Study of a valorisation process for biomass industrial waste involving acid cooking and enzymatic hydrolysis

(1)

IN

DEGREE PROJECT BIOTECHNOLOGY,

SECOND CYCLE, 30 CREDITS STOCKHOLM SWEDEN 2020,

Study of a valorisation process for biomass industrial waste involving acid cooking and enzymatic

hydrolysis

NICOLAS BRUNET

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF ENGINEERING SCIENCES IN CHEMISTRY, BIOTECHNOLOGY AND HEALTH

(2)

BRUNET Nicolas

Study of a valorisation process for biomass industrial waste

involving acid cooking and enzymatic hydrolysis

External supervisors:

COLIN Julien (LGPM, CentraleSupélec, Fr)

NGUYEN Dang-Mao (LGPM, CentraleSupélec, Fr) REMOND Caroline (FARE, URCA, Fr)

Internal supervisor:

VILAPLANA Francisco (CBH, KTH, Se)

Examiner:

ZHOU Qi (CBH, KTH, Se)

(3)

2

Acknowledgements

Many people were of crucial importance for good progress of this project. Therefore, I would like to thank them, from the bottom of my heart.

First, I would like to express all my gratitude to Emmanuel Fredon (ENSTIB), the researcher who dedicated some of his time to the manufacture of the particleboards used in the conduct of this project. Obviously, without this valuable help, none of the experiments would have had any sense.

I thank Cedric Guerin, engineer in analytical chemistry (CEBB), for kindly explaining me how to use the analytical devices, which physical principles are behind them all, and for supporting me when implementing them. Furthermore, all the valuable help provided about quantifications of compounds urea, formaldehyde and cyanates, both for theory and practice, would be sufficient reason to be rewarded a quotation here. Thanks again for all the nice time we spent together.

Similarly, I would like to thank François Gaudard (FARE) for all his help in implementing of the equivalent devices dedicated to monosaccharide quantification at FARE, including preparation of standard and eluents. I am also very grateful to your communicative cheerfulness at all times.

In relation with this last device, I thank Anouck Habrant (FARE) as well for the time spent in setting the device and for the help I could find at any critical time.

I thank Julia Parlatore Lancha (CEBB) for explaining me how device WAVE^T v2 works. Good courage again for the final touches to your thesis report and for your thesis defence!

I am also very thankful to Mahamadou Mounkaila (CEBB) for all the time spent around the pre- treatment device, be it for the modifications related with pressurized air connection or for adjunction of the temperature acquisition. Thank you for explaining me what happens behind the pictures.

In the same field, I thank Joel Casalinho or his precious feedback during modifications planning and for order checking.

I also show gratitude to Rayen Filali (CEBB) for his priceless support and reactivity along the fermentation experiments. Planning, preparing, and managing such experiments in so little time and unclear substrate conditions would have been absolutely out of reach without such a solid aid. Special thoughts go for the hours spent in Malassez chamber counting, which would have been impossible without you. Thanks as well for your advice about future.

I would like to thank Flora Korovine (FARE) for her explanations about acid hydrolysis and all her help with analysis tables dedicated to monosaccharides and cocktail Cellic cTec2. I take the opportunity of these lines to wish you a swift recovery for your knee.

They are too numerous to all be named, but I would like to thank all the members of both CEBB (without distinction between the two units) and FARE labs for all the good time, meal and conversation we spent together, be it in or out of the workplaces. Special thoughts go to the COCA cell members (FARE) and the other trainees in CEBB office, notably Elise Viau, Héloïse Giordana and Zineb el Hdiy for their presence during dark late autumn days. I am sure I will keep in touch with you.

Last, and most important, I would like to specially thank my tutors Francisco Vilaplana (KTH), Dang- Mao Nguyen and Julien Colin (CEBB) and Caroline Rémond (FARE), first for taking me in charge and second or all their valuable help in report reading and correcting, as late as necessary. I thank you deeply as well for all your advice, and I really hope we will keep in touch in the future.

(4)

3

Abstract

Lignocellulosic biomass has potential to chip in the chemical and biofuels supplies in future societies, even though lignocellulose is a recalcitrant structure that has to be treated in several steps. After their proper life cycle, wood-derived materials such as particleboards have few outcomes today apart from energy recovery for heat production. Then, they may be used as lignocellulosic biomass sources in the production of molecules of interest. Fermentation from wood-derived monosaccharides imposes preliminary sugar retrieval, for instance through pre-treatment and enzymatic hydrolysis. This study focuses on the potential of particleboards waste for chemical and biofuel production by comparing saccharification through simulated steam explosion pre-treatment and enzymatic hydrolysis between native and particleboard-derived wood, with an insight in subsequent fermentation by Saccharomyces cerevisiae. Urea-Formaldehyde bound particleboard was investigated, as well as some aspects of Melamine-Urea-Formaldehyde bound particleboard.

Pre-treatment resulted in apparition of lignocellulosic degraded compounds in a much larger extent in native wood than in particleboard, which seemed to be only superficially impacted. Formation of degraded compounds from sugars – furfural and 5-hydroxymethylfurfural – was enhanced when pre- treatment was prolonged. Removal of a substantial fraction of the adhesive contained in the particleboards was observed, leading to comparable concentrations in free urea, its degraded products, and formaldehyde between native wood and particleboards during enzymatic hydrolysis.

Enzymatic hydrolysis with cellulases and hemicellulases highlighted a critical role of pre-treatment to enhance final yields, both in native wood and in Urea-Formaldehyde particleboard. Adding 20 minutes steam-explosion type pre-treatment at 160 °C resulted in glucose yields increase from 18.5 % to 32.8

% for native wood and from 15.6 % to 37.4 % for particleboard. Prolonging pre-treatment residence time to 35 minutes resulted in much better glucose extraction for native wood but only slight progress for the particleboard, as glucose yields reached 64.5 % and 41.1 % respectively. Maximal concentrations achieved were 277 and 184 mg/gbiomass respectively.

Fermentation brought to light high inhibition from both native wood and particleboard sources of media, which were attributed to components or degraded products of lignocellulose that were not analysed in this project. Ethanol was formed during fermentation, with reduced productivity but increased yields as compared with the control sample. Inhibition was so strong that no difference could be given between native and particleboard wood. In this situation, no inhibition potential of resin or its degradation products could be proved.

