FULL PAPERS
`
`Synthesis of 2,5-Diformylfuran and Furan-2,5-
`Dicarboxylic Acid by Catalytic Air-Oxidation
`of 5-Hydroxymethylfurfural. Unexpectedly Selective
`Aerobic Oxidation of Benzyl Alcohol
`to Benzaldehyde with Metal/Bromide Catalysts**
`
`Walt Partenheimer,* Vladimir V. Grushin
`
`Central Research and Development, E. I. DuPont de Nemours & Co., Inc., Experimental Station, Wilmington, Delaware
`19880–0328, USA
`Fax: +1 3 02-6 95–83 47; e-mail: Walter.Partenheimer@usa.dupont.com
`
`Received August 3, 2000; Accepted October 16, 2000
`
`Abstract: The alcohol group of hydroxymethylfur-
`fural (compound 1, HMF) is preferentially oxidized
`by dioxygen and metal/bromide catalysts [Co/Mn/
`Br, Co/Mn/Zr/Br; Co/Mn=Br/(Co+Mn) = 1.0 mol/
`mol]
`to form the dialdehyde, 2,5-diformylfuran
`(compound 2, DFF) in 57% isolated yield. HMF can
`be also oxidized, via a network of identified inter-
`mediates, to the highly insoluble 2,5-furandicar-
`boxylic acid (compound 5, FDA) in 60% yield. For
`
`Keywords: cobalt; di-
`oxygen; green chem-
`istry;
`homogeneous
`catalysis; hydroxyme-
`thylfurfural; oxidation
`
`comparison, benzyl alco-
`hol gives benzaldehyde in
`80% using the same cata-
`lyst
`system. Over-oxida-
`tion (to CO2) of HMF is
`much higher than that of the benzyl alcohol but can
`be greatly reduced by increasing catalyst concentra-
`tion.
`
`Introduction
`
`At the current rate of consumption, proven crude oil
`reserves are estimated to last for less than four dec-
`ades.[1] Therefore, in recent years serious considera-
`tion has been given, in both academia and industry, to
`alternative feedstocks for the chemical industry of the
`future. The use of renewable resources, i. e., natu-
`rally occurring carbohydrates and oils produced by
`various plants, would result in the development of be-
`nign, environmentally friendly processes,
`the so-
`called green chemistry.[2]
`5-Hydroxymethylfurfural (HMF; compound 1) is
`one of the few individual organic compounds that
`can be prepared directly from various carbohydrates
`in up to 98% yield. While the best yields of HMF have
`been obtained from fructose, other abundant, low-
`cost mono-, di-, and polysaccharides can be used,
`such as glucose, sucrose, and starch.[3]
`Selective oxidation reactions of HMF are presently
`viewed as attractive routes to 2,5-furandicarboxylic
`
`** Contribution No. 8095
`
`acid (FDA) and/or 2,5-diformylfuran (DFF; com-
`pound 2), monomers for furan-containing polymers
`and materials with special properties.[4] While a vari-
`ety of oxidants have been used for oxidation of HMF
`to 2,5-furandicarboxylic acid and DFF, only few re-
`ports describe catalytic oxidations of HMF with oxy-
`gen or air, the most economical oxidants. Thus, HMF
`has been oxidized with O2 to 2,5-furandicarboxylic
`acid in the presence of heterogeneous Pt catalysts
`with stoichiometric amounts of alkali[5,6] and to DFF
`with TEMPO radicals[7,8] or supported vanadium cat-
`alysts.[9,10] Although homogeneously catalyzed oxida-
`tion reactions of alcohols have received much atten-
`tion in recent years,[11–18] no reports have appeared
`in the literature, describing the oxidation of HMF with
`O2 and soluble metal complex catalysts.
`In this paper, we report the first examples of aero-
`bic HMF oxidation reactions, catalyzed with homoge-
`neous metal/bromide systems. The easily prepared,
`low-cost metal/bromide catalysts, the most common
`being a mixture of Co/Mn/Br, are widely used for the
`selective and efficient autoxidation reactions of hy-
`drocarbons,[19] e. g., the large scale industrial synth-
`esis of terephthalic, isophthalic, and trimellitic acids
`
`102
`
`(cid:211) WILEY-VCH Verlag GmbH, 69451 Weinheim, Germany, 2001
`
`1615-4150/01/34301-102–111 $ 17.50-.50/0
`
`Adv. Synth. Catal. 2001, 343, No. 1
`
`Petitioners' Exhibit 1003, Page 1 of 10
`
`

`
`from p-xylene, m-xylene, and pseudocumene respec-
`tively.[1,19] Surprisingly little is known, however,
`about oxidation of alcohols using the metal/bromide
`catalysts.[19] In this work, we found that, depending
`on reaction conditions, hydromethylfurfural can be
`oxidized to DFF or 2,5-furandicarboxylic acid with
`unexpectedly high selectivity. Furthermore, the se-
`lective formation of DFF in the metal bromide-cata-
`lyzed oxidation of HMF prompted us to study the oxi-
`dation of benzyl alcohol under similar conditions.
`Remarkably, it was found that under controlled con-
`ditions this oxidation can afford benzaldehyde in high
`yield.
`
`Results
`
`Products Formed
`GC/MS studies were performed on two selected sam-
`ples during HMF autoxidation at 70 bar, which are
`consistent with the products and pathways given on
`Figure 1. In addition, the usual products from the
`autoxidation of acetic acid were observed, i. e., formic
`acid, acetoxyacetic acid, glycolic acid, maleic acid, fu-
`maric acid, succinic acid, and bromosuccinic acid in
`trace amounts. A side reaction is the esterification of
`the alcohols to form the more oxidatively stable acet-
`ate, see compounds 6 and 7 in Figure 1 and benzyl
`acetate in Figure 2. DFF and FDA have been isolated
`and characterized by elemental analysis and NMR
`spectra. The 2-carboxy-5-formylfuran was identified
`and quantified by 1H NMR spectroscopy of isolated so-
`lid samples that were either 2,5-furandicarboxylic
`acid or 2,5-furandicarboxylic acid/2-carboxy-5-for-
`mylfuran mixtures. The oxidation of benzyl alcohol
`gives the expected benzaldehyde, benzyl acetate,
`and benzoic acid products (see Figure 2).
`
`Figure 1. Products from the autoxidation of hydroxymethyl-
`furfural
`
`FULL PAPERS
`
`Figure 2. Products during autoxidation of benzyl alcohol
`
`Formation of Diformylfuran from Hydromethyl-
`furfural and Benzaldehyde from Benzyl Alcohol at
`Atmospheric Pressure (Table 1)
`In experiments 1 and 5, the attempt to initiate the re-
`action at the lower temperature failed, hence the
`temperature was raised to the higher given value.
`The formation of the reaction products, as deter-
`mined by GC and LC, from HMF and benzyl alcohol
`is illustrated in Figures 3 and 4. Maximum observed
`yield of the aldehydes is 57% for DFF and 80% for
`benzaldehyde. Maximum aldehyde yields occur as
`the conversion of the alcohol approaches 100%. The
`selectivity decreases as the conversion of the alcohol
`increases with the values for HMF (51–90%) being
`lower than for benzyl alcohol (80–93%), see Figure 5.
`Doubling the catalyst concentration during the oxy-
`genation of HMF (i) increases the reaction rate by a
`factor of 2.1, (ii) increases the yield and selectivity to
`DFF by 5 and 10%, respectively, and (iii) decreases
`the ‘overoxidation’ to CO and CO2 by a factor of 4
`(see experiments 2, 3, and 4 in Table 1 and Figure 6).
`Further increase in catalyst concentration does not
`further improve DFF yield (see experiments 4, 5, and
`6). For benzyl alcohol, the maximum benzaldehyde
`yield of 80% occurs at a conversion of 85% with only
`4.6% benzoic acid and 3.8% benzyl acetate formed
`(Table 1, experiment 7). Based on their rates of disap-
`pearance, benzyl alcohol is 2.0 times more reactive
`than HMF (see examples 4, 7, and 8 in Table 8).
`The rate of disappearance of the alcohol and the
`rate of disappearance of the aromatic aldehyde are
`consistent with first order kinetics and the rate con-
`stants are given on Table 1. The rate of disappearance
`of HMF is 8.1 times faster than the rate of disappear-
`ance of 2-carboxy-5-formylfuran, suggesting that the
`dialdehyde is quite stable, as seen from the kinetic
`data on Table 1. This is consistent with the subse-
`quent work-up of the reaction mixture and isolation
`of DFF in a yield close to that previously determined
`by GC. The benzyl alcohol, however, reacts to form
`benzaldehyde faster by only a factor of 1.1 than ben-
`zaldehyde reacting to benzoic acid (Table 1). Re-
`
`Adv. Synth. Catal. 2001, 343, 102–111
`
`103
`
`Petitioners' Exhibit 1003, Page 2 of 10
`
`

`
`asc.wiley-vch.de
`
`Table 1. Oxygenation of hydroxymethylfurfural and benzyl alcohol at ambient atmospheric pressure
`
`2
`
`HMF
`
`3
`
`HMF
`
`75
`0.794
`6.6
`0.15
`9.68(0.18)
`[0.997]
`119
`1.22(0.34)
`[0.862]
`450
`51
`92
`55
`5.9
`7.4
`
`75
`0.804
`6.6
`0.15
`8.12(0.61)
`[0.972]
`142
`–
`
`642
`50
`95
`53
`7.5
`8.5
`
`4
`
`HMF
`
`75
`0.797
`13.5
`0.15
`16.6(1.4)
`[0.999]
`69
`
`–
`
`5
`
`HMF
`
`50 then 75
`0.796
`26.8
`0.15
`10.8(0.5)
`[0.992]
`106
`–
`
`6
`
`HMF
`
`75
`0.806
`27.3
`0.15
`15.1(0.5)
`[0.994]
`76.5
`
`–
`
`310
`57
`91
`63
`7.2
`2.1
`
`550
`51
`95
`54
`6.1
`1.8
`
`430
`52
`97
`54
`5.7
`2.6
`
`7
`
`benzyl
`alcohol
`75
`0.793
`13.4
`0.15
`40.4(3.9)
`[0.988]
`28.5
`30(3)
`[0.943]
`100
`80
`85
`93
`5.6
`0.05
`
`Exp.
`
`Temp, (cid:176)C
`Reagent, M
`Co, mM
`Zr, mM
`Rate, s–1 [a]
`
`1
`
`HMF
`
`50 then 95
`0.725
`2.6
`0.0
`
`–
`
`Alcohol, half-life, min –
`Rate, s–1 [b]
`–
`
`414
`41
`98
`42
`8.4
`
`–
`
`[d]
`
`Time,min [c]
`Yield [c]
`Conv.,% [c]
`Select.,% [c]
`Acetate,%[c]
`Alcohol to COx
`[a] Rate of disappearance of aromatic alcohol · 105. Standard deviation in parenthesis ( ), correlation coefficient in brackets [ ].
`[b] Rate of disappearance of aromatic aldehyde.
`[c] When maximum alkylaromatic aldehyde is observed.
`[d] Loss of alcohol due to carbon monoxide and carbon dioxide formation. Assumes no COx formation from the solvent.
`
`Figure 3. Autoxidation of hydroxymethylfurfural at 75 (cid:176)C
`
`Figure 4. Autoxidation of benzyl alcohol at 75 (cid:176)C
`
`markably, the oxidation is catalyzed in such a way
`that essentially all the benzyl alcohol reacts first. The
`benzaldehyde formed starts to undergo further oxida-
`tion to benzoic acid only after the oxidation of the
`
`Figure 5. Benzaldehyde selectivity as function of catalyst
`concentration and type of alkylaromatic alcohol
`
`benzyl alcohol is close to completion, despite the fact
`that PhCH2OH and PhCHO exhibit very similar meas-
`ured reactivities in the same experiment.
`
`Formation of DFF from HMF at 70 Bar Air (Table 2)
`Experimental error, as determined by 5 replicate ex-
`periments, is given in entry 7 of Table 2. As can be
`seen, 50 and 75 (cid:176)C for 2 h are sufficient conditions for
`obtaining good yields of DFF, up to 63%. By compar-
`ing experiments 1 and 2, 3 and 4, 5 and 6, and 8 and
`9, one finds that increasing catalyst concentration
`leads to (i) increased activity as evidenced by higher
`conversions, (ii) higher selectivity for DFF (except
`for experiments 5 and 6), and (iii) higher yield. Com-
`paring experiments 1 and 3, 2 and 4, 5 and 8, and 6
`and 9, one finds that the Co/Mn/Zr/Br catalyst is
`
`104
`
`Adv. Synth. Catal. 2001, 343, 102–111
`
`Petitioners' Exhibit 1003, Page 3 of 10
`
`

`
`FULL PAPERS
`
`Formation of 2-Carboxy-5-formylfuran and 2,5-
`Furandicarboxylic Acid at 70 Bar (Table 3)
`The initial amount of HMF used was only 0.2–0.75 g
`and the yields are based on isolated and washed so-
`lids which were analyzed by NMR. When the tem-
`perature is increased from 75 to 100–125 (cid:176)C, precipi-
`tation of poorly soluble 2,5-furandicarboxylic acid
`commences. 2-Carboxy-5-formylfuran is also either
`fairly insoluble or is prone to co-crystallization with
`2,5-furandicarboxlic acid, which results in their co-
`precipitation. The yield increases with catalyst con-
`centration (Figure 7), with temperature (entries 1
`and 2 and 3 and 4 of Table 3), but not with the addition
`of Zr to the Co/Mn/Br catalyst (entries 1 and 3 and 2
`and 4). Extrapolation from Figure 7 suggests that the
`maximum obtainable 2,5-furandicarboxylic acid
`yield is about 70% using the Co/Mn/Zr/Br catalyst at
`the specified molar ratios of these elements. It is be-
`lieved that variation of the molar amounts of the Co,
`Mn, Zr, and Br could well improve the yield of 2,5-
`furandicarboxylic acid. Since the oxidation proceeds
`through three steps from HMF to 2,5-furandicar-
`boxylic acid (steps 1, 3, and 5 in Figure 1) and the re-
`activity of the HMF is probably higher than 2-car-
`boxy-5-formylfuran one would expect that staging
`the temperature would increase yield.[19] This was
`not observed however, since staging the temperature
`from an initial value of 50 (cid:176)C for 1 h and then 125 (cid:176)C
`for 2 h gave no better results than the oxygenation at
`125 (cid:176)C for 3 h (Figure 7).
`
`Figure 6. Carbon oxide selectivity as function of catalyst
`concentration and type of alkylaromatic alcohol
`
`more active, giving higher conversion, than Co/Mn/
`Br, with the only exception being experiments 2 and
`3 where the conversions are similar. The addition of
`zirconium not only affects conversion, but can also
`profoundly increase the selectivity (Table 2). This
`point is illustrated by experiments 1 and 3 where the
`addition of Zr results in a much higher yield of DFF
`(67 vs. 38%) at the same conversion of ca. 60%. Ex-
`periments 6 and 8, for which the conversions vary sig-
`nificantly, represent the only exception. Under com-
`parable conditions, the conversion increases with
`temperature, as expected.
`
`Table 2. Oxidation of hydroxymethylfurfural (HMF) to diformylfuran (DFF) at 70 bar air
`
`Exp.
`
`Catalyst
`
`[Co], mM
`
`HMF, M
`
`Temp, (cid:176)C
`
`Time, h
`
`HMF, conv. % DFF select. % DFF, yield %
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`Co/Mn/Br
`Co/Mn/Br
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`Co/Mn/Br
`Co/Mn/Br
`
`3.44
`6.82
`3.44
`6.82
`3.44
`6.82
`6.82
`3.44
`6.82
`
`0.375
`0.372
`0.375
`0.377
`0.375
`0.375
`1.12
`0.377
`0.377
`
`50
`50
`50
`50
`75
`75
`75
`75
`75
`
`2
`2
`2
`2
`2
`2
`2
`2
`2
`
`60.4
`69.2
`60.6
`61.7
`82.5
`99.7
`74.1(1.0)
`71
`92.2
`
`66.6
`65.3
`38.4
`54.6
`73.2
`61.6
`67.5(1.4)
`54.3
`68.3
`
`40.2
`45.2
`23.3
`33.7
`60.4
`61.4
`49.9(0.6)
`38.6
`63.0
`
`Table 3. Oxidation of hydromethylfurfural to 2-carboxy-4-formylfuran (CFF) and furan-2,5-dicarboxyfuran (FDA) at 70 bar
`air
`
`Exp.
`
`catalyst
`
`[Co], mM
`
`[HMF], M
`
`temp, C
`
`time, h
`
`CFF, mol %
`
`FDA, mol %
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`Co/Mn/Br
`Co/Mn/Br
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`Co/Mn/Br/Zr
`
`3.44
`3.44
`3.44
`3.44
`3.44
`6.82
`13.7
`20.5
`3.41
`6.82
`13.7
`20.5
`
`0.377
`0.371
`0.377
`0.374
`0.758
`0.753
`0.749
`0.755
`0.781
`0.774
`0.0753
`0.768
`
`100
`125
`100
`125
`50, 125
`50, 125
`50, 125
`50, 125
`125
`125
`125
`125
`
`2
`2
`2
`2
`1, 2
`1, 2
`1, 2
`1, 2
`3
`3
`3
`3
`
`3.1
`2.1
`4.1
`1.8
`1.6
`2.5
`0.0
`0.0, 0.0
`1.7
`0.0
`0.0
`0.0, 0.0
`
`18.7
`36.5
`29.7
`35.2
`28.3
`28.1
`55.4
`58.4, 63.1
`27.7
`41.6
`54.6
`60.9, 58.6
`
`Adv. Synth. Catal. 2001, 343, 102–111
`
`105
`
`Petitioners' Exhibit 1003, Page 4 of 10
`
`

`
`asc.wiley-vch.de
`
`Figure 7. Effect of catalyst concentration and temperature
`staging on FDA yield. See details in Table 3.
`
`Discussion
`
`General Considerations
`The reaction network in Figure 1 is consistent with
`the detailed studies of the oxidation reactions of
`many substituted methylaromatic species, aromatic
`alcohols, and benzaldehydes using metal/bromide
`catalysts.[19] The latter is thought to operate via a
`modified free radical chain mechanism (see be-
`low). The free radical chain mechanism gives the
`oxidizability of toluene, benzyl alcohol, and benzal-
`dehyde as 0.05, 0.85, and 290 respectively.[22] It is
`clear from these values that the steady state con-
`centration of benzaldehyde is expected to remain
`low in metal/bromide catalyzed systems. We find
`however that the oxygenation of HMF gives prefer-
`entially DFF rather than 5-(hydroxymethyl)furan-
`2-carboxylic acid (compare steps 1 and 2 on Fig-
`ure 1). Extending this work to benzyl alcohol gave
`even higher yields and selectivity to aldehydes. The
`kinetics of the Co/Br catalyzed oxygenation of ben-
`zyl alcohol has been reported[23] albeit without a
`comment on potentially high selectivities and
`yields of benzaldehyde.
`The advantages of the catalytic oxidation described
`herein is that the catalyst is composed of inexpensive,
`simple metal acetate salts and a source of ionic bro-
`mide (NaBr, HBr, etc.). The reaction times are within
`a few hours at easily accessible temperatures. The
`acetic acid solvent is inexpensive and nearly all alco-
`hols are highly soluble in it. Although acetoxylation of
`the alcohols with the acetic acid solvent does occur,
`this side-reaction results in only a 5–8% yield loss. It
`is noteworthy that there are other solvents available
`for metal/bromide catalyzed systems, which could
`potentially eliminate this problem.[19] Due to the high
`activity of the metal/bromide catalysts the aldehyde
`formed in high yield can undergo further oxidation.
`To obtain high yields of aromatic benzaldehydes from
`
`the corresponding aromatic alcohols the catalytic
`process should be carefully monitored, so that subse-
`quent oxidation of
`the aldehyde formed can be
`avoided.
`
`Structure of the Catalyst
`Addition of the simple acetate salts into acetic acid re-
`sults in a complex mixture which is only partially un-
`derstood. A brief synopsis based on available infor-
`mation follows. The structures of Co(II) and Mn(II)
`in acetic acid/water mixtures can be summarized by
`the equation:
`
`where the square brackets indicate the ligands in the
`inner coordination sphere. In acetic acid, the cationic
`metal species are largely associated, with the small
`quantities of the dissociated species existing as ion
`pairs[24]
`in both monomeric and dimeric forms
`(n = 1, 2).[25–27] Upon addition of water, equilibrium
`is established between various metal aquo acetic acid
`complexes. Using reported equilibrium constants[28]
`one can calculate the distribution of these complexes
`and demonstrate that these aquo/acetic acid metal
`species exist in 10% water/acetic acid mixtures.[29]
`The weakly bound AcOH ligand (5.9 kcal/mol) is la-
`bile, exchanging with water and acetic acid instanta-
`neously at room temperature.[30] The addition of per-
`acids, peroxy radicals, oxygenated intermediates, etc.
`to a mixture of Co(II)/Mn(II) in acetic acid may there-
`fore result in fast ligand exchange to form the transi-
`ent catalytic species. Addition of hydrogen bromide to
`Co(II) or Mn(II) or a Co(II)/Mn(II) mixture in anhy-
`drous acetic acid results in the majority of the brom-
`ide being coordinated to the metal. However, addition
`of water (5% or greater) results in almost complete
`ionization of the M–Br bond.[29] In the presence of
`water the addition of bromide results in outer-sphere
`ligand exchange processes, as shown in the equation
`below.
`
`It is possible that the ion-paired bromide forms hy-
`drogen bonds to the aquo ligands (Figure 8, struc-
`ture b). The lability of the ligands, the known di-
`meric
`structure
`of
`Co(II)
`acetate,
`and
`polynuclearity of Zr(IV) in water suggests that poly-
`nuclear Co(II)–Mn(II) and Co(II)–Zr(IV)–Mn(II)
`may exist (Figure 8). Such mixed-metal polymeric
`species have been isolated from acetic acid.[31] Re-
`cent observations[32] suggest that acetic acid/water
`solutions may be more complex, containing water-
`rich microphases. It is proposed that Co(III) aquo
`
`106
`
`Adv. Synth. Catal. 2001, 343, 102–111
`
`Petitioners' Exhibit 1003, Page 5 of 10
`
`

`
`FULL PAPERS
`
`Figure 8. Suggested structures for Co,Mn,Br mixtures in 5% H2O/HOAc. M’=M’’=Co(II), Co(III), Mn(II), Mn(III)
`
`acetate and Co(III) aquo acetate bromide have
`structures similar to those shown in Figure 8 (a and
`b, respectively).
`
`Theory and Models of Metal/Bromide Catalysis
`Different aspects of metal/bromide catalysis have
`been discussed with emphasis on high reactiv-
`ity,[19,27,33–36] superior selectivity over a broad tem-
`perature range,[19,34] and the synergy and antagonism
`of the metals.[34,36,37] Important new observations in
`this field have been recently reported.[38] Kinetic stu-
`dies suggest that oxidation of Co(II) by peracids (to
`give carboxylic acids)[27,37] and peroxy radicals (to
`give peroxides)[33] initiates the series of reactions
`shown in Figure 9. The rapidity of the peroxide reac-
`tions with Co, followed by the subsequent redox cas-
`cade leading to the generation of the selective bro-
`mide atom or the dibromide radical[38] accounts for
`the properties of these catalysts. Initiation of the hy-
`drocarbon RH to the radical R. via the Co/Mn/Br re-
`dox cascade is faster than Co/Br, which in turn is fas-
`ter than Co. Co(III)a, Co(III)s, Co(III)c are different
`Co(III) compounds, with structures suggested in Fig-
`ure 8 possessing different reactivity.[26]
`
`Rationale for High Yields of Aromatic Benzalde-
`hydes from Aromatic Alcohols
`High yields of benzaldehydes are observed despite
`the fact that benzyl alcohol and benzaldehyde react
`at nearly the same rate in the metal/bromide cata-
`lyzed system. In particular, for benzyl alcohol oxida-
`tion the benzoic acid yield remains under 1% at 60%
`conversion and is only 4.6% when the maximum
`yield of benzaldehyde is obtained (80%) at 85% con-
`version. These observations are certainly unexpected
`and hence merit a comment. There are at least three
`factors which would account for the clean and selec-
`tive formation of benzaldehyde under the conditions
`employed.
`There may be a rapid, preferential bonding of the
`aromatic alcohol with either or both Co(II) and
`Mn(II), which initiates their oxidation in preference
`to the benzaldehyde. The formation of benzyl alcohol
`metal species might occur via replacement of the la-
`bile, weak acetic acid or aquo ligands or hydrogen
`bonding to the coordinated AcOH or water molecules
`(similar to the bromide in Fig. 9). Once all of the aro-
`matic alcohol has been oxidized the catalyst initiates
`the benzaldehyde oxidation.
`There is experimental evidence that acetic acid re-
`tards autoxidation by hydrogen bonding to the peroxy
`
`Adv. Synth. Catal. 2001, 343, 102–111
`
`107
`
`Petitioners' Exhibit 1003, Page 6 of 10
`
`

`
`asc.wiley-vch.de
`
`Figure 9. Summary of chemistry of Co, Co/Br, and Co/Mn/Br autoxidation catalysts. Half-lifes are at 60 (cid:176)C in 10% water/
`acetic acid
`
`Figure 10. Suggested structure of hydrogen bonded acetic
`acid to benzyl alcohol intermediates
`
`radicals at the a-position (Figure 10).[39] It is concei-
`vable that the acetic acid is more effective at inhibit-
`ing the carbonyl functionality than the benzylic hy-
`droxy group. This explanation seems less likely
`because the oxidation of the aromatic alcohol and
`benzaldehyde would have been already initiated,
`and even if they proceed further at different rates, sig-
`nificant amounts of aromatic acids should be formed
`from the benzaldehyde. However, the actual amount
`of benzoic acid formed from benzaldehyde does not
`exceed 1–5% (see above).
`It is possible that one or more coordination com-
`pounds in the reaction mixture specifically inhibits
`the benzaldehyde oxidation or promotes the aromatic
`alcohol reaction. We have found that sodium bromide
`strongly inhibits the oxidation of benzaldehyde,
`whereas Co enhances this reaction. The rate of oxy-
`gen uptake is 12.0 mL/min without catalyst, 0.3 mL/
`min with sodium bromide and 13.2 mL/min with
`Co(II) acetate at 80 (cid:176)C in acetic acid.[40]
`Obviously, further experimentation is required to
`confirm these conjectures.
`
`Effect of Zirconium on Selectivity
`In the high pressure experiments, we found that the
`selectivity to DFF increased in the presence of Zr in
`the Co/Mn/Br catalyst. The effect of Zr on cobalt me-
`tal/bromide catalysts is generally thought to increase
`its activity[19,35,41] and Zr does not affect the rate de-
`
`termining step because the q values in a Hammett
`plot of Co/Mn/Zr/Br and Co/Mn/Br are the same
`within experimental error.[35] However, there is a
`brief report that addition of Zr to a Co/Br catalyst de-
`creases the rate of benzyl alcohol formation and in-
`creases the rate of benzaldehyde formation.[42] We
`have duplicated this effect with a Co/Mn/Zr/Br cata-
`lyst in 10% water/acetic acid at 95 (cid:176)C with the oxyge-
`nation of p-xylene. The increase in the rate of reac-
`tion is proportional to the Zr concentration which in
`turn is directly proportional to the observed reduction
`of the benzyl alcohol/benzaldehyde ratio.[40] One pos-
`sibility is that the rate of alcohol oxidation (Figure 2,
`step 3) is increasing relative to step 2. It is also possi-
`ble that step 1 is becoming more important than
`step 2. We suggest that the new catalytic species form
`when Zr is added to Co/Mn/Br (see Figure 8), which
`goes through similar redox cascades as shown in Fig-
`ure 9, changing some of the relative rates presented
`in Figure 2. More details on the function of Zr are
`available.[19]
`
`Overoxidation to COx
`A weakness of most published work on oxygenations
`using air as the primary oxidant is the lack of meas-
`urement of COx formation during the reaction. The
`potential for the formation of the highly reactive per-
`oxides and consequently peroxy, hydroxyl, etc. radi-
`cals always exists when mixtures of transition metals,
`dioxygen, hydrocarbons, and organic solvents are
`present and hence ‘overoxidation’ to COx nearly al-
`ways occurs.
`Much higher amounts of carbon monoxide and car-
`bon dioxide (COx; x = 1, 2) form during HMF oxyge-
`nations as compared with the benzyl alcohol oxida-
`tion under similar conditions (Table 1, Figure 6).
`Tracer studies indicate that the origin of COx is from
`both the aromatic substrates and the acetic acid sol-
`vent.[43] The formation of COx in the HMF reaction is
`
`108
`
`Adv. Synth. Catal. 2001, 343, 102–111
`
`Petitioners' Exhibit 1003, Page 7 of 10
`
`

`
`FULL PAPERS
`
`Figure 11. Important pathways in hydroxymethylfurfural oxidation
`
`apparently not predominately from the solvent be-
`cause only trace amounts of acetoxyacetic acid, an
`oxidatively stable by-product formed from the acetic
`acid, is observed by GC. We determined the amount
`of COx formed by numerical integration and then as-
`sumed that 100% came from the total destruction of
`HMF. A significant yield loss of 1.8 to 8.5% is calcu-
`lated. This significant loss, as compared to the benzyl
`alcohol oxidation (ca. 0.05% lost to COx) is accounted
`for by at least two reasons.
`Decarbonylation. By-product formation during p-
`xylene[44] and alkylnaphthalene oxidation[19] is con-
`sistent with hydrogen atom abstraction from the alde-
`hyde followed by decarbonylation and eventual aro-
`matic ring loss via the formation of phenol
`(Figure 11). HMF is initially a di-functional molecule
`already containing a formyl functionality in contrast
`to benzyl alcohol. Hence higher HMF loss via this me-
`chanism is anticipated.
`Enhanced ring attack due to reduced resonance en-
`ergy. The resonance energy values for benzene,
`naphthalene, and furan are 36, 31, and 17 kcal/
`mol.[45] The difference between metal/bromide cata-
`lyzed alkylbenzene and alkylnaphthalene oxygena-
`tions has been discussed[19], [46] and is largely due to
`the enhanced ring bromination and enhanced peroxy
`radical ring attack that occurs in the naphthalene de-
`rivatives. This forms intermediates (e. g., phenols)
`which quickly undergo exhaustive oxidation to COx.
`Since the resonance energy of furan is even lower
`than naphthalene (31 vs. 17 kcal/mol), the higher rate
`of COx formation is not surprising.
`
`Effect of Catalyst Concentration on Activity and
`Overoxidation to COx
`We find that increasing catalyst concentration in-
`creases activity at the early stages, but then remains
`constant or decreases slightly (Table 1), consistent
`with previous observations.[19] Kinetic studies show a
`second order dependence of cobalt concentration for
`a Co/Br catalyst.[33]
`Remarkably, overoxidation to COx is suppressed at
`higher catalyst concentrations. This has been ob-
`served in metal/bromide catalyzed systems pre-
`viously.[47] The non-selective, thermal pathways are
`(i) decarbonylation in step 5 and subsequent by-pro-
`duct formation in step 6, (ii) peroxy radical attack on
`the furan ring in step 7, and (iii) thermal dissociation
`of the peroxide (step 9), leading to the carboxylate ra-
`dical and the highly reactive OH radical. Step 9 is fol-
`lowed by ring addition of the hydroxyl radical to furan
`(step 10), which will eventually lead to by-products
`including COx. The carboxylate radical can decar-
`boxylate (step 11), leading to the same products as in
`the decarbonylation process (step 12). As the catalyst
`concentration increases at least two selective, metal
`catalyzed pathways become increasingly important.
`At [Co] > 0.01 M, kinetic and chemiluminescence data
`provide evidence that the direct oxidation of Co(II) by
`peroxy radicals becomes important[33] (step 13). This
`reaction will increasingly supplant step 7 as the cata-
`lyst concentration increases, hence reducing ring at-
`tack and deaccelerating the COx formation. Because
`step 14 is 400,000 times faster than step 9, displaying
`a 18 kcal/mol lower activation energy barrier,[27] hy-
`droxyl radical formation and decarboxylation are
`
`Adv. Synth. Catal. 2001, 343, 102–111
`
`109
`
`Petitioners' Exhibit 1003, Page 8 of 10
`
`

`
`asc.wiley-vch.de
`
`greatly diminished as the cobalt concentration in-
`creases. The reason both selectivity and activity are
`often enhanced in metal/bromide systems is that re-
`actions 13 and 14 produce Co(III) which quickly goes
`through the redox cascade shown in Figure 9 to con-
`tinue to initiate the reaction.
`In the future, it is planned to extend the methodol-
`ogy described herein to substituted benzylic alcohols,
`aliphatic alcohols, and a variety of other alkylaro-
`matic systems such as naphthalene, pyrrole, and thio-
`phene derivatives.
`
`Experimental Section
`
`Aldrich cobalt(II) and Fluka manganese(II) acetate tetrahy-
`drates, Alfa cerium(III) acetate hydrate, EM Science sodium
`bromide, benzyl alcohol and acetic acid, Baker hydrobromic
`acid, Aldrich zirconium(IV) acetate and biphenyl, and Lan-
`caster hydromethylfurfural were used as received. Catalysts
`were prepared by dissolving the above compounds into
`acetic acid in the amounts specified on Table 1–3.
`
`Autoxidation at Ambient Atmospheric Pressure
`A glass cylindrical reactor, as previously described,[20] was
`used. Initial weight of acetic acid was 100 g, with Co/
`Mn = 1.0 mol/mol, Br/(Co+Mn) = 1.0 mol/mol in all cases.
`We found that HMF, but not benzylic alcohol, required addi-
`tion of 0.5–1.0 g of acetaldehyde to initiate the reaction at
`75 (cid:176)C. The rate of oxygen uptake was continually monitored
`by measuring the flow rate into the reactor and the concen-
`tration of dioxygen in the vent gases. The vent gases (O2, N2,
`CO, CO2) were measured using an automated GC system.
`Liquid samples were removed during the reaction and ana-
`lyzed via GC as soon as possible.
`
`Autoxidation at 70 Bar in Air
`These reactions were performed in a 20-mL cylindrical
`glass reactor. The samples were analyzed after removal
`from the reactor. Caution: The use of high pressures and the
`use of dioxygen/nitrogen mixtures is potentially explosive
`and dangerous. They should be performed only with ade-
`quate barriers for protection.
`The rate of dioxygen uptake, in mL/min, is given by the
`equation: R(O2) = F(20.9 – [O2]) where F = flow rate of air
`into the reactor, [O2] = concentration of oxygen in the exit
`gas stream and 20.9 is the concentration of dioxygen in air
`(in %). The carbon dioxide (x = 2) and carbon monoxide
`(x = 1) selectivity is defined by:
`SCOx = rate of formation of COx/rate of dioxygen reacted.
`SCOx = RCOx/Ro = F[COx]/(F(20.9 – Vo)) = [COx]/(20.9 – Vo)
`where [COx] is the vent carbon oxide concentration, ex-
`pressed as percent.
`DFF from example 5, Table 1, was isolated by evaporation
`of the solvent and vacuum sublimation of the residual solid
`(90 (cid:176)C at 10–50 millitorr). The sublimed material (5.2 g;
`51% in agreement with the GLC yield) was 95% pure DFF
`containing ca. 3–5% 5-acetoxymethylfurfural (NMR). The
`crude DFF was further purified by filtration of its dichloro-
`methane solution through silica, followed by partial eva-
`poration of the filtrate and precipitation with hexanes. 1H
`
`NMR (CDCl3, 20 (cid:176)C): (cid:14) = 7.4 (s, 2 H, furan H), 9.8 (s, 2 H,
`CHO); 13C NMR (CD2Cl2, 20 (cid:176)C): d = 120.4 (s, CH), 154.8 (s,
`qC), 179.7 (s, CHO); Mass spectrum: m/z = 124.
`2,5-Furandicarboxylic acid was isolated in the following
`manner. The solubility of 2,5-furandicarboxylic acid is
`6.6 · 10–4 g/g in 3% H2O/HOAc at room temperature. Hence
`when the reaction solutions are cooled to room temperature
`99% of the 2,5-furandicarboxylic acid precipitates. The so-
`lids after reaction were filtered, washed with acetic acid,
`then water, and air-dried. If insufficiently oxidized, the so-
`lids contained both 2-carboxy-5-formylfuran and 2,5-furan-
`dicarboxylic acid in varying amounts. All of the reported 2-
`carboxy-5-formylfuran and 2,5-furandicarboxylic acid yields
`are based on the precipitated solids only. The composition of
`the isolated solids containing 2-carboxy-5-formylfuran and
`2,5-furandicarboxylic acid in the solids were determined
`from their NMR spectra in DMSO.
`For 2-carboxy-5-formylfuran; 1H NMR (DMSO): d = 7.4 (s,
`1 H, furan CH), 7.7 (s, 1 H, furan CH), 9.8 (s, 1 H, CHO); 13C
`NMR (DMSO): d = 122.3 (s, CH), 153.2 (s, CH), 172.0 (s,
`COOH), 179.7 (s, CHO).
`For 2,5-furandicarboxylic acid; 1H NMR (DMSO): d = 7.3
`(s, 2 H, furan CH); 13C NMR (DMSO): d = 118.5 (s, CH), 148.1
`(s, C), 158.8 (s, COOH); anal. calcd. for C6H4O5, %: C, 46.16;
`H, 2.59; found for solids obtained in experiments 6, 7, 11, 12
`in Table 3, %: C, 45.93; 45.93; 45.45; 45.79; H, 2.57; 2.43; 2.44;
`2.43.
`
`Acknowledgments
`
`We thank Dr. James