Current Situation of the Challenging Scale‐Up Development of Hydroxymethylfurfural Production

Abstract Hydroxymethylfurfural (HMF) is a high‐value platform chemical derived from renewable resources. In recent years, considerable efforts have been made to produce HMF also at industrial scale, which still faces some challenges regarding yield as well as sustainable and economic process designs. This critical Review evaluates the industrial process development of sustainable biomass conversion to HMF. Qualitative and quantitative guidelines are defined for the technological assessment of the processes described in patent literature. The formation of side products, difficulties in the separation and purification of HMF as well as catalyst regeneration were identified as major challenges in the HMF production. A first small‐scale, commercial HMF production plant with a capacity of 300 tHMF per year has been operating in Switzerland since 2014.


Introduction
The limited amount of fossil resources and rising environmental concerns relatedt oC O 2 emissions have drawn public and scientifica ttentiont om ore sustainable ways of chemical production. For as ustainable development, the use of hazardous materials and fossil resources shouldb em inimized or avoided, whereas the use of renewable resources should be enhanced. [1] Hydroxymethylfurfural( HMF) is ap romising molecule derived from renewable resources. It is ak ey intermediate between biomass and biochemicalsa nd has the potentialt oreplacea range of conventionally produced buildingb locks. It has been referred to as "sleeping giant" [2] due to the anticipated enormous market potential of HMF and its derivatives. The US Department of Energy listed 2,5-furandicarboxylica cid (FDCA), a HMF derivative, aso ne of twelve top value-added chemicals in 2004. It is ap romising starting block for polyethylene 2,5-furandicarboxylate (PEF) synthesis, which is ab io-derived alternative to polyethylene terephthalate (PET). [3] The production of bioplastics from bio-based chemicals has come into the focus of several industries. PEF is ap romisingb ioplastic withe xcellent gas-barrier performance, recyclability,a nd extended mechanical properties. [4] As hort overview of chemical compounds derived from HMF and their potentialapplication fields is given in Figure 1.
HMF is ak ey intermediate for valuable chemicals out of C 6carbohydrate (hexose)b uilding blocks, for example, levulinic acid, 1,6-hexanediol, or adipic acid. Its derivatives, for example, 2,5-bis(hydroxymethyl)furan [5] or 2,5-diformylfuran, [6] are promising cross-linkers in the resin production. These biobased resins are capable replacementso fc urrently used, fossil-based adhesives in variousindustries, such as foundryo rw ood industries. In addition, the potentialo fH MF derivatives as solvents [7] and fuels [8] has been reported. 2,5-Dimethylfuran is an alternative biofuel, with av olumetric energy density comparable to gasoline. It is insoluble in water,s table in storagea si tw ill not be contaminated through water absorption from atmosphere, and has ah igh research octanen umber. [9] The commercial production of chemicals concerns economic, environmental, andi ndustrial aspects. For as ustainable development the chemical product should be derived from biobased chemicals such as HMF;i na ddition, the negative environmental impact of the processing and manufacture must be minimized. As ustainable chemical product must satisfy both sides, the producer andt he consumer.I th as to be ac ommercial successf or the producer and still be affordable fort he end-user. [10] Various reviewsw ere published on the laboratory-scale synthesis of HMF.T he reviewsm ainly focused on solvents [4,11] and catalytic systems. [12] Reviewso nt he used feedstocks, [13] biological properties, and its synthesis and applications [14] can be found in the literature as well. Ar eview on the chemistry of HMF,p rocess technologies, and its application as platform chemicalw as published by van Putten et al. [15] in 2013. Since then, the implementation of industrial-scale HMF production processes has gained much more attention, and an increasing number of HMF production methods have been patented in the last couple of years. Ar eview with am ore holistic point of view,w hich connects the work from academia and industry,i s still needed. [16] The aim of this Review is to bridge this gap and identify the biggest challenges researchers face in the development of HMF production methods and to critically assess the developed process technologiesa nd the applicability for industrial systems. It also provides an overview of the main improvementsi np rocess development, especially with regard to green processing.
Hydroxymethylfurfural( HMF) is ah igh-value platform chemical derived from renewable resources. In recent years, considerable efforts have been made to produce HMF also at industrial scale, which still faces some challenges regarding yield as well as sustainable ande conomic process designs. This critical Review evaluates the industrial process development of sustainable biomass conversion to HMF.Q ualitative and quantita-tive guidelines are defined for the technological assessment of the processes described in patent literature. The formation of side products, difficulties in the separation and purification of HMF as well as catalyst regeneration were identified as major challenges in the HMF production. Af irst small-scale, commercial HMF production plant with ac apacity of 300 t HMF per year has been operating in Switzerland since 2014.

Principles of monosaccharide dehydrationtoH MF
HMF combines the functionalities of furfural and furfuryl alcohol. Characteristics of HMF are its hydroxyl and aldehyde group as well as the furan ring, as can be seen in Figure 2.
The thermal, acid-catalyzed dehydration of hexoses, for example,g lucose or fructose, resultsi nt he formationo fH MF. Severalk inetic studies on theH MF formation from various biomass feedstock were summarized by van Putten et al. [15] Kinetic studies [17] can not only be used to get insight into the mechanisms of the HMFf ormation on molecular level but serve also as input for the development of optimum reactorc onfigurations and process conditions. Several mechanismsh ave been proposed for the formation of HMFf rom hexoses. [18] The direct formation of HMF by acidcatalyzed dehydration is generally described as the removalo f three water molecules from the sugar molecule. Depending on the structure of the formed intermediates, the proposed mechanismsc an be divided into cyclic and acyclic routes. [18] There have also been 13 Ci sotopic labellings tudies for fructose dehydration, [19] but no definite proof for either of the mechanistic routesh as yet been published for the HMF case. However,f or the structurally closely relateds ystem of hexeneuronic acid( 4deoxy-b-l-threo-hex-4-enopyranosiduronic acid), leading to 5formyl-2-furoic acid by triple dehydration,t he occurrence of both acyclica nd cyclic intermediates has been demonstrated by acombination of 13 Cisotopiclabelling and NMR spectroscopy. [20] In addition, solvente ffects make the comparison of kinetic parameters for dehydration reactions in biphasic waterorganic solventm ixtures and monophasic systems difficult. [17c] 1.2. Challenges of HMF synthesis:Side reactionsa nd isolation HMF can be derived from hexoses, preferably from hexoketose d-fructose. The formation of HMF from fructose often entails some side reactions, such as isomerization, fragmentation, and condensate formation. [18] The HMF yield obtained from d-fructose is higher than from d-glucose under the same reaction conditions. Since d-glucose is cheaper,s everal studies also focused on the HMF synthesis from glucose. [15] Isomerization of Catherine Thoma received her master's degree in Material Science from Te chnical University Vienna in 2018. She now works as aj unior researcher at Kompetenzzentrum Holz GmbH in the area Wood Material Technologies. She is pursuing aP hD at the University of Natural Resources and Life Sciences. Her research interests involve carbohydrate conversion and sustainable production of carbohydrate-based resins for wood-based panels. ChemSusChem 2020, 13,3544 -3564 glucoset of ructoses eems to be ar equired step in the synthesis of HMF from glucose, making ane fficient isomerization catalyst necessary.S ince the glucose-fructose isomerization is best base-catalyzed and the following dehydration of fructose is acid-catalyzed,t his has spurred some research on the catalytic systems, especially on bifunctionalcatalysts. [12a] HMF reacts in aqueous mixtures with two water molecules in ar ehydration reaction, forming levulinic acid and formic acid ( Figure 2), sometimes referred to an Achmatowicz-type process. [21] This degradation decreases the overall HMF yield and makes expensive purification processesn ecessary.T he rehydration of HMF is suppressed in non-aqueous systems. The dehydration reactioni sa ccompanied by condensation reactions, which form ab lack tarry by-product consisting of complex furanico ligomers called humins. [22] They have recently been shown to consist of quinoid-furanoid ladder-type oligomers rather than of linear polymers as previously assumed. [23] Their extremelyh igh extinction coefficients accountf or their black appearance. From an economical and technological perspective, the formation of huminsi sh ighly undesired. In general, it lowers the efficiency of the dehydration process, renders purification and decolorationd ifficult, and decreases catalyst efficiency.R ecent publications focused on finding new ways for humins valorization to turn those drawbacks into an advantage. [24] The presence of condensation products causes major problems, especially for HMF separation and purification. The recovery of HMF is associated with difficulties due to its thermal lability under long-term heatingi nb oth alkaline and acidic conditions. Thus, separation of HMFf rom the reactionm ixture,f or example, by distillation, is challenging. In ar ecent publication, Gomes et al. [25] described the enhanced thermals tability of HMF during synthesisi nb iphasic systems and distillation in the presence of sodium dithionite. Without the addition of sodiumd ithionite, the formation of degradation products, mainly tarry carbonaceous materialw as formed. Furthermore, HMF is difficult to store due to its relative instability and sensitivity towards acids, alkali, and oxygen even under mild conditions. [26] Galkin et al. [27] showedt hat during two weeks of storage of aH MF oil with 97-99 %p urity decomposition took place, leadingt ot he formation of dimers and larger oligomers.
The separation and purification of HMF is one of the most important challenges in the scale-up of HMF production.

