An integrated biore ﬁ nery to produce 5-(hydroxymethyl)furfural and alternative fuel precursors from macroalgae and spent co ﬀ ee grounds †

5-Hydroxymethylfurfural (HMF) is a promising platform chemical produced from the dehydration of C 6 sugars, that is a precursor for a range of renewable fuels and polymers. In this study, an integrated macroalgal biore ﬁ nery was designed to produce an array of products including HMF, hydrothermal liquefaction (HTL) biocrude and biochar. In this process two di ﬀ erent species of macroalgae, Ulva lactuca and Chorda ﬁ lum , were investigated and co-processed with spent co ﬀ ee grounds to assess if such blends could be e ﬀ ectively used, with the spent co ﬀ ee grounds mitigating for lower macroalgae availability throughout the year. U. lactuca and the spent co ﬀ ee ground blends were e ﬀ ectively used in a biore ﬁ nery design for the production of HMF. Interestingly, blends yielded higher amounts of HMF (35 – 47 g per kg of dry biomass processed) than the separate components alone. This is presumably due to the elevated amount of C 6 sugars being available from the macroalgae, coupled with the presence of lipids from the co ﬀ ee grounds. The lipids likely form a separate organic layer in the dehydration reaction, into which the HMF migrates after being formed in the aqueous fraction, halting further dehydration reactions to levulinic acid. The HTL on the resultant solids from dehydration yielded a relatively similar amount of biocrude (68 – 78 g per kg of dry biomass) compared to spent co ﬀ ee grounds (SCG) (90 g per kg of dry biomass). However, the C. ﬁ lum biore ﬁ nery yielded far lower biocrude and HMF, presumably due to the lower lipid and C 6 sugar content in this feedstock. Overall, an HMF biore ﬁ nery from macroalgae is plausible, with spent co ﬀ ee grounds being a highly suitable material to make up for seasonal availability. However, the large di ﬀ erence in yields from macroalgal species demonstrates the importance of high lipid content, alongside higher C 6 sugar composition, in the macroalgal feedstock.


