 °C under nitrogen flow. As a reference a blank sample containing only distilled water was dried and used. The dried samples were silylated by addition of 150-200 µl of pyridine, 200 µl of HMDS silylation reagent (hexametyldisilazane) and, finally, 100 µl of TMCS reagent (chloromethylsilane). After half an hour of stirring at room temperature the samples were ready for analysis. The samples were transferred into small ampoules and analyzed with GC-MS. The GC-MS was equipped with an HP-1 column (25 m x 0.2 mm x 0.11 µm) and a flow of 0.8 ml/min (helium) was used and the split was 15 ml/min. The heating program started at 60 °C with a heating rate of 8 °C/min up to 325 °C.

Liquid chromatography

Although gas chromatography is widely used, it is limited to samples that are thermally stable and easily volatilized. Nonvolatile samples, such as peptides and carbohydrates, can be analyzed by GC, but only after they have been made more volatile by a suitable chemical derivatization. For this reason, the various techniques included within the general scope of liquid chromatography are among the most commonly used separation techniques. In high-performance liquid chromatography (HPLC), a liquid sample, or a solid sample dissolved in a suitable solvent, is carried through a chromatographic column by a liquid mobile phase. Separation is determined by solute/stationary-phase interactions, including liquid–solid adsorption, liquid–liquid partitioning, ion exchange and size exclusion, and by solute/mobile-phase interactions. In each case, however, the basic instrumentation is essentially the same [86]

Over the past three to four decades, high-performance liquid chromatography (HPLC) has been the method of choice for the determination of carbohydrate species, and as a result a large number of HPLC methods have been developed to determine a wide variety of

carbohydrate samples. One of the most challenging areas of analytical chemistry is the development of selective and sensitive methods for analyses of carbohydrates. The main difficulties in the analysis of carbohydrates arise from their considerable number of isomeric forms due to the various possible configurations of the monosaccharides. Oligo- and polysaccharides are composed of different combinations of them, forming linear or branched polymeric species, many of which may differ only in the position of attachment and anomeric configuration of the glycosidic linkages. RI detection is the most popular detection technique for carbohydrates in HPLC but a serious drawback is that it can only be used to monitor isocratic separations [87].

Gas chromatography

In gas chromatography (GC) the sample, which may be a gas or liquid, is injected into a stream of an inert gaseous mobile phase (carrier gas). The sample is carried through a packed or capillary column where the sample’s components separate based on their ability to distribute themselves between the mobile and stationary phases. The most common mobile phases for GC are He, Ar, and N2, which have the advantage of being chemically inert toward both the sample and the stationary phase. The choice of which carrier gas to use is often determined by the instrument’s detector [86].

Mass spectrometry

A mass spectrometer consists of three major parts: the ion source, the mass analyzer, and the detector. The first mass spectrometer was used in England by JJ Thompson in 1912, and by F. W. Aston in 1919, but the instrument that resembles today's mass spectrometers was first constructed in 1932 [88]. Over the last two decades mass spectrometry has become one of the central techniques in analytical chemistry. Its importance is now comparable to that of the more traditional electrophoresis, liquid and gas separations techniques, and it is often used in conjunction with them as so-called “hyphenated” techniques, such as CE-MS, LC-MS and GC-MS. Mass spectrometry is used for both quantitative and qualitative analysis and is based on production of unstable ions from neutral species, that is, atoms, molecules or clusters and the study of the decomposition (fragmentation) of these. The separation should also be independent of the chemical conformation of the species. When a compound is ionized in vacuum the compound decays whereby new ions are formed. The apparatus must be kept under vacuum (typically <10-6 mmHg) to prevent reactions between the unstable fragment ions and gases, resulting in mass spectra that cannot be interpreted. The ions are separated and deposited in a chart relative to the

proportion of ions towards the ion mass, resulting in a so-called mass spectrum. Each compound decomposes in a manner which is unique precisely for this compound and thus a mass spectrum is obtained that is specific for that molecule. The decomposition gives us information about the association structure, which allows the mass spectrum can be interpreted and compound can be identified. All mass analyzers presently in use are based on electromagnetism so ions are required to obtain separation. The different types of mass spectrometers differ in regard to the mechanism by which ions are separated. There are several types of mass analyzers used in mass spectrometric research and they can be divided into different categories, such as magnetic or pure electric, scanning/non-scanning (pulse based), and trapping/non-trapping analyzers [89].

Derivatization for GC/MS

Silylation is the most widely used derivatization procedure for gas chromatography and mass spectrometry analysis and have been used since the late fifties. A wide variety of products (and functional groups) like alcohols, carboxylic acids, amines, thiols, and phosphates etc. can be derivatized by silylation. The process involves the replacement of an active proton with a trialkylsilyl group, typically trimethylsilyl (-SiMe3). Compared to their parent compounds, a silyl derivative generally are more volatile, less polar, and more thermally stable and enables the GC-MS analysis of many involatile and too unstable compounds [90].

Definitions for the yield calculations of levulinic acid (LA) and formic acid (FA)

The yield of levulinic acid (LA) on molar basis (풀푳푨) was defined as the ratio between the LA concentration in the reaction product (푪푳푨) and the concentration of the available C6-sugars (푪푪ퟔ ) in cellulose and hemicelluloses fractions in the pulp:

퐶퐿퐴 푌퐿퐴(푚표푙 %) = × 100 % (1) 퐶퐶6

It is also possible to define the yield of LA on a weight base (풀푳푨,풘풕), corresponding to the mass ratio between the LA and the available C6-sugars in the pulp:

퐶퐿퐴 × 푀퐿퐴 푌퐿퐴,푤푡(푤푡 %) = × 100 % (2) 퐶퐶6 × 푀퐶6

In eq. (2), the terms and represent the molecular weight of LA (116 g mol–1) –1 and C6-sugars (180 g mol ), respectively.

The yield of LA can also be defined as the ratio between the mass of LA and the total mass of the oven-dried pulp:

퐶퐿퐴×푀퐿퐴 푌퐿퐴,푡표푡푎푙(푤푡 %) = × 100 % (3) 푚푝푢푙푝

where 풎풑풖풍풑 represent the mass of the oven-dried pulp.

The yield of formic acid (FA) on molar basis (풀푭푨 ) was calculated in the same manner except that the sugar concentration was based on the available

C6- and C5-sugars in cellulose and hemicelluloses fractions in the pulp. The theoretical yield of LA from C6-sugars is 100 mol %, or 64.5 wt % due to the co-production of formic acid.

Modelling software and techniques

For the modeling of the reaction kinetics, the software Modest® was used [91]. The software solves the reactor model equations (system of ordinary differential equations) with the backward difference method and optimizes the parameter values using a hybrid method involving Simflex and Levenberg-Marquardt methods.

The software minimizes the following objective function

2 푄 = ∑ ∑(푐푖,푡,푒푥푝 − 푐푖,푡,푚표푑푒푙) 푤푖,푡 (4) 푡 푖

Results and Discussion

Catalyst Characterization

In Fig. 2 we find the N2 adsorption–desorption isotherms of the fresh and spent Amberlyst 70 catalysts. The results exhibit that the materials have a sharp inflection at higher relative pressures in the range of 0.7–1.0p/p0. This corresponds to type-V curves according to the IUPAC classification, characteristic to large pore materials. The hysteresis loop observed at the inflection, reflects the irregularity of the pore sizes over the surface. The results indicate that the surface areas of fresh and spent catalysts are 0.165 and 35.72 m2 g−1, respectively; at the same time, the pore volumes were measured to reside at 0.0015 and 0.168 cm3 g−1; and the average pore sizes at 37.5 and 18.16 nm, respectively. The results of the spent catalysts revealed that the morphological properties are clearly altered during the reaction. The surface area and pore volume of the spent catalyst were increased enormously; this might be due to swelling nature of the catalyst in aqueous reaction conditions.

