5-Hydroxymethylfurfural (HMF). A Review
`Focussing on its Manufacture*
`B. F. M. Kuster,
`Eindhoven (The Netherlands)
`
`The acid-catalysed dehydration of hexoses results in the formation of
`5-hydroxymethylfurfural (HMF). Fructose and inulin are especially
`good starting materials. A review is given of the many methods to
`produce HMF. The reaction kinetics are dealt with in the sections:
`raw material, hydrolysis and reversion, catalysts, reaction tempera-
`ture and time, the concentration and the risk of polymerization and
`the solvent and HMF stability. Manufacturing processes are illus-
`trated for fructose as the starting material: aqueous systems with
`homogeneous acid catalysis or those using ion-exchangers as catalyst,
`systems using dimethyl sulfoxide as the solvent and those using other
`organic solvents. A short treatment is given on processes starting
`with glucose, work up procedures and the formation of levulinic acid
`and other in situ-formed HMF-derivatives.
`
`5-Hydroxymethylfurfurall. Ein Uberblick uber seine Herstellung.
`Die saurekatalysierte Dehydratisierung von Hexosen resultiert in der
`Bildung von 5% -Hydroxymethylfurfural (HMF). Fructose und Inulin
`sind besonders gute Ausgangsmaterialien. Es wird ein Uberblick
`iiber die zahlreichen Methoden zur Herstellung von HMF gegeben.
`Die Reaktionskinetik ist eingeteilt in die Abschnitte ,,Rohmaterial,
`Hydrolyse und Reversion, Katalysatoren, Reaktionstemperatur und
`-zeit, Konzentration und das Risiko der Polymerisierung und das
`Losungsmittel und die Stabilitat". Die Herstellungsverfahren werden
`fur Fructose als Ausgangsmaterial illustriert: WaRrige Systeme mit
`homogener Saurekatalyse oder unter Verwendung von Ionenaustau-
`schern als Katalysatoren, Systeme unter Verwendung von Dimethyl-
`sulfoxid als Losungsmittel und solche, die andere Losungsmittel ver-
`wenden. Ein kurzer Hinweis wird auf Verfahren gegeben, die mit
`Glucose beginnen, Aufarbeitungsverfahren sowie auf die Bildung
`von Livulinsaure und anclere in situ gebildeten HMF-Derivaten.
`
`1 Introduction
`
`Earlier reviews on HMF include that of Moye [3] in 1964 (161
`ref.), which covers preparation procedures as well as physical
`and chemical properties and industrial uses. In the series
`Advances in Carbohydrate Chemistry two reviews have
`appeared, one by Newth [4] in 1951 and the other by Feather and
`Harris [5] in 1973. Two French reviews appeared in 1981 : one
`by Gaset [6] on HMF manufacture (88 ref.) and the other by
`Faury [7] on further HMF chemistry (150 ref.). Furan polymers
`have been reviewed in 1986 by Gandini [8] (209 ref.). There is
`an abundant amount of literature available on HMF (over 1000
`references), but this review will focus on H M F manufacture
`covering most of the relevant data of the last thirty years. Some
`overlap with Gaset's review cannot be avoided, but the data arc
`treated in a different way and the most recent data are included.
`
`1.1 The dehydration reaction
`
`HMF, shown in Fig. 1, originates from hexoses by the loss of
`three molecules of water in an acid catalysed reaction. In
`aqueous mixtures HMF enters into a consecutive reaction
`taking up two molecules of water, to yield levulinic and formic
`acid. In non-aqueous systems HMF hydrolysis can be suppres-
`sed. Under all circumstances cross-polymerization reactions
`occur leading t o coloured soluble polymers and insoluble,
`brown to black, humins.
`The dehydration passes at least through two intermediate
`stages: ChHIOOS and C6H804. Mechanisms that involve either
`cyclic or acyclic intermediates have been proposed as shown in
`
`*) This is a short version. Additional material including a more
`comprehensive reference list, can be obtained from the author.
`This review has been written on the occasion of the third symposium
`of "Lawine", (Dutch Working Group on Inulin), and the renewed
`activities on HMF in the Netherlands in general. Dutch work on
`HMF, partly financed by the IOP-k (Dutch National Innovation
`Oriented Program on Carbohydrates) is going on at Delft University
`of Technology [ 11, TNO Zeist [2] and in the Eindhoven group of the
`author.