Introduction
Macroalgae is a promising feedstock for the next generation of bioreneries as it does not compete with food crops, has a higher rate of carbon dioxide xation than land crops, 1 has no freshwater requirement and is simple to process. 2 In addition, its cultivation can help to alleviate the eutrophication in seas and oceans and aid in carbon capture and sequestration. 3 There are over 30 million tonnes currently cultivated worldwide, 4,5 predominantly in China and India. The UK has one of the most extensive coastlines in Europe (approximately 12 500 km) and the ideal water temperature for seaweed production. These conditions and the underdeveloped market present a large opportunity for further development. 6 There are a wide range of seaweed species growing around the UK, that inevitably, have different properties and components depending on the type, habitat, cultivation method and harvest time. 7,8 However, the predominant feature across almost all species are the elevated carbohydrate levels (65-75%). 9 Several authors have demonstrated that macroalgae can be used in the hydrothermal liquefaction process. 10 Raikova et al. presented a comprehensive study on a broad range of macroalgae species from the UK and the ideal hydrothermal liquefaction conditions to convert these into bio-crude and nutrient partition into the aqueous phase. 11 In a further study the authors also combined plastics present in the ocean with the macroalgae demonstrating that this led to a higher biocrude heating value. 12 However, the use of macroalgae in an HTL-based industry presents two key issues. The rst is the relatively low production of crude from the majority of species when compared to microalgae. 13 This is primarily because of the elevated carbohydrates levels (relative to the more proteinaceous microalgae), that predominantly breakdown into insoluble biochar, a lower quality fuel than the biocrude. [14][15][16] Indeed, the majority of research on macroalgal valorisation has focussed on the conversion of the carbohydrate fraction, mainly through fermentation to ethanol 17 and alternative cellular products. 18 However, the saccharides must rst be extracted and processed into fermentable sugars. There are a high number of competing pathways for carbon in any fermentation and the breadth of sugars that any one organism would need to metabolise make these routes challenging. A simpler, more targeted approach, is the acid catalysed breakdown of the macroalgal saccharides to produce 5-hydroxymethyl furfural (HMF). HMF is a highly promising chemical building block, with a forecasted market of 61 million USD by 2024. 19,20 Its production from carbohydrates and the potential to be converted into high value biofuels make this molecule an important intermediate between the carbohydrate and petroleum-based industry. 21 In addition the chemical processing of biomass tends to have far lower capital costs than the biochemical processing of a similar size. 22 HMF from biomass is usually produced through a sequence of steps: (1) depolymerisation of the feedstock macromolecules into sugars; (2) isomerisation of glucose to fructose; (3) acid-catalysed dehydration of fructose into HMF. 23 HMF has the potential to be converted into dimethylfuran (DMF), a biofuel with high energy density; and 2,5-furandicarboxylic acid (FDCA) a building block used in the production of polyethylene 2,5-furandicarboxylate (PEF), an emerging biobased polymer proposed to replace polyethylene terephthalate (PET). 24,25 The production of HMF from macroalgae has been demonstrated including from isolated agar 26 and the red macroalga Kappaphycus alvarezeii. 27 In the latter, Lee et al. demonstrated the production of glucose, galactose, levulinic acid and HMF from the red seaweed in an acid-catalysed hydrothermal process. The authors reached an HMF concentration of 3.02 g L À1 . Jeon and Park studied the optimal conditions for the production of glucose, galactose and levulinic acid from the red algae Gelidium amansii. 28 In one of their many different studied conditions, the highest HMF concentration achieved was 4.49 g L À1 using an acid-catalysed hydrothermal process at 130 C for 30 min. Similarly, Kholiya et al. described the extraction of agar/agarose from seaweed producing an aqueous extract followed by its conversion into HMF and levulinic acid. 29 In an alternative approach, Gonzales et al. demonstrated that HMF can be sequestrated using granular activated carbon as an adsorbent from hydrolysates from lignocellulose and algal biomass. 30 Another key issue is the seasonal growth of macroalgae, which does not lend itself to an effective all year supply, thereby limiting the size or scope of a potential biorenery. [31][32][33][34] Recent studies have shown that blending with alternative biomass sources can be used to even out supply, and produce products all year round. For example, Jin et al. showed that coliquefaction of microalgae and macroalgae was possible, and even increased the bio-crude energy heating value compared to when the feedstocks were processed separately. 35 Similarly, we recently demonstrated that microalgal seasonality can be effectively addressed by blending this feedstock with spent coffee grounds (SCG) in periods where the microalgae production is lower during the colder parts of the year. 36 In this bio-renery set-up the saccharides were extracted and fermented before the resulting stillage was processed through HTL. Spent coffee grounds are a promising material for bioprocessing, which are available all year round, relatively stable to store, with the worldwide production of SCG being approximately 10 million tonnes in 2019. 37 The composition of SCG varies but, similarly to macroalgae, they contain high carbohydrates (42-55 w/w %) with a similar C 6 composition. 38 We recently demonstrated the suitability of producing HMF from spent coffee grounds in an integrated biorenery design using an organosolv fractionation to isolate cellulose. 23 In this investigation therefore, we aimed to combine these approaches and address two fundamental issues impeding the development of macroalgal bioreneries. To this end, an integrated approach to increase the atom efficiency and produce both HMF and HTL products was demonstrated, co-processing the biomass with SCG to even out seasonality issues and allow steady production all year round.