Fig. 2. Nitrogen adsorption–desorption curves for (a) fresh Amberlyst 70 and (b) spent Amberlyst 70.

The changes in the surface properties were further evidenced with SEM images of the catalysts. The scanning electron micrographs (SEM) show sponge like porous surface over the spent catalyst (see Fig. 3).

The exchange capacity of the fresh resin determined by the dry weight titration method resulted in a capacity of 2.50 mequiv. g-1. This value was rather well in line with the manufacturer’s data (2.55 mequiv. g-1). However, when the catalyst was exposed to the reaction temperature of 190 °C, a reduction in the number of acid sites at around 70 % of the original value was obtained whereas 200 °C reaction temperature resulted in a reduction in the number of acid sites of around 60 %, in comparison to the fresh catalyst.

Figure 3. Scanning electron micrographs of fresh (top) and spent (below) Amberlyst 70 catalyst (scales correspond to 500, 20, and 2 respective 0.5 µm).

Substrate characterization

Typical scanning electron micrographs of the wood pulps from Metsä-Serla (M-S) and Domsjö (D), respectively, are depicted in Fig. 4. When looking at the sample M-S 0.35 AD with fibres from Metsä-Serla, one can see that the fibre surface is much smoother than the fibre surface of the sample from Domsjö, D 0.35 AD. A probable reason for this is that all the hemicellulose contained within the pulp was not dissolved from the kraft pulp from Metsä- Serla; however, the sulfite pulp from Domsjö turned out to be almost hemicellulose-free, resulting in a less glossy surface appearance. Upon a comparison between air-dried and freeze-dried pulp, it was a bit more difficult to draw any further conclusions. In theory, freeze drying should be less destructive to the fibers; however, in this case it was difficult to see any differences between the micrographs obtained for the wood fibre samples.

M-S 0.35 AD

D 0.35 AD

D 0.35 CPD

D CPD

D AD

Figure 4. Scanning electron micrographs of wood pulps. Scales correspond to 200, 20, and 10 µm. M-S (Metsä-Serla), D (Domsjö), AD (air-dried), CPD (critical-point-dried), 0.35 (grinded to; mm).

Auto-catalysis

Initially, reactions were also carried out without catalyst until to the reaction temperature of 200 °C containing 10, 30, 50 and 80 mM formic and levulinic acid, which corresponds to the concentration range we work within. Consequently, we could observe that we have very little or negligible autocatalysis occurring during our experiments. What could be observed was that the hemicellulose and a very small part of the cellulose were converted to intermediates and final products before the reaction subsided. It appeared as if the solution was not sufficiently acidic to react the crystalline part of the cellulose, but enough acidic (LA pKa = 4.59, FA pKa = 3.75 [92, 93]) to react some of the amorphous part. However, the reaction rates were very slow (particularly at low concentrations) and approximately around 5 wt% of the total mass of pulp was converted before reaction died. In an experiment with catalyst (heterogeneous catalysis), the contribution of autocatalysis in the system is likely to be very negligible. The reason is that all the hemicellulose and probably also the most part of the amorphous cellulose is already converted before the concentration of formic acid and levulinic acid becomes so high that it could contribute to autocatalysis. A blank reaction without the catalyst was also performed at 200 °C and under higher CO2 pressure than at normal reaction conditions (90 bars instead of 50 bars). The goal was to determine whether the carbonic acid formed catalyzes the reaction. No pulp conversion was observed; i.e. no. reaction products or intermediates (or humins visually) were observed upon analysis.

Reusability of the catalyst, catalyst deactivation

The catalyst reusability was studied at a reaction temperature of 190 °C and 200 °C, respectively. After the first hydrolysis reaction, the resin was separated from the reaction solution and washed with distilled water and dried at 100 °C for 24 h. At 190 °C, the yield of LA an FA was approximately similar as upon the first batch, but the reaction time was significantly longer. At 200 °C reaction temperature, the catalyst activity reflects the results of the acidic capacity measurements, i.e. lower product yields, lower reaction rate and a higher formation of humins. The manufacturer, Rohm and Haas, has suggested a maximum operating temperature of 190 °C, but some investigators have reported that the resin could be considered as quite stable even above this temperature [83]. Consequently, the loss of sulfonic groups as a result of extended reaction times leads to the lower activity of the catalyst. Nevertheless, an acid treatment should at least partially restore the catalytic performance, but it has never been examined in practice.

Reaction network

A simplified reaction network for the acid-catalyzed hydrolysis of wood pulp to levulinic and formic acids is introduced in Fig. 5.

Wood pulp H O 2 H2O

OH OH OH OH O O OH OH

OH OH OH OH Glucose Xylose Humins (or Tar)

3 H2O 3 H2O

O O

HO O O decarbonylation Furfural HMF

2 H2O

OH + O OH O OH H3C Levulinic acid O Formic acid Formic acid

Decomp. prod. Figure 5. The reaction network upon production of the platform chemicals levulinic and formic acid from wood pulp.

In fact, the one-pot synthesis of levulinic acid from wood is a demanding reaction since the whole process most conveniently starts with acid hydrolysis of pulp thus giving rise to various issues in terms of waste produced as well as the occurrence of multiple side reactions. Two reaction pathways are involved: degradation of cellulose results into formation of glucose, while the degradation of hemicelluloses generates xylose. The glucose formed is under the experimental conditions further transformed to 5-hydroxymethylfurfural (HMF) which, in turn, is further converted in a consecutive hydration-dehydration reaction sequence to levulinic (LA) and formic acids (FA). The xylose pentoses originating from the hemicellulose fractions transform to furfural which is finally decomposed to formic acid. Furthermore, HMF can be decarbonylated into furfural [78]. The desired raw material, lignocellulosic biomass, is mainly composed of cellulose, hemicelluloses and lignin of which the first two can yield the desired hexoses, the key intermediates in the production of levulinic acid.

Thereafter, sugars need to be dehydrated and rearranged to 5- hydroxymethylfurfural and furfural depending on whether hexoses or pentoses, respectively, are to be dehydrated. Consequently, hexoses can then further rehydrate to formic and levulinic acids also requiring two equivalents (moles) of water.

Since two different organic carboxylic acids are formed, the theoretical yield of levulinic acid from hexoses and cellulose has been reported to reach 64.5 wt % and 71.5 wt % only, respectively [32]. It should, however, be stated that the real yields of levulinic acid are typically lower than the theoretical yields due to the harsh reaction conditions required for hydrolysis of macromolecules. Thus, the formation of humins is hardly to be avoided and throughout all kinetic experiments, black insoluble-substances known as humins/tar were also found on the reactor walls and the amount was visibly increased with reaction time. These are well known products of side- reactions upon acid-catalyzed decomposition of pulp. Girisuta et al. [94] confirmed by elemental analysis that the solids are a mixture consisting mainly of humins (typical composition: C, 63.1; H, 4.2) and some unreacted cellulose (C, 42.2; H, 6.1). In addition to the anticipated products, Girisuta et al. [42] also detected small amounts of glucose-reversion products (e.g., levoglucosan, isomaltose and gentiobiose). The formation of the reversion products was also observed in this study, but the maximum amount of these products was very low and not accurately quantified in this work. A typical example of an HPLC chromatogram of a sample taken in the beginning during a reaction is given in Fig. 6.