(5)

4

Sammanfattning

Lignocellulosic biomassa har potential att bidra till kemikalier och biobränsletillförsel i framtida samhällen, trots att lignocellulosa är en rekalcitrant struktur som måste behandlas i flera steg. Idag trämaterial som spånskivor bara används för energiåtervinning och värmeproduktion efter deras livscykel. De kan därför användas som råvara för framställning av värdefulla molekyler.

Fermenteringsprocesser behöver frisättningen av trä monosackarider genom förbehandlingsprocesser och enzymatisk hydrolys. Studien fokuserar på potentialen för avfall från spånskivor för kemisk och biobränsleproduktion. Vi har jämfört sackarifiering mellan nativt trä och spånskivor genom simulerad ångaxplosion och enzymatisk hydrolys, med en inblick i efterföljande fermentering av Saccharomyces cerevisiae. Spånskivor bunden av urea-formaldehyd undersöktes, liksom vissa aspekter av spånskivor bundna med melamin-urea-formaldehyd.

Förbehandlingen producerade högre koncentration av lignocellulosa nedbrytningsprodukter från nativt trä jämfört med spånskivor. Bildningen av nedbrytningsprodukter från sockerarter - furfural och 5-hydroxymethylfurfural - ökade med längre förbehandlingar. En väsentlig fraktion av limmet borttogs från spånskivorna, vilket ledde till jämförbara koncentrationer i fri urea, dess nedbrytningsprodukter och formaldehyd mellan naturligt trä och spånskivor under enzymatisk hydrolys.

Enzymatisk hydrolys med cellulaser och hemicellulaser avslöjade den kritiska rollen av förbehandling för att förbättra utbytet, både i naturligt trä och i urea-formaldehyd spånskiva. Längre (20 minuter) ångexplosion vid 160° C resulterade i högre glukosutbytet (från 18,5% till 32,8% för naturligt trä och från 15,6% till 37,4% för spånskivor). Förlängning av uppehållstiden före behandlingen till 35 minuter resulterade i mycket bättre glukosekstraktion för nativt trä (64,5%) men endast liten framsteg för spånskivan (41,1%). Detta resulterade i maximalt utbyte av 277 mg Glc/g biomassa och 184 mg Glc/ g biomassa för nativt trä och spånskivor, respektive.

Fermentering visade hög hämning från lignocellulosa nedbrytningsprodukter som inte analyserades i projektet för både nativt trä och spånskällor för media. Etanol bildades under fermentering med reducerad produktivitet men ökade utbyten jämfört med kontrollprovet. Hämningen var så stark att ingen skillnad kunde ges mellan naturligt trä och spånskivor. I denna situation kunde ingen hämningspotential för lim eller dess nedbrytningsprodukter bevisas.

(6)

5

Introduction

The resource challenges in the future

Mankind is currently going through an era of change and challenges. According to the World Energy Outlook 2018, the need for fossil energy sources would go on increasing between 2017 and 2040, mostly because developing economies cannot only rely on renewable sources for their booming power, industry and transportation requirements. Among these needs, oil-derived petrochemicals account for roughly 150 Mtoe, a number that emphasises the crucial importance of developing alternative sources for such chemicals [1]. Yet, on the other side, carbon emissions must be quickly decreased to keep climate change under control, with targets ranging from a 25 % reduction by 2030 to target a 2 °C increase, and even a 45 % reduction to keep temperature evolution below a 1.5 °C increase [2].

To curb the massive use of fossil resources in the chemical and transportation sectors, new production routes have been surveyed focusing on crop and biomass instead of oil as a resource. Among these new opportunities, waste from the wood and agricultural industries attract most of the attention as they do not compete with other applications such as food or feed and they often do not get valorised in the current most common processes. In these extents, they prove more interesting than so-called first generation substrates, even though the latter have been widely studied and used in the industry since they result in the best yields. Still, second generation substrates such as wheat and rice straws [3, 4], corn stover [5], sorghum and sugar cane bagasse [6, 7] or waste wood residues have been quite well studied since the beginning of the twenty-first century and a range of applications using such resources can be found in literature.

Using wood as a material for production of sugars and biofuels 1.2.1. Wood composition

Wood is a lignocellulosic resource, meaning that its main component is lignocellulose, which is a structural assemblage of polymers including cellulose, hemicelluloses, lignin and other extractives.

Proportions in lignocellulosic components in a range of wood species have been reported for long [8]:

these results were averaged in Figure 1 to show the composition variability between softwoods and hardwoods.

(9)

8

Figure 1 – Average chemical composition of softwood (left) and hardwood (right) species. Adapted from Sjöström et al.

[8]

The three main components of lignocellulose have very different properties, which are related to their structures. First, cellulose is formed out of a large number of glucose dimers cellobiose to make a chain of glucan linked in b-1,4 (see Figure 2). This bonding pattern results in stretched molecules. 36 of these glucan chains further stack into a microfibril, that may agglomerate into fibrils throughout a tight intra- and inter-molecular H-bonding pattern. The overall assemblage reaches up to 200 glucan chains that form a very tough and organized fibrous structure [8]. Cellulose may fall into two categories according to this organization degree: it is crystalline where the fibrils are the most tightly linked, and amorphous where they are less organized. Disrupting the structure is easier at amorphous than at crystalline areas [9].

Hemicelluloses are made of heterogeneous polysaccharides: xylans, mannans, xyloglucans and mixed b-glucans. These heteropolysaccharides are constituted by a variety of monomers: D-glucose, D- mannose, D-galactose, D-xylose, L-arabinose, as well as other less common monomers and acids like L-rhamnose and D-glucuronic acid (Figure 2). They assemble into main chains that are substituted and they differ in composition and proportion from one species to another, and still more between softwoods and hardwoods. They share a common point as they are really likely to be hydrolysed, particularly under acid conditions [10].

Finally, lignins are complex polymers containing essentially phenyl groups (Figure 2). Again, their composition widely varies between coniferyl alcohol-rich softwoods and sinapyl- and para-coumaryl alcohol-rich hardwoods. Lignins contain numerous types of linkages (C-C, ether, …) that render these polymers recalcitrant for fractionation. Cellulose fibrils as well as hemicelluloses chains are embedded in the lignin matrix within plant cell walls. Lignin solubility in water is very low and removing the lignin matrix from cellulose thus proves much more difficult than treating the hemicellulose fraction [11].