`
`HEXOSES
`
`- 3 H20
`
`HOH2C'($Y3-lO
`
`HMF J + 2 H20 '4
`
`+
`FORMIC ACID
`LNULlNlC ACID
`Products from the acid catalysed dehydration and rehydration,
`Fig. 1.
`of hexoses.
`
`Fig. 2. Although the acyclic route seems to be most favoured
`nowadays, real evidence is still lacking. Mechanistic studies are
`difficult because of the lability of the intermediates which,
`therefore, occur in tiny amounts only.
`During reaction the hexoses form reversion products by self-
`condensation. From aldoses, e.g. glucose, oligosaccharides are
`formed, whereas ketoses, e.g. fructose, give rise chiefly to the
`tricyclic dianhydrides. These compounds can be hydrolysed
`back to hexoses. Some of the possibilities are given in Fig. 3.
`2 Kinetics
`
`Kinetic studies have been carried out by Teunissen (1930) [9, 101, M c
`Kibhins et al. (1958) [ I I , 121 Breen (1964) [13], Kuster et al. (1975)
`[14-201, Mercadier et al. (1981) [21-231 and Van Darn et al. (1986)
`[24]. Maximalisation of HIMF yield has been the main objective of these
`and many other studies. It appears from Fig. 4, showing the overall
`reaction network, that it is not obvious to obtain high yields. In addition
`to a solvent and a catalyst the reaction mixture consists of hexoses and
`reversion products (raw materials), HMF, levulinic- and formic acid
`(hydrolysis products). intermediates, hydroxyacetylfuran (HAF, see
`Fig. 2), other side products, soluble brown polymers and insoluble
`
`314
`
`starchlstarke 42 (1990) Nr. 8. S. 314-321 0 VCH VerkigSgeSekhaft mbH, D-6940 Weinheim. 1990
`
`0038-9056/90/0X08-031493.50+.2.5/0
`
`Petitioners' Exhibit 1018, Page 1 of 8
`
`

`
`F=O -
`
`7H20H
`
`7
`
`7
`
`OH
`CH,OH 4
`HOCH-HCOH
`b
`
`HOH,C-
`HXo\
`HOFH 7
`HOCH-HCOH
`b
`HFOH
`a or 8 HEXULOFURANOSE
`
`HOH2C-FH/o\f-
`
`HO
`
`HOCH-HC
`
`HOHzC-jH 'O\
`
`T=CHOH
`
`HOCH-HCOH
`
`YHO
`FHO
`F="
`F" + fiH
`FH
`HtoH
`H7°H
`
`fiOH
`
`HFOH
`
`a
`
`CH20H
`
`CHzGH
`
`HMF
`
`H?oH
`CH20H
`
`FRUCTOSE
`
`.lr
`
`fiHOH
`
`?OH
`HOtH
`
`HFoH
`
`H?oH
`CH20H
`
`1,2-ENEDIOL dl
`YHO
`
`HFOH
`HOFH
`H$OH
`HVOH
`CH20H
`
`GLUCOSE
`
`2.3-ENEDIOL
`
`W?
`
`HO
`
`HO CH2
`
`HO QOH
`
`HO
`a-1.6-DIGLUCOSE
`(ISOMALTOSE)
`
`OH
`
`*VH
`
`p-p: a-f
`
`8-P: B-P
`
`HO
`I HO
`
`HO
`
`HO
`
`HO
`
`8 - P : a - p
`
`HO
`HO
`Fig. 3. Some reversion products.
`
`starchhtarlte 42 (1990) N r X. S 314-321
`
`0
`II
`
`ficcHzoH
`
`Fig. 2. Possible mechanisms for the dehydration reaction.
`a) acyclic route. sec e.g. Feurher and Harris [ S ] and references cited
`therein.
`b) cyclic route. see e.g. N m , t h [4] and references cited therein.
`c) such route could also be envisaged.
`
`HAF
`
`C12H1809
`
`C1zHzoolo
`C,H,,O,
`
`CSH803
`
`CH,O,
`
`REVERSION PRODUCTS
`
`1
`
`HEXOSES
`
`SOLUBLE
`
`-C+
`R
`0
`--;+ POLYMERS
`S
`INTERMEDIATES - P+
`0
`INIERMEDIATES
`L
`Y
`HAF - M - 3
`E
`-R+
`I
`S
`-A+ T
`I
`0 N
`
`INSOLUBLE
`
`HUMINS
`
`C,(H,0)2.5
`
`..2.0
`
`I
`J.