Fig. 6. Typical HPLC chromatogram for the acid-catalyzed hydrolysis of soft wood pulp.

any gas addition. CO, CO2 and H2 are typically detected. The amounts observed were less than 0.1 wt% of the cellulose intake. This implies that the formation of gas-phase compounds is only a minor reaction pathway under the investigated reaction conditions, at lower temperatures. However, we could not quantify the presence of any formed carbon dioxide since it was used to pressurize the reactor system during most of our experiments. On the other hand on the basis of the graphic illustrations of the kinetic experiments, we can clearly see that formic acid decomposes at higher temperatures. Most likely increases the levels of CO and CO2 with increasing reaction temperatures, but no conclusive investigations were performed.

Formic acid decomposes primarily to CO and H2O in the gas phase (dehydration), but to CO2 and H2 (decarboxylation) in the aqueous phase [95] and indicating an increase in the total pressure of the reaction vessel.

The influence of reaction conditions and raw material on product yields

Comparison between different wood pulps

In Fig. 7, one can see the results from the two experiments performed with different wood pulps. One can clearly see that the yield of formic acid compared to levulinic acid in the wood pulp from Metsä-Serla is much higher than in the wood pulp from Domsjö. The underlying reason lies in the high hemicelluloses content found in the wood pulp from Metsä-Serla. The main difference when comparing the catalytic transformation of pulp from Domsjö and pulp from Metsä-Serla over Amberlyst 70 is the absence of xylose and, consequently, lower formation of furfural from the former raw material.

60 LA Domsjö

50 FA Domsjö

40 LA Metsä-Serla % - FA Metsä-Serla 30

Yield / mol / Yield 20

0 0 10 20 30 40 50 60 Reaction time / h Fig. 7. The comparison of formic and levulinic acid yields for soft wood pulp from Domsjö (sulfite pulp) and Metsä-Serla (sulfate pulp) (the reaction atmosphere is CO2, the catalyst and substrate amount is 3 g respectively and the temperature is 180 ◦C).

Reaction with D-glucose

The conversion of wood pulp to LA and FA is a complex multistep process. It is generally believed that the polymer chains of cellulose are broken down into low molecular weight fragments and finally to glucose. Then, glucose is decomposed to HMF, which is further converted into LA and FA in a molar ratio of 1:1 [9]. However, HMF has also been reported to react into furfural (carbonylation) which is further converted to formic acid [78]. Fructose is also a known intermediate in acid catalyzed decomposition of glucose, but could not be detected in this study. The epimerization step is very rapid and, consequently, fructose was omitted as an individual chemical species. To find out if we also have the same plausible reaction mechanism for the Amberlyst 70 catalyzed conversion of wood pulp we did an experiment with glucose. We can clearly see (Fig. 8) that LA and FA are not formed in a molar ratio of 1:1. It is even clearer if we take into account the decomposition of FA during time. All above-mentioned products (except fructose) were detected in this study, which means we also detected furfural in our product samples taken from the reactor during reaction with glucose. Furfural formed in experiments with D-glucose originated most probably from the decarbonylation of HMF, but also most certainly from the transformation of some of the impurities, C5 sugars, in the D-glucose.

100 90 D-glucose 80

70 %

- 60 Levulinic acid 50 40

Yield / mol / Yield Formic acid 30 20 10 0 0 5 10 15 20 25 30 Reaction time / h

Fig. 8. Experiment with d-glucose (the reaction atmosphere is CO2, the catalyst and substrate amount is 3 g respectively and the temperature is 180 ◦C).

The influence of reaction temperature and catalyst to substrate dosage

Hydrolysis of wood pulp by Amberlyst 70 is highly dependent upon the reaction temperature. The effect of reaction temperature on the cellulose conversion into LA and FA was investigated, and the experiments were conducted at 150, 170, 180, 185, 190 and 200 °C, respectively. At lower temperatures (150 °C), there is little conversion of cellulose to levulinic acid as well as to formic acid. Note however, that the experiments at 150 and 170

°C were interrupted before all the wood pulp was consumed because of low reaction rate. As the temperature was increased, the conversion rates improved very rapidly. The increased conversion is likely related to increased solubility of cellulose at higher temperatures. The conversion at 190 and 200 °C is very fast as can be seen in Fig. 9.

60 60 200 C 190 C 180 C 50 185 C 50 200 C 190 C 185 C 180 C

40 40 %

% 170 C

- - 170 C

30 30

FA / mol/ FA

LA / mol / LA Ƴ Ƴ 20 20

10 150 C 10 150 C

0 0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Reaction time / h Reaction time / h Fig. 9. The influence of reaction temperature on levulinic and formic acid yields (the reaction atmosphere is CO2 and the catalyst and substrate amount is 3 g respectively).

It has been reported that when the reaction temperature was increased above 200 °C, the yield of LA began to fall. Higher temperatures could thus accelerate the rate of chemical reactions, but unwanted side reactions also appear at the same time. For example, LA is unstable above 230 °C, which would be dehydrated to unsaturated lactones. Additionally, more humins may occur during the reaction [10]. Therefore, an increase in the temperature is unfavorable for the extent of reaction, and, consequently, the reaction temperature was limited to 200 °C in this work. Another reason that the maximum reaction temperature was limited to 200 °C is that the manufacturer of Amberlyst 70 recommends a maximum reaction temperature of 190 °C. However, it has been reported that the catalyst could be considered as stable in some conditions even above this temperature [84]. Fig. 9 also shows the effect of the temperature on the decomposition of FA. It can be observed that temperature has a significant effect on the decomposition rate of FA. When the reaction temperature was increased from 180 to 200 °C, there was a huge increase in the decomposition rate of FA. The amount of FA decreased quickly very likely forming carbon monoxide, carbon dioxide and hydrogen gas [95] indicated by a small but notable rise of the pressure in the autoclave. The effect of catalyst dosage on the yields of products is shown in Fig. 10.

60 60

50 185 C, 3 g cat 50 185 C, 3 g cat 180 C, 3 g cat 180 C, 3 g cat

40 40

% - - 180 C, 2 g cat 180 C, 2 g cat

30 30

FA / mol/ FA

LA / mol / LA Ƴ Ƴ 20 20

10 10

0 0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70 Reaction time / h Reaction time / h Fig. 10. The effect of temperature and catalyst amount on levulinic and formic acid yields (the reaction atmosphere is CO2 and substrate amount is 3 g).

The amount of catalyst in the reaction mixture was varied from 2 to 4 g at 180 °C and 1 to 3 g at 185 °C, respectively. As can be seen, the amount of catalyst affected the extent of cellulose conversion, but not as much as the reaction temperature does. Fig. 11 illustrates the maximum levulinic acid yields obtained upon kinetic experiments conducted at different acid and substrate concentrations. Theoretically, a higher liquid-to-solid ratio should result in a higher yield of levulinic acid. However, it was further observed that higher substrate loadings had an adverse effect in terms of the final levulinic acid yields detected in the reaction mixture. These observations are in line with other studies on the acid-catalyzed decompositions of HMF and glucose, whereupon low substrate concentrations favored high LA yields. Generally, lower concentration of cellulose resulted in higher yields of levulinic acid. For each levulinic acid molecule generated, two molecules of water are needed; otherwise 5-hydroxymethylfurfural (HMF) may be converted into humins [42]. As also depicted by Fig. 11, evidently there exists an optimal catalyst loading. Consequently, the addition of excess catalyst has little or even negative effect on the LA yield. In fact, Yan et al also observed that low concentration of hydrochloride acid resulted in too low concentration of H+ ions thus being unable to facilitate cellulose hydrolysis and, consequently, a low reaction rate was observed. Using low acid concentrations is problematic, because the lignin and other impurities present in the biomass consumes a fraction of the acid during the reaction. In our case a load of 9 g substrate and 1 g catalyst resulted in only 5.9 mol % yield of levulinic acid, at a reaction temperature of 190 °C.

