Reactivity of metal catalysts in glucose-fructose conversion

DOI: 10.1002/chem.201402437

Full Paper

&

Transition Metals

Reactivity of Metal Catalysts in Glucose–Fructose Conversion

Claudia Loerbroks, Jeaphianne van Rijn, Marc-Philipp Ruby, Qiong Tong, Ferdi Schꢀth, and

Walter Thiel*^[a]

Abstract: A joint experimental and computational study on

the glucose–fructose conversion in water is reported. The re-

activity of different metal catalysts (CrCl₃, AlCl₃, CuCl₂, FeCl₃,

and MgCl₂) was analyzed. Experimentally, CrCl₃and AlCl₃

achieved the best glucose conversion rates, CuCl₂and FeCl₃

were only mediocre catalysts, and MgCl₂was inactive. To ex-

plain these differences in reactivity, DFT calculations were

performed for various metal complexes. The computed

mechanism consists of two proton transfers and a hydro-

gen-atom transfer; the latter was the rate-determining step

for all catalysts. The computational results were consistent

with the experimental findings and rationalized the ob-

served differences in the behavior of the metal catalysts. To

be an efficient catalyst, a metal complex should satisfy the

following criteria: moderate Brønsted and Lewis acidity

(pK_a=4–6), coordination with either water or weaker

s donors, energetically low-lying unoccupied orbitals, com-

pact transition-state structures, and the ability for complexa-

tion of glucose. Thus, the reactivity of the metal catalysts in

water is governed by many factors, not just the Lewis acidi-

ty.

Introduction

Biomass provides one of the renewable alternatives to fossil re-

sources. The depolymerization of cellulosic and lignin-contain-

ing biomass yields platform molecules such as 5-hydroxyme-

thylfurfural (HMF), which can be converted to other small or-

ganic molecules, such as levulinic acid (LA) and the biofuel di-

methylfurfural (DMF), the latter in almost quantitative yield

(Scheme 1).^[1]An important step in these processes is glucose–

fructose isomerization.^[2]In technical applications, it is prefera-

ble to start from the educt glucose rather than fructose be-

cause of the lower price of glucose.

The isomerization was carried out experimentally in different

solvents (ionic liquids,^[3]organic solvents,^[4]water)^[5]with vari-

ous catalysts (acids,^[3a]metal salts)^[3,5a]and continues to be the

topic of ongoing investigations. Among all systems tested,

chromium(III) salts in ionic liquids were the most promising

catalysts. The subsequent step of Brønsted acid catalyzed HMF

formation from glucose has been studied thoroughly, both ex-

perimentally and computationally.^[6]

Scheme 1. Possible pathways for the formation of platform molecules from

biomass.

active species, a glucose–chromium(III)–water complex, was

also investigated by Car–Parinello molecular dynamics (CPMD).

An activation barrier of 15.5 kcalmol^À1was estimated from an

Arrhenius plot of the initial rate of glucose conversion with

a 3:100 CrCl₃/glucose ratio at 413 K.

Recently, detailed experimental kinetic studies led to the

conclusion that glucose–fructose conversion is catalyzed not

only by the Lewis acidity of metal cations, but also by the

Brønsted acidity of the water complexes of the cations.^[6g]The

Additionally, the formation of HMF from glucose in ionic liq-

uids was studied by DFT. At the B3LYP/LANL2DZ//B3LYP/6-31+

G** level of theory, the glucose–fructose isomerization was

found to be endothermic by 6.8 kcalmol^À1, with barriers of 20–

40 kcalmol^À1depending on the chosen catalyst (WCl₃, MoCl₃,

CrCl₃, FeCl₃).^[7]Hensen and co-workers explored the isomeriza-

[a] C. Loerbroks, J. van Rijn, M.-P. Ruby, Q. Tong, Prof. Dr. F. Schꢀth,

Prof. Dr. W. Thiel

Max-Planck-Institut fꢀr Kohlenforschung

Kaiser-Wilhelm-Platz 1, 45470 Mꢀlheim an der Ruhr (Germany)

Fax: (+49)208-306-2996

E-mail: thiel@kofo.mpg.de

[9]

tion by using CrCl₂,^[8]CrCl₃,^[8b]FeCl₂,^[8b]CuCl₂,^[8b]and SnCl₄as

Supporting information for this article is available on the WWW under

http://dx.doi.org/10.1002/chem.201402437.

catalysts (PBE0/6-311+ +G**//PBE0/6-31+G*). They proposed

Chem. Eur. J. 2014, 20, 1 – 13

1

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

These are not the final page numbers! ÞÞ

Full Paper

that a dimeric complex catalyzes the reaction better than

a monomeric catalyst and that chloride and hydroxyl ligands

facilitate proton transfers during the reaction. The rate-deter-

mining step was identified as a hydrogen shift with a barrier of

about 16 kcalmol^À1. According to their calculations and experi-

ments, FeCl₂and CuCl₂are not active because the direct coor-

dination of the corresponding active species to glucose is un-

favorable. Reaction mechanisms for proton-catalyzed isomeri-

zations were also the topic of extensive studies.^{[5b,c,6d,e,10]}

Despite all this previous work, the role of the metal salt in

the reaction is still not fully understood, and to the best of our

knowledge mechanisms have been studied in ionic liquids not

water, which would most probably be the solvent of choice for

any commercial process for cost-related reasons. To shed light

on the activity of different metal salts in the glucose–fructose

conversion in water, we have carried out a combined experi-

mental and theoretical study, which made it possible to identi-

fy criteria for efficient catalysts that give high-conversion rates

for glucose–fructose isomerization in water.

which had been developed for this purpose.^[20]The role of the

second hydration shell for the stabilization and constitution of

the complexes had been examined previously: no exchange

between the first and second hydration shell and no influence

of the second hydration shell on the spin ordering could be

detected, and therefore it was concluded that only the first hy-

dration shell needs to be treated explicitly.^[21]

Different spin states were investigated for transition-metal

complexes with different ligands (H₂O, Cl^À, OH^À) to identify

the lowest-energy state. For all ligand systems, iron(III) cations

prefer high-spin complexes (sextet), and chromium(III) cations

prefer quartet rather than doublet spin states by about 26–

35 kcalmol^À1. Copper(II) complexes favor the doublet state.

These results are in agreement with the literature and are pre-

sented in the Supporting Information.^[22]

A natural bonding orbital (NBO) analysis for selected struc-

tures was performed by using the NBO 3.1 package imple-

mented in Gaussian 09.^[23]The interactions of localized orbitals

were quantified by the occupation of the orbitals and the nat-

ural charges.

All of the proposed mechanisms feature protonation and de-

protonation reactions, which were treated by thermodynamic

cycles (see the Supporting Information). This approach was

also used for protonation reactions in a previous study on glu-

cose protonation pathways.^[10]

Computational details

DFT calculations were performed by using the PBE0 density

functional^[11]and the 6-31+G(d,p) basis set provided in the

Gaussian 09 program suite.^[12]The basis set contains diffuse

functions that are required for the correct description of intra-

molecular hydrogen bonds.^[13]The PBE0 hybrid functional has

previously been found appropriate for the treatment of hydro-

gen-bonding complexes (mean unsigned error (MUE)=

0.34 kcalmol^À1) and their structures (angles: MUE=1.118, dis-

tances: mean average error=0.0103 ꢂ).^[14]Benchmark studies

have documented the high accuracy of PBE0 for various transi-

tion-metal-catalyzed reactions with^[15]and without sugar mole-

cules.^[16]For the sake of comparison, we carried out single-

point calculations with other functionals and larger basis sets

(see the Supporting Information), which gave results that

agree well with those from the PBE0/6-31+G(d,p) calculations.

Geometry optimizations were carried out without any con-

straints, unless noted otherwise. Local minima (no imaginary

mode) and transition states (one imaginary mode) were char-

acterized by harmonic force constant analysis. Thermal correc-

tions to obtain Gibbs free energies were carried out at 1 bar

and 298 K for all stationary points and at 70 bar and 413 K for

rate-determining transition states and other important steps

(see the Supporting Information). We observed only minimal

changes in free energy at the higher temperature and pres-

sure. The connectivity of transition states and minima was veri-

fied by intrinsic reaction coordinate (IRC) calculations,^[17]unless

mentioned otherwise. To estimate the energy of elusive transi-

tion states, scans of the potential-energy surface (PES) were

performed.

Several possible mechanistic pathways were explored by

consideration of different catalyst complexes, transition states

in different protonation states, and also keto–enol tautomeriza-

tion as a mechanistic alternative to hydrogen-transfer reac-

tions. In the Results section, only the lowest pathways are

shown. Data for all calculated pathways can be found in the

Supporting Information.

Various metal–ligand systems were used in the calculations

for iron(III), magnesium(II), aluminum(III), chromium(III), and

copper(II) cations. If different choices were possible for the po-

sitioning of a ligand, different possibilities were examined and

the lowest-energy structure of the starting glucopyranose–cat-

alyst complex was selected (see the Supporting Information).

Examples of metal catalysts and their coordination to glucose

are shown in Figure 1. According to the literature, coordination

of the O1/O2 hydroxyl groups is essential for the reaction.^[8b]

Results and Discussion

The experimental results are summarized in Table 1. The yields

of fructose, HMF, and LA do not add up to the total conversion

of glucose because acid-catalyzed side reactions of glucose

with the products can occur. Due to the large variety and low

yields of the side products, these were not analyzed.

The performance of the catalysts can be divided into three

groups (Table 1). The first group consists of catalysts with high

conversion rates: aluminum(III), zirconium(IV), and chromium-

(III). They also have reasonable yields of fructose, HMF, and/or

LA. The second group includes bismuth(III)-, copper(II)-, and

iron(III)-catalyzed reactions, which show low conversion and

even lower product yields. The third group contains the reac-

Solvent effects were treated by a polarizable continuum

model (CPCM)^[18]with universal forcefield (UFF) cavities and

water as the solvent. This model was applied because it gave

the best results for cationic species in solution.^[19]The calcula-

tion of solvation free energies in thermodynamic cycles was

performed with the solvent model density (SMD) approach,

&

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

2

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Full Paper

tions with little or no conversion, i.e., the magnesium(II) and

acetic acid catalyzed reactions and the uncatalyzed reaction.

The experimental data indicate that Lewis acidity plays

a role in this reaction because acetic acid is not a reactive cata-

lyst, even though it has almost the same pK_aas chromium(III)

or aluminum(III). But if Lewis acidity is important for the reac-

tion, why do iron(III) and magnesium(II) not perform well,

which are known as good Lewis acids? These experimental

data already qualitatively suggest that the interplay between

different factors is decisive for catalyst performance in the glu-

cose–fructose conversion.

To answer this question in detail, the reaction was investi-

gated computationally. Two examples from each group (Cr³⁺

and Al³⁺; Fe³⁺and Cu²⁺; Mg²⁺and no catalyst) were used to

calculate the pathway from glucose to fructose. First, we pres-

ent the general reaction mechanism, which was found to be

lowest in energy for the glucose–fructose isomerization. Be-

cause multiple pathways and a variety of catalysts were investi-

gated we will not dwell on the details of all pathways, but

rather give an overview of the critical mechanistic steps within

each group of catalysts. The focus will be on electronic, ener-

getic, and geometrical trends that distinguish these groups

and explain their reactivity.

Figure 1. Glucose complexes of different metal cations with an octahedral

ligand sphere: a) Hydrogen-bonding coordination, b) direct complexation of

glucopyranose, and c) bonding of a dimeric catalyst.

Table 1. Experimental results for the reaction of glucose (413 K, 80 bar)

with the given catalyst in water.

The choice of cation environment for the single metal cat-

ions, alternatives to the general mechanism, and detailed de-

scriptions of selected mechanistic pathways can be found in

the Supporting Information.

Catalyst

Conv. [%] HPLC yield [%]

Fructose HMF LA

pK_a

pH of reaction

Before

After

CrCl₃·6H₂O

AlCl₃·6H₂O

ZrCl₄

CuCl₂·2H₂O

BiCl₃

FeCl₃·6H₂O

MgCl₂

CH₃COOH

uncatalyzed

99

88

69

23

17

12

2

0

11

7

1

0.6

13

19

5

6

5

13

6

4.1 3.4

5.5 3.1

1.9

1.4

1.1

2.1

1.3

1.6

3.3

3.2

–

13 À0.3 1.2

2

3

0

8.0 4.5

À3.5 1.3

2.2 2.0

11.4 7.4

4.8 3.1

0

1

3

0.8

Reaction mechanism

0.1

0

0.8

0

0.6

0

15.7

–

The lowest-energy pathways found consist of the steps shown

in Scheme 2.

Scheme 2. Proposed mechanism for glucose to fructose isomerization. The example shown has a-glucose as a ligand.

Chem. Eur. J. 2014, 20, 1 – 13 www.chemeurj.org ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

3

&

These are not the final page numbers! ÞÞ

Full Paper

Starting with a catalyst–glucopyranose complex, the first

The metal cations investigated will be discussed separately

step is a proton transfer from O1 to O5 (PT₁) followed by C1À in three groups, according to their activity (see the Experimen-

O5 bond cleavage (TS₁), which yields an open-chain glucose–

catalyst complex. PT₁can be separated into deprotonation of

O1 and protonation of O5. In some cases PT₁and TS₁are con-

certed (uncatalyzed, magnesium(II)- and copper(II)-catalyzed

pathways) and C1ÀO5 bond cleavage is, in most cases, a barri-

er-free process. In this initial step, the catalyst influences the

C1ÀO1 and C1ÀO5 bond distances.

tal section): group one (high conversion: Al³⁺, Cr³⁺), group two

(low conversion: Cu²⁺, Fe³⁺), and group three (no conversion:

uncatalyzed, Mg²⁺). For each group, we give a general over-

view over the cations then discuss the crucial proton transfers,

hydrogen-atom transfers, and catalyst separation/reaction en-

ergies.

To form the open-chain fructose–catalyst complex, a hydro-

gen atom is transferred from C2 to C1 (TS₂) and a proton is

transferred from O2 to O1 (PT₂). These steps were also identi-

fied from NMR experiments as part of the reaction pathway for

solid Lewis acid catalysts in water.^[24]For the steps TS₂and PT₂

there is an alternative: deprotonation of O2 (DP₂), hydrogen-

atom transfer (TS₂), and re-protonation of O1 (RP₂). The depro-

tonation of O2 (DP₂) may be lower in energy as the negative

charge on O2 eases the hydrogen transfer (TS₂) due to its posi-

tive inductive effect. This alternative pathway (DP₂, TS₂, RP₂) is

favorable for the uncatalyzed pathway and for some of the

[MO¹O²(H₂O)₄]ⁿ⁺complexes and the dimeric catalysts. In the

transition state TS₂electron density is delocalized between the

transferred hydrogen atom, the s*(CÀO) orbitals, and the orbi-

tals of the metal or the metal–oxygen bonds.

Group one: aluminum(III), chromium(III)

Group one comprises the best catalysts under the reaction

conditions, for example, Al³⁺and Cr³⁺, which yield 13–19%

HMF and have high conversions of 88–99%. Analysis of their

reactions should help us with finding criteria for good conver-

sion.

For Al³⁺and Cr³⁺, the catalysts [M(H₂O)₆]³⁺, [MO¹O²(H₂O)₄]³⁺,

[MO¹O²(H₂O)₃OH]²⁺, [MO¹O²(H₂O)₃Cl]²⁺, and [M₂O¹O²(H₂O)₇OH]⁵⁺

were investigated. Reaction pathways with species [MO¹O²-

(H₂O)₂Cl₂]⁺and [MO¹O²(H₂O)Cl₃] were only calculated for the

chromium(III) cation. Under the reaction conditions aluminum-

(III) complexes with water ligands and chromium(III) complexes

with one to three chloride ligands are most likely to appear.

Deprotonated and dimeric complexes are also common for

both metal cations in water.

To complete the fructose–fructofuranose transformation, the

C2ÀO5 bond is formed (TS₃) and a proton transfer from O5 to

O2 takes place (PT₃). All proton transfers proceed by a deproto-

nation followed by a protonation step. The C2ÀO5 bond for-

mation is, in most cases, a barrier-free reaction.

Both metal complexes have similar Brønsted acidities (pK_a=

4.1 (Cr³⁺) and 5.5 (Al³⁺) in water) and both are regarded as

hard Lewis acids.

Rate-determining steps

Proton transfer

In all isomerization pathways, the first proton transfer (PT₁), the

hydrogen-atom transfer (TS₂), and the separation of the cata-

lyst from the product are possible critical steps. This section fo-

cuses on the comparison of these steps when different cata-

lysts are used. Tables 2–4 below list the lowest energies found

for these critical steps in each group. The rate-determining

step for each mechanism is defined, and energies, NBO data,

and natural charges are compared. Because the reaction takes

place at 413 K, energies up to 36 kcalmol^À1are assumed to be

feasible; higher barriers would be difficult to overcome under

these conditions.

The sum of the energies of [M(H₂O)₆]ⁿ⁺and a-glucopyranose

was taken as reference for all energies (DG) for a given metal

M. No attempt was made to establish a common free-energy

scale for all metals by calculations of the standard chemical po-

tential at a given pH. The energy of the glucopyranose–catalyst

complex (or, if lower, the sum of the energies of glucopyranose

and the modified [M(H₂O)₆]ⁿ⁺complex) serves as reference to

determine the barriers for transition states TS₂and PT₁

[DDG(TS₂/PT₁)]. DDG(separation) is the energy needed to sepa-

rate fructofuranose from the catalyst; in this case, the refer-

ence is the energy of the fructofuranose–catalyst complex.

DDG(glucose-fructose) denotes the energy difference between

the fructofuranose–catalyst complex and the glucopyranose–

catalyst complex (relative to the latter).^[25]

The proton transfer PT₁was feasible for all Al³⁺and Cr³⁺com-

plexes (Table 2). Low energies of 13–20 kcalmol^À1were found

for the [M(H₂O)₆]³⁺, [MO¹O²(H₂O)₄]³⁺, and [MO¹O²(H₂O)_aCl_b]ⁿ⁺

pathways. In the [MO¹O²(H₂O)₃OH]²⁺system, the barriers rise to

25 and 30 kcalmol^À1for M=Cr³⁺and Al³⁺, respectively. In gen-

eral, PT₁is higher in energy for the unbound catalyst

Table 2. Relative free energies [kcalmol^À1] at 298 K and 1 atm for critical

steps (PT₁, TS₂, product/catalyst separation (sep.), and reaction energy for

the complexed glucopyranose–fructofuranose (gluc.–fruc.) conversion) in

the glucose–fructose isomerization with Al³⁺and Cr³⁺catalysts. PBE0/6-

31+G**, CPCM (water).

Catalyst

DDG(PT₁) DDG(TS₂) DDG(sep.) DDG(gluc.–fruc.)

[Al(H₂O)₆]³⁺

17.0

13.3

29.2

23.0

41.0

23.1

40.1

26.9

18.0

34.2

25.9

32.4

31.7

36.2

9.4

15.4

17.7

13.0

–

À3.1

À16.2

À14.1

À11.8

–

[AlO¹O²(H₂O)₄]³⁺

[AlO¹O²(H₂O)₃OH]²⁺30.1

[AlO¹O²(H₂O)₃Cl]²⁺

[Al₂O¹O²(H₂O)₇OH]⁵⁺

[Cr(H₂O)₆]³⁺

13.2

–

13.9

16.2

13.0

18.3

14.4

17.0

12.7

6.8

À7.2

À17.5

À10.9

À13.5

À8.0

À4.3

–

[CrO¹O²(H₂O)₄]³⁺

[CrO¹O²(H₂O)₃OH]²⁺25.4

[CrO¹O²(H₂O)₃Cl]²⁺

[CrO¹O²(H₂O)₂Cl₂]⁺

[CrO¹O²(H₂O)Cl₃]

16.2

17.4

19.7

–

[Cr₂O¹O²(H₂O)₇OH]⁵⁺

–

&

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

4

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Full Paper

[M(H₂O)₆]³⁺and for complexes with non-water ligands com-

pared to [MO¹O²(H₂O)₄]³⁺

.

This can be explained by the diminished Brønsted acidity of

O1 due to negatively charged ligand L^Àin the metal cation

([MO¹O²(H₂O)₃OH]²⁺) and/or by the longer distance between

the metal cation and the transferred proton ([M(H₂O)₆]³⁺). The

hydroxyl ligand has a stronger influence on the energy than

the chloride ligand. The hydroxyl group oxygen atom should

be a better s- and p-electron donor than Cl^Àbecause of the

shorter MÀL bond length (MÀL=1.79 and 2.19 ꢂ for [AlO¹O²-

(H₂O)₃OH]²⁺and [AlO¹O²(H₂O)₃Cl]²⁺, respectively). Therefore, hy-

droxyl ligands diminish the Brønsted acidity more effectively

than chloride ligands.

Hydrogen transfer

TS₂is the rate-determining step for all catalysts in group one

with DDG(TS₂)=18–41 kcalmol^À1. Only for [CrO¹O²(H₂O)₄]³⁺is

the product/catalyst separation energy in the same energy

range as the TS₂barrier.

The [AlO¹O²(H₂O)₃OH]²⁺and [Al₂O¹O²(H₂O)₇OH]⁵⁺catalyzed

reactions are not feasible, with barriers around 40 kcalmol^À1

for TS₂; all other aluminum(III) systems and the chromium(III)-

catalyzed reactions are feasible. Only the TS₂barrier for the

dimer [Cr₂O¹O²(H₂O)₇OH]⁵⁺is rather hard to overcome

(DG(TS₂)=36 kcalmol^À1). With more chloride ligands, the TS₂

barrier increases: 18.0 ([CrO¹O²(H₂O)₄]³⁺) < 25.9 ([CrO¹O²-

(H₂O)₃Cl]²⁺< 32.4 ([CrO¹O²(H₂O)₂Cl₂]⁺ꢀ31.7 kcalmol^À1(CrO¹O²-

(H₂O)Cl₃]). Even more extreme is the rise in energy for the OH^À

ligand, which increases the energy of TS₂for aluminum(III) and

chromium(III) catalysts by 16–18 kcalmol^À1relative to

[MO¹O²(H₂O)₄]³⁺.

Figure 2. Hydrogen-atom transfer (TS₂) for (top) [CrO¹O²(H₂O)₄]³⁺

,

(middle) [CrO¹O²(H₂O)₃OH]²⁺and (bottom) [CrO¹O²(H₂O)₃Cl]²⁺. PBE0/6-

31+G**, CPCM (water).

accepting negative charge from the broken C2ÀH bond be-

cause it has already taken up electron density from the ligands

[natural charge on Cr: 1.18e ([CrO¹O²(H₂O)₄]³⁺)>0.89e ([CrO¹O²-

(H₂O)₃Cl]²⁺)>0.59e ([CrO¹O²(H₂O)₂Cl₂]⁺)].

The TS₂barriers increase with the electron density in the un-

occupied d orbitals of the chromium(III) center (due to dona-

tion from the non-water ligands) and the d orbitals are desta-

bilized, indicated by their orbital energies: 0.11e/À1.93 eV

([CrO¹O²(H₂O)₄]³⁺), 0.14e/2.23 eV ([CrO¹O²(H₂O)₃OH]²⁺), 0.12e/

À0.27 eV ([CrO¹O²(H₂O)₃Cl]²⁺), and 0.15e/0.49 eV ([CrO¹O²-

(H₂O)₂Cl₂]⁺). For aluminum(III), which does not have low-lying d

orbitals, the p*_AlÀOorbitals are used instead (electron density

in the p*_AlÀOorbitals rises by 0.01–0.07e for all bound catalysts

[MO¹O²L₄]ⁿ⁺relative to the unbound catalyst [Al(H₂O)₆]³⁺).

Apparently, these explanations do not apply to the [M₂O¹O²-

(H₂O)₇OH]⁵⁺catalyzed reaction, which is not feasible even

though the electronic structure is similar to [MO¹O²(H₂O)₄]³⁺. Its

TS₂structure (Figure 3) has a lengthened MÀO1 bond to both

metal centers (Cr: Dd(CrÀO1)=+0.08 ꢂ compared to [CrO¹O²-

(H₂O)₄]³⁺) and a slightly shortened MÀO2 bond (Cr: Dd(CrÀ

Compared to water ligands, the higher TS₂barrier in the

case of OH^Àand Cl^Àligands can be attributed to their negative

charge and stronger s- and p-electron donation, which dimin-

ishes the Lewis acidity of the metal complex. The lower acidity

leads to less stabilization of the hydrogen-atom transfer, which

can also be seen from geometry and NBO data (see below).

We discuss three major effects: MÀO1/2 bond elongation, re-

duced charge at M, and increased occupancy of the orbitals of

M.

The introduction of OH^Àand Cl^Àligands results in longer

M···H₂O and M···O1/O2 distances, which is expected due to the

inductive effect of the negatively charged ligand(s) (Figure 2).

For the Cl^Àligand, an increase of the Al···O2 distance by 0.1 ꢂ

is observed relative to the [AlO¹O²(H₂O)₄]³⁺complex. Introduc-

tion of the OH^Àligand in [CrO¹O²(H₂O)₃OH]²⁺leads to an in-

crease of Dd(Cr···O2)=0.2 ꢂ compared to [CrO¹O²(H₂O)₄]³⁺. This

elongation can be explained by the trans effect and the

change in the nature of the ligand: water ligands are only

weak s-bonding ligands, whereas Cl^Àand OH^Àare s- and po-

tential p-donors. The latter anionic ligands shift more electron

density to the metal center and weaken preferably the bonds

in the trans position. With non-water ligands in axial or multi-

ple positions the effect on the bond length is less pronounced,

but the TS₂energy still rises due to the higher electron density

on the metal cation. Additionally, the cation is less capable of

Figure 3. Hydrogen-atom transfer (TS₂) for [Al₂O¹O²(H₂O)₇OH]⁵⁺. PBE0/6-

31+G**, CPCM (water).

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

5

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

These are not the final page numbers! ÞÞ

Full Paper

O2)=À0.01 ꢂ compared to [CrO¹O²(H₂O)₄]³⁺). Unlike all other

TS₂structures discussed above, the transferred hydrogen atom

is closer to C2 than C1, which hints at an early transition state.

Because the two positive metal centers repel each other, their

distances to O1 must become longer. Therefore, the electron

density is less stabilized in the C1ÀO1 orbitals and the hydro-

gen atom is bound to C2 instead of C1. The charge on both

metal centers and the electron density in the p*_AlÀOorbitals or

d orbitals are nearly the same as for [MO¹O²(H₂O)₄]³⁺. Given

these similarities in the electronic structure, the higher TS₂

energy does not stem from use of the dimer as the catalyst,

but from the energy that is needed to form this dimer. For ex-

ample, the glucose–catalyst complex is 26 kcalmol^À1lower in

energy with a monomeric rather than dimeric catalyst. The

average dimer-formation energies for different doubly bridged

aluminum(III) complexes were previously calculated to be 25–

For iron(III), we investigated the systems [Fe(H₂O)₆]³⁺,

[FeO¹O²(H₂O)₄]³⁺, [FeO¹O²(H₂O)₃OH]²⁺

,

[FeO¹O²(H₂O)₂(OH)₂]⁺,

[FeO¹O²(H₂O)₃Cl]²⁺, [FeO¹O²(H₂O)₂(OH)Cl]⁺, [FeO¹O²(H₂O)₂Cl₂]⁺,

[FeO¹O²(H₂O)(OH)Cl₂], [FeO¹O²(H₂O)Cl₃], [FeO¹O²(OH)Cl₃]^À, and

[Fe₂O¹O²(H₂O)₇OH]⁵⁺. Most likely, iron(III) is present in the chlo-

ride complexes, the dimers, and the deprotonated species. For

copper(II), different catalysts were found in water: [Cu(H₂O)₆]²⁺

,

[CuO¹O²(H₂O)₄]²⁺, [CuO¹O²L₂], and [CuO¹O²(H₂O)Cl]⁺. The li-

gands depend on the concentration of CuCl₂·2H₂O. Experimen-

tally, a concentration of 19.9 gL^À1was used, which seems low

compared to the maximum solubility of 1150 gL^À1in water at

208C.^[27]Therefore, copper(II) complexes with water ligands are

expected to be preferred.

As can be seen from the diverse complexation preferences,

iron(III) and copper(II) differ in their properties. Iron(III) com-

plexes are hard Lewis acids with high Brønsted acidity (pK_a=

2.2), whereas copper(II) complexes are soft Lewis acids with

low Brønsted acidity (pK_a=8.0). Despite these differences both

show low reactivity for the glucose to fructose conversion.

50 kcalmol^{À1 [26]}

and a CPMD simulation led to the separation

,

of chromium(III) dimers into monomeric species.^[6g]Also, no

Al³⁺dimer species was detected in our NMR experiments.

Catalyst separation, reaction energy

Proton transfer

The catalyst/fructose separation is feasible for all catalysts (en-

dothermic by 9–18 kcalmol^À1). It becomes more favorable

when more chloride ligands are attached: 18.3 ([CrO¹O²-

(H₂O)₄]³⁺)>17. ([CrO¹O²(H₂O)₃Cl]²⁺)>12.7 ([CrO¹O²(H₂O)₂Cl₂]⁺)>

6.8 kcalmol^À1([CrO¹O²(H₂O)Cl₃]).

The differences in Brønsted acidity become evident in the en-

ergies for PT₁(Table 3). This step is feasible for most of the

Table 3. Relative free energies [kcal mol^À1] at 298 K and 1 atm for critical

steps (PT₁, TS₂, product/catalyst separation, and reaction energy for the

complexed glucopyranose–fructofuranose conversion) for the glucose–

fructose isomerization with Fe³⁺and Cu²⁺catalysts. PBE0/6-31+G**,

CPCM (water).

The energy difference between the glucopyranose–catalyst

and fructofuranose–catalyst complexes is negative (exother-

mic) for all catalysts, but becomes less so as the number of

chloride ligands increases: À17.5 ([CrO¹O²(H₂O)₄]³⁺)>À13.5

([CrO¹O²(H₂O)₃Cl]²⁺)>À8.0

([CrO¹O²(H₂O)₂Cl₂]⁺)>À4.3 kcal

Catalyst

DDG(PT₁) DDG(TS₂) DDG(sep.) DDG(gluc.–fruc.)

[Fe(H₂O)₆]³⁺

16.9

14.9

18.0

25.5

20.3

32.3

53.9

30.2

42.0

35.4

51.6

31.7

60.4

31.1

36.4

36.5

38.2

43.3

42.7

13.0

17.2

13.1

16.0

13.8

11.5

9.5

3.0

–

4.5

À7.1

À17.9

À9.6

À4.3

À10.7

À4.9

À4.8

À1.4

0.0

mol^À1([CrO¹O²(H₂O)Cl₃]). The separation energies and the

energy differences for glucopyranose–fructofuranose conver-

sion show that the complexes become less stable when non-

water ligands are present.

[FeO¹O²(H₂O)₄]³⁺

[FeO¹O²(H₂O)₃OH]²⁺

[FeO¹O²(H₂O)₂(OH)₂]⁺44.8

[FeO¹O²(H₂O)₃Cl]²⁺

[FeO¹O²(H₂O)₂(OH)Cl]⁺30.1

[FeO¹O²(H₂O)₂Cl₂]⁺

15.9

[FeO¹O²(H₂O)(OH)Cl₂] 39.1

15.1

Experimentally, the aluminum(III) and chromium(III) catalysts

mediate the glucopyranose–fructofuranose transformation

with 88–99% conversion. This is consistent with the computed

barriers of 18–34 kcalmol^À1, which are high but still feasible at

413 K. According to the calculations, chloride or hydroxyl li-

gands have a significant impact on the reaction barriers. The

more non-water ligands are attached to the catalyst, the

higher the barriers PT₁and TS₂and the less exothermic the re-

action. Another effect of non-water ligands is destabilization of

the product complex.

[FeO¹O²(H₂O)Cl₃]

[FeO¹O²(HO)Cl₃]^À

[Fe₂O¹O²(H₂O)₇OH]⁵⁺

[Cu(H₂O)₆]²⁺

[CuO¹O²(H₂O)₄]²⁺

[CuO¹O²(H₂O)₂]²⁺

[CuO¹O²(H₂O)Cl]⁺

[CuO¹O²Cl₂]

19.3

–

20.1

20.2

28.5

–

À2.3

À2.1

À1.6

0.1

6.1

10.5

6.6

8.3

0.6

Because Al³⁺and Cr³⁺are not prone to attract hydroxyl li-

gands (due to medium Brønsted acidity) and the reaction is

still feasible with some chloride ligands (due to good Lewis

acidity and empty d orbitals), the isomerization has high con-

versions with these catalysts.

iron(III) catalysts (uphill by 15–30 kcalmol^À1) and unfeasible

only for [FeO¹O²(H₂O)₂(OH)₂]⁺and [FeO¹O²(H₂O)₂(OH)Cl₂] (uphill

by 39–45 kcalmol^À1). Iron(III) behaves similarly to chromium(III)

with regards to energy trends and their explanations for PT₁:

the more non-water ligands attached to the cation, the lower

the Brønsted acidity and the higher the PT₁barrier (see group

one, above).

Group two: iron(III), copper(II)

The criteria we found for group one were applied to less pro-

ductive cations: the group two catalysts copper(II) and iron(III)

with 3–6% HMF yield and a conversion rate of 12–23%.

The PT₁energies for copper(II) complexes range from 20–

30 kcalmol^À1and are, therefore, slightly higher than those for

the previously discussed Al³⁺, Cr³⁺, and Fe³⁺catalysts with simi-

&

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

6

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Full Paper

lar ligands, due to the generally lower Brønsted acidity of the

copper(II) complexes. PT₁could not be calculated for

[Cu(H₂O)₆]²⁺and [CuO¹O²(H₂O)₄]²⁺, because water ligands disso-

ciated during the calculation of the thermodynamic cycle (see

the Supporting Information).

orbitals in the x–y plane and higher d orbitals with z-axis con-

tributions. Because each iron(III) d orbital is occupied by one

electron, additional electron density from the hydrogen trans-

fer is partly shifted to the 4s* and 4p* orbitals. However, unlike

the group one case, catalysts with chloride ligands do not

have higher electron density in the antibonding orbitals. In-

stead, the population in the s* orbitals is higher for the

[FeO¹O²(H₂O)₄]³⁺than the [FeO¹O²(H₂O)₃Cl]²⁺catalyst (0.15e/

0.24 eV and 0.10e/0.16 eV, respectively). The reason for the

higher population might be that the MÀO2 bond is still pres-

ent in [FeO¹O²(H₂O)₄]³⁺but is broken in [FeO¹O²(H₂O)₃Cl]²⁺. The

orbital energies show that it is energetically expensive to put

electron density into virtual orbitals. Additionally, the charge

on the metal cation is decreased [natural charge on Fe: 1.39e

([FeO¹O²(H₂O)₄]³⁺)>1.31e ([FeO¹O²(H₂O)₃Cl]²⁺)>0.61e ([FeO¹O²-

(H₂O)₂Cl₂]⁺)]. The situation for copper(II) complexes is analo-

gous to that for iron(III) complexes with axial ligands (or with

both OH^Àand Cl^Àligands).

Hydrogen transfer

As for group one, TS₂is the rate-determining step. For iron(III)

complexes with only water ligands ([Fe(H₂O)₆]³⁺and [FeO¹O²-

(H₂O)₄]³⁺) the energy for the isomerization is as low as for

aluminum(III) and chromium(III) complexes (DDG(TS₂)=20–

25 kcalmol^À1). For catalysts with chloride or hydroxyl ligands,

the TS₂energy is in the range 30–35 kcalmol^À1, slightly higher

than for the corresponding group one catalysts with one or

two chloride ligands ([MO¹O²(H₂O)₃Cl]²⁺: 25.9 (Cr) versus

30.2 kcalmol^À1(Fe); [MO¹O²(H₂O)₂Cl₂]⁺: 32.4 (Cr) versus

35.4 kcalmol^À1(Fe). As for chromium(III) catalysts, the TS₂

energy rises further when more chloride or hydroxyl ligands

are present. If the complex contains two hydroxyl ligands or

both hydroxyl and chloride ligands, the TS₂barrier exceeds

40 kcalmol^À1and the reaction becomes unfeasible.

The substitution of water ligands with hydroxyl or chloride

ligands may have a larger effect on iron(III) and copper(II) than

on chromium(III) because for copper(II) and iron(III) all d orbi-

tals are filled with one or two electrons according to ligand

field theory for quadratic planar and octahedral complexes, re-

spectively (high-spin complexes). Therefore, there is no vacant

low-energy metal orbital that can accept additional electron

density from the transferred hydrogen atom and, hence, the

MÀO2 bond is broken.

In the case of copper(II) catalysts, the TS₂energies with one

or two chloride ligands are also too high (38–43 kcalmol^À1). Re-

actions with the catalysts [Cu(H₂O)₆]²⁺and [CuO¹O²(H₂O)₄]²⁺are

barely feasible because they have barriers of 36–37 kcalmol^À1

.

The reasons for the increased TS₂energies for catalysts with

non-water ligands and for the dimeric catalyst are analogous

to those discussed for the chromium(III) catalysts. We find an

elongated MÀO2 bond for the monomeric catalysts if chloride

or hydroxyl ligands are attached (Figure 4). Here, the MÀO2

The broken MÀO2 bond in complexes with non-water li-

gands creates another problem for the group two catalysts. Be-

cause the reaction is performed under high pressure, a transi-

tion state with a higher volume should be less favorable.

Under high pressure the M···O2 bond distance will probably

shrink, which will tend to raise the barrier because more elec-

tron density is donated into the already filled d orbitals and

the high-energy s* orbitals of the metal cations.

For the iron(III) dimer, the TS₂energy is as low as for the

monomeric complex. Unlike the cases of aluminum(III) and

chromium(III), complex formation through deprotonation of

water ligands is favorable for iron(III) due to the high acidity of

its complexes. Iron(III) dimers bound to glucose molecules

were also found experimentally.^[28]

Catalyst separation, reaction energy

The separation of the catalyst from fructofuranose is endother-

mic in all cases. The separation becomes easier for the iron(III)

catalysts when more chloride and hydroxyl ligands are at-

tached: 17.2 ([FeO¹O²(H₂O)₄]³⁺)>16.0 ([FeO¹O²(H₂O)₃Cl]²⁺)>

11.5 ([FeO¹O²(H₂O)₂Cl₂]⁺)>3.0 kcalmol^À1([FeO¹O²(H₂O)Cl₃]). No

such trend was found for copper(II) complexes. Analogously to

chromium(III), the difference in free energy between the gluco-

pyranose and fructofuranose complexes was negative for most

iron(III) catalysts. However, with an increasing number of chlo-

ride ligands, the reaction becomes less exoergic (for [FeO¹O²-

(H₂O)Cl₃] and [CuO¹O²(H₂O)Cl]⁺DDG(glucose–fructose)=0 kcal

mol^À1).

Figure 4. Hydrogen-atom transfer (TS₂) for (top) [FeO¹O²(H₂O)₄]³⁺and

(bottom) [FeO¹O²(H₂O)₃Cl]²⁺. PBE0/6-31+G**, CPCM (water).

bond is not simply weakened, but broken with M···O2 distan-

ces of 2.81 ([FeO¹O²(H₂O)₂Cl₂]⁺), 2.97 ([CuO¹O²L₂]), and 3.38 ꢂ

([FeO¹O²(H₂O)₃Cl]²⁺). After cleavage of the MÀO2 bond, the

metal cation no longer has octahedral coordination but adopts

a roughly pentacoordinate structure, which leads to lower d

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

7

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

These are not the final page numbers! ÞÞ

Full Paper

The lower HMF yields and conversions for group two cata-

lysts can be explained on the basis of this data. The iron(III)-

catalyzed isomerization does not have similar barriers to the

chromium(III)-catalyzed reaction because the higher Brønsted

acidity of the Fe^IIIcomplexes prevents the formation of advan-

tageous ligand systems. In water, a pK_avalue of 2.2 and strong

Lewis acidity favors coordination of multiple hydroxyl and

chloride ligands to the metal cation, which raises the TS₂barri-

er. Furthermore, the high-spin nature of the iron(III) catalysts

leads to breaking of the MÀO2 bond in the rate-determining

transition state TS₂, which will disfavor the reaction at high

pressure.

Proton transfer

PT₁is high in energy for the uncatalyzed reaction (38 kcal

mol^À1) and for the two magnesium(II) catalysts with (26 and

34 kcalmol^À1for [MgO¹O²(H₂O)₄]²⁺and [Mg(H₂O)₆]²⁺, respec-

tively). Due to the low Brønsted acidity, this step requires more

energy than for other catalysts with water ligands. Thus, the

proton transfers are already close to the limit of what is feasi-

ble under the reaction conditions.

Hydrogen transfer

For copper(II) the problem is rather the low Lewis acidity.

Even though only water ligands are present at low concentra-

tions of copper(II) complexes, the TS₂barrier is high (around

36 kcalmol^À1).

Even higher barriers were found for the rate-determining step

via TS₂: 78 (uncatalyzed), 40 ([Mg(H₂O)₆]²⁺), and 35 kcalmol^À1

([MgO¹O²(H₂O)₄]²⁺).

In the [MgO¹O²(H₂O)₄]²⁺catalyzed reaction, the structure of

TS₂is similar to those for the iron(III)- and copper(II)-chloride

catalysts, and different from those for the aluminum(III) and

chromium(III) cases (Figure 5). The MÀO2 bond is broken

In summary, the disadvantages of the group two relative to

the group one metal cations arise from their Lewis acidity

(either too high or too low) and from their high-spin electronic

structure.

Group three: magnesium(II), uncatalyzed

For group three we may expect similar reasons for the low re-

activity as for the group two cations. We examined the uncata-

lyzed reaction and the magnesium(II)-catalyzed reactions.

For magnesium(II), the catalysts [Mg(H₂O)₆]²⁺and [MgO¹O²-

(H₂O)₄]²⁺were investigated (Table 4). Magnesium(II) complexes

Table 4. Relative free energies [kcalmol^À1] at 298 K and 1 atm for critical

steps (PT₁, TS₂, product/catalyst separation, and reaction energy for the

complexed glucopyranose–fructofuranose conversion) for the glucose–

fructose isomerization catalyzed by Mg²⁺complexes and for the uncata-

lyzed reaction. PBE0/6-31+G**, CPCM (water).

Catalyst

DDG(PT₁) DDG(TS₂) DDG(sep.) DDG(gluc.–fruc.)

[Mg(H₂O)₆]²⁺

33.7

40.1

35.1

78.2

À0.4

7.4

1.5

À0.3

À5.2

1.5

[MgO¹O²(H₂O)₄]²⁺25.5

uncatalyzed

38.1

Figure 5. Hydrogen-atom transfer (TS₂) for the (top) hydrogen-bonded cata-

lyst [Mg(H₂O)₆]²⁺and (bottom) glucose-bonded catalyst [MgO¹O²(H₂O)₄]²⁺

.

PBE0/6-31+G**, CPCM (water).

are hard Lewis acids, but weak Brønsted acids (pK_a=11.4).

Most likely the [Mg(H₂O)₆]²⁺complex is present; it was found

experimentally that magnesium(II) does not attach glucose as

a ligand.^[29]The uncatalyzed reaction has only water (pK_a=

15.7) to stabilize its intermediates.

(2.26 ꢂ), whereas the MÀO1 bond is still intact (2.00 ꢂ). This TS₂

structure should also be less favorable under the reaction con-

ditions of 60 bar because of the large M···O2 distance. The

donation of electron density into antibonding orbitals is even

higher than for [FeO¹O²(H₂O)₄]³⁺(0.18e in 4s*): 0.22e are found

in the s* orbital (7.07 eV) of [MgO¹O²(H₂O)₄]²⁺, with a similar

value for [Mg(H₂O)₆]²⁺. There is no analogous occupation of an

s* orbital in the aluminum(III) catalysts, which also lack d orbi-

tals. Apparently, the acidity is too low to stabilize the electron

density in this case, which explains why the magnesium(II) cat-

alyst with only water ligands has the highest TS₂barrier of all

the [M(H₂O)₆]²⁺complexes studied.

In all of the group three reaction pathways, the reactions via

PT₁/TS₁are concerted because the O5-protonated intermediate

is not sufficiently stabilized and the C1ÀO5/C2ÀO5 bond opens

immediately upon protonation of the pyranic oxygen. Whereas

a proton is detached in the uncatalyzed reaction before reach-

ing TS₂, it is energetically preferable for the [Mg(H₂O)₆]²⁺and

[MgO¹O²(H₂O)₄]²⁺catalyzed pathways to retain the proton on

O2 until after TS₂.

Moreover, this might rationalize why the association of glu-

cose hydroxyl groups or other non-water ligands is unfavora-

ble for magnesium(II) relative to the other metal catalysts. If

&

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

8

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Full Paper

water ligands with their weak s donation can already fill anti-

bonding orbitals, ligands with stronger s-donation ability will

destabilize the complex even more so that the complexation

of glucose by magnesium(II) becomes unfavorable. However,

this is not reflected in the energy of the glucopyranose–

[MgO¹O²(H₂O)₄]²⁺complex, which lies below the separated cat-

alyst and glucopyranose thanks to the entropic contribution

(DG=À0.7 kcalmol^À1, DH=5.9 kcalmol^À1), whereas the forma-

tion of glucopyranose–[Mg(H₂O)₆]²⁺is endoergic (DG=2.1 kcal

mol^À1, DH=À9.8 kcalmol^À1).

inefficient. Unlike for the group two catalysts, non-water li-

gands are not a problem in this case.

Analysis of PT₁and TS₂energies

Further to the discussion of the individual metal cations (and

their complexes) and explanation of their reactivity, we present

a general comparison between the three groups. To provide

a better overview of the influence of the acidity and the ligand

systems on the critical steps, PT₁(deprotonation of O1 and

protonation of O5) and TS₂, the computed relative free ener-

gies are related to the experimental pK_avalues (Figure 6).

Catalyst separation, reaction energy

Proton transfer

The separation of the fructose complex is barely exoergic for

[Mg(H₂O)₆]²⁺and is endoergic for [MgO¹O²(H₂O)₄]²⁺. For the

latter complex, the glucose–fructose conversion is slightly

exoergic (DG=À5.2 kcalmol^À1), whereas the uncatalyzed reac-

tion is endoergic by 1.5 kcalmol^À1. Thus, the metal catalyst af-

fords stabilization of the final product complex, which may

help prevent the formation of side products.

The pK_avalue of the metal cation in water is plotted against

*

the deprotonation energy of O1 in PT₁(Figure 6, ) for the

[MO¹O²(H₂O)₄]ⁿ⁺mechanism. As expected, lower energies were

found for lower pK_avalues. However, comparison of iron(III) to

aluminum(III) (DpK_a=3.3) shows that the deprotonation

energy is only slightly lower for iron(III) (DDG(PT₁-deprot)=

0.7 kcalmol^À1). The corresponding energy gain from copper(II)

to aluminum(III) (DpK_a=2.5) is distinctively larger (DDG(PT₁-

deprot)=6.2 kcalmol^À1).

In summary, the main problem of the magnesium(II) com-

plexes is their low acidity and the missing association of the

hydroxyl groups of glucopyranose, which renders the reaction

1

2

Figure 6. Relative free energy of the PT₁(O1 deprotonation ( ), O5 protonation ( )) and TS₂steps ( ) for [MO O (H₂O)₄]_n⁺(top, left) and [MO¹O²(H₂O)_4-aCl_a]ⁿ⁺

,

&

*

^

a=0–3 (top, right and bottom). pK_aof metal complex: Fe³⁺=2.2, Cr³⁺=4.1, Al³⁺=5.5, Cu²⁺=8.0, Mg²⁺=11.4, water=15.7. PBE0/6-31+G**, CPCM (water).

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

9

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

These are not the final page numbers! ÞÞ

Full Paper

The influence of the metal complex on the protonation of

Zirconium(IV) is a group one cation (high conversion, but

only 5% HMF) with a pK_avalue of À0.3 and empty d orbitals.

In aqueous solution, zirconium(IV) has the coordination

number 8 and is coordinated by water and hydroxyl ligands.^[22]

The high Brønsted acidity should hinder the reaction, whereas

the empty d orbitals appear perfectly suited for the reaction.

Because zirconium(IV) has a high conversion rate but only

a low HMF yield [similar to copper(II) and iron(III)] and a high

yield of LA, we assumed, in view of the strong Brønsted acidity

of the cation complex, that the reaction is acid catalyzed and

thus takes a pathway different from the one described in this

study; there may also be subsequent acid-catalyzed reactions

that reduced the yield of HMF.

&

O5 in PT₁was also plotted (Figure 6, ). Compared to the un-

catalyzed reaction, the protonation energy rises with decreas-

ing pK_avalues, but the influence of the metal cation is less ap-

parent than in the case of deprotonation. Acidic metal cations

that also have a high Lewis acidity better stabilize the depro-

tonated O1 atom of the glucopyranose complex. This results in

a longer C1ÀO1 and a shorter C1ÀO5 bond, which makes it

harder to protonate O5 [(C1ÀO5 bond lengths=1.41 (uncata-

lyzed)>1.38 ([MgO¹O²(H₂O)₄]²⁺)>1.35 ([AlO¹O²(H₂O)₄]³⁺)ꢀ1.35

([CrO¹O²(H₂O)₄]³⁺)ꢀ1.36 ꢂ ([FeO¹O²(H₂O)₄]³⁺). Again, the differ-

ence between the cations iron(III) and aluminum(III) is rather

small when the difference in their pK_avalues is considered

(DDG(PT₁-prot)=2.6 kcalmol^À1), whereas 5.9 kcalmol^À1are

gained for aluminum(III) versus copper(II).

Bismuth(III) is a group two cation. Experimentally, the cation

is coordinated in water with three hydroxyl and three water li-

gands at pH 7.^[22]Its strong Brønsted acidity makes it likely that

the reaction again follows proton-catalyzed pathways.

Hydrogen transfer

Comparison with previous computational studies

Analogously to the deprotonation of O1, the TS₂barrier is

The only previous study performed in water^[6g]employed

CPMD simulations to investigate two Cr³⁺species [Cr-

(H₂O)₅OH]²⁺together with one glucose molecule in a box of

water. Similar to our case, the first solvation shell around the

metal cation consisted of six oxygen atoms, with the O1 and

O2 hydroxyl groups of glucose able to bind to the metal

cation. For the monomeric complex, an association with glu-

cose was found to be more favorable than a hydrogen-bond-

ing interaction. The dimeric catalyst complex did not coordi-

nate to glucose, but separated into two monomers during the

course of the simulation. These results are in agreement with

our findings of an unfavorably high energy for dimer formation

and higher TS₂barriers for chromium(III) dimers compared to

monomers. However, there are also differences. In the previous

study,^[6g]a metal complex with one hydroxyl ligand was pro-

posed as the active catalyst. The hydroxyl ligand was reported

to deprotonate O2 before TS₂and to protonate O1 after TS₂,

analogous to our findings for the proton transfers DP₂and RP₂.

From an Arrhenius plot, the activation energy was estimated

to be 15 kcalmol^À1at 413 K. In our calculations for a reasonable

analogue, the [CrO¹O²(H₂O)₃OH]²⁺catalyst, the deprotonation

step PT₂(after TS₂) was found to require less energy than the

steps DP₂, TS₂, and RP₂. Using thermodynamic cycles to treat

the proton transfers, we arrived at an activation barrier of

34 kcalmol^À1for the rate-determining step, with the hydrogen

transfer alone already requiring an activation of 27 kcalmol^À1

(relative to an open-chain glucose molecule). Because our ex-

periments detected only slow or no conversion at tempera-

tures lower than 1208C, barriers around 30 kcalmol^À1would

seem reasonable, whereas a barrier of 15 kcalmol^À1could al-

ready be easily overcome at room temperature. There is no

clear explanation for the large difference between the two

computed activation energies (except for noting the obvious,

namely that the chosen computational procedures are quite

different).

^

lower for metal complexes with a lower pK_avalue (Figure 6, ).

For pK_a<6 the change in transition-state energy is less pro-

nounced than for higher pK_avalues. For example, 13.5 kcal

mol^À1are gained when going from Cu²⁺to Al³⁺(DpK_a=2.5),

whereas only 2.7 kcalmol^À1are gained between Al³⁺and Fe³⁺

(DpK_a=3.3).

The same analysis has been performed for ligand systems

with 0–3 chloride ligands (Figure 6). Chloride ligands increase

the energy for the deprotonation of O1 and for TS₂, but slight-

ly lower the energy for the protonation of O5. This is to be ex-

pected because chloride ligands lower the Brønsted and Lewis

acidity of the catalysts. The same holds for hydroxyl ligands.

It seems that once a certain Lewis and Brønsted acidity is

reached (pK_a=4–6), the influence of the metal cation on

a given ligand system is small and a limit of improvement is

approached. Low yields for acidic catalysts below this pK_a

threshold must be explained by other factors, for example,

their ligand system or electron configuration. Catalysts with

lower pK_avalues can also access other pathways (mainly

proton catalyzed).

We conclude that aluminum(III) and chromium(III) cations

work successfully for the reaction, because their Brønsted and

Lewis acidity is high enough to catalyze the PT₁and TS₂steps

well, but not so high that they would exist mainly in their de-

protonated form and attract multiple chloride ligands. These

cations also have empty orbitals, which allows them to accept

electron density during the reaction (when needed). On the

other hand, iron(III) seems to be too acidic, thus favors com-

plexes with too many hydroxyl and chloride ligands, and its

prevalent high-spin complexes are not well suited to act as

electron acceptors. Copper(II) and magnesium(II) do not have

sufficient Lewis acidity to be effective.

To conclude this report, we return to the experimental re-

sults. More metal cations were investigated experimentally

than were explored in the computational study, thus we

checked if our qualitative notions could rationalize the results

for these additional cations.

All other previous computational studies were carried out in

ionic liquids.^[7,8]In these solvents, there are different active spe-

&

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

10

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Full Paper

cies (MCl₄^2À), and imidazole cations and chloride ligands are as-

sumed to support the reaction, unlike the situation in water.^[16]

Therefore, the chemistry and the barriers are expected to be

different in these systems (see the Supporting Information for

further comments).

periment was repeated several times with different catalysts and

the yields were reproduced within analytical accuracy.

The products were analyzed by HPLC by using a Shimadzu instru-

ment equipped with a Shimadzu CBM-20A controller, an organic

acid resin column (300ꢃ8 mm), two pumps (Shimadzu LC-20AB &

LC-20AD), an auto injector (Shimadzu SIL-20AC), and a column

oven (Shimadzu CTO-20A). HMF was detected with a UV detector

(Shimadzu SPD-M20A & SPD-M30A). All other compounds were an-

alyzed with a refractive index (RI) detector (Shimadzu LC-10A). Tri-

fluoroacetic acid (2 mm) in water was used as the eluent at 408C

There have also been many studies on the proton-catalyzed

reactions.^{[5b,c,6d,e,10]}They conclude that under highly acidic con-

ditions different hydroxyl groups of glucopyranose, which are

more basic than the pyranic oxygen, can be protonated.^[30]

After such protonation, the water moieties formed can leave

as neutral species, followed by a cationic rearrangement of the

pyranose ring to give different products. These results are rele-

vant for metal complexes that act as a Brønsted acid rather

than a Lewis acid because these proton-catalyzed pathways

can lead to HMF without a fructose intermediate.^[8b]

with a flow rate of 1.0 mLmin^À1

.

Acknowledgements

Financial support from the research cluster Sustainable Chemi