Materials
Spent coffee grounds were acquired from a local café at the University of Bath. A sample was weighed and placed in the oven at 60 C. Aer two days the sample was weighed, and moisture content determined.
In addition to the 'pure' feedstocks of U. lactuca, C. lum and spent coffee grounds, two blends were prepared for each macroalgae. The blend with 40% macroalga and 60% SCG simulates a season when there is less production of macroalgae (possibly winterdepending on the strain) and the blend with 60% macroalgae and 40% SCG simulate an intermediate season between maximum and minimum macroalgae production (possibly spring and/or autumn). Therefore, the seven feedstocks studied are as follows: Pure Ulva lactuca -UL. 60% UL + 40% SCG -UL 0.6 + SCG 0.4 . 40% UL + 60% SCG -UL 0.4 + SCG 0.6 . Pure spent coffee grounds -SCG.
Acid dehydration 20 g (dry weight) of feedstock were added to a 300 mL Parr reactor (Parr Company Moline, IL, USA, 4560 mini reactors), followed by the addition of a mixture of 2% (w/w) sulphuric acid (from Sigma-Aldrich) in deionised water to make up a total of 100 g. The reactor and its contents were weighed before the reaction. Agitation was initiated and the reactor heated to 155 C, at which point the temperature was held for 15 minutes. The reaction mixture was cooled rapidly over 20 minutes using the in-built cooling system operating at À4 C. The reactor and its contents were weighed to determine gas losses. All the seven feedstocks/scenarios studied were repeated at least three times and the standard deviation calculated.

HMF extraction
The solids from the reactor were separated through ltration.
The HMF was extracted from the ltrate in a 500 mL separatory funnel, in a 100 mL using a mixture of dichloromethane (DCM) and 2-butanol (50 : 50 w/w %). Both phases (aqueous and organic phase containing the HMF fraction) were collected separately for further analysis. The recovery ratio of HMF in the extraction is calculated as follows: where C HMF,org and C HMF,aq are the concentration of HMF in the organic and aqueous fraction, respectively, in g L À1 . V org and V aq are the volume of the organic and aqueous fraction, respectively.

Second acid dehydration
The aqueous phase collected from the HMF extraction was combined with the solids obtained from ltration upstream from the original HMF extraction. This slurry was then added to the reactor, H 2 SO 4 was added and the same reaction procedure was used under the same conditions to produce a further amount of HMF. The aqueous, organic and solids were separated following the procedure given above. The solid fraction was removed, washed in solvent and dried overnight in an oven at 40 C to reduce moisture content to residual quantities before being used in the hydrothermal liquefaction.

Hydrothermal liquefaction (HTL)
HTL reactions were performed in a 50 mL stainless steel batch reactor, equipped with a pressure gauge, pressure relief valve and a needle valve. 3 g of dried solids were weighed, loaded and mixed with 15 g of deionised water into an HTL reactor. The reactor was sealed and placed in a furnace pre-heated to 800 C. Temperature was closely monitored using a thermocouple until it reached 350 C (approximately between 150 and 180 bar). At this temperature the reactor was removed from the furnace and le to cool. Gas phase was measured and collected from the needle valve (using water displacement technique). The contents were separated through a pre-weighed lter paper. The ltrate (aqueous phase) was poured through the funnel into a pre-weighed vial. The vial and the ltrate obtained were weighed for total aqueous fraction weight determination. An aliquot of the aqueous fraction was oven-dried at 60 C to determine the aqueous fraction residue yield. The reactor and the ltered solids were then thoroughly washed (using the same lter paper) with chloroform into a pre-weighed round bottom ask until the ltrate ran clear. The chloroform was removed in vacuo. On solvent removal, the ask was weighed and the biocrude fraction gravimetric yield was obtained. Solids were dried in an oven at 60 C. Filtered solids were weighed for biochar gravimetric yield determination.

Carbohydrates and levulinic acid analysis
Aqueous and organic fractions obtained aer the rst and second extractions were ltered and analysed for carbohydrate content using a high performance liquid chromatography (HPLC), from Agilent Technologies, equipped with an Aminex HPX-87H organic acids column (300 mm Â 7.88 mm, Bio-Rad Laboratories) and a refractive index detector (RID) was used to quantify the carbohydrates in this study. 5 mM H 2 SO 4 solution was used as mobile phase at a ow rate of 0.6 mL min À1 . Column was heated up to 65 C. Mobile phase was prepared using sulphuric acid provided from Sigma-Aldrich.