67.8 70 65.4 66.1 60.2 70 62.1 62.7 57.4 59.9 61.2 51.5 59.2 60 54.3 57.0 55.5 60 51.7 43.1 43.9 50 50.7 46.9 50 48.2 1 1 3 5.9

Levulinic acid yield [mol %] [mol yield acid Levulinic 3 Levulinic acid yield [mol %] [mol yield acid Levulinic 40 6 Substrate 40 6 Substrate 1 9 [g] 1 9 3 3 [g] 180 °C 6 190 °C 6 Catalyst [g] 9 Catalyst [g] 9 Figure 11. Maximum levulinic acid yields formed at 180 °C and 190 °C at different acid and substrate concentrations (the reaction atmosphere is CO2).

The influence of reaction gas and pressure

In this study, carbon dioxide was used as the reaction atmosphere. However, to investigate whether various gases affect the conversion of wood pulp, some tests with other gases and pressures were also performed. Fig. 12 illustrate the results of these experiments.

60 No gas addition 60 Ar 3 bar Ar 3 bar

50 CO2 50 bar 50 Ar 50 bar Ar 50 bar CO2 50 bar

40 40

% - - No gas addition

30 30

FA / mol/ FA

LA / mol / LA Ƴ Ƴ 20 20

10 10

0 0 0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80 Reaction time / h Reaction time / h

Fig. 12. The influence of reaction atmosphere and pressure on levulinic and formic acid yields (the catalyst and substrate amount is 3 g respectively and the temperature is 180 ◦C).

As can be observed, argon as the reaction atmosphere seems to promote the reaction rates, whereas the presence of carbon dioxide acted as a mild inhibitor. This was surprising since we expected the significantly better soluble CO2 to yield carbonic acid that would enhance the performance of our system. However, the much less soluble and supposedly inert argon (by a factor of roughly 103 [96]) seemed to yield better performance. The reason for this is at present not clear to us. Our hypothesis is that the argon gas and formic acid are capable of forming relatively strong argon–HCOOH complexes [97] resembling hydrogen bonding networks. Thus, the equilibrium in between the gas and the liquid phase is presumably shifted toward the gas phase (low solubility of noble gases in water) and, consequently, any possible product inhibition contribution to the additional

formation of HCOOH from cellulose (in liquid phase) is diminished, resulting in increased amounts of the final desired products. Similar phenomena might also contribute in the case of levulinic acid. In addition it is well known that CO2 can be hydrogenated to formic acid (reversible reaction). Further, particularly at higher temperatures, formic acid tends easily to decompose. Thus, it might be that since these reactions are typically metal catalyzed that our reactor setups (stainless steel and zirconium) could to certain extent interact with the species emerging. A system containing at least water, CO2 and HCOOH, in the presence of metals could provide conditions that inhibit the initial HCOOH formation from cellulose.

Selectivity

Overall, we can conclude that the catalysis reaction is highly selective, and there are very little by-products formed (except humins). Nevertheless, if we compare our main products formic and levulinic acid, we can perhaps discern small differences among the selectivity under different reaction conditions. The molar yield of formed levulinic acid as a function of the molar yield of produced formic acid in the catalytic transformation of wood pulp over Amberlyst 70 resin is presented in Fig. 13, respectively.

60 No gas addition 50 CO2 50 bar Ar 50 bar

40 Ar 3 bar

% -

30 LA / mol/ LA 20

0 0 10 20 30 40 50 60 FA / mol-% 60 60 200 C 185 C, 3 g cat 190 C 185 C, 1 g cat 50 185 C 50 180 C, 4 g cat 180 C 180 C, 3 g cat

40 170 C 40 180 C, 2 g cat %

150 C %

- -

30 30

LA / mol/ LA LA / mol/ LA 20 20

10 10

0 0 0 10 20 30 40 50 60 0 10 20 30 40 50 60 FA / mol-% FA / mol-% Fig. 13. The influence of the reaction atmosphere, temperature and catalyst amount on the molar yield of levulinic acid as a function of the molar yield of formic acid (the reaction atmosphere is CO2, the substrate and catalyst amount is 3 g respectively and the temperature is 180 ◦C).

The most striking difference is perhaps that an atmosphere with low pressure of argon seems to favor formic acid transformation. However, up to 40% yield level, the differences are extremely small. It seems likely that in the end of a reaction the decomposition of formic acid affects to a greater degree the selectivity. Ultimately, more experiments are needed to draw more accurate conclusions.

Kinetic modelling

Upon modelling of the reaction kinetics, Modest® was used [91]. The software uses the backward difference method to solve the reactor mass balances (a system of ordinary differential equations, ODEs) and a hybrid method involving Simflex and Levenberg-Marquardt methods for the estimation of the kinetic parameters (rate constants and activation energies) defined in the kinetic model. The Levenberg-Marquardt method [98], is a more effective modification of the well-known Newton-Raphson method and the Simplex algorithm is based on the Simplex method introduced by Spendley et al. [99]. If the two methods are compared, one can conclude that the Simplex algorithm is slightly better at finding the way to the optimum from a poor initial guess of parameters, while the sharp optimum is easier and faster achieved by the Levenberg-Marquardt method, provided that the initial guess is sufficiently close to the optimum. Consequently, the rate equations for the proposed elementary reaction kinetic model and batch reactor model were implemented in the software which also provides the graphical interface.

Wood pulp H O 2 H2O A1 / A2 r r1 2

OH OH OH OH O O OH OH

OH OH OH OH Glucose Xylose Humins

(or Tar) r r3 4 3 H2O 3 H2O

O O

HO O O decarbonylation Furfural HMF

r 5 r6 2 H2O

OH + O OH O OH H3C Levulinic acid O Formic acid Formic acid

r14

Decomp. prod. Figure 14. The reaction network upon production of the platform chemicals levulinic and formic acid from cellulose as well as the reactions used in the mathematical model.

For the parameter estimation the following objective function was minimized

2 푄 = ∑ ∑(푐푖,푡,푒푥푝 − 푐푖,푡,푚표푑푒푙) 푤푖,푡 (5) 푡 푖

where ci,t,exp and ci,t,model are the experimentally recorded concentrations and the concentrations predicted by the model, respectively. The weight factors, wi, were at this time set equal to unity (1) for all experimental data.