`INTERMEDIATES
`\1
`LEWLINIC +
`FORMIC ACID
`
`The dehydration reaction network.
`Fig. 4.
`(HAF = hydroxyacetylfuran. see also Fig. 2)
`
`black humins. Due to the ratio of formation and disappearance rates for
`the intermediates and HAF. these occur in verly low C O I I C C I I ~ ~ B ~ ~ ~ I I ~
`only and do generally not show up in the analysis.
`The kinetics will be treated in the following subsections:
`- raw material, hydrolysis and reversion.
`- catalyst
`reaction temperature and time,
`- the concentration and polymer formation,
`- the solvent and HMF stability.
`
`~
`
`2.1 Raw material, hydrolysis and reversion
`
`HMF can be produced from all hexoses. and also from those oligo- and
`polysaccharides which can yield hexoses on hydrolysis. as well as from
`waste materials containing such compounds. However, it appears to be
`more selectively produced from keto-hexoses. The cheapest source
`
`31s
`
`Petitioners' Exhibit 1018, Page 2 of 8
`
`

`
`nowadays amply available is fructose, obtained from sucrose hydrolysis
`or glucose isomerisation processes. However inulin, a fructan, shows
`promising [25].
`In aqueous systems hydrolysis of sucrose and inulin is generally much
`faster than dehydration. Using sucrose the glucose unit often can be
`recovered unchanged [26].
`Two reasons can be mentioned for the higher yields of HMF from
`fructose when compared to glucose.
`Glucose forms a very stable ring structure, the fraction of open chain
`forms in solution and the enolisation rate are consequently low.
`Fructose on the other hand forms less stable ring structures, there is
`more of the open chain form in solution and the enolisation rate is
`comparatively high. Because enolisation is believed to be the rate
`determining step HMF yields will be higher starting from fructose than
`from glucose.
`Fructose forms di-fructose-di-anhydrides [27-291 in an equilibrium
`reaction in such a way that the most reactive groups for cross-
`polymerisation are internally blocked, with a consequently positive
`effect on selectivity. Glucose on the other hand forms true oligo-
`saccharides which still contain reactive reducing groups, leaving more
`risk for cross-polymerisation with reactive intermediates and HMF.
`The importance of reversion during the dehydration reaction has been
`pointed out by Van Dam et al. [24] (see Fig. 3).
`Because glucose is cheaper than fructose, the use of glucose as a raw
`material will get some special attention in the section on manufacturing
`processes. Most of the further data will refer to fructose dehydration.
`
`2.2 Catalysts
`
`In Table 1 the catalysts mentioned in the literature are summarized.
`All reactions presented in Fig. 4 are catalysed by acids, proton as well as
`Lewis acids. In the first preparations oxalic acid has been used. (The
`
`Table I.
`Catalysts Used in the Dehydration Reaction.
`
`Catalyst
`
`Oxalic acid
`None
`H ~ S O J , H3PO,+, HCI
`Levulinic acid
`p-Toluene sulfonic acid
`''12''
`NH4+ sulfate/sulfite
`Pyridinium phosphate
`Pyridinium HCI
`ZnC12, AICli
`Al salts
`Th, Zr-ions
`Zr-phosphate
`Cr, Al, Ti, Ga, In-ions
`ZrOC12, VO (SO,), Ti02
`V-porph yri ti
`Zr, Cr, Ti-porphyrin
`BF?
`Ion-exchangers
`Zeolite
`
`References
`
`e.g. [13, 26, 301
`[31-341
`e.g. [11, 14, 28, 35-43]
`[74l
`~ 4 , 4 1 1
`[44-461
`[471
`[48 - 501
`[511
`~521
`[53-551
`~561
`[571
`~581
`[59, 601
`[611
`[62, 631
`~641
`e.g. [I, 21, 65-72]
`P I
`
`
`reaction can also be carried out catalysed by in situ formed acidity, but
`this generally does not lead to optimal yields). Most commonly used
`cheap acids are H2SOJ, H3POJ and HCl. Iodine also has been mentio-
`ned as a catalyst but the catalytic action appeared to be due to
`hydroiodic acid formed [39]. The use of Zn, Al, Cr, Ti, Th, Zr and V, as
`ion, as salt or as complex, in soluble as well as in insoluble form, is
`mentioned in literature. However, it does not result in significant
`improvement in yield although sometimes a higher activity is observed.
`In DMSO, BF3-etherate has been used.