HMF and furfural analysis
An HPLC (Agilent Technologies) equipped with a diode-array detector (DAD) and an Aminex HPX-87H column (300 mm Â 7.88 mm, Bio-Rad Laboratories) was used in the HMF and furfural analysis on the aqueous and organic samples obtained aer both extractions. The mobile phase (5 mM H 2 SO 4 solution prepared in house using sulphuric acid provided from Sigma-Aldrich) was owing at 0.6 mL min À1 . Column was heated up to 65 C. DAD signal set at 280 nm.

Lipid analysis
Lipid composition was determined using 1 H nuclear magnetic resonance (NMR), detailed information is given in the ESI. †

Results and discussion
The biorenery approach was designed to produce HMF from the C 6 sugar fraction in an acid dehydration. By using a strong acid such as H 2 SO 4 , the cleaving of the biomass composite structure, the depolymerisation of polysaccharides into oligo and monosaccharides and the subsequent dehydration of these monomers into HMF is possible. 39 This process was followed by an extraction of the produced HMF, where the solvent (dichloromethane : 2-butanol (50 : 50 w/w %) was recycled. The extracted stillage from this extraction was then fed into a second acid dehydration to produce further HMF from the unused carbohydrates of the rst reaction. The reactor effluent was subject to a second HMF extraction with the solvent being recycled. The resulting waste stream from the entire process was then submitted to hydrothermal liquefaction to produce biocrude and biochar (Fig. 1). SCG and U. lactuca have a carbohydrate content of approximately 47.8% and 45.6%, respectively. Most of the carbohydrates present in these feedstocks were C 6 sugars (89% and 80% of the total carbohydrates, respectively - Table 1). Such results suggest that these feedstocks can have a high potential for conversion into HMF. 23,40 On the other hand, C. lum has a 24.8% carbohydrate content, from which only 25% of the total carbohydrates were C 6 sugars.

Acid dehydration for HMF production
The initial sulphuric acid treatment depolymerised a portion of the saccharide feedstock, and some HMF production was observed. The SCG produced the most with, 2.69 g L À1 produced, a yield of 2.8% from the saccharide portion. U. lactuca produced 1.76 g L À1 (1.9%) whereas C. lum produced 0.24 g L À1 (0.5%) Fig. 2a. Yields of 6.9% and 7.0% were achieved for UL 0.6 + SCG 0.4 and UL 0.4 + SCG 0.6 , respectively, while 6.7% and 3.9% were obtained for the blends of SCG with C. lum (CF 0.4 + SCG 0.6 and CF 0.6 + SCG 0.4 , respectively) ( Table 2).
Interestingly, the blends of Ulva and SCG produced far more HMF than either of the raw materials when processed separately. This was also observed for blends of SCG and C. lum. This is potentially due to the macroalgal species having more glucose, which is more readily released than in SCG, while the far higher content of lipids in SCG forms an organic layer. 9,38,42,43 It has been previously observed, that a biphasic system can increase stability and yields by partitioning HMF as it is formed away from the aqueous phase. 44 This is supported by a reduction of approximately 50% on the HMF produced from the blend when defatted SCG were used in the same series of experiments (Fig. 2c).
Apart from the C. lum system, all the other systems examined have low C 5 sugar content, and accordingly lower furfural production. The reactor effluent was analysed for carbohydrates (Fig. 2b), and still showed a relatively high content in carbohydrates (approximately 14-23 g L À1 depending on feedstock). The analysis also showed that galactose/mannose is the monosaccharide most abundant in the SCG-rich feedstocks, while higher concentrations of glucose are observed in U. lactuca-rich feedstocks and arabinose and fucose in C. lum. Levulinic acid was also present in these slurries, mostly in the feedstocks containing SCG, suggesting that there are signicant levels of HMF to dehydrate with the systems containing lipids, with far less observed for the two macroalgae species on their own. The presence of such high concentrations of monosaccharides in this stream indicated that these can be further processed in a second acid dehydration to produce more HMF.  . 2 (a) HMF and furfural production in 1st dehydration; (b) monosaccharides concentration after 1st dehydration; (c) comparison of the HMF produced when using the raw feedstocks and when using a defatted feedstock.
Prior to the second dehydration the organic fraction, containing HMF, was removed. This was to prevent further dehydration of the HMF product while also allowing for more HMF to migrate to the organic layer formed by the lipids. The extraction system, a mixture of DCM and 2-butanol in a ratio of 1 : 1, showed relatively high recovery ratios for the HMFapproximately 80 to 90% of the HMF was recovered into the extraction solvent, while the rest remained in the aqueous fraction ( Table 2) this is similar to the results reported by Tan et al. 45 Subsequent acid dehydration to HMF A second acid dehydration was undertaken on the combined solid and aqueous phase once the HMF had been removed. This led to higher production of HMF for the aqueous phase produced from the pure feedstocks of U. lactuca (6.1 g L À1 ) and SCG (5.1 g L À1 ), though the HMF from the C. lum (0.2 g L À1 ) and blends of this macroalgae with SCG were greatly reduced (Fig. 3a). This is presumably because there is a large pool of glucose and mannose that can be dehydrated in the SCG and U. lactuca and far less in the C. lum. In addition, the lack of lipids in C. lum does not allow for a large production of HMF, similarly to what happened in the rst dehydration. However, the HMF produced in the blends of C. lum is considerably lower than what was achieved in the rst dehydration. This might be due to the low amounts of C 6 sugars le in solution aer the rst dehydration (b). When compared to the sugars present in solution aer the second dehydration (b), sugars such as arabinose and fucose were barely consumed. The sugar analysis demonstrates that most of the glucose and mannose are consumed in the dehydration into HMF, while fucose and arabinose are somewhat more stable.