Process Assessment Criteria
Based on the number of patent applications related to HMF production there has been ac ontinuous growth of interest in this topic as can be seen in Figure 3. Severala djustments to existing HMF production methods have been made to improve the chemical and economic efficiency of potential HMF productionprocesses. The related research can roughlyb ed ivided into the main fields given in Ta ble 1.
Methodsh ave been developeda nd adjusted startingf rom basic operational variations, for example, different operation modes and reactor designs for ab etter control of temperature and reactiont ime, to the testing of single-phase and biphasic reactionm ixtures and different catalytic systems to reduce the side reactions. In terms of feedstock selection,t he use of isomerization enzymesh as been proposed to increase the HMF yield from saccharides. Several production methods are based on setting ap artial conversion endpoint to limit the formation of follow-up products and increase the efficiency of HMF production. The production of HMF from agricultural side prod-ucts was developed to make the process more economic. The production methods described in patent literature each have its benefits and drawbacks, the most criticalf actor being the HMF yield.
The process assessment of the upscaledH MF production methods is limited due to the lack of data provided in the patents. The product yield and the reactionm ass efficiency (RME) are the parameters that describe the efficiency of ap rocess and are thusn eededf or aq uantitative process assessment. Yet neither product yield nor RME consider byproducts, wastes, solvents, catalysts, or energy issues. Aq ualitative analysiso f the used catalyst, solvents ystems, and feedstock is, however, still possible and can point out options and directions for future process developments.
One strategy for the development of greener processes is the appropriate selection of the solvent. [10] With regard to the "greenness" of aprocess, solvents are usually an environmental concern due to the typically large quantities used. TheI nnovative MedicinesI nitiative (IMI)-Chem21 [28] published ac omparative surveyo fd ifferent solvents resulting in selection guides.
The Chem21 classification of the solvents usedi nt he described processesi si ncluded in Ta ble 2. As can be seen, the most sustainable solvents for HMF production are alcohols, water,and methyli sobutyl ketone (MIBK).
Several reviews [11b, 12a, 14, 15] were published on the catalytic systemsu sed in HMF synthesis. For as ustainable process, high selectivity of the catalystt owards HMF generationi sp referred. Up to 93 %H MF yields were obtained using ionic liquids (ILs) and acidic ion-exchanger esins.
Menegazzo et al. [13] recently summarized the publications on the direct synthesis of HMF from raw biomass,i ncluding edible biomass,n on-ediblel ignocellulosic biomass,a nd food wastes. Hexoses,f or example, fructoseo rg lucose, have been used preferably as feedstock for HMFs ynthesis. The HMF yield is higher when fructose is useda sf eedstock, but in general glucose is more easily available and cheaper.I ng eneral, lignocellulosic biomass consisting of cellulose, hemicellulose, and lignin are ap romising feedstock for the conversion to HMF since cellulose and hemicellulose can be degraded into hexoses and pentoses. Even thoughm onosaccharides are the easiest starting materials to be convertedi nto HMF,t he additional step of obtaining monosaccharides out of polysaccharides is a drawback. Still, lignocellulosic biomass as HMF feedstock is favored from an economic ands ustainable point of view. [11b, 15] 3. Early Work on HMF ProductionP rocess Development (Until 2006) In 1895, Düll and Lintner [29] were the first to synthesize HMF from inulin using 0.5 %o xalic acid as ac atalyst. In the same year,K iermayer [30] dehydrated fructoseu nder pressure using 0.3 %o xalic acid solution. Given the analytical methodso ft hat time, it is admirablet hat Kiermayer identified the structure of HMF almost correctly.H aworth and Wiggins [31] modified Kiermayers processa nd found that an enhanced HMF yield was obtainedw hen saccharose is dehydrated at higher temperatures of 162-167 8Cw ithoutt he use of an additional catalyst. The acidic substances formed in the conversion of saccharose were found to sufficiently catalyze HMF formation.
The Food Chemical and Research Laboratories Inc. [32] reported the formation of HMFi napatent in 1956. The reactions were performed under pressure in the presence of an acidic catalyst, for example, HCl, HBr,H 3 PO 4 ,H 2 SO 4 ,Z nCl 2 ,o rA lCl 3 .A n aliphatic mono-ol, for example, butanol,w as used as the reaction medium, along with saccharosea nd fructose as the feedstocks. The overall yield was rather moderate. At 150 8C, the conversion of saccharose gave the highest yield of 68.6 %i na butanol/water mixture after 20 min. The reactionw as performed in an autoclave. In additional experiments in glass tubes,t he conversion of fructose was 68.0 %a t1 70 8Ca fter 8min. Based on the experimental data, ap seudo-first orderr eaction kinetics for the HMF formation was proposedi nt he case of low sugar concentrations.
Dentrol Inc. [33] filed ap atenti n1 958 on ap rocess for HMF production from cellulosic raw material, such as small pieces of hardwood, for example, oak wood chips. This feedstock was dispersed in 0.6 %H 2 SO 4 and then charged in ar eactionv essel at high temperature (285.6 8C) and pressure( 6.9 MPa). The dehydration reactioni sp erformed using high-pressure steam. The liquid condensation product contains about 8% HMF, which is about 20 %o ft he theoretic yield based on the cellulose charge (based on 40 %c ellulose in wood). However,t he low HMF yield appeared unattractive when comparing the results to those of other studies at that time.
In 1960 Merck Co. Inc. [34] patented ac ontinuous process for carbohydrate conversion to HMF in aqueous solution at temperatures between 250-380 8C. AH MF yield of 37 %w as obtained from the conversion of saccharose at 270 8Ca nd ar eaction time of 45 s. The formation of ab lack, soluble tar was also reported. In 1969, an improved method was patented, [35] in which aluminums alts were used as the catalysts. Theh ighest obtained yield was 58 %H MF from the conversion of fructose. In this experiment,a mmonium aluminums ulfate (NH 4 Al(SO 4 ) 2 ·12 H 2 O), wasu sed as catalyst with ar eaction time of 9min at 270 8C. The efficiency of the catalyst for the conversion of sorbose and galactose was significantly lower (27.4-37.7 %H MF). It is interesting to see that many of the early contributionst oH MF process developments also tested different feedstocks, for example, wood and lignocellulosics, saccharose, galactose or sorbose, whereas later work often focusedo nt he conversion of fructose andg lucose.
Atlas Chemical Industries [36] filed ap atento nt he acid-catalyzed dehydration of hexoses to HMF in 1963, such as sorbose or glucose, or hexose disaccharides,such as saccharose. The reaction mediumc onsisted of water anda no rganics olvent, for example, MIBK or dioxane. Mineral acids, such as HCl or H 2 SO 4 , were used as catalysts together with salts, for example, AlCl 3 or CrCl 3 .T he separation and recovery of HMF from the reaction mixture was not coveredinthe patent. The reactions were performed at 150, 180, and 210 8C. The HMF yield was related to the hexose charged and hexose consumed in the process. Overall,t he obtained yields were rather moderate( 40-67 %). The highest HMF yield was 66.9 %( hexosec harged) and 80.0 % (hexose consumed). In this example, sorbose was reacted in triethylene glycol using 0.13 %H Cl as the catalystf or 3min at 180 8C.
Ap atent of Roquette Freres [37] disclosed ap rocess for the decomposition of hexoses in ab iphasic reaction mixture at temperatures between 85-90 8C. An ion-exchange resin with a Table 1. Researchtopics and defined process assessment criteria.