The reaction system was described with the following reactions

푟1 푟3 푟5 퐴1 → 퐺푙푢푐표푠푒 → 퐻푀퐹 → 퐿푒푣푢푙푖푛푖푐 푎푐푖푑 + 퐹표푟푚푖푐 푎푐푖푑 (6)

푟2 푟4 푟6 퐴2 → 푋푦푙표푠푒 → 퐹푢푟푓푢푟푎푙 → 퐹표푟푚푖푐 푎푐푖푑 (7)

푟14 퐹표푟푚푖푐 푎푐푖푑 → 퐷푒푐표푚푝표푠푒푑 푝푟표푑푢푐푡푠 (8)

In the initial approach, only the experimental runs at higher temperature range, i.e. 180–200 °C, were concerned since the reaction proceeds extremely sluggishly at lower temperatures and batch experiments lasting for several days had to be carried out. Nevertheless, some preliminary modeling revealed that a first order reaction kinetics model was able to describe the reactions in a reasonable and simple way. Thus, the reaction rates were determined as follows:

푟푗 = 푘푗푐푖 (9)

where rj denote the individual reaction rates for the reaction j, kj denote the reaction rate constants and ci the concentration of component i. On the other hand, the temperature influence on the reaction rate was taken into account with the Arrhenius equation

퐸퐴푗 1 1 − ( − ) 푅 푇 푇 (10) 푘푗 = 푘0푗푒 푚푒푎푛

and thus, kj denote the reaction rate constants, koj the frequency factor, R is the ideal gas constant, T is the temperature and EAj the activation energy. The modeling approach chosen verified that the experimental results obtained at 180–200 °C, in the heterogeneously catalyzed reaction displayed prominent Arrhenius-type temperature dependence demonstrated for the reaction products identified.

For the reactor, the ideal batch reactor model was used.

푑푐 푖 = 푟 푚 (11) 푑푡 푖 푐푎푡

where ci denote the concentration of component i, ri the generation rate for component i and mcat the catalyst amount.

The parameters (frequency factors k0j, activation energies EAj) were estimated using the parameter estimation software. Estimation of the

reaction kinetics parameters were split into two parts, whereupon first only parameters for the reactions r2, r4 and r6 were estimated, after which the parameters for r2, r4 and r6 were fixed and the rest of the parameters were estimated. Initially an estimation of all parameters simultaneously was tried but it did not give good results for the reactions r2 and r4 (formation of xylose and furfural), mainly due to large differences in the concentration levels of the compounds present thus leading to unstable behavior of the solver. The values of the estimated parameters are enlisted in Table 1.

Table 1. Numerical values for the estimated parameters (degree of explanation 97.8 %). Parameter / (for units see Value Parameter error / % below) k01 2.7 1.5 k02 15.3 11.8 k03 18.9 4.3 k04 30.8 8.4 k05 291 130 k06 40.7 14.2 k014 0.429 2.7

EA1 163 2.8

EA2 33.3 82.7

EA3 163 7.2

EA4 64.5 34.1

EA5 8.83 2705

EA6 16.7 203

EA14 134 7

-1 Tmean = 185 °C, Frequency factor (koj) in h and activation energy (EAj) in kJ/mol.

Fit of the model to experimental data

The most common measure for the goodness of fit is the degree of explanation, the R2 value. The idea is to compare the residuals given by the model to the residuals of the simpliest model one may think of, the average value of all the data points. The R2 value is given by the expression

2 2 (푦푚표푑푒푙 − 푦푒푥푝푒푟푖푚푒푛푡) 푅 = 100 2 (12) (푦푚표푑푒푙 − 푦̅푒푥푝푒푟푖푚푒푛푡) where the y values denotes the estimated vs. measured values of the respective parameters. The closer the R2 value is 1, the better is the fitting of the model. For the parameter estimation the degree of explanation of 97.8 % was obtained.

The distribution and sensitivity of the parameters were controlled with the Markov Chain Monte Carlo method (MCMC) [100], see Figure 16. These figures also show low correlations between parameters and confirm well identified optimal values.

Figure 15. Sensitivity analysis for the parameters.

Figure 16. Fit of model to experimental data for one of the experiments performed at 190 °C with 6 g catalyst and 6 g substrate. Levulinic acid (○), formic acid (+), glucose (*), 5-HMF (□), xylose (∆) and furfural (◊) vs. time.

25 0.7

0.6 Levulinic acid 20

0.5 Formic acid Xylose

15 0.4

mmole mmole mmole 0.3 10

0.2 Furfural

5 0.1

Glucose HMF 0 0 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 time (h) time (h) Figure 17. Fit of model to experimental data for one of the experiments performed at 180 °C with 6 g catalyst and 9 g substrate. Levulinic acid (○), formic acid (+), glucose (*), 5-HMF (◊), xylose (∆) and furfural (□) vs. time.

7 0.8

6 0.7

0.6 5 Formic acid 0.5 4 0.4

mmole mmole 3 mmole Levulinic acid 0.3

2 0.2

1 0.1 Xylose HMF Glucose Furfural

0 0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 time (h) time (h) Figure 18. Fit of model to experimental data for one of the experiments performed at 200 °C with 1 g catalyst and 3 g substrate. Levulinic acid (○), formic acid (+), glucose (*), 5-HMF (◊), xylose (∆) and furfural (□) vs. time.

The figures depicting the fit of model to experimental data (Figs. 16-18) show that the model can predict the behavior of the reaction rather well at

different conditions as displayed in the parity plot Figure 19. Nevertheless, the fit could be better for the components present in low concentrations.

Figure 19. Fit of model to all experimental data (levulinic acid (○), formic acid (+) and glucose (*)).

Conclusions

Levulinic acid can be successfully produced from biomass using either homogeneous or heterogeneous catalysts albeit with variations in performance. Besides more traditional approaches like taking advantage of mineral acids, in particular hydrochloric and sulfuric ones, many solid acids have shown comparable performance. New, unorthodox processing alternatives such as application of microwave dielectric heating or various mechanical commutator designs have also been demonstrated for this process. In general, the beneficial conditions involve relatively low acid concentrations, low initial concentrations of feedstock and moderate temperatures. In this thesis solid acid catalyzed degradation of bleached Nordic industrial dissolving pulp (a mixture of softwoods Norway Spruce and Scots Pine) was investigated under mild batch wise reaction conditions and the goal was to obtain versatile platform chemicals – levulinic and formic acids. The optimal reaction conditions were determined and a simple, first-order kinetic approach was developed to describe the experimental data with satisfactorily accuracy. Sulphite-cellulose, being lean on pentose sugar containing hemicelluloses was revealed as a good raw material compared to sulphate pulp (birch kraft pulp), thus resulting in about 59 and 68 mol % theoretical yields of levulinic vs. formic acid, respectively, obtained at a reaction temperature of 180 °C and at a low initial cellulose concentration. It was also observed that excessively high reaction temperatures contribute to thermal degradation of the formed formic acid, since formic acid tends to degrade to carbon dioxide, carbon monoxide, hydrogen and water, which in turn led to lower overall yields of the desired products. Further it was observed that a low substrate load favors a high levulinic acid yield and that the catalyst loading has an optimum, i.e. whether the amount of catalyst is too high or too low reduces the yield. One can also observe that a low acid concentrations is problematic, because the lignin and other impurities present in the biomass consumes a fraction of the acid during the reaction. Consequently, the loss of sulfonic groups as a result of extended reaction times leads to the lower activity of the catalyst. Nevertheless, an acid treatment should at least partially restore the catalytic performance, but it has never been examined in practice. More surprisingly was that argon as the reaction atmosphere seems to promote the reaction rates, whereas the presence of carbon dioxide acted as a mild inhibitor. We expected the significantly better soluble CO2 to yield carbonic acid that would enhance the performance of our system. The reason for this is at present not clear to us.

Also we have observed very little auto-catalysis occurring during our experiments. What could be observed was that the hemicellulose and a very small part of the cellulose were converted to intermediates and final products before the reaction subsided. It appeared as if the solution was not sufficiently acidic to react the crystalline part of the cellulose, but enough acidic to react some of the amorphous part. In an experiment with catalyst (heterogeneous catalysis), the contribution of auto-catalysis in the system is likely to be very negligible.