`Except for some metal salts most of the above mentioned catalysts are
`in dissolved form and are difficult to regenerate, the cheaper acids are
`just neutralized and disposed off as salt after reaction. To avoid
`
`316
`
`regeneration and disposal problems use can be made of acidic ion-
`exchangers. Using ion-exchangers the reaction temperature is generally
`limited to below 130"C, resulting in a limited rate per amount of
`catalyst. Furthermore, for a long lifetime, the formation of insoluble
`humins should strictly be atoided. Therefore, ion exchangers have only
`been used in DMSO or in aqueous systems with continuous HMF
`removal by extraction or adsorption.
`
`2.3 Reaction temperature and time
`
`Within the practical limits for catalyst concentration, there is a recipro-
`cal relationship between the reaction temperature and the time needed
`for proper conversion. Working in water around 150°C reaction times
`are usual 1 to 5 h [26,43]. Increasing the temperature to 270°C brings
`reaction time back to 10 s [74]. Working in DMSO with ion-exchangers
`as catalyst at 80"C, 8 h are needed [57]. Because catalyst activity is
`higher in certain non-aqueous systems shorter reaction times and lower
`temperatures suffice in such systems, e.g. working in polyethylenegly-
`col at 16O"C, 30 s are usual [17, 201.
`It appears that the activation energy for HMF formation is higher than
`for HMF disappearance with the result that the maximum obtainable
`concentration increases with increasing temperature [ 11, 14, 351.
`Formation rate of HMF is increased by a higher enolisation rate as well
`as by a higher proportion of acyclic and furanose forms of fructose at
`higher temperature.
`
`2.4 The concentrat:ion and the risk of
`polymerization
`
`For economical reasons the reaction mixtures should be as concentra-
`ted as possible. However, apart from handling problems, the main
`disadvantage of concentrated solutions is the resulting low selectivity.
`The higher the chance that reactive compounds collide with each other.
`the higher is the rate of cross-polymerization and humin formation [14,
`23, 241.
`In aqueous systems losses due to humin formation amount up to 35%
`for 1~ fructose solutions going down to 20% for 0.25 M . It is not clear if
`polymerization losses can be completely avoided by working in very
`dilute solutions. In this respect it is interesting to know that a large
`portion of certain humins isolated from sorbose have been stated to
`consist of HAF-polymers [75]. If this would also be true for fructose
`humin, apparently humin also forms via a parallel reaction which
`cannot be avoided by dilution.
`In certain non-aqueous systems less problems occur with polymeriza-
`tion using high concentrations. In mixtures of e.g. DMSO or polyethyl-
`eneglycol containing up to 50% by weight fructose, yields around 70%
`can be obtained [20,42, 641. Heating an equimolar mixture of crystal-
`line fructose with pyridinium hydrochloride for 30 min at 120"C, has
`been reported to yield 70'% H M F [51].
`
`2.5 The solvent and HMF stability
`
`The solvent appears to have a very pronounced effect on the course of
`the reaction. The primary task of the solvent is to induce fluidity and
`enable contact of reactant and catalyst. Good solvation properties for
`the solutes are needed if working in concentrated mixtures is preferred
`and solute interactions leading to humin formation must be prevented.
`Water has been most frequently used. It is an excellent solvent for as
`well fructose as HMF. However, water is also a reactant: it plays a role
`in the reversion equilibrium and in the hydrolysis of HMF. Decreasing
`water concentration by mixing in organic solvents, shifts the reversion
`equilibrium to the dianhydrides thereby diminishing the risk of humin
`formation [24] and suppresses HMF hydrolysis resulting in high yields
`[9-11, 171.
`The use of aqueous solvent mixtures has been the subject of several
`patents, Peniston [35] added butanol, Hales et al. [38] dioxane. If
`solubilities are high enough also completely anhydrous systems can be
`used: dimethylformamide [44,45], acetonitrile [71], quinoline [46] and
`most favourite, dimethyl sulfoxide (DMSO) [33,57,64,68]. However,
`these solvents have several drawbacks: they are high-boiling and
`therefore product removal and purification give some additional pro-
`blems. Apart from being rather expensive they also have some risk for
`health and environment.
`
`atarch/starke 42 (1990) Nr. 8, S. 314-321
`
`Petitioners' Exhibit 1018, Page 3 of 8
`
`

`
`Srnyrhe and Moye [42] gave an extensive treatment on non-aqueous
`sugar solvents, several of which. e.g. polyglycol-ethers, also were
`witable for the dehydration reaction.