Process assessment criteria
Research topics operational aspectso perating mode reactor design solvent system single-phase systems biphasic systems catalytic systems alts acid catione xchange resin metal halides mineral acids feedstock selection + conversion isomerase enzymes partial conversion endpoint production of HMF from by-products cationic functionalization was used as the solid catalyst. The highest obtained yield was 89 %H MF,p roduced in MIBK as organic phase at 85 8C, catalyzed by the cation-exchange resin Lewatit SPC 1008. Contrary to earlier studies,l arge volumes (40 Ls olvent + 1kgf ructose) were used in the reaction, but the conversion rate was rather low (21 %). It is important to point out that most early studies-as well as most current work-performed the dehydration to HMF in very small quantities, with limited conclusivenesstolarger-scale processes.T his makes the research of Roquette Freres ag ood starting point for furtheri nvestigations. Increasing the catalytic efficiency and improving the yield had proven to be an important area for future work. Roquette Freres [38] also patented ac ounter-current process technology for the synthesiso fH MF.T he sugar-containing starting materiali sd issolved in ap olar aprotics olvent, for example, dimethyl sulfoxide( DMSO), in the presence of as olid catalysta tt emperatures between 75-80 8C. The formed HMF is then extracted to another solvent, for example, MIBK, in ac ontinuous counter-current setup. In one example, the ion-exchange resin Lewatit SPC 108 was used as the catalyst with DMSO as the reactionm edium at 80 8C, giving an HMF yield of 97.5 %a sd etermined by gas chromatography.T he high yield of HMF was ac lear advantage of systemsu sing DMSO, although its role in the conversion is still not completely understood and is the subject of considerable debate. [39] Ts ilomelekis et al. [40] analyzed the molecular structure, morphology,a nd generation of humins in a system that used DMSO as co-solvent. Their analytical data supported the postulated mechanism of humin growth by van Zandvoort et al., [41] in which humins are formed through electrophilica ttack of HMF carbonyl moieties at the a-o rb-position of furan rings. In their experiments, they showed that this pathway is significantly suppressed in polar aproticc o-solvents, such as DMSO.I naprevious study using frontier molecular orbitalt heory by Tsilomelekis et al. [42] they found that DMSO minimizes the susceptibility to nucleophilic attack and thus rehydration and humin formation due to the reduction of the LUMO energy.T hey also stated that the hydrogen bond acceptors trength of DMSO is higher than that of the HMF carbonyl group.I th as also been shownb yR en et al. [43] that the isomer distribution of fructose in DMSO-containing mediai s different from that in water.I nD MSO, b-d-fructofuranose is the most stable form of fructose, whereas in water the b-pyranose is dominant. In addition, Ren et al. [43] postulated that in the presence of a Brønsted acids the catalytically active sulfonium species [DMSOH] + is formed, which interacts with the fructofuranose isomer (see Figure 4). In 1988, Südzucker AG [44] patented ab atch process for producing HMF in aqueous media. Oxalic acid was used as catalyst. Fructose or inulin from chicoryr oots were used as feedstock, the obtained product solution was purified by column chromatography.A nH MF yield of 33 %c ould be reached from the dehydrationo ff ructosea fter 130 min at 135-142 8C. When using inulin and H 2 SO 4 ,anH MF yield of 13 %a nd 30 %f ructose was obtained at 140 8Ca fter 120 min. Humins were formed as side product and filtered off before the purification process. HMF (99 %p urity) was obtained after crystallization. The yield is significantly lower than those of processes in previously described literature. The formation of humins was reported. [44]

More Recent Process Development for HMF Production (2006 to Present)
Since 2006, the number of publications and patents on HMF production methods has increased steadily.S everalc ompanies developed processes for carbohydrate conversion to HMF,a nd many of them patented multiple process methods and also improvedt heir process concepts. Details on operational aspects, solvents election, catalytic systems, and feedstock selection are discussed in this sectiona sf ar as available in literature and separately for each process. For ab etter comparison of the specific details on the conversion to HMF,t he information is summarized in Ta ble 3. The table is structured in accordance with the text based on the company that filed ap atent for the respective process concept.

BASF SE
BASF SE (Ludwigshafen am Rhein, Germany) workedo nt he fundamental understanding of the influence of the operating mode on the HMF yield by comparing reactor systems with differento peration modes (semi-batch, continuouslys tirred tank reactor,a nd pipe reactor). In 2013 and 2014, BASF SE patented [45] at wo-step HMF production methodi nacontinuously stirred tank reactor.F irst, dehydration of fructose occurs at comparativelyl ow temperatures (100-160 8C) using the IL 1-  4 ]a nd CrCl 3 as the catalyst. An HMF yield of 63 %w as obtained after the second reactor.T he use of metal chlorides as effective Lewis-acid catalysts for the synthesiso f HMF from fructose or glucose has already been described in literature. [12a] In ar ecent publication, Zhou et al. [ For these studies, aB rønsted acid with an anion corresponding to the anion of the IL was used as the catalyst. BASF SE reported the highest HMF yield of 86.5 %a t9 7.6 %c onversion for the reaction in as emi-batch process. Using ac ontinuously stirred tank reactor,t he HMF yield was reduced to 75.8 %a t 93 %c onversion.T he yield of this experiment lies in the range of previously patented dehydrationsi nb atch reactors by BASF SE. In the pipe reactor,a no verall yield of 71.4 %H MF at 97 % conversion was achieved. Interestingly, the HMF yield in the continuous process was significantly lower than in the semibatch processes. The separationo fHMF from the product solution is done by short-path evaporation. However,ac loser look to the used fructosec oncentration points out some problems regarding the efficiency of the reaction. Even though ah igh HMF yield (86.5 %) was obtainedi nt he semi-batch process, only a2 0wt% fructose solution was used as the feedstock. When using am ore highly concentrated fructose solution of 65 %, the HMF yield dropped to only 50.9 %. Ah igher concentration of the feedstock would increaset he efficiency and sustainability of the process since less solvent and less energy are  Besides varying the operating mode, the utilization of different reactor designsw as reported.Areactor design that has gained some attention is the continuous extraction of water from the reaction medium in a wiper-blade evaporator.I n2 012, BASF SE [47] reported a continuous method to dehydrate ac arbohydrate-containing feedstock in the presence of an IL and an organic co-solvent. The reactions olution was then evaporated and the HMF-containing, gaseous discharge of 421.1 gh À1 was condensed and separated from the organic solvent. In the given examples, a continuous feed of [BMIm]Cl( 300 gh À1 )a nd am ixture of fructose/methanol/water (1:1:1) at 22.3 gh À1 are merged and evaporated at 200 8Ct og ive an HMF yield of 8%.U sing 200 gh À1 1-hexyl-3-methylimidazolium chloride ([HMIm]Cl) feed and 44 gh À1 fructose/ methanol/water at 170 8C, the yield increased to 10.1 %. Ac riticalo pen question is what factor exactly influences the increase in yield, whether it is the change of the IL or the lower temperature and shorter reactiont ime in the evaporator,o racombination of those. The short reaction time in the evaporator,t he implementation as continuousm ethod and the abundance of ac atalysta re stated as advantages of the method. However,t hese advantages cannot compensate the low yield of HMF, which makes optimization of the process necessary.I ngeneral, the large amounts of organic solvent mixed with ILs pose disposal problems that call for elaborated recyclingo ft he reaction media, which in turn leads to more complex processes and higherproduction costs.