Lipid analysis
The streams of HMF (both aer rst and second dehydrations) and the stream of stillage aer the second dehydration were analysed for lipid content. Lipids were only found in the stream of stillage aer the second dehydration.
This demonstrated that no lipids were extracted with the HMF, which conrms the lipids form an organic phase in both acid dehydrations, corroborating the results obtained by Wang et al. and Prates Pereira et al. 23,40 This double layer presumably allows for the HMF to move into the organic phase, once produced in the aqueous phase. The presence of lipids in the stillage aer the second dehydration would also potentially allow for a lipid extraction prior to the HTL reaction. The extraction of the lipids as a product in a separate stream could add further value to this biorenery approach. The analysis on the lipids present in the stillage aer the second dehydration Table 3 Type and content of fatty acids found in the lipids in the extracted stillage after the second acid dehydration. MUFA stands for mono-unsaturated fatty acids. Analysis and quantification of fatty acids in U. lactuca and C. filum was difficult due to the low content of fatty acids in these samples. However, in addition to the fatty acids shown below, C. filum revealed the presence of an unusual component, most probably stearidonic acid   show a high content of saturated fatty acids as well as linoleic and linolenic acids (Table 3).

Hydrothermal liquefaction (HTL)
On the extraction of HMF, the resulting solid was converted into further products through hydrothermal liquefaction (Fig. 4). The U. lactuca, SCG and respective blends yielded relatively similar bio-crude and biochar yields, with a slight increase in biocrude with higher percentage of SCG in the blends. This is presumably due to the higher content of lipids in SCG, which is converted to biocrude in this process, as described by Madsen et al. 46 . These results demonstrate that blends of U. lactuca with SCG can be used in an effective HTL process without substantial reduction in main product yields (biocrude and biochar). The same trend was observed with SCG and C. lum, where a higher percentage of SCG in the blend associated with a higher biocrude yield. This is presumably due to the higher lipid content in SCG (converted into biocrude in HTL). On the other hand, the biochar yield increases with higher percentages of C. lum. This is presumably due to the breakdown of the carbohydrates not used in any of the dehydration processes and were broken down into biochar in the HTL. In contrast to U. lactuca, C. lum blends with SCG led to considerable differences in the biocrude and biochar yields. Therefore, if SCG were to be blended with C. lum in a biorenery, the processes downstream to HTL would need to be prepared to handle different volumes of biocrude and biochar, depending on the season, thus on the percentages of the blends used.