`A special position have the free glycols. Mixing in ethylene glycol in
`nater. Varr DM et al. [24] showed that this had a strong destabilizing
`effect on HMF: apparently by acetal formation, the hydrolysis rate of
`HMF 15 speeded up. That also the initial sugar can have a similar effect
`has been shown by Kirsrerand Van der Baan [14,16] who found that rate
`constants for HMF hydrolysis are substantially higher in the presence
`than in the absence of sugars.
`Whcrc compounds containing free u-diol groups induce instability of
`HMF hy acetalisation of the aldehyde group, compounds with isolated
`hydroxyls generally form stable HMF ethers with the hydroxymethyl
`qroup. If no solvent hydroxyls are available. HMF can easily form the
`Gi-HMF-ether.
`The solvent system also influences the formation and appearance of
`polymeric materials and humins. Non-aqueous polar solvents generally
`give no problems with solid humin formation.
`When working in aqueous systems the solvation of the acid protons
`highly diminishes the catalytic aciticity. Changing to partly anhydrous
`\ystemc the acids are generally much more able to protonate the
`rcactants. Fructose dehydration ist speeded up at least 10 times going
`from water to an aqueous mixture with 70% polyethyleneglycol. In
`spite of this effect on the catalytic activity, the hydrolysis rate of HMF
`decrease\ upon reducing the water content [14. 171.
`
`3 Manufacturing Processes
`
`The first preparations of HMF used sucrose as feedstock [26,
`30. 31.37],, and while the glucose part stayed largely unconver-
`ted, molar yields of only 25% (50% based on the fructose part)
`were obtained. Yields could be improved using high temperatu-
`res IS3.741. aqueous solvent mixtures [35] or aluminium sulfate
`catalyst [54. 551, but especially by continuous removal of HMF
`from the reaction mixture by extraction [32]. Methyl isobutyl
`ketone (MIBK) appeared to be an ideal extraction solvent and
`has been used in many studies.
`This section will focus on procedures which may have good
`prospects for economical HMF production. For that reason we
`will not treat anyfurther the procedures using “special” cata-
`lysts as mentioned in Table 1, but only those using cheap acids
`or ion-exchangers. The use of glucose or fructose as a feedstock
`demands different conditions and will be treated separately.
`The use of sucrose resulting in an equimolar mixture of glucose
`
`Table 2.
`Processes Starting from Fructose
`
`and fructose will probably never lead to interesting yields and
`will therefore not be treated further. This process section will be
`concluded by giving some details on work-up procedures.
`
`3.1 Processes starting from fructose
`The best raw material for HMF manufacture is fructose,
`crystalline fructose. fructose rich syrups or inulin hydrolysate.
`The presence of glucose and other carbohydrate compounds as
`well as protein impurities will lower the yield.
`Processes can be subdivided into aqueous and non- or mixed-
`aqueous systems, into those using simultaneous product remo-
`val or not and into those using homogeneous acids or acid ion-
`exchangers. Furthermore processes can be carried out batch-
`wise or in a continuous mode. In order to obtain reasonable
`yields, in aqucous systems simultaneous product removal is
`always necessary.
`In Table 2 the most interesting processes are presented. They
`will be shortly commented on in the following sections. An
`important parameter is the HMF load in the final solution which
`will be given as an average weight percentage.
`
`3.1.1 Aqueous systems, homogeneous acid
`Using aqueous sucrose solutions, Cope [32] introduced the
`method of simultaneous extraction with MIBK, in order to
`increase HMF yield. With a similar method Kirster et al. [lY]
`obtained selectivities up to 75% at reasonable HMF concentra-
`tions, 10% in MIBK, using a continuously operated stirred tank
`reactor. The aqueous phase could be further extracted and
`recycled with the unreacted fructose.
`The patent of Rapp [43] gives the only procedure solely using
`water which obviously is an advantage. The reaction mixture is
`separated on calcium loaded ion-exchanger columns. Unreac-
`ted fructose can be reused. The HMF-fraction is pure enough to
`be concentrated and crystallized. Disadvantages are the high
`dilution with water resulting in 1% aqueous HMF solution and
`a low selectivity of 55%.
`
`3.1.2 Aqueous systems, ion-exchangers
`
`The Toulouse group together with Roquette Frkres [23. 67,691
`did extensive studies using ion-exchangers and continuous
`
`Solvent
`
`Catalyst
`
`“C
`
`Time
`
`Batch
`continuous
`
`Yield**
`selectivity
`
`Remarks“
`
`Ref.