ArcherDanielsM idlandCompany
In ap atent assigned to Archer Daniels Midland Company (ADM) (Chicago, USA), [48] severale xamples (see Ta ble 3) are listed to illustrate the effect of temperature, solvent, and distillation on the HMFy ield. The highest yield of 80.6 %a taconversion of 94.1 %w as obtained in ab atch reactor using N-methylpyrrolidinone (NMP)a st he solventa nd the commercial ion-exchange resin Amberlyst 35 WET as the catalyst at 115 8Cf or 300 min. At emperature reduction of 10 8C led to ar eductioni nH MF yield to 71.6 %. When changing the solvent to N,N-dimethylacetamide (DMAc) at 105 8Cu nder the same reaction conditions as above,the HMF yield dropped to only 62.1 %. When performing an in situ distillation at 105 8Cf or 120 min using NMP and Amberlyst 35, the HMF yield was 75.7 %. Purification processes for each examplea re given in the patent as well. Although there are many studies using ion-exchange resins for the dehydration to HMF,t here is al imitation to the applicable reaction temperature.T ypically,t emperatures below 150 8Ca re tolerable for these catalysts. The review of Qiao et al. [12b] showedt hat high HMF yields could be obtained with ion-exchange resins andb iphasic systems, organic solvents, and ILs. Another apparent drawback of this method is the use of hazardoussolvents, such as DMAc and NMP.
ADM [49] also developed aH MF production method employing microwavei rradiation as heating source.T he highest HMF yield was 77.7 %a tafructose conversiono f8 0.1 %. Further, 12.4 %o fs ide-products were formed. The synthesis was performed in NMP with 1.8 %H 2 SO 4 at 160 8Cf or 20 min.
In another ADM [50] process, the fructose-containingf eedstock, water,ahomogeneous acid catalyst, and as olvent are added in ar eactor and converted to ad efined partial conversion endpoint that did not exceed 80 mol %o ft heoretical HMF yield. The reaction mixture is then quenched and neutralized. Separation and purification of HMF is done by liquid-liquid extraction,p hase separation, and filtration of humins. The large amountso fo rganic solvents typically needed in liquid-liquid extractionp rocesses makes them less sustainable. The maximum HMF yield reached in this process was 49 %, using HCl and 2-butanol as the catalyst/solvent couple. The reactionw as performed at 120 8Cf or 30 min. NaCl was also added to increase the partitionc oefficiento fH MF in the biphasic water/ organic solvent system.T he salting-out effect, induced by NaCl, resulted in an increased immiscibility of the aqueous and organic phase, which improved the extraction of HMF from the aqueous phase and consequently reduced unwanted side reactions in water.I nt he patent, ethyleneglycoldimethylether (glyme), 1,4-dioxan, bis(2-methoxyethyl)ether,a nd THF were also tested as solvents, but the HMF yield was even lower.T his is in accordance with previous work by Romµn-Leshkova nd Dumesic, [51] who extensively studied the impact of different salts on the HMF yield as well as the impact of the solvent in biphasic systems saturated with NaCli nl aboratory experiments.T hey concluded that within the studied solventc lasses (primary and secondary alcohols, ketones, and cyclic ethers in the C 3 -C 6 range), C 4 solvents gave the highest HMF yield. In addition, NaCl and KCl lead to the highest extraction power and HMF selectivity.
In ar ecent publication, [52] it has been shown that hexafluoroisopropanol, al ow-boiling extraction solvent, has ap artition coefficient superior to solvents such as MIBK or n-butanol. The easy isolation of HMF,t he good selectivity,a nd easy recyclability are clear advantages of this solvent. This highlightsa gain the importance of proper solvent and catalyst selection.
ADM [53] designed ap rocess for HMF production, in which unreacted sugarsa re directly fermentedt oe thanol. High fructose corn syrup with 42 %f ructosec ontent (HFCS-42), was used in the acid-catalyzed dehydration.T he HMF yield was set to about 20 %, which is too low to be of interestf or an economic process.

Mikromidas Inc.
Ar emarkable process technology with regardt ot he reactor type was disclosed by Micromidas Inc. (West Sacramento, USA). [54] They described ap rocess utilizing am ultiphase reactor,f or example, af luidizedb ed reactor,f or the conversion of cellulosic feedstock to chloromethylfurfural( CMF). The production of HMF was also mentioned. Biomass and dried gaseous acid, for example, HCl gas, is continuously fed into am ultiphase reactor.T he reactioni sp erformed at temperatures between 200 8Ca nd 250 8C. As can be seen in Figure 5, the separation of the gaseous acid and the reactionm ixture is done by using as olid-gas separator,f or example, ac yclone, af ilter or a gravimetric system. Suitable solvents for the purification are, among others, dichloromethane or hexane. With regard to sustainability and commercial-scale application, the use of ac ellulosic feedstock is preferable. The advantages of performing the reactioni nafluidized bed reactor are the rapid mixingo ft he suspended solid aroundt he bed, the uniform heat transfer and the elimination of hot spots within the reactionm ixture. The mixing also reduces the need for pretreatment of the biomass. CMF and furfural are reported as main products. For the production of CMF,atotal yield of 35 %w as given. Unfortunately,n oH MF yield was disclosed in the patent. The uniform mixing and temperature gradienta re clear advantages of fluidized bed reactors. Kinetic studies [17a,c] indicate that the reaction time and temperature are criticalp arameters for the dehydra-