(10)

9

Figure 2 - Structures of cellobiose in a chain of cellulose (top), a possible example of fragment of lignin (middle), and a glycoxylan exemplifying hemicellulose. Illustrations from Alonso et al. [12] and Sette et al. [13]

1.2.2. Wood residues valorisation

To date, lots of options to valorise wood residues issued from the forestry and wood industries already exist [14]. This raw material consists in an amount of wood chips and particles of small size, ranging from sawdust to bigger fragments that are still too small to be used as primary wood resource for top applications. Traditional uses of such material include particle or fibre recycling into composite materials, pulping for paper crafting, or burning for energy recovery. However, cascade applications can be imagined between sawdust or primarily recycled materials and mere energy recovery. These applications are major subjects of research today. Sugar retrieval for fermentation, phenol production, or still more recent applications such as a project aiming at producing lactic acid [15] are examples of such intermediary applications. Retrieval of second generation (2G) sugars from lignocellulosic biomass seems like a credible alternative to energy recovery, be it for such residues or for larger biomass sources. Several methods have been implemented to recover the monosaccharides from biomass, although industrial scaling up is far from being as advanced as that oriented to first generation (1G) sugars.

1.2.3. Application for biofuel production

In most applications, sugars are not the final product interesting for industrial applications. The glucose retrieved from lignocellulosic resource is often used in subsequent fermentative processes (Figure 3).

Yeast Saccharomyces cerevisiae is the organism used in most of the applications aiming at producing bioethanol from glucose. Using xylose is also interesting since it is an essential building block of

(11)

10

hemicellulose for most crop species. Consequently, yeast strains are being evolved to be able to take xylose as well as glucose for substrate in ethanol production.

Several other technologies aiming at producing higher added-value molecules from sugars have been well-documented since their first implementations. Now many metabolic routes are known to lead to overproduction of molecules in alternative ways to petrochemicals.

Production of bioethanol as a biofuel or chemical platform is probably one of the most documented employs for glucose – as well as other sugars, hence the choice to focus on this molecule. The metabolic balance of alcoholic fermentation is:

!_"#_$%&_"+ 2 *+, + 2 ,_- → 2 !_%#_/&# + 2 !&_%+ 2 *0,

However, before any fermentation step to be performed, some treatment is necessary to turn the sugars contained in lignocellulosic biomass to glucose or other monosaccharides. Several chemical and biological options exist to degrade biomass. Enzymatic hydrolysis is viewed as a promising and environmental-friendly way to retrieve glucose. Therefore, enzymes attract research subjects in the field of lignocellulosic valorisation today.

Two main strategies may be set for the overall process regarding whether enzymatic hydrolysis is performed prior or simultaneously to fermentation. Separated Hydrolysis and Fermentation (SHF) consists in performing enzymatic hydrolysis and fermentation in successive steps, which allows appropriate tuning for each step and is thus easier to control, for instance at the lab scale. Another important strategy for ethanol production, is known as simultaneous saccharification and fermentation (SSF). It consists in starting fermentation at the same time as production of monosaccharides by adding both enzymes and yeast simultaneously with the pre-treated biomass. SSF is interesting at the industrial scale since it limits inhibition of cellulases by their product glucose and reduces the number of tanks necessary to complete the process. Operation parameters must then be balanced between the requirements of enzymes and micro-organisms which have different optimal pH and temperatures [16].

Figure 3 - Transformation chain from lignocellulosic biomass to bioethanol, including pre-treatment and enzymatic hydrolysis for retrieval of the glucose and fermentation of the glucose to ethanol

(12)

11

Enzymatic hydrolysis

1.3.1. The enzymatic system degrading cellulose and hemicellulose

The enzymatic action allowing degradation of lignocellulose results of a large cooperation of enzymes.

Access to the cellulose requires the synergy of a range of enzymes, aiming at disrupting lignin and hemicellulose first, as displayed in Figure 4 [17]. Considering cellulose alone, the enzymatic degradation system includes three types of enzymes. Endoglucanases, first, are designed to break the long cellulose strands into shorter fractions. Second, cellobiohydrolases degrade these fragments to form cellobiose units. Third, these glycosylic dimers are cleft by b-glucosidases yielding glucose monomers. These enzymes are all classified as glycoside hydrolases, EC.3.2.1.x.

Systems such as cellulose binding modules (CBMs) co-operate with these enzymes to form very complex and diverse systems to improve digestibility of cellulose by facilitating enzymatic binding to specific or unspecific sections of the cellulose [16]. Another class of accessory enzymes are the recently discovered copper-dependent lytic polysaccharide monooxygenases (LPMOs) that promote disruption of the crystalline parts in cellulose structure, through oxidative mechanisms (Figure 5) and enabling further catalytic action by proper cellulases (Figure 4) [18, 19]. The roles of LPMOS have been speculated to be wider again in helping cellulose degradation since this discovery.

Figure 4 - Typical enzymatic system involved in lignocellulose disruption. Degradation of lignin and hemicellulose is required to enhance action of proper cellulases. Figure from Champreda et al. [17]

(13)

12

Figure 5 - Oxidative cleavage of cellulose by LPMOs as proposed by Horn et al. [19]

Of course, similar systems are at stake in the case of hemicellulases (Figure 4). The family scene of enzymes degrading hemicellulose is even more complicated to draw since hemicelluloses vary with the lignocellulosic source. Complexity of the enzymatic system involved in hemicellulose degradation is higher than that designed for cellulose, essentially because of the number of possible bonds as compared with the mere glycosylic b-1,4 linkages found in cellulose. An extensive review over the variety of enzymatic systems degrading hemicellulose was published in Current Opinion in Microbiology and Figure 6 summarizes the variety of bonds found in hemicellulose and the dedicated enzymes [20]. Again, CBMs [16] and accessory enzymes LPMOs [21] have their role in accelerating the process of hemicellulose degradation.

Figure 6 - The variety of hemicellulose structures found in nature and the dedicated enzymes allowing their deconstruction. Figure taken from Shallom and Shoham [20]

But even with this very sophisticated system of enzymes, the overall result is still highly dependent on the substrate properties. Particularly, tailored enzymes are often needed since the surrounding

(14)

13

environment is as much crucial in enzymatic action as the type of bond cleaved by an enzyme. In the case of cellulose, even when the specific surface area is quite large, the fibre can only be bound to molecules of small dimensions, that were estimated to be lower than 51 Å [22, 23]. This makes reaching of the active sites tricky for most enzymes. The enzymes behaviour can be modified by the presence of other molecules in the medium which may act as inhibitors. Cellobiose and glucose are responsible for product inhibition of the cellulolytic system, particularly in SHF because cellobiose and glucose accumulate in the medium with this strategy (Figure 3). Phenols and polyphenols from lignin, as well as mono- and oligosaccharides (notably xylo-oligosaccharides), were shown to have a strong impact towards cellulases depending on their concentration, as described before [24, 25, 26]. As a result of all these interactions and constraints, the overall kinetic is almost impossible to report properly [16].