`
`H: 0
`H: 0
`H20
`H2O
`HzO
`DMSO
`DMSO
`DMSO
`DMSO
`H20/butanol
`H20/dioxane
`THFAigl ycolethers
`PG-600
`acetone
`
`H;POd
`HzSOd
`I.E.
`I.E.
`I.E.
`none
`none
`I.E.
`I.E.
`H2SOI
`HCI
`misc .
`HCI
`HzSOi
`
`170-220
`140
`78-88
`90
`90
`1 so
`I00
`60- 130
`76
`170
`180
`boiling point
`120 - 200
`210
`
`1 - 10 min
`2 h
`8-24 h
`15 h
`48 h
`2 h
`16 11
`2-8 h
`5 h
`8 min
`4 min
`2.5 s
`1-10 min
`28 s
`
`C
`b
`blc
`b
`b
`b
`b
`b/c
`
`C
`b
`b
`b
`C
`C
`
`76 ( 5 )
`s5 (s)
`78 (s)
`“YO”
`70 (s)
`92
`100
`80-96
`97
`68
`69 ( 5 )
`“82“
`65
`81
`
`A
`B
`A
`A
`C
`
`A
`
`I9
`43
`23. 67. 69
`72
`1
`33
`71
`66, 70
`68
`35
`38
`42
`20
`I07
`
`A : simultaneous extraction with MIBK B: fractionation of reaction mixture using coluntn chromatography C: continuous HMF removal by
`selective adsorption.
`** (s) selectivity, .. ” method of analysis. e.g. UV-absorption, questionable.
`
`ctarch/stiirke 42 (1990) Nr. 8. S. 314-321
`
`317
`
`Petitioners' Exhibit 1018, Page 4 of 8
`
`

`
`extraction with MIBK. Disadvantages are rather low yields,
`low catalyst activity and a high dilution with MIBK resulting in
`2% solutions. El Hajj et al. [72], trying out a similar procedure,
`reported a high selectivity (90%).
`Recent work in Delft has been carried out by Vinke et al. [ 11 on
`a process with simultaneous product removal by selective
`adsorption on activated carbon. The carbon can adsorb as much
`as 0.3 g HMF/g in equilibrium with a 0.5% aqueous HMF
`solution. Up to now a selectivity of 70% has been obtained.
`HMF could be removed from the carbon by extraction with
`ethanol. Selective adsorption would especially be interesting if
`good recovery without excessive dilution is garanteed.
`
`3.1.3 Systems using dimethyl sulfoxide (DMSO)
`
`The use of DMSO for the dehydration reaction is especially
`known from the work at the Noguchi Institute [66]. High HMF
`yields are reported using ion-exchanger catalysts. Brown et al.
`[71] and later Museau et al. [33] mentioned almost quantitative
`yields of HMF by just heating fructose in DMSO solution
`without any catalyst. Gaset et al. [68] studied the combination
`DMSO as a solvent, ion-exchanger as the catalyst and simul-
`tancous extraction with MIBK. These investigators reported a
`yield of 97% but the final MIBK solution had only a 2% HMF
`content and contained all the DMSO used (over 10% of total
`weight).
`Main disadvantage of processes using DMSO is that this solvent
`is difficult to separate from HMF. (See also work-up proce-
`dures).
`
`3.1.4 Other organic-solvent systems
`
`Good examples of the early work, demonstrating the beneficial
`effect of the addition of organic solvents to aqueous fructose
`solutions, are given by Peniston [35] adding butanol and by
`Hales et a1 [38] adding dioxane. Similar trials were repeated
`recently by Stenzenberger et al. [49, SO]. Smythe and Moye [42]
`tested a range of anhydrous solvents for the dehydration
`reaction, especially solvents containing the grouping R-O-C-C-
`OH e.g. tetrahydrofurfuryl alcohol and mono-methyl ethers of
`(di-)ethyleneglycol. They demonstrated that reasonable yields
`(up to 80% according to UV-absorption) could be obtained,
`HMF-ethers being formed as by-products. Kuster [20] tested
`polyethyleneglycol-600, which can be used as a flow agent in the
`vacuum destillation of HMF [76], as a solvent for the dehydra-
`tion reaction. He found that a 1/1 w/w mixture of fructose and
`PG-600 became homogeneous on adding a small amount of acid
`and heating for a few minutes at 85°C. While fructose itself
`hardly dissolves in PG-600 it apparently did so by the formation
`of soluble dianhydrides (see Fig. 3). On passing this mixture
`through a tube reactor reasonable HMF yields are obtained at
`high temperatures and short reaction times. Also some ethers
`are formed.