Wisconsin Alumni Research Foundation
Wisconsin Alumni Research Foundation( WARF) (Madison, USA) [55] studied the production of HMF with and withoutc atalysts from various carbohydrate feedstocks in DMAc/LiCl in 2009. The conversion of fructosew ithout additional catalyst gave moderate yields of 55-65 %a tt emperatures of 80-140 8C. The yield was significantly improved by the use of Brønsted acids, for example, H 2 SO 4 ,a nd Lewis acids, for example, CuCl 2 . The addition of ILs was also beneficial. The highest HMF yield (94 %) startingf rom fructosew as obtained in as ystem using H 2 SO 4 as catalysta nd DMAc/LiCI as the solventa t1 20 8Ca nd 12 min reactiont ime. Ap roblem that has been overlookedi s the limited stability of DMAc at elevated temperatures, especially in the presence of acidic catalysts, which causes hydrolysis and formation of dehydracetic acid-typec ondensation products. [56] In addition, DMAci sc lassified as" hazardous"a ccordingt o the Chem21 Initiative. [28] In general,t he proposed cyclic pathway of HMF formation involves the formationo fafructofuranosyl oxocarbenium ion, which is then deprotonated at C1 to form an enol ands ubsequently an aldehyde( see Figures 2a nd 6). Based on their experimental data, WARF proposed two variations of the fructose conversion mechanism,t aking the effect of metal halidesi nto account (see Figure 6).
In the base pathway,t he halide ion (X À )d eprotonates the C1 and forms the enol. This pathway must be considered highly unlikelyd ue to the low basicity of chloride. In the nucleophilic pathway,t he halide ion forms a2 -deoxy-2-halo inter-mediate that was proposedt ob em ore stable and less prone to side reactions. Upon HX elimination, the enol is formed. By adding ILs, for example, [EMIm]Br,ahigher HMF yield of 94 % was achieved( see Ta ble 3). Thec onversion of glucose to HMF was studied asw ell. The highest yield of 81 %w as obtained using CrCl 2 and NaBr as the catalysta nd DMAc as the solvent at 100 8Cf or 300 min. The results of the experimentsi ndicated that bromide was the most effective halide ligand for the conversion of glucose. It was also stated that halides served as ligands for the chromium atom and influenced the selectivity of the reaction.
The HMF yields from the conversion of cellulose (54 %) and mannose( 69 %) were comparably lower.M oreover,a lthough the described experiments have provided mechanistic insights into carbohydrate conversion in DMAc, the approach has weaknesses also with regard to the low amountso fr eaction solution used in addition to the above-mentioned instability of DMAc under the appliedc onditions. The experiments have reportedly been done in small glass vials, leaving much doubt whether as uccessful scale-upcould be possible.
In 2016, WARF [57] patented ap rocessf or HMF generation from cellulose in polar aprotics olvents, for example, THF,i n the absence of water.C ellulose is first degraded to levoglucosan and then dehydrated to HMF.T he highest HMF yield was 44 %. Ac omparison of the described processes is given in Ta ble 3. The utilizationo falow-boiling-point aprotic polar solvent facilitates the separation of HMF.C ompared to the previously described cellulose conversion, the HMF yield is rather low.T hism ight be attributable to the fact that levoglucosan formation from cellulose generally needs higher temperatures of about 180 8C. In thesee xperiments, 60 mL reaction solution was used for the conversion in ab atch reactor.
Besides monophasic systems, biphasic water-solvent systems weres tudied. Monophasics ystems have am ajor drawback regarding the separation of the catalyst from the solvent. This can be overcomeb yb iphasic systems, in whicht he catalyst remains in the water or IL phase and HMF is transferred into the organic solvent. This also reduces the risk of rehydration of HMF to levulinic acid and formic acid. There have been numerous studies on biphasic systems described in literature for lab-scale synthesis of HMF. [11a] Besides the previously described process by ADM, [50] some major contributions regarding biphasic systemsf or the development of industrial scale processes came from WARF [58] Biphasic reaction systems in the presence of an acid catalyst and ac hemical modifier were studied, the latter comprising an inorganic salt, for example, metal halides, and ad ipolar aprotic additive. Kazi et al. [59] published at echno-economic analysiso f the process. In this theoretical analysisf ructose was used as feedstock and water and butanol as solvents in the biphasic reactor.H Cl and NaCl are used as the catalysts. The annualc apacity of the hypothetic HMF production was set to 61 000 metric tons, operating with 300 metric tons of fructosep er day.K azi et al. [59] calculated the investment costs for ap lant with aH MF productiony ield of 61 kt per year and al ifetimeo f 20 years at approximately 110873 000 E*( *at ac urrencyr ate of 0.7 E/$ [10.11.2010]). Thisvalue is based on HMF production investment of 1302 000 E*, HMF separation of % 24 759 000 E*, and fructosea nd levulinic acid recovery of 45 598 000 E*a s equipmentc osts and % 39 214 000 E*f or additional costs (e.g., engineering, legal expenses, etc.). Operational costs were calculated with % 45 668 000 E*p er year including 22 064 000 E* per year for feedstock. The minimum selling price of HMF was approximately 0.9 E*L À1 .T hey concludedt hat ab etter performance of the process, for example, by increasing the yield, is necessarytoo vercome economic uncertainties.
In 2014, WARF [60] developed as imilarm ethodf or glucose conversion to HMF in ab iphasic reactor.T he dehydration of glucosew as performed using ah omogeneous Brønsted acid, such as mineral acids, and Lewis-acidic metal halides, such as AlCl 3 ,S nCl 4 ,V Cl 3 ,I nCl 3 ,G aCl 3 ,L aCl 3 ,D yCl 3 ,o rY bCl 3 .T he yield of HMF was 62 %u sing AlCl 3 and HCl as the catalysts and sec-butylphenol (SBP) as the organic extraction solvent. The utilization of expensive solvents such as DMAc is avoided. In addition, the separation and purification of previous processes was cost and time consuming. The low yield compared to previous processes from WARF was as ignificant drawback for industrialscale application.I nasimilarm ethod, [61] variousl actones, furans, and pyransw ere used as organic extraction solvents for the conversion to HMF.S ystemsu sing g-valerolactone (GVL), 5butyloxolan-2-one (GOL),5 -propyloxolan-2-one( GHL),a nd 5heptyloxolan-2-one (GUL) were given in the examples. Reacting fructose to HMF in ab iphasic system with GVL gave ac onversion of 94 %a nd as electivity of 84 %. The existing process technologiesw ith biphasic systems still faces ome challenges regarding low HMF yield that need to be overcome for an economic large-scale production. As can be seen in Figure 1, GVL can be derived from HMF or more precisely from levulinic acid. GVL is classified as "problematic" according to the Chem21 Initiative. [62]

Battelle Memorial Institute
Not only WARF [55] demonstrated that the addition of ILs to the reaction mediumi ncreased the HMF yield. Other patents by BattelleM emorial Institute (Columbus, USA) [63] also indicated such improvement. They described am ethod for the conversion of fructose to HMF using ILs and metal halide catalysts, such as CrCl 2 .T he formationo fs idep roducts,s uch as levulinic acid, formic acid,o rh umins, were major drawbacks when performing the HMF synthesis in aqueous solutions. This problem was overcome by using ILs, leading to higherc onversion rates and HMF yields. Experimentsw ere performed with fructose and glucose at 80 8C. An HMF yield of 63-83 %w as obtained from the conversion of fructose and 68-70 %w hen glucose was used as the feedstock. Previous studies by WARF [55] demonstrated that as ignificantly higherH MF yield for fructose (81-94 %) and glucosec onversion (78-81 %) is possible when using DMAc and metal halides.T he influence of metal halides as catalysto nt he glucose conversion is schematically given in Figure 6( bottom). Am ajor drawback of the utilization of ILs is typicallyt he costly recycling of the solvent. BattelleM emorial Institute [63b] also developeda na dsorption separation process for ILs, describing am ethod and an apparatus for the separation of reaction products, such as HMF,from ILs, thus providing aw ay to reuse the costly IL medium.