For the laboratory and industrial applications, enzyme cocktails have been developed by companies such as Novozymes to combine these activities in appropriate concentrations, providing user-friendly solutions. Such cocktails e.g. Celluclast and Cellic CTec2, have been extensively reviewed [27, 16, 26, 25].

1.3.2. Improving the enzymes performance and cost

As the enzyme production is still a significant burden in terms of costs, much investment was carried out in order to improve the enzymes efficiency as well as their production cost [28, 16].

To improve the yield of this step, it is still possible to add other elements to the medium, such as accessory LPMO enzymes that could participate in the enzymatic action in addition to the proper cellulolytic mixture. For example, adding LPMO AA9 to the medium improved mixture Celluclast’s efficiency by affecting the poorly accessible crystalline fraction of the cellulose and thus providing improved yields in glucose [9]. Metabolic engineering could result in the synthesis of fermentation microorganism strains with improved skills at degrading directly cellulose, which would be an opportunity to cut costs by performing all the degradation steps at once and by increasing yields through enzymatic thermostability and activity [29]. Moreover, more recent enzymes mixtures were developed to improve the performance of former mixtures, such as CTec3 resulting in better results than Celluclast [25]. Finally, great efforts were made to better understand cellulose enzymatic degradation and to potentially engineer more performant enzyme producers [30].

Wood pre-treatment

1.4.1. Wood structure hinders sugar retrieval

As detailed above, wood is a composite material, the components of which have very different properties. This specificity of lignocellulosic structure renders the dedicated enzymatic cleavages very hard to be performed in untreated biomass because the access of cellulose fibres is hindered by hemicellulose and lignin, which form a matrix inhibiting enzymatic degradation (Figure 4). Use of undisrupted cellulose results in increased enzyme consumption and lower cellulose digestibility [31], which has to be improved for an industrially viable process.

Some initial pre-treatment is thus needed to make the following enzymatic action more efficient.

Hemicelluloses, lignin and acetyl groups which hinder the cellulose are partially removed to increase the accessible surface area (ASA), which determines the ability for the enzymes to reach the cellulose

(15)

14

strands [32, 33]. As explained in part 1.2.1, crystallinity of the cellulose is also limiting to its enzymatic degradation [28]. That is why reducing crystallinity is also one of the roles of pre-treatment. In brief, the pre-treatment operation aims at making the cellulose more accessible for subsequent operations.

1.4.2. The variety of pre-treatments

In order to reach the expected effects described above, several methods can be implemented and are well-covered by literature, notably in large reviews [34, 35]. These techniques include mechanical treatment of biomass such as grinding, pyrolysis or irradiation; chemical treatment using for instance dilute acid, alkali, ozone, ionic liquids or organic solvents; physico-chemical action, for example through hydrothermal, steam, oxidative, sulphite, CO2 or ammonia fibre explosion (AFEX) or ammonia recycle percolation (ARP) pre-treatment [34, 35]. Some examples of such treatments are given in bibliography for complementary information [36, 27, 37, 38, 39, 40, 14, 41]. It is also possible to use microorganisms to degrade biomass through biological pre-treatments [34, 35]. The results obtained after the pre-treatments quoted before were surveyed by several studies [6, 31, 22, 42, 43] and their performances are reported as Figure 7. When directed at enzymatic hydrolysis enhancement, pre- treatment often involves partial to total removal of the hemicellulose, degradation of the lignin matrix, and preliminary attack of the cellulose molecules to disrupt their strong organization [28].

It was also advocated that combining various treatments could prove useful to benefit from their combining effects towards biomass [6, 31]. For instance, as dilute acid and lime efficiently treat efficiently respectively the hemicellulose and lignin fractions of sweet sorghum bagasse, it was suggested that a double-pre-treatment would result in purer cellulose retrieval [6].

Variation in the responses to pre-treatment between softwoods and hardwoods, that bear guaiacyl- lignin and syringyl-lignin respectively on top of different hemicellulose composition, were extensively investigated to review and compare these main substrate classes [44, 37, 25, 27]. Softwoods are generally recognized as more recalcitrant than hardwood substrates.

Figure 7 - Comparison between wood pre-treatments as adapted from Zhao et al. [35]. Symbols are meant regarding how cellulose digestibility is impacted by pre-treatment: (+) improved, (-) decreased, (/) no clear impact, (?) unknown. For

(16)

15

instance degrading hemicellulose or lignin is favorable to cellulose digestibility (+), whereas degrading cellulose during pre-treatment causes lower digestibility for the enzymes (-).

1.4.3. Pre-treatment impacts: focus on steam explosion

Steam explosion is one of the available techniques for disruption of lignocellulosic biomass, which leads to results relatively close to those of hydrothermal or dilute acid pre-treatments (Figure 7). It consists in cooking of biomass in high pressure and temperature conditions, followed by a brutal reduction of pressure to atmospheric pressure. The characteristic time of this second step is below 1 s. At the pilot scale, this can be performed by an outlet in the cooking device through which biomass can be pushed out. Consequences of such brutal changes in environmental conditions include explosion of the wood cells, hence the name of steam explosion technique.

In such conditions, hemicellulose is no strong structure. It is actually quite easy to hydrolyse into smaller oligomers or monomeric sugars. As compared to hemicellulose, lignin is more recalcitrant and it is much more difficult to decompose into a liquid fraction, although kraft pulping for paper production is a relevant example of acid-based process that has been employed at industrial scale for a long time to treat lignin [45]. Several studies have shown that lignin is much easier to degrade under alkaline treatments, notably using lime [6].

That being said, pre-treatment must be well-tuned since too strong treatment of the lignocellulose may also result in overproduction of degraded products. For steam explosion, those mostly include furfural and 5-hydroxymethylfurfural (HMF), formed throughout dehydration of pentoses and hexoses respectively. These compounds can be converted to other chemical species afterwards, namely formic acid for furfural and levulinic and formic acids for HMF [37] (Figure 8). Phenolic compounds coming from the degradation of lignin are also present; they are even overproduced under alkaline conditions.