Acknowledgements

First of all, I would like to express my deepest gratitude to my supervisor and closest colleague, Prof. Jyri-Pekka Mikkola for giving me the opportunity to work at the laboratory.

I would like to give special thanks to Prof. Johan Wärnå for his contribution in the computational modelling and Dr. William Larsson for his assistance with laboratory equipment.

Also my assistant supervisors Prof. Leif Jönsson and Doc. Torbjörn Lestander are acknowledged for their invaluable advices and discussions. Markku Reunanen and Dr. Carina Jonsson for carrying out the GC–MS analysis and Per Hörstedt are acknowledged for carrying out the SEM analysis.

I also would like to thank all the co-authors and co-workers and all the rest of my collages.

Finally, I would like to express gratitude to my parents Gunvor and Runar for their love and encouragement throughout my education. Special thanks go also to my two sisters, Karin and Ingrid, and their families for their support.

It would not have been possible without you all!

The Bio4Energy programme, Knut and Alice Wallenberg and Kempe Foundations are gratefully acknowledged for the financial support. Further, COST action CM0903 (UBioChem) and the Academy of Finland are acknowledged and greatly appreciated. Collaboration with Processum Biorefinery Initiative AB is acknowledged. This work is also a part of activities at the Åbo Akademi Process Chemistry Centre, a center of excellence financed by the Åbo Akademi University.

References

[1] P. Gallezot, Process options for converting renewable feedstocks to bioproducts, Green Chemistry, 9 (2007) 295-302.

[2] D. Mohan, C.U. Pittman, P.H. Steele, Pyrolysis of wood/biomass for bio-oil: A critical review, Energy & Fuels, 20 (2006) 848-889.

[3] S.N. Naik, V.V. Goud, P.K. Rout, A.K. Dalai, Production of first and second generation biofuels: A comprehensive review, Renewable & Sustainable Energy Reviews, 14 (2010) 578-597.

[4] G.W. Huber, S. Iborra, A. Corma, Synthesis of transportation fuels from biomass: Chemistry, catalysts, and engineering, Chemical Reviews, 106 (2006) 4044-4098.

[5] R. Rinaldi, F. Schuth, Acid hydrolysis of cellulose as the entry point into biorefinery schemes, Chemsuschem, 2 (2009) 1096-1107.

[6] W.L. Faith, Development of the scholler process in the United-States, Industrial and Engineering Chemistry, 37 (1945) 9-11.

[7] A.A. Efremov, G.G. Pervyshina, B.N. Kuznetsov, Production of levulinic acid from wood raw material in the presence of sulfuric acid and its salts, Chem Nat Compd, 34 (1998) 182-185.

[8] C. Chang, P. Cen, X. Ma, Levulinic acid production from wheat straw, Bioresource Technology, 98 (2007) 1448-1453.

[9] B. Girisuta, L.P.B.M. Janssen, H.J. Heeres, Kinetic study on the acid- catalyzed hydrolysis of cellulose to levulinic acid, Industrial & Engineering Chemistry Research, 46 (2007) 1696-1708.

[10] L. Yan, N. Yang, H. Pang, B. Liao, Production of levulinic acid from bagasse and paddy straw by liquefaction in the presence of hydrochloride acid, CLEAN – Soil, Air, Water, 36 (2008) 158-163.

[11] J.-P. Lange, W.D. van de Graaf, R.J. Haan, Conversion of Furfuryl Alcohol into Ethyl Levulinate using solid acid catalysts, Chemsuschem, 2 (2009) 437-441.

[12] D.W. Rackemann, W.O.S. Doherty, The conversion of lignocellulosics to levulinic acid, Biofuels, Bioproducts and Biorefining, 5 (2011) 198- 214.

[13] Amberlyst 70 high temperature strongly acidic catalyst http://www.amberlyst.com/literature/a4/70.pdf, (accessed 6.11.2011).

[14] R. Rinaldi, R. Palkovits, F. Schüth, Depolymerization of cellulose using solid catalysts in ionic liquids, Angewandte Chemie International Edition, 47 (2008) 8047-8050.

[15] I. Agirrezabal-Telleria, A. Larreategui, J. Requies, M.B. Güemez, P.L. Arias, Furfural production from xylose using sulfonic ion-exchange resins (Amberlyst) and simultaneous stripping with nitrogen, Bioresource Technology, 102 (2011) 7478-7485.

[16] X. Hu, C.-Z. Li, Levulinic esters from the acid-catalysed reactions of sugars and alcohols as part of a bio-refinery, Green Chemistry, 13 (2011) 1676-1679.

[17] J.F. Saeman, Kinetics of wood saccharification-hydrolysis of cellulose and decomposition of sugars in dilute acid at high temperature, Industrial & Engineering Chemistry, 37 (1945) 43-52.

[18] R.D. Fagan, H.E. Grethlein, A.O. Converse, A. Porteous, Kinetics of the acid hydrolysis of cellulose found in paper refuse, Environmental Science & Technology, 5 (1971) 545-547.

[19] D.R. Thompson, H.E. Grethlein, Design and evaluation of a plug flow reactor for acid hydrolysis of cellulose, Industrial & Engineering Chemistry Product Research and Development, 18 (1979) 166-169.

[20] J.-P. Franzidis, A. Porteous, J. Anderson, The acid hydrolysis of cellulose in refuse in a continuous reactor, Conservation & Recycling, 5 (1982) 215-225.

[21] M. Green, S. Kimchie, A. Malester, B. Rugg, G. Shelef, Utilization of municipal solid wastes (MSW) for alcohol production, Biological wastes, 26 (1988) 285-295.

[22] I.A. Malester, M. Green, G. Shelef, Kinetics of dilute acid hydrolysis of cellulose originating from municipal solid wastes, Industrial & Engineering Chemistry Research, 31 (1992) 1998-2003.

[23] Q. Fang, M.A. Hanna, Experimental studies for levulinic acid production from whole kernel grain sorghum, Bioresour Technol, 81 (2002) 187-192.

[24] T. Frost, E. Kurth, Tappi, 1951.

[25] Q. Jing, X. LÜ, Kinetics of non-catalyzed decomposition of glucose in high-temperature liquid water, Chinese Journal of Chemical Engineering, 16 (2008) 890-894.

[26] J. Shen, C.E. Wyman, Hydrochloric acid-catalyzed levulinic acid formation from cellulose: data and kinetic model to maximize yields, AIChE Journal, 58 (2012) 236-246.

[27] A. Onda, T. Ochi, K. Yanagisawa, Selective hydrolysis of cellulose into glucose over solid acid catalysts, Green Chemistry, 10 (2008) 1033- 1037.

[28] G.J. Mulder, J. Prakt. Chem., 21 (1840).

[29] T. Werpy, G. Petersen, A. Aden, J. Bozell, J. Holladay, J. White, A. Manheim, D. Eliot, L. Lasure, S. Jones, in, Top value added chemicals from biomass. Volume 1-Results of screening for potential candidates from sugars and synthesis gas, DTIC Document, 2004.

[30] A.M.R. Galletti, C. Antonetti, V. De Luise, D. Licursi, N. Nassi, Levulinic acid production from waste biomass, BioResources, 7 (2012) 1824-1835.

[31] B. V. Timokhin, V. A. Baransky, G. D. Eliseeva, Levulinic acid in organic synthesis, Russian Chemical Reviews, 68 (1999) 73-84.

[32] R.H. Leonard, Levulinic acid as a basic chemical raw material, Industrial & Engineering Chemistry, 48 (1956) 1330-1341.