Agency for Science, Technology,and Research
HMF yields comparable to the ones from Battelle Memorial Institute were obtained by the Agencyf or Science, Te chnology, and Research (A*STAR, Singapore), which developed [64] ap rocess for the dehydration of carbohydrates to HMF in alcoholic solvents. The highest HMF yield of 83 %o btained using isopropanol and HCl as catalysta tatemperature of 100 8Ca nd ar eaction time of 4h.E thanol, 1-propanol, and 1-butanol were tested as solvents as well, but the HMF yield was reduced due to ether formation (see Ta ble 3). In the work up procedure, NaOH was added for neutralization and the solvent was removed by vacuum distillation. Then, the product was dissolved in water and extracted with ethyl acetate, which in turn was removed by distillationt og ive the crude product. Isopropanol is av ery good solvent regarding sustainable processing,i ti s classified as "recommended" according to the CHEM21 solvent-selection guidelines. Lai and Zhang, the inventorso ft he patent,a lso discussed their finding of the fructose conversion in isopropanol in ap ublication. [65] The main by-product in the conversion of fructose in isopropanol was high-boiling humins, which were removed by filtration. In addition to the formation of the target product A, the formation of by-products B-D (see Figure 7) was reported for methanol as the solvent. In ethanol, the main products formed were A( HMF) and by-product B, and in isopropanol and tert-butanolm ainly HMF was formed. In this publication, also the influence of different Brønsted acids wasa ddressed;t he highest yield was reportedf or HCl, followed by H 2 SO 4 .H NO 3 ,H 3 PO4, HCOOH, CH 3 COOH,a nd B(OH) 3 were also tested but gave only traces of HMF or no HMF at all in isopropanol.A mberlyst 15 was tested as as olidacid catalystind ifferent alcohols and caused increased etherification and acetalization.
The conversion of carbohydratest oH MF in alcoholics olvents should certainly be further developed.I na ddition, the suppression of humin formation must be ac entral topic to be exploredi nf uture research.

Sartec Corporation
Another HMF-generation approachu sing alcohols as solvents was developed by Sartec Corp.( Anoka, USA) [66] In this process, as accharide solution is in contact with am etal oxide catalyst (TiO 2 )a tt emperatures between 180-200 8C. The single-step reaction is continuously performed in ac olumnr eactor packed with TiO 2 particles. Monophasic solvent systems (isopropanol, ethanol, methanol) were tested in the conversion of saccharose using ZrO 2 or TiO 2 and gave rather low HMF yields of 10-14 %. Biphasic solvent systems( butanol/watero rM IBK/water) were also used for the conversion of glucose, fructose, and saccharose. The highest HMF yield was 46 %f or the conversion of glucose in MIBK/water (10:1) at 180 8Cf or 2min in aT iO 2packed column. HCl was used as ac o-catalyst. No further work-up procedure wasg iven in the patent; the yields were analyzed by HPLC from the organic and aqueous phase. Interestingly, the overall HMF yield was rather lowi nb utanol/water systems( < 20 %). Previous studies by Romµn-Leshkov andD umesic [51] revealed that the HMF selectivity of 1-butanol systems is rather low,e specially compared to 2-butanol. In the Sartec Corp. process, HMF was detected in both the organic and the aqueous phase in considerable amounts. Thiss hows that the describedm ethodh as some apparent problems with the extraction efficiency of the organicp hase. It has been shown in previouss tudies [50,51] that the extraction sufficiency can be improvedb yt he additiono fi norganic salts (see above). In the process by Sartec Corp.,t his was not considered. Another apparentl imitation is the lack of information on possible workup procedures.E ven though the continuous production in a packed columnr eactori sa ni nteresting approach, evidently furtherd evelopments and improvements of the methoda re needed.

Novamont S.P.A.
Besidest he solvents ystem, also the used catalysto bviously impactst he HMF yield. Patents assigned to Novamont S.P.A. (Novara, Italy) [67] disclose ad ehydration process of saccharides to HMF using various catalytic systems. The flowcharto ft he process is given in Figure 8. The reaction mixture consists of a catalyst( eitherT iO 2 supportedo ni mmobilized SiO 2 ,p hosphotungstica cid supported on SiO 2 (HPWO/Si50O), a-Zr(HPO 4 ) 2 or Ti(HPO 4 ) 2 ,aq uaternary ammonium salt, water,a nd as accharide as feedstock. The quaternarya mmonium salt can be tetramethylammonium chloride (TMAC),t etraethylammonium chloride (TEAC), tetraethylammonium bromide( TEAB), or tetrabutylammonium bromide (TBAB).A no rganic solventi su sed for the extractiono fH MF from the aqueous reactionm ixture. In the given examples, saccharose wasd ehydrateda t8 0 8Cf or 15 min and then the temperature was increased to 100 8C. TEAB and HPWO/Si50O as catalysta nd 2-butanone as extraction solvent were used. HPWO/ Si50O was synthesized from Figure 7. Conversion of fructose to HMF( product A) and its derivatives (products B-D) in alcoholic solvents (R = methyl,e thyl). [65]  T he HMF yield was 67 %a nd the purity 94.1 %. When this reaction was performed with fructose, the yield was 80 %a t9 9.6 %p urity.I na nother example, fructosew as dehydrated in the presence of TiSi50O as catalyst and TEAC as salt. The synthesis of the TiSi50O catalyst from dioxane, SiO 2 and titaniumi sopropoxid [Ti(iPrO) 4 ]w as also described in the patent.E thanol, chloroform,a nd THF wereu sed for the separation andp urificationo fH MF.T he yield was 93 % and the purity of HMF was 97.6 %. The solid catalysts werer ecycled afterwards. After the third cycle, the HMF yield dropped to 82 %. When using a-Zr(HPO 4 ) 2 and TEAB as catalyst, 87.6 % HMF yield with 99 %p urity was obtained.T he good HMF yields as well as the relativelyl ow reaction temperatures are advantages of the method. The applicability of these results to large quantities is certainly an interesting topic for future research. However,t he use of toxic and water-contaminating quaternary ammonium salts should be avoided.

Novozymes A/S
Severalp atentsa ssigned to Novozymes A/S (Bagsvaerd, Denmark) [68] disclosed ac ontinuous methodf or the salt-catalyzed dehydration of fructoset oH MF.T he reaction mediumc onsists of an organic phase anda na queous phase containing the salt. Besides fructose conversion,a lso the utilization of mannose or glucosei ss tated in the patents. The enzyme glucose isomerase converts glucose to fructose, and correspondingly mannose isomerase converts mannose to fructose. The highest HMF yield obtained was 70 %a tafructose conversion of 98 % and as electivity of 72 %. KCl was used as as alt catalysti nt his experiment. These findings clearly indicate that the selection of ap roperc atalyst is essential.T he glucose isomerase enzyme showedg ood stability in the presence of NaCl, KCl, and Na 2 SO 4 .Anumber of interesting research questions, for example, the effect of HMF and salt on the activity of the isomerase enzyme and the conversion of glucose, fructose, andm ixtures of both in MIBK/water systemsw ere investigated. Unfortunately,acomplete process startingf rom monosaccharide to HMF using isomerasee nzymes was not included in the patent, leaving the question of the scalability of this system open.