Overall yields
The overall product yields were compared across the different blends and feedstocks ( Table 4). The total HMF presented is the sum of the HMF extracted aer the rst and second dehydrations. Yields were extrapolated to 1 kg of dry biomass fed into the process. The biocrude and biochar amounts were calculated based on the amount of solids obtained aer 2nd dehydration and on the HTL yields presented in Fig. 4. These were also extrapolated to 1 kg of dry biomass supplied to the entire system.
A higher amount of HMF was produced from UL 0.6 + SCG 0.4 (46.6 g kg biomass

À1
) and UL 0.4 + SCG 0.6 (35.3 g kg biomass À1 ). These are the blends intended to replace U. lactuca in periods of intermediate and lower supply of this seaweed. However, when processed separately, U. lactuca and SCG only produce 31.2 and 30.2 g kg biomass À1 , respectively, demonstrating that the blending of U. lactuca and SCG creates improved conditions for HMF production. A higher level of bio-crude was produced for the SCG (89.5 g kg biomass

À1
) presumably due to the higher content of lipids in this feedstock, though a similar conversion was observed for all biomass blends. However, the higher the percentages of U. lactuca in the blends, the higher the production of biochar. This is mainly due to the higher content of ash in the macroalgae. Overall, both UL 0.6 + SCG 0.4 and UL 0.4 + SCG 0.6 demonstrate that SCG is a good replacement of U. lactuca in periods of lower seaweed supply.
Unlike the U. lactuca blends, when taking both extractions into account, the highest level of HMF was produced from the SCG and the lowest from the C. lum. The blends of SCG and C. lum were proportional to this. Such low production of HMF from this macroalgae is due to the low content of C 6 sugars and lipids in this feedstock. This result has a high impact on the HTL results as the carbohydrates that were not dehydrated into HMF were then broken down into char, resulting in a high production of char of 757.2 g kg biomass

À1
. In addition to this, the low content of lipids in this feedstock led to a lower biocrude production (14.6 g kg biomass

À1
) when compared to SCG (89.5 g kg biomass

Conclusions and future perspectives
In this study a biorenery producing 5-(hydroxymethyl)furfural (HMF), biocrude and biochar from macroalgae and spent coffee Fig. 4 Mass balance of the extracted stillage used in HTL reaction. Gravimetric yields calculated based on the dry weight of the extracted stillage fed into the reaction. Gas phase calculated assuming a 100% content in CO 2 , while the aqueous gravimetric yield was calculated considering the solids in this phase. Table 4 Overall yield of the fuel products and precursors including total produced HMF, lipid production, biocrude and biochar from the proposed biorefinery, all values are given as kg per tonne and calculated on a dry ash free basis grounds blends was developed. U. lactuca was found to be a suitable species, containing elevated C 6 sugars that could be converted into HMF. Interestingly, the addition of spent coffee grounds increased the production substantially, in comparison to the macroalgae or spent coffee grounds alone. This was presumably due to the formation of a lipid layer in the aqueous phase that reduced the decomposition of HMF to levulinic acid. The stillage from the reaction was also further converted into fuel products through HTL, yielding 46.6 g kg À1 HMF, 78.2 g kg À1 biocrude, 390 g kg À1 biochar in the optimised system. The same system with C. lum was less productive, presumably due to lower lipid and C 6 sugars in the macroalgae. The aim of this work was achieved as it demonstrates that an integrated HMF biorenery is possible using U. lactuca, and that the addition of spent coffee grounds not only would allow all year round production, but could also increase the yield of specic target products, such as HMF and HTL products. Indeed, blending with spent coffee grounds has potential to improve process versatility making biomasses considered unsuitable alone (such as Chorda lum) into viable feedstocks. Furthermore, the current work demonstrates that the production of HMF can be performed in a single-step reaction rather than in three-steps as widely presented in the literature. Further perspectives to be considered in future work include the optimisation of the HMF production in the dehydration process. A techno-economic analysis is also suggested to determine the protability of the second acid dehydration process, possibly a third acid dehydration to use the unused sugars, and the inclusion of a lipid extraction process upstream to the hydrothermal liquefaction.

Conflicts of interest
There are no conicts to declare.