[33] S.W. Fitzpatrick, Production of levulinic acid from carbohydrate- containing materials (US Patent 5,608,105, 1997).

[34] D.B. Bevilaqua, M.K.D. Rambo, T.M. Rizzetti, A.L. Cardoso, A.F. Martins, Cleaner production: levulinic acid from rice husks, Journal of Cleaner Production, 47 (2013) 96-101.

[35] M. Kang, S.W. Kim, J.-W. Kim, T.H. Kim, J.S. Kim, Optimization of levulinic acid production from Gelidium amansii, Renewable Energy, (2012).

[36] D.-m. Lai, L. Deng, Q.-x. Guo, Y. Fu, Hydrolysis of biomass by magnetic solid acid, Energy & Environmental Science, 4 (2011) 3552- 3557.

[37] P. Wang, S. Zhan, H. Yu, Production of levulinic acid from cellulose catalyzed by environmental-friendly catalyst, Advanced Materials Research, 96 (2010) 183-187.

[38] L. Peng, L. Lin, J. Zhang, J. Zhuang, B. Zhang, Y. Gong, Catalytic conversion of cellulose to levulinic acid by metal chlorides, Molecules, 15 (2010) 5258-5272.

[39] B. Girisuta, B. Danon, R. Manurung, L.P.B.M. Janssen, H.J. Heeres, Experimental and kinetic modelling studies on the acid-catalysed hydrolysis of the water hyacinth plant to levulinic acid, Bioresource Technology, 99 (2008) 8367-8375.

[40] V.S. Minina, A.E. Sarukhanova, K.U. Usmanov, Preparation of fural and levulinic acid by hydrolysis of pressed cotton stems, Fiz. i Khim. Prirodn. i Sintetich. Polimerov, Akad. Nauk Uz. SSR, Inst. Khim. Polimerov, 1962, pp. 87-93.

[41] A. Szabolcs, M. Molnar, G. Dibo, L.T. Mika, Microwave-assisted conversion of carbohydrates to levulinic acid: an essential step in biomass conversion, Green Chemistry, 15 (2013) 439-445.

[42] B. Girisuta, L. Janssen, H. Heeres, Green Chemicals: A kinetic study on the conversion of glucose to levulinic acid, Chemical Engineering Research and Design, 84 (2006) 339-349.

[43] A. Thompson, Method of making levulinic acid (US patent 2,206,311, 1940).

[44] W.W. Moyer, Preparation of levulinic acid (US patent 2,207,328, 1942).

[45] Z. Yang, H. Kang, Y. Guo, G. Zhuang, Z. Bai, H. Zhang, C. Feng, Y. Dong, Dilute-acid conversion of cotton straw to sugars and levulinic acid via 2-stage hydrolysis, Industrial Crops and Products, 46 (2013) 205-209.

[46] H. Lin, J. Strull, Y. Liu, Z. Karmiol, K. Plank, G. Miller, Z. Guo, L. Yang, High yield production of levulinic acid by catalytic partial oxidation of cellulose in aqueous media, Energy & Environmental Science, 5 (2012) 9773-9777.

[47] L.J. Carlson, Process for the manufacture of levulinic acid (US patent 3,065,263, 1962).

[48] H. Mehdi, V. Fábos, R. Tuba, A. Bodor, L. Mika, I. Horváth, Integration of homogeneous and heterogeneous catalytic processes for a multi-step conversion of biomass: From sucrose to levulinic acid, γ- valerolactone, 1,4-pentanediol, 2-methyl-tetrahydrofuran, and alkanes, Top Catal, 48 (2008) 49-54

[49] Y. Takeuchi, F. Jin, K. Tohji, H. Enomoto, Acid catalytic hydrothermal conversion of carbohydrate biomass into useful substances, J Mater Sci, 43 (2008) 2472-2475.

[50] A.A. Efremov, G.G. Pervyshina, B.N. Kuznetsov, Thermocatalytic transformations of wood and cellulose in the presence of HCl, HBr, and H2SO4, Chem Nat Compd, 33 (1997) 84-88.

[51] W.O.S. Doherty, P. Mousavioun, C.M. Fellows, Value-adding to cellulosic ethanol: Lignin polymers, Industrial Crops and Products, 33 (2011) 259-276.

[52] R.A. Schraufnagel, H.F. Rase, Levulinic acid from sucrose using acidic ion-exchange resins, Product R&D, 14 (1975) 40-44.

[53] J. Jow, G.L. Rorrer, M.C. Hawley, D.T.A. Lamport, Dehydration of d- fructose to levulinic acid over LZY zeolite catalyst, Biomass, 14 (1987) 185-194.

[54] K. Lourvanij, G.L. Rorrer, Dehydration of glucose to organic acids in microporous pillared clay catalysts, Applied Catalysis A: General, 109 (1994) 147-165.

[55] J. Hegner, K.C. Pereira, B. DeBoef, B.L. Lucht, Conversion of cellulose to glucose and levulinic acid via solid-supported acid catalysis, Tetrahedron Letters, 51 (2010) 2356-2358.

[56] R. Weingarten, W.C. Conner, G.W. Huber, Production of levulinic acid from cellulose by hydrothermal decomposition combined with aqueous phase dehydration with a solid acid catalyst, Energy & Environmental Science, 5 (2012) 7559-7574.

[57] R.W. Thomas, H.A. Schuette, Studies on levulinic acid. I. It’s preparation from carbohydrates by digestion with hydrocloric acid under pressure, Journal of the American Chemical Society, 53 (1931) 2324-2328.

[58] D.M. Alonso, S.G. Wettstein, M.A. Mellmer, E.I. Gurbuz, J.A. Dumesic, Integrated conversion of hemicellulose and cellulose from lignocellulosic biomass, Energy & Environmental Science, 6 (2013) 76-80.

[59] J.Y. Cha, M.A. Hanna, Levulinic acid production based on extrusion and pressurized batch reaction, Industrial Crops and Products, 16 (2002) 109-118.

[60] H. Chen, B. Yu, S. Jin, Production of levulinic acid from steam exploded rice straw via solid superacid, Bioresource Technology, 102 (2011) 3568-3570.

[61] Y. Bin, C. Hongzhang, Effect of the ash on enzymatic hydrolysis of steam-exploded rice straw, Bioresource Technology, 101 (2010) 9114- 9119.

[62] V. Chang, M. Holtzapple, Fundamental factors affecting biomass enzymatic reactivity, M. Finkelstein, B. Davison (Eds.) Twenty-First Symposium on Biotechnology for Fuels and Chemicals, Humana Press, 2000, pp. 5-37.

[63] V.M. Ghorpade, M.A. Hanna, Method and apparatus for production of levulinic acid via reactive extrusion (US Patent 5,859,263, 1999).

[64] S. Van de Vyver, J. Thomas, J. Geboers, S. Keyzer, M. Smet, W. Dehaen, P.A. Jacobs, B.F. Sels, Catalytic production of levulinic acid from cellulose and other biomass-derived carbohydrates with sulfonated hyperbranched poly(arylene oxindole)s, Energy & Environmental Science, 4 (2011) 3601-3610.

[65] G. Banerjee, S. Car, T. Liu, D.L. Williams, S.L. Meza, J.D. Walton, D.B. Hodge, Scale-up and integration of alkaline hydrogen peroxide pretreatment, enzymatic hydrolysis, and ethanolic fermentation, Biotechnology and Bioengineering, 109 (2012) 922-931.