UC Regents
Previous studies by Novozymes A/S did not discuss the applicability of ac omplete process using isomerase enzymes. Research by UC Regents (Oakland,U SA) investigated this issue further. Ap rocess in which glucosei se nzymatically converted to fructosew as assigned to UC Regents [69] in ap atent on the production of HMF in ILs, for example, EMIm-based ILs. Glucose was used as as tartingm ateriala nd was enzymatically converted to fructose with glucosei somerase and borate salts. An acid catalyst, for example, AlCl 3 ,w as used for the dehydration of the fructose to HMF,t he highest yield of HMF in these conversion examples wasi nt he range of 55-60 %, which is rather lowc ompared to other processes.A nother example of the conversion of cellulose to glucosei n[ C 4 mim]Cl using HCl at 140 8Cf or 60 min was included.I nt he given example, glu-cose was treated with glucosei somerase and sodium borate to yield fructose, whichw as then reacted in [C 2 mim]Cl for 30 min at 100 8Ct oy ield HMF.D ata on the isomerization efficiency is missing, it would be interesting to compare the turnover frequency and the total turnover number of the described enzymes to those of industrial isomerization processes.

Südzucker AG
The production of HMF in aqueous systems faces major challenges due to the formation of side-products, especially when mineral acids are used as catalysts. The main challenge in the continuous production of HMFi na queous systemsa re the solid side-products.
The setting of ap artial conversion endpoint is an interesting methodt oincreasethe HMF yield and to reduce the formation of side-products.Acontinuous process that uses this method was patented by Südzucker AG (Mannheim, Germany). [70] The Südzucker process is given in Figure 9. Mineral acids, such as HCl, H 2 SO 4 ,o rH 3 PO 4 (1-2.5 wt %), were used for the dehydration of fructose to HMF in aqueous media at temperatures in the range of 80-165 8Ci naplug flow reactor (PFR). The maximum conversion of fructose was 40 %, HMF was purified by columnc hromatography.F or as ustainable process, all produced streamsm ust be valorized and resupplied to the process. Duet ot he low HMF yield, additional process steps,f or example, additional purification steps to isolate the unreacted carbohydrates, are needed, which also affects the cost and efficiency of the process.
Previous studies also investigated this HMF formation setup (aqueous system,f ructose as feedstock, mineral acids as catalyst) in continuous processes.M any of the described continuous systems applied am icroreactors ystem due to the better mixing and heat transfer.T uercke et al. [71] used 10 %f ructose solution mixed with 0.1 mol L À1 HCl (1:1) to obtain 54 %H MF yield in am icroreactor.H igher fructose concentrations were only used when organic solvents were added to the system. 60 %H MF yield was obtainedb yM uranaka et al. [72] (5 min at 180 8Ci namicroreactor). Very low fructose concentrationso f 1wt% in phosphate-buffered saline were used. The monophasic experiments showed an increaseinside-products compared to biphasic systems.
Consequently,t he continuous production in am icroreactor is also limited by the formation of solidside-products. Typically, the reactions are performed at very low fructose contents to minimize humins formation and reduce the risk of clogging.

AVALON Industries AG
The previously described processes focused on the conversion of mono-or disaccharides to HMF.A VALON Industries AG (Zug, Switzerland) [73] developed ap rocess for the conversion of lignocellulose to HMF.T hey filed severalp atentso nt he hydrothermalc arbonization (HTC) of lignocelluloset op roduce HTC char,w hich can be used as energy source. [74] In this process, lignocellulose is decomposedu nder high temperature and pressure to glucose and then fructose, which is dehydrated to HMF.U tilization of fructosea sf eedstock would enhancet he formationo fH TC char and reduce the HMF yield. The lignin from the lignocellulose slows down this formation and ensures ac onstant dehydration of fructoset oH MF.T he utilization of a lignocellulosic feedstock is ac lear advantage of this process. The HMF-containing process water is then extracted with as olvent, for example, supercritical CO 2 ,i nacountercurrent mixsettler columnp rocess. The HMF-enriched solvent is then subjected to another separation step. In case of supercritical CO 2 , ad edusting technology is used for the separationo fH MF and the supercritical CO 2 .T he company also issued patents [75] on carbon-linked HMF oligomers that contain at least two HMF units. The proposed structure is given in Figure 10. The main advantage of this HTC process is that HMF is produced as a side-product that is dissolved in the process water.T hisg enerates additional revenues.U nfortunately,n oH MF yield is given to evaluate the efficiency in more detail. In contrastt ot he previously described processes, the HTC technology has already been scaled up to small-scale commercial production of HMF.
An overview of relevant factorso ft he described processes is given in Ta ble 3.
The conversion of cheap lignocellulosic materials to HMF is ah ighly relevant topic, especially with regard to ap roduction at larger scale. The industrialization of many of the described processes is hindered by the formation of hydrochar (humins that are polymerized in different ways than the humins formed by the acid-catalyzed process). The structure depends on many parameters, for example, used feedstock and synthesis parameters. Many modelsh aveb een proposedf or the hydrochar formationf rom hydrothermal carbonization of carbohydrates. Patil and Lund [76] suggested that the initial step in the hydrochar formation is the hydrolytic ring opening of HMF and the  Figure 10. Hydrothermal process by AVABiochem (left) [73] and proposed HMF oligomerf ormation (right). [ formationo fa ldol condensation products,s uch as 2,5-dioxo-6hydroxy-hexanal (DHH). Shi et al. [77] proposed that a-carbonyl aldehydes, such as DHH, pyruvaldehyde, and 3-deoxyglucosone are key primary precursors for the formation of hydrochars.

Commercial HMF Production
Since 2014, AVAB iochem produces HMF at ac ommercial small scale. AVAB iochem is as ubsidiary of AVALON Industries AG, which announced that they will focus on the global implementation of the HTC technology for industrial-scale HMF production. [78] Even though the AVAB iochem plant claims to have reachedt echnology readiness level (TRL) 9, corresponding to full commercial applicationw ith the developed system proven in operational environment, [79] it is still not operating al argescale, industrial production-the operating capacity of the plant in Muttenz, Switzerland, being 300 tp er year.I n2 019, AVAB iochem announced that it is currently planning the next scale up to 5000-10000 tp er year. [80] In October 2018, an onfarm biorefinery technology centeropened at Stuttgart-Hohenheim in Germany,t he core of the small-scale plant being an HMF module. In the EU project "Grace" (2017-2022), Miscanthus biomass is used as startingm aterial. [81] Up to present, no commercial, large-scale HMF plants are running,w hich is reflected in the still rather high price of HMF (Sigma-Aldrich, 3500 E kg À1 ).

Analysis of Issues of Scaled Production
Severalp rocesses for the production of HMFh ave been developed and patented by companies.
AVAB iochem already set the first steps towardi ndustrial production with as mall-scaleo perating plant. Many companies contributed to the development in patenting variousp rocess technologies, butt here are still many problems that have to be dealt with. The main challenges in the upscaled HMF production are the formation of side-products,e specially of solid humins, the separation of HMF from the reactionm edia, and its subsequentp urification. In general,b y-product formation depends on the reaction parameters and affects the purification and ultimately also the economic efficiency of the HMF production.
Many of the described processesa re still at av ery early stage of development, in which media and parameter optimization is the main focus.