Pre-treatment severity enhances sugar retrieval as well as the transformation towards these degraded products [37, 28, 36, 38]. A balance must be found between degradation of hemicellulose, degradation of lignin, and relatively little formation of degraded products during pre-treatment.

Figure 8 - Summary of monosaccharides main degradation products and phenolic compounds generated during acidic pretreatment. Image from Jönsson et al. [46]

(17)

16

At regular levels of furfural and HMF, early studies reported no effect towards the enzymatic activity of cellulase and b-glucosidase, although these degraded products could affect fermentation [47, 48].

Their removal during downstream processing is usually a necessary step to achieve satisfying growth of microorganisms. Phenolic compounds, on the other side, impact the enzymes and microorganisms even more than lignin itself [49, 50]. Therefore, they must be treated for industrial sugars production by fermentation. It is still possible that the microorganisms used for fermentation grow some tolerance towards these inhibitors if adaptive evolution time step allows it [4], thus lowering the necessity for heavy efficient downstream processing steps.

Other physical factors are to be taken into account when considering cellulose accessibility. As the ASA was shown to be directly connected to the enzymatic activity [49], the crystallinity of cellulose fibres also plays a role in their accessibility [3, 9]. Treatments such as drying and pressing may then cause structural change in cellulose and could affect enzymes binding [51, 52, 41].

From an economical consideration, it is also necessary to check the possible yields in degraded products. Indeed, instead of preserving mild conditions to get sugar monomers, it could be also interesting to force strong conditions and degrade most of the sugars into HMF or furfural, which may be used as key molecules for green chemistry [53].

1.4.4. Consequences of pre-treatment for enzymatic hydrolysis and fermentation

As explained in 1.4.1, pre-treatment is used to degrade lignocellulose prior to addition of enzymes, ultimately improving the cellulose content by removing partially lignin and hemicellulose from the residue and increasing the ASA for the enzymes.

After steam explosion pre-treatment for instance, the biomass treated is divided between a liquid fraction containing mostly saccharides from the hemicellulose, and a solid fraction that includes most of the cellulose and lignin. Usually, these two phases are then separated by filtration, since the liquid fraction contains mono- and oligosaccharides and phenols that are inhibitive to the enzymatic activity according to several studies [25, 30, 54]. The residue resulting of this process is then hydrolysed enzymatically to retrieve the glucose contained in cellulose in presence of cellulases. When hemicellulases are present, the remaining hemicelluloses are also hydrolysed into monosaccharides.

The results of this enzymatic treatment are strongly dependent on the pre-treatment severity, as stated before [38, 37, 39]. Again, it must be noted that this stands for acid-type pre-treatments. Pre- treatments of other kinds could lead to other environmental conditions for the enzymes.

Degradation products formed during pre-treatment (Figure 8) may also impact the results of enzymatic hydrolysis, although their main effect is against the organisms used for subsequent fermentative processes according to literature [46]. As for the saccharides and phenols released by pre-treatment, these molecules should then be considered when studying the variety of potential adverse effects to the enzymes and the microorganisms. Furthermore, oligomeric phenols, lignin-carbohydrate complexes as well as tannic acid strongly inhibit the action of cellulases, sometimes even causing precipitation with cellulose. Some other polymers could have a positive impact for enzymatic hydrolysis and fermentation though [24, 55].

Using the liquid fraction of pre-treatment may enhance the productivity of the whole process since it also contains sugars for potential fermentation, but proper downstream processing is often necessary then to remove the inhibitors that are present in larger amounts in this fraction.

(18)

17

Wood-based boards, another possible source of biomass 1.5.1. A promising source of 2G sugars

The literature cited so far mostly focuses into wood wastes that have not undergone any specific treatment with chemicals to form proper wood-based products. Yet, those may prove to be a promising source for feedstock, as they form a large amount of biomass which calls for recycling at the end of their proper life and that are only poorly valorised today. Recent data about panel production amounts was released by the Food and Agriculture Organization of UNO. It shows that the amount of wood susceptible to be valorised reaches stunningly high figures, and that these numbers were increasing (Figure 9) [56].

Figure 9 - Wood-based panel balance, Europe, 2015-2017 (thousand m³). Figure from UNECE/FAO [56]

1.5.2. Wood-based boards life cycle

In addition to high-value products, wood processing must be accounted for retrieval of large quantities of wood particles that are no longer suitable for high-value products manufacturing. However, these side-products can be used for other applications, notably those of the wood-based board industry.

Figure 10 shows some figures of panel production in Europe. It is worth noting that particleboards count for half of the total panel production.

Wood chips are collected, blended with other products and processed into medium-density fibreboards (MDF), oriented-strand boards (OSB) or particleboards for instance. Figure 11 compares some characteristics of the different categories of wood panels. The chemicals added to wood particles are structurally relevant to panels and may include adhesives, hardeners, coatings or flame-retardants, in higher or lower proportions selected considering the application the panels are designed to [57].

The fraction of structural adhesives is much higher in MDF than in particleboard and OSB, owing to the typical size of the particles used and the aimed final density. The proportions of each of the kinds of panels produced is regularly displayed by the FAO [56].

(19)

18

Figure 10 - Wood-based panel production, Europe, 2016 [56]

Figure 11 - Comparison of the different types of wood-based panels. Data extracted in [45]

Boards are then used for domestic or industrial applications, e.g. the building industry. For instance, they are crucial for making wood-based objects of specific dimensions, as plain wood would certainly not be used to make large and light panels. Their treatment against parasites or resistance to fire can also be improved compared to traditional wood.

But when they come to the end of their life cycle, the additives that were once useful to maintain panel characteristics become a burden for their recovery, since they may not be compatible with the processes designed for other purer wood sources (Figure 12). The content in additive chemicals impedes similar treatment as for other cleaner wood sources, at least a priori, and data in that field is still lacking.

(20)

19

Figure 12 - Illustration of wood-based boards life cycle

1.5.3. Concerns regarding wood-based boards recycling operations

Indeed, very few options exist regarding the recycling of spoiled wood sources such as particleboards.

This is mostly due to the content of such materials. There are several possible origins for the wood, and a general and user-friendly recycling process must work independently of the composition [14].