[66] G. Banerjee, S. Car, J.S. Scott-Craig, D.B. Hodge, J.D. Walton, Alkaline peroxide pretreatment of corn stover: effects of biomass, peroxide, and enzyme loading and composition on yields of glucose and xylose, Biotechnol Biofuels, 4 (2011) 16.

[67] M.F. Zaranyika, M. Madimu, Heterogeneous dilute acid hydrolysis of cellulose: A kinetic model for the hydrolysis of the difficultly accessible portion of cellulose based on donnan's theory of membrane equilibria, Journal of Polymer Science Part A: Polymer Chemistry, 27 (1989) 1863-1872.

[68] L. Liu, J. Sun, C. Cai, S. Wang, H. Pei, J. Zhang, Corn stover pretreatment by inorganic salts and its effects on hemicellulose and cellulose degradation, Bioresource Technology, 100 (2009) 5865- 5871.

[69] J. Potvin, E. Sorlien, J. Hegner, B. DeBoef, B.L. Lucht, Effect of NaCl on the conversion of cellulose to glucose and levulinic acid via solid supported acid catalysis, Tetrahedron Letters, 52 (2011) 5891-5893.

[70] S.G. Wettstein, D.M. Alonso, Y. Chong, J.A. Dumesic, Production of levulinic acid and gamma-valerolactone (GVL) from cellulose using GVL as a solvent in biphasic systems, Energy & Environmental Science, 5 (2012) 8199-8203.

[71] I.T. Horvath, H. Mehdi, V. Fabos, L. Boda, L.T. Mika, [gamma]- Valerolactone-a sustainable liquid for energy and carbon-based chemicals, Green Chemistry, 10 (2008) 238-242.

[72] J.-P. Lange, J.Z. Vestering, R.J. Haan, Towards 'bio-based' Nylon: conversion of [gamma]-valerolactone to methyl pentenoate under catalytic distillation conditions, Chemical Communications, (2007) 3488-3490.

[73] M. Chalid, H.J. Heeres, A.A. Broekhuis, Ring-opening of γ- valerolactone with amino compounds, Journal of Applied Polymer Science, 123 (2012) 3556-3564.

[74] J.Q. Bond, D.M. Alonso, D. Wang, R.M. West, J.A. Dumesic, Integrated catalytic conversion of γ-valerolactone to liquid alkenes for transportation fuels, Science, 327 (2010) 1110-1114.

[75] J.C. Serrano-Ruiz, D.J. Braden, R.M. West, J.A. Dumesic, Conversion of cellulose to hydrocarbon fuels by progressive removal of oxygen, Applied Catalysis B: Environmental, 100 (2010) 184-189.

[76] S.K.R. Patil, C.R.F. Lund, Formation and growth of humins via aldol addition and condensation during acid-catalyzed conversion of 5- hydroxymethylfurfural, Energy & Fuels, 25 (2011) 4745-4755.

[77] V.G. Devulapelli, H.-S. Weng, Synthesis of cinnamyl acetate by solid– liquid phase transfer catalysis: Kinetic study with a batch reactor, Catalysis Communications, 10 (2009) 1638-1642.

[78] M. Käldström, N. Kumar, D.Y. Murzin, Valorization of cellulose over metal supported mesoporous materials, Catalysis Today, 167 (2011) 91-95.

[79] A. Sundberg, K. Sundberg, C. Lillandt, B. Holmbom, Determination of hemicelluloses and pectins in wood and pulp fibres by acid methanolysis and gas chromatography, Nordic Pulp and Paper Research Journal, 11 (1996) 216-219.

[80] P. Strunk, T. Öman, A. Gorzsás, M. Hedenström, B. Eliasson, Characterization of dissolving pulp by multivariate data analysis of FT-IR and NMR spectra, Nordic Pulp and Paper Research Journal, 26 (2011) 398.

[81] S. Willför, A. Sundberg, A. Pranovich, B. Holmbom, Polysaccharides in some industrially important hardwood species, Wood Science and Technology, 39 (2005) 601-617.

[82] A. Sundberg, S. Emet, P. Rhen, l. Vähäsalo, B. Holmbohm, Estimation of proportions of different pulp types in recycled paper and deinked pulp by acid methanolysis, Proc. of 13th International Symposium on Wood, Fibre and Pulping Chemistry, Auckland, New Zealand, 2005, pp. 361-365.

[83] P.F. Siril, H.E. Cross, D.R. Brown, New polystyrene sulfonic acid resin catalysts with enhanced acidic and catalytic properties, Journal of Molecular Catalysis A: Chemical, 279 (2008) 63-68.

[84] R. Bringué, M. Iborra, J. Tejero, J.F. Izquierdo, F. Cunill, C. Fité, V.J. Cruz, Thermally stable ion-exchange resins as catalysts for the liquid- phase dehydration of 1-pentanol to di-n-pentyl ether (DNPE), Journal of Catalysis, 244 (2006) 33-42.

[85] Z.P. Xu, K.T. Chuang, Effect of internal diffusion on heterogeneous catalytic esterification of acetic acid, Chemical Engineering Science, 52 (1997) 3011-3017.

[86] D. Harvey, Modern analytical chemistry, McGraw-Hill Boston, (2000).

[87] C. Corradini, Liquid Chromatography, (2000).

[88] Willard, H.H., Dean, J.A., Settle, F.A., and Merritt, L.L., Instrumental methods of analysis, 7th ed.; Wadsworth: Belmont, CA (1988).

[89] A. Kraj, D.M. Desiderio, N.M. Nibbering, Mass spectrometry: instrumentation, interpretation, and applications, John Wiley & Sons, (2008).

[90] Silylating agents, http://www.sigmaaldrich.com/content/dam/sigmaaldrich/docs/Aldrich/General_Information/silylation_overview.pdf, (accessed 10.2.2014).

[91] H. Haario, Modest user´s guide, Profmath Oy, Helsinki, 1994.

[92] M. Saeed, M. Depala, D. Craston, T. Lynch, A standard addition method for the quantitative determination of succinic acid and levulinic acid in processed acid stream extracts employing indirect UV capillary electrophoresis, Chromatographia, 47 (1998) 709-715.

[93] A. Gildberg, J. Raa, Properties of a propionic acid/formic acid preserved silage of cod viscera, Journal of the Science of Food and Agriculture, 28 (1977) 647-653.

[94] B. Girisuta, L.P.B.M. Janssen, H.J. Heeres, A kinetic study on the decomposition of 5-hydroxymethylfurfural into levulinic acid, Green Chemistry, 8 (2006) 701-709.

[95] N. Akiya, P.E. Savage, Role of water in formic acid decomposition, AIChE Journal, 44 (1998) 405-415.

[96] W. Hayduk, W. Landie, A.I.Ch.E.J, 19 (1973) 1233.

[97] J.J. Panek, P.K. Wawrzyniak, Z. Latajka, J. Lundell, Interaction energy decomposition analysis for formic acid complexes with argon and krypton atoms, Chemical Physics Letters, 417 (2006) 100-104.

[98] D.W. Marquardt, An algorithm for least-squares estimation of nonlinear parameters, Journal of the Society for Industrial & Applied Mathematics, 11 (1963) 431-441.

[99] W. Spendley, G.R. Hext, F. Himsworth, Sequential application of simplex designs in optimisation and evolutionary operation, Technometrics, 4 (1962) 441-461.

[100] N. Metropolis, A.W. Rosenbluth, M.N. Rosenbluth, A.H. Teller, E. Teller, Equation of state calculations by fast computing machines, The journal of chemical physics, 21 (1953) 1087.