Formation of side-products
The increase of efficiency and reduction of side reactions is essentialf or an economic and sustainable process. In general,i t is better to prevent the formationo fs ide-products from the beginningu nless they can be valorized.F ormation of levulinic acid andf ormic acid does reduce the HMF yield, but this presents only am inor challenge because of their usability.
The formation of condensation products-humins-is much more problematic. Recently,t he valorization of humins has come into focus as they are considered ak ey factor for an economically feasible process. This valorization of humins also presents several challenges, as their chemical structure and yield is process dependent and their separation rather demanding.
Many publications focused on the analysiso fH TC humins, which are also referred to as hydrochar.U nlike the acid-catalyzed dehydration of carbohydrates,t he HTC process does not involved acids. Consequently, the resultings tructure of the HTC humins is different from the humins produced according to acid-catalyzed dehydration.
The molecular structure of hydrochar as well as its formation kineticsa re still under debate. Shi et al. [77] summarized af ormation route for hydrochar as follows: 1. Biomass (cellulose and hemicellulose) are hydrolyzed to monosaccharides. 2. Monosaccharides are dehydrated to furanoic compounds, for example, HMF. 3. The formed compounds undergo aseries of polymerizationpolycondensation reactions leading to the formationo fpolyfuranic compounds. 4. These polymers further undergo aromatization to form a polyaromatic hydrochar structure. Humins are still mainly used for energyo rh eat generation, even though higherv alue-addeda pplicationsa re desired. Hoang et al. [82] analyzed the applicationo fh umins in gasification or steam-reforming processes for hydrogen production. The pyrolysis of humins for the production of liquid fuels and biochari sd escribed in the literature as well. [83] The pyrolysis of humins led to awide range of compounds, the primary volatile compounds being furans and phenols. An increasedp yrolysis temperature caused aromatization of the products. Further research is still required to understand mechanism andk inetics of humin pyrolysis.
The potential of humins in higher-value applications,such as thermoset materials, was investigatedb yM ija et al. [24a] as well as Sangregorio et al. [24b] To si et al. [84] reportedt he production of macroporous foamsf rom humins. Its application as feedstock for the production of activatedc arbons and the utilization as potentiale lectrode materials for supercapacitora pplications was also studied. [85] The valorization of side products-humins in our specific case-is one approach to make ap rocess more efficient. Nevertheless, humins formation might reduce the HMF yield drastically.

Separation and purification of HMF
Many earlier publications dealt with the prevention of condensation reactions since they cause major problemsi nH MF separation and purification. An increasing number of publications focus on the kinetics of the dehydration to HMF.D ashtban et al. [4] reviewed the kinetic analysis of HMF for aqueous, biphasica nd IL systems.
There have been numerouss tudies on the fructosec onversion to HMFi np olar aprotic co-solvents, such as DMSO.T ypically,high HMF yields and fewer by-products are formed in sys-tems using DMSO as solvent. This has been also shown in the process developed by Roquette Freres. [38] Some aspectso ft he role of DMSO in the conversion to HMF are still debated, for example, the ability to increaset he reaction rate in the absence of an acid catalyst.R ecently,W hitaker et al. [39] showed that the fructose conversion in DMSO is mainly promoted by solvatione ffects and does not originate from H 2 SO 4 catalysis. This is in line with generalk nowledge from synthetic carbohydrate chemistry that often uses DMSO as advantageous solvent for water elimination reactions from furanoid and pyranoid systems. So far the following roles of DMSO in the fructose conversion have been proposed: [39,41,43] 1. DMSO alters the isomer distribution of fructose. 2. DMSO promotes conversion through solvation effects. 3. DMSO reduces HMF susceptibility to nucleophilic attack. 4. DMSO stabilizes HMF in solution.
Even thought he formation of side-products is significantly reduced in DMSO and other solvents with high boiling points (T b ), the recovery of HMF from these solvents is am ajor issue in large-scale production. Typically,t he purification processes involved istillation or extraction, for example, with ethyl acetate. Dependingo nt he boilingp oint of the solvent and its affinity to HMF,t hese separation and purificationp rocesses can be very cost intensive and might require large amounts of organic solvents.
The same holds true for biphasic systems containing organic solvents and water.I tn eeds to be critically mentioned that literature accountso ften do not describe purificationo rs eparation processes. In many cases the HMF yield is determined by HPLC, such as in ap rocess [66] using MIBK/water as the biphasic reactionsystem.
Processes using alcohols as solvents, have been described as well. The HMF yield in alcoholic solvents is reducedd ue to side reactions, such as ether formation anda cetalization.I n general,a lcohols are favorable solvents since they are environmentally friendly and give good HMF yields.S eparation from alcoholicr eactionm edia can be done by low temperature vacuum distillation as has been shown by A*STAR.T he easy separation is ac lear advantage of alcoholic and low-boiling point systems.
From an environmental perspective,w ater would also be a good solvent, but the pronouncedf ormation of side products is am ajor issue here. The continuous production of HMF in aqueous systemsw as reported as well but was impeded by the formationo fs olid side-products,w hichr emains ag eneral problem in the continuous productiono fH MF in aqueous systems catalyzed by mineral acids.

Catalyst regeneration
Catalyst separation andr ecycling are additional aspects to be considered for as ustainable process. The separation of as olid catalystf rom the reaction media can be done quite easily by filtration.M ore challenging is the separation of catalyst from solid by-products and the risk of catalyst inactivation.N ovamont S.P.A [67] described the production,s eparation, and recycling of variousc atalysts as well as their performance after reuse. They reported ad rop in HMF yield after several cycles, although they were still rather high (82 %).

Economic considerations
Most of the described processes are still in av ery early development stage, in whichm edia and parameter optimizationa re in focus.T he influence of solvent, catalyst, and feedstock on HMF yield and selectivity were investigated.
Many publications focus on the use of monosaccharides, especiallyf ructose, as the starting materials for the conversion reaction( see Ta ble 3). From an economic point of view,l ignocellulosic biomass is much cheaper than neat fructose. The HMF production process developed by AVAB iochem [73] is based on hydrothermalc arbonization of lignocellulosic feedstocks. As it is one of the few implemented small-scale processes, it can be assumed that the HMF yield is sufficient for an economic process. Although for the utilization of HMF as buildingb lock in the chemical industry,as ignificant lower price must be targeted. [39,[82][83][84][85] Evaluation of economic feasibility and thel ikelihood of the implementation of ap rocessd esign at larger scale are important future research topics. Only af ew techno-economic analyses of HMF production processesh ave been published so far. At echno-economic analysiso faprocessd eveloped by WARF [59] gave ar ather high selling price of HMF.I tw as concluded that ab etter performance of the process, that is, a higherH MF yield, is needed to bring about guaranteed economic feasibility.F or an economic feasibility,s ufficient material must be availablef or an acceptable price. As ar eference, the production of particle boardsa lone requires some 5-7 million tons of low priced standard adhesive each year. [86] Further techno-economic analysis on HMF production processes are necessary for ab etter comparison of the economic feasibility of different technological options. In ar ecent publication, Motagamwala et al. [87] reported ap rocessf or the production of HMF from fructose using acetone/water as solventa nd HCl as catalyst. They obtained 89 %H MF yield with an HMF selectivity of 95 %, after evaporation of the solvent under reduced pressure HMF was recovered with 99 %p urity.T hey also developed ap rocess model and performed at echno-economic analysis of the process. They obtained am inimum selling price of HMF of 1567.2 E*( *at ac urrencyr ate of 0.9 E/$ [12.04.2020]), assuming af ructose cost of 745.7 E.

Summary and Outlook
There has been ac ontinuous growth of interest in large-scale HMF production. Severalp rocesses were described in literature, but they still face severalchallenges that need to be overcome beforehand. Several aspects ranging from the choice of startingm aterial to the recycling of solventa nd catalyst to the formation of side-products need to be considered for ac omprehensive process assessment. Many processes just target a specific problem insteado ff ocusingo na ll challenges. HMF yield and reaction mass efficiency (RME) are good indicators of the efficiency of aprocess.