`Recent work in our laboratory [ 1071 has shown that it is also
`possible t o use acetone as a solvent and a crude mixture of
`fructoseacetonides as the feedstock for a continuous high
`pressure tube reactor. However, up to now high yields only
`have been obtained at relativily low concentration (1-2%).
`
`3.2 Processes starting from glucose
`
`To valorize ligno-cellulosic waste materials, it often has been
`tried to use these materials for the production of HMF.
`However, due to the low enolization rate of glucose, low yields
`are obtained. Therefore, these processes will be given a short
`treatment only. An extensive kinetic study on glucose dehydra-
`tion has been carried out by McKibbins et al. [ l l , 121. In the
`
`318
`
`older patents several data can be found on the use of glucose.
`Aluminium salts were used as catalyst in wood hydrolysate
`conversion by Bolcs [S2] and Garber and Jones [53,74], H2S04,
`short reaction times and temperatures up to 290°C in wood
`chips and glucose conversion by Snyder [36] and Tokarea and
`Sharkoa [77], while Hales [38] used mixed-aqueous solvents.
`Sometimes yields up to 80% have been mentioned but in view
`of later results, yields probably will never have been higher than
`50%.
`With the purpose to increase yields, acid-base mixtures have
`been tried as the cata1:yst; the base supposed to catalyze the
`enolization or intermediate formation of fructose. Smith [47]
`used ammonium sulfate or sulfite and Mednick [48] a mixture of
`pyridinium and phosphoric acid, the latter procedure being
`repeated recently by St'enzenberger et al. [49, SO]. The highest
`yield obtained was 50%. The combined action of amines and
`acids results in complicated product mixtures containing furans,
`together with phenols, pyrroles and pyridines [78, 791.
`Recent studies on H M F production from cellulosics by combi-
`ned hydrolysis and dehydration are given by Dadgar and Foutch
`[SO] using MIBK-extraction, Carves [81, 831 using mixed
`aqueous solvent systems and Durand-Pinchard [84] using a
`mixlure of solids and ,acidified MIBK. Again yields did not
`exceed SO%. Szmant and Chundury [64] dehydrated starch and
`glucose using DMSO and BF3-catalyst. Simkouic et al. [34]
`obtained a 31% yield in supercritical water.
`
`3.3 Work-up proceldures
`In order to make proper cost estimates for HMF not only data
`must be available on reaction conditions and yields but also full
`details should be known about how to isolate HMF in a more or
`less pure form.
`To obtain pure H M F from a reaction mixture the following
`procedure has been gi,ven by Middendorp [30] and later by
`Haworth and Jones [26]:
`- filtration of humin, -- neutralization with CaC03,
`- addition of Pb (OPLC)~ and filtration, -extraction of the
`filtrate with EtOAc, -drying the extract with Na2S04,
`- evaporation of EtOAc and - high vacuum distillation of the
`dark brown syrup obtained.
`The high-vacuum distillation has been the subject of two
`patents. For a good distillation yield the feed should be free of
`acid and moisture, be degassed and have a minimum contact
`time at high temperature [8S]. High-vacuum film evaporation is
`most suitable and a non-volatile flowing agent, e.g. Polyethyle-
`neglycol-600, should be added for proper handling [76]. Reco-
`veries of over 90% can be realized. Using vacuum distillation it
`is also possible to separate HAF from HMF, HAF being slightly
`more volatile, and to further purify a crude distillate to over
`99% purity [37].
`While distillation is the oldest procedure in use, chromatogra-
`phic methods are the latest development. Many systems com-
`mon in laboratory practice can be applied, e.g. fractionation on
`silica or alumina columns, but the materials used are expensive
`and probably difficult to regenerate. However, the use of ion-
`exchangers and activated carbon, common techniques in sugar
`industry practice, offer some interesting possibilities. Fractio-
`nation of reaction mixtures is possible using ion-exchanger
`columns in the Ca-form, yielding HMF fractions pure enough
`for direct crystallization [43]. Also selective adsorption/desorp-
`tion on activated carbon could offer good possibilities [l]. The
`generally high dilution of the resulting HMF solutions is a
`disadvantage, while up to now it is not yet clear if the reusability
`of the materials suffices.
`The procedures given sofar all have been in use to work up
`aqueous reaction mixtures. While volatile solvents easily can be
`removed by distillation, the use of high-boiling solvents posses-
`
`starchktarke 42 (1990) Nr. X, S. 314-321
`
`Petitioners' Exhibit 1018, Page 5 of 8
`
`

`
`ses some special problems, e.g. it is difficult to use vacuum
`distillation to separate HMF from DMSO [71. 721. As yet it is
`impossible to make proper cost evaluations because too little i s
`known about efficient work-up procedures for such mixtures.
`Final purification of crude HMF can be done by vacuum
`distillation, by fractional melting and washing, and by recrystal-
`lization from miscellaneous solvents e.g. water, diethyl ether,
`petroleum ether [37, 431.
`4 In situ Formed Derivatives
`
`HMF is generally rather unstable under the conditions of its
`formation. It often also is not used as such but in a derivatised
`form. Therefore. it is equally well interesting to consider
`manufacture of HMF-derivatives. especially if those derivatives
`can easily by formed in sitic. are more stable, can easier be
`worked up or are even more useful compounds than HMF
`itself.
`We will shortly deal with the following derivatives: levulinic
`acid. HMF-ethers, esters and halogen derivatives, and also with
`in situ oxidation and hydrogenation. All the compounds men-
`tioned are often as useful as HMF itself.
`
`4.1 Levulinic acid
`
`Carrying the acid catalyzed dehydration in aqueous systems to
`completion results in an equimolar mixture of levulinic and
`formic acid along with a certain amount of insoluble humins.
`The kinetics of HMF hydrolysis have been studied [9, 10, 141.
`Because levulinic acid is a stable product, yields up to 70% can
`be obtained from fructose as well as from glucose if not too high
`concentrations are used. An extensive kinetic study on the
`formation of levulinic acid from glucose has been given by
`I2.lcKihhen.s et al. [ 1 1. 121 and most recent mechanistic studies
`by Horrut et al. [86, 871. Sclzrnufnagel and Ruse [65] used ion-
`exchangers as catalyst and sucrose as the raw material: Jow et
`al. [73] obtained a 67% yield from molten fructose using a
`zeolite catalyst. Because of its stability, levulinic acid can also
`be prepared from hexose containing wastes. Reviews on levuli-
`nic acid have been written by Leonard (1956) [88]. Kztano et al.
`( 1975) [89] and Thornm and Barile (1984) [90, 911. The latter
`authors mention thc use of levulinic acid derivatives as liquid
`fuel extenders.
`
`4.2 HMF ethers
`
`In anhydrous systems HMF forms the di-HMF-ether or in the
`presence of alcohols 5-alkoxymethylfurfurals [40, 921. The
`ethers are more stable than HMF [71] and can often more easily
`be isolated, the lower alkyl ethers by distillation, the di-HMF-
`ether by crystallization. Early data on di-HMF-ether are given
`by Midtieizdorp [30] and Aso [93]. High yields of this compound
`also can be obtained from DMSO solutions [33,94]. Garces [83]
`obtained a low yield of methoxymethylfurfural. upon heating
`starch. cellulose or liqnocelluloses with acidic methanol.
`
`4.3 HMF esters
`
`HMF can easily be esterified in sitrr and, although esters are less
`stable than ethers. doing so can be very helpful in the isolation
`and purification. I n situ procedures in DMSO are given in two
`patents [70. 951.
`
`4.4 Halomethylfurfurals
`
`The halomethylfurfurals are most popular amoungst the HMF-
`derivatives, probably because of their usefulness in further
`
`organic synthesis. If the dehydration reaction is carried out in
`non- or mixed-aqueous systems containing at least an equimo-
`lar amount of transferable halogen, the compounds are formed
`in high yield. It has even been suggested to prepare HMF by
`hydrolysis of the chloromethyl compound [96]. Early prepara-
`tions are given by Huworth and Jones [26]. More recently much
`work has been done by Hamada et al. [97. 981. These authors
`obtained 5-chloromethylfurfural by heating a mixture of
`hydrochloric acid, a surface active agent and aromatic or
`halogenated solvents with sugars for 1 h at 70 to 80°C applying
`high speed stirring. The reported yields. as a brown syrup, were
`60% from glucose, 62% from sucrose and 77% from fructose.
`Using MgCI2.6H2O (one equivalent per mol o