Although recent data for total European sawn wood shows that softwood is used in amounts 5 times larger than for hardwood, with almost 200 million against 40 million m³, the actual composition of panels is very difficult to deem since it depends on the supply chain for each factory, that collect timber from far away and from various origins. Interestingly, spruce and pine alone make as much as two thirds of the total round timber amount used in particleboard crafting in Europe [58]. Softwood and hardwood difference in structure should be accounted for a first limitation to such projects since substrate-tailored treatment is necessary to enhance production but unrealistic in actual industrial applications.

What is more, particleboards are composite materials containing not only wood, but also adhesives and other additives (plasticizers, hardeners, flame-retardants…), the behaviour of which is poorly known under the same treatment as clean wood wastes [43]. These compounds could contaminate the resource and ruin the techno-economic feasibility of such applications. Several patented techniques were described to valorise such wood composite materials, but they often need uneasy processes or expensive hardware [59].

No doubt, companies that gather particleboards and other contaminated sources could be interested in easier treatment for such resources. Today, the landfilling of organic waste including wood panel composites is banned in some countries such as Sweden. Incineration of chemically contaminated wood must be operated in dedicated plants as it releases chemicals that may corrode the devices on top of yielding environmental-unfriendly ashes and effluents [57]. No other valorisation method has yet been developed at the industrial scale.

(21)

20

1.5.4. The adhesive fraction

Among the different existing adhesives, urea-formaldehyde (UF) resins are of the most commonly used to make particleboards, accounting for more than 90 % of total adhesives, although other mixtures are used for other applications, notably Oriented Strand Boards [57, 60]. The extensive UF-adhesives use is due to their great properties as low cost, non-flammable, rapidly cured and aesthetic coloured compounds [57]. Their main disadvantage for most applications is their poor resistance to hydrolysis, contrary to other resins designed for outdoor applications such as Melamine-Urea-Formaldehyde (MUF) resin. Typical linkages in these resins are shown on Figure 13.

The UF-mixture is obtained through mixing urea with formaldehyde at varying proportions to reach the required texture. The molecules then react to monomethylolurea through a reversible reaction [61]. Their degradation may also result into production of degraded molecules from hydroxymethyl and methylene groups [62, 63] (Figure 14).

Figure 13 - Typical linkages in UF and MUF resins [63, 64]

Figure 14 - Possible hydrolysis reaction pattern from UF resins [62]

Several environmental factors could affect hydrolysis operations, such as the amount of resin, the formaldehyde per urea ratio which is closely related to the crystalline or amorphous nature of the material [65, 66, 67, 68], or the severity of treatment conditions, including temperature and chemical species in the medium [69].

Such resins were given extra attention in the past because they could result into health issues, which call their use into question when they are used into indoor environments. UF resins usually naturally

(22)

21

emit formaldehyde to the air, which is a carcinogenic molecule [70]. The emissions are stronger again when hydrolysis occurs and when the glue is exposed to higher temperatures [71, 66], that seem to play a key role in the emissions.

1.5.5. Using wood-based boards as feedstock for high-value products generation

The huge majority of studies that have been released considering the valorisation of lignocellulosic biomass into fermentation sugars take into account unprocessed substrates e.g. wood chips. That is why, the conclusions of these documents may not be the same considering particleboard-type biomass, which does not only include lignocellulosic biomass but also other fractions such as adhesives, hardeners, flame-retardants and other molecules.

These additives which were blended with wood to perform crafting and provide properties to the panels may now well make traditional treatment processes obsolete, indeed impacting for instance pH, cellulose accessibility or the enzymes behaviour. However, these molecules could also have no impact or even promote some of the steps necessary to degrade particleboards into sugars or further degraded products. Good example is the case of urea, a molecule widely used to make the adhesive fraction of particleboards. Urea has been known for long to cause protein denaturation at high concentrations [72], but besides it is a H-bond breaker that may support saccharification in the range of concentrations observed in the panels [73]. Furthermore, urea degrades into ammonia and cyanate, which in turn degrades to carbon dioxide and ammonia, under acid conditions [74]. That could change the pH conditions [71] and prove useful as a nitrogen source in subsequent fermentation operations.

From that point, two strategies are possible regarding the additives. Either they can be kept in the reactional environment, which would of course be much easier from an industrial point of view; or they need appropriate treatment for removal, ideally hardly impacting the additive fraction but preserving the lignocellulosic material from too strong degradation.

When literature references concerning the impact of such chemicals on wood recycling are missing, at least to our knowledge, it becomes necessary to investigate the wood-additives-treatment interactions to figure it out. The type of wood panel probably being a key element and time being limited, this project focuses on the case of particleboards, widely used in indoor environments.

Materials and Methods

Materials: particleboard panels and their components

The available materials include two particleboards manufactured using UF and MUF resins respectively, that were provided by ENSTIB (Épinal, FRANCE). They were manufactured through blending of 925 g wood residues with 4 % water content, 108 g UF or MUF adhesive at a concentration of 66 % in resin (the remaining part consisting in water), and 1.4 g hardener ammonium sulphate.

Resins used in the panels were Sadecol L 3096S for UF resin and Sadecol L 3118 for MUF resin, which are adhesives designed for industrial use, provided by society Sadepan chimica (ITALY). Ammonium sulphate was purchased from Sigma. This corresponds to roughly 8.0 % resin content in the finished product. In this study and contrary to industrial panels, in order to focus on the effect of resin on subsequent steps, no other additive was incorporated in the panels.

(23)

22

Adhesive and wood particles were first mixed for 1 mn. Then the mixture was pressed using a single- opening press to be shaped into a panel at T = 200 °C through a three-steps process:

1. 120 s at 70 bar pressing;

2. 60 s at 24 bar pressing;

3. 120 s at 13 bar pressing.

They were then cooled down in open air. The steps of panel preparation are displayed on Figure 15.

Figure 15 - Preparation of the panels. From left to right: gluing; pre-pressing; pressing in single-opening press; and cooling down.

The result was two square panels of roughly 30x30x1 cm in each dimension (Figure 16). The first one was bond with UF resin; the second with MUF adhesive.

To support the experiments, untreated sawdust of the same nature as that of the boards was retrieved from the same source. These wood particles were provided by ENSTIB and are assumed to come from the Vosges area in north-eastern France, in which softwood species such as spruce are dominant. UF and MUF resins of the same nature were also provided to allow modifications of the resin content.

Figure 16 - Available material include unprocessed native wood, UF- and MUF-bond particleboards manufactured with the same type of biomass, and UF and MUF resins.

Hydrothermal pre-treatment 2.2.1. Updated pre-treatment system

In this study, an acid cooking system derived from device WAVE^T v2 [75] was used. This system, which was initially designed to perform rheological studies, was modified to be employed as an acid cooking device for the acid hydrolysis step. Preliminary operations leading to the modified system were performed in a previous project carried out at CEBB [76]. In the present work, several modifications

(24)

23

were performed in order to reduce the time necessary for heating and cooling of the device: the acid cooking tool was connected to the pressurised air system, so that the water inside the lower chamber was already set into high-pressure conditions (Figure 17). This way, the only energy to be provided to the system is used for water heating, whereas water vaporisation and pressurization had to be taken into account in the first system.

Figure 17 - Comparison between the hydrothermal treatment system as it was in September (left, picture from [76]) and as it is now (right). The main differences include pressurized air input and output as well as computer acquisition of

temperature; test tubes are used to perform the cooking and easily retrieve the treated biomass

A schematic view of system WAVE^T v2 shown on Figure 17 is depicted on Figure 18 (left). The initial system being designed to perform rheological studies, some of the elements of the mechanical chain of instrumentation are absent in the device used in this project, which is only designed to pre-treat samples to be characterized. Further modifications include a system for set and retrieval of the sample, involving an attach and a hook to suspend a 40 mL test tube containing the substrate to pre-treat, as displayed on Figure 18 (right). During experiments, test tubes were filled with 2g substrate completed with 20mL ultrapure (UP) water. A 1 mg precision scale was used to record the weight of biomass and water. Control of temperature is ensured thanks to a proportional-integral-derivative (PID) controller;

temperature in the lower chamber is measured with a K thermocouple.

(25)

24

Figure 18 – (left) Schematic view of experimental device WAVE^T v2. 2, pressure gauge; 3, safety valve; 5, gas leakage; 6, series of fins; 9, electric heater. Elements 1, 4, 7 & 8 are absent in the modified version displayed on Figure 16 [75].

(right) System of attach of test tubes. A test tube with a thread is filled with water and substrate particles; then an inox attach is screwed to the tube and suspended to the lowest fin (6 of left figure) thanks to a hook. Testing chamber is then

filled with 100 mL ultrapure water to heat up the sample.

In the modified system proposed to the left of Figure 17, temperature can be acquired on a computer for further analysis. PID controller must be set with appropriate coefficients for proportional, integral and derivative corrections to ensure satisfying control of temperature during the hydrothermal pre- treatment. Figure 19 proposes two sets of parameters, as were acquired with the Autotune function of PID controller during this work and from the original device WAVE^T v2 during previous rheological experiments. A priori, the best choice would be to keep the Autotune coefficients which were obtained with the modified device described before (Figure 17 and Figure 18).

Comparison between the two sets of parameters exemplifies the dilemma between speed to reach the temperature plateau and stability of the temperature during this plateau (Figure 20 and Figure 21).

For the same input schedule, the parameters provided by the Autotune program caused a fast heating- up, but serious instability of temperature during temperature plateau. By contrast, the higher parameters from the original device WAVE^T v2 resulted in longer time to reach the target temperature, but improved stability during the steady phase (Figure 21).

As repeatability of pre-treatment conditions was key in this project, the coefficients from the original device WAVE^T v2 were selected since they provide average temperatures closer to target and lower standard deviation, even though heating-up phase is slower in such conditions, as showed in Figure 20.

Figure 19 - PID controller parameters obtained with Autotune function on the new device (3 top lines), and with Autotune

from original device WAVE^T v2 (3 last lines)

Figure 20 - Comparison between the behaviour of the device with coefficients obtained with Autotune for the new device

and for device WAVE^T v2. Time before plateau, temperature average and dispersion (standard deviation) are compared.

New device coefficients WAVE ^T v2 coefficients

Time before plateau (s) 350 990

Average T on plateau 160.9 160.3

S.D. on plateau 2.175 0.742

(26)

25

Figure 21 – Comparisons of heating-up and plateau profiles obtained for a temperature input of 160°C. Coefficients provided for device WAVE^T v2 result in slower heating-up but more precise and stable temperature plateau.

2.2.2. Samples preparation

The panels were first sawed, leading to panel cubes of 2 to 5 cm in their two main dimensions. The panel fragments and the particles were then grinded using mill Retsch SM300 and a 4 cm-grid. The aim of this operation was to get particle size roughly comparable to that of native wood (Figure 22). By obtaining similar granulometries, such a treatment ensures that the following steps are only dependent on the initial composition.

(27)

26

Figure 22 - (left) View of mill Retsch SM 300. (right). Granulometries obtained after milling in Retsch SM 300 for native wood (top) and UF particleboard (bottom).

2.2.3. Simulation of steam pre-treatment

One purpose of the project was to review steam explosion pre-treatment of lignocellulosic biomass.

However, no steam explosion device was available, be it at CEBB or at FARE. Instead, it was decided to combine hydrothermal treatment with grinding: this last process aims at simulating the explosion step.

Still, some bias is possible since the wood components accessible to enzymes in the enzymatic hydrolysis step is not the same between grinding and explosion (Figure 23 and Figure 24). Explosion of cells exposes their inner layers (inside the lumens, mainly composed of cellulose) to enzymes, whereas grinding causes split between cell and exposes the outer layers (the compound middle lamella, mainly composed of lignin).

Figure 23. Specific breaking zones of wood cells for (a) explosion and (b) grinding [77]

Figure 24. Distribution of the main constituents in (a) hardwood and (b) softwood cell walls [78]

(28)

27

2.2.3.1. Hydrothermal cooking step

Figure 25 summarizes the various types of pre-treatments which were applied to each of the substrates, using the system described in 2.2.1.

Figure 25 - Matrix of experimental conditions tested in pre-treatment

Planning numerous experiments without a strategy seems to be an expensive and time-consuming process. That is why some criteria are often used to reduce the number of trials to determine the pre- treatments to be applied to the substrates.

The combined severity factor (CSF) was proposed by Overend and Chornet in 1987 [79]. It sets up a relationship between the experimental temperature, duration and initial pH. It is established on the observation that increasing temperature by 10 °C yields results equivalent to doubling the final time:

1₂ 3, 0 = 1₂ 3

2 , 0 + 10 ; Hence the relationship: 1² 3, 0 = 3 ∗ 9:;:<=>

?@,AB

Then, the combined severity factor is defined to take into consideration the acid species present in the medium, then employing the expression of pH with a log-scale:

!CD = log (1₂) − K#

As no chemical species were added to the medium to change pH in this project, 12 seemed to be the designed candidate for comparison between experiments. However, according to step-by-step calculations of this criterion, pre-treatments NW-PT-180-20 and NW-PT-160-60 would lead to similar results, which proved not to be the case in practice (data not shown). Severity parameters indeed rely on one major hypothesis: that polymers decomposition to monomers is much faster than the monomers’ further degradation. This assumption is probably not checked in the experiments conducted in this project.

As a consequence, it was decided to set the temperature to 160°C and to restrict variations to the time parameter in the following of this project.

2.2.3.2. Simulation of explosion step

Both native wood and panel-derived particles were milled to smaller particles in mill IKA M20 (Figure 26). The grinder was filled to half the container with wood particles, and programmed to work for four times 15 seconds, each separated by 15-second breaks to prevent both particles and grinder from heating. The resulting particles are displayed on Figure 26.

(29)

28

Figure 26 – (left) view of mill IKA M20. (right) Granulometries obtained after milling in IKA M20 for native wood (top) and UF particleboard (down).

Simulation of steam explosion was supposed to be made by first hydrothermal step and second simulation of explosion by milling. Actually, this was difficult to implement since the solid fraction is hard to retrieve alone from the test tubes. Size of particles from Retsch SM 300 also sometimes limited their setting in the test tubes. Therefore, milling simulating the explosion step was performed prior to hydrothermal treatment. Limits of this method are discussed in the discussion part of this report.

Enzymatic hydrolysis treatment 2.3.1. Preparation of enzymatic hydrolysis

After pre-treatment, the samples pre-treated according to the matrix of experiments given in Figure 25 were split between a gross liquid fraction and the remaining solid. Liquid fraction was retrieved by pouring the supernatant of pre-treated tubes in separate containers. UP water was then added inside the test tubes to facilitate removal of the solid: the resulting solution was poured in a separate container. Both prehydrolysate and solid containers were preserved by conservation at -20 °C.

The samples were then taken to FARE. In both containers, the solid fraction was separated from the liquid by centrifugation at 5000 rpm for 10 mn using a Thermo Scientific SORVALL ST8R Centrifuge. The solids were then placed in a ventilated thermal dryer set to 50°C for desiccation during at least 60 h.

Enzymatic hydrolysis was performed in 10 mL bottles, closed by corks with Polytetrafluoroethylene (PTFE) seals. Immediately after the desiccation step, 100 mg of substrate (dried solid fraction) was weighed and poured in each bottle, together with a magnetic stirrer. The remaining substrate powders were saved into hermetic tubes to undergo carbohydrate contents analyses (2.5.2). Water content in the substrates was then assumed to be 0%, except for the un-pre-treated substrates NW-NT and UF- NT. Two sets of enzymatic hydrolyses were performed. The experiments were performed in duplicate for the first set, which included NW-NT, UF-NT, as well as a preliminary overview of conditions NW- PT-160-35, rinsed UF-PT-160-35 and unrinsed UF-PT-160-35. Experiments were performed in triplicate in the second set, which included NW-PT-160-20, UF-PT-160-20, NW-PT-160-35, UF-PT-160-35 and MUF-PT-160-35. Description of the codes is provided in the experimental matrix (Figure 25).

(30)

29

Different solutions containing antimicrobial compounds and buffer were then added into the bottles containing the substrates: 100 µL of 0.5 mg/L chloramphenicol dissolved in ethanol; 50 µL of 0.2 % sodium azide dissolved in UP water; 4.67 mL 0.5M citrate-phosphate buffer at pH 5.5. Chloramphenicol was purchased from Eurobio. Sodium azide was purchased from Riedel-de Haën. Chemicals used for preparation of the buffer were purchased from Sigma. Last, enzymatic cocktail Cellic cTec2 with cellulase activity measured to 84 FPU/mL (Figure 27) was diluted 10 times, and 178 µL were added to each bottle to reach a cellulase activity of 15 FPU/mgsubstrate in solution. Final enzymatic hydrolysis volume was then 5 mL. The cocktail used also contained hemicellulose activities.

Enzymatic activities in the cocktail were tested between 30/10/2019 and 11/12/2019 with FPU method in citrate-phosphate buffer. Data was provided by FARE (Figure 27)

Figure 27 - Summary of some enzymatic activities in citrate-phosphate buffer

Previously autoclaved Erlenmeyer flasks containing respectively substrates NW-PT-160-35 and UF-PT- 160-35 were also prepared for further fermentation experiments. The components described before were added in similar proportions but in 30 times bigger amounts, resulting in 150 mL enzymatic hydrolysis medium.

2.3.2. Proper enzymatic hydrolysis and sampling

All preparations were placed in a ventilated cabinet thermically controlled to 50 °C, on a stir plate allowing mild agitation of the media under magnetic stirring.

Samples were taken after 0h, 24h, 48h and 96h in each bottle. For the 0h, 24h and 48h, 500 µL were taken from each bottle and placed into 2 mL Eppendorf tubes. These were placed in boiling water for ca. 10mn to perform denaturation of the enzymes, then centrifuged at 5000 rpm for 10 mn. The supernatant was finally retrieved and saved in other Eppendorf tubes at -20 °C until further analyses.

After 96 h, the solution remaining in the bottles was poured into small tubes and placed in boiling water for enzymes denaturation. 1 mL was then taken from these tubes and placed into Eppendorf tubes to undergo the same treatment as 0h, 24h and 48h samples. The Erlenmeyer flasks were left for enzymatic hydrolysis 185h in total, then placed into ice to stop the reaction and taken to CEBB for fermentation experiments. 1 mL samples were retrieved after 96h and 185h, with similar treatment as described above.

It must be noted that the 0h sample was made similarly to the 24h and 48h ones in the first set of experiments, but that sample was taken prior to enzymes addition in the second set of experiments.

This was made to ensure that the enzymatic hydrolysis is not begun in this 0h sample. Volumes and monosaccharide content of the enzymatic cocktail were corrected in the final results thanks to a complementary composition analysis for the enzymatic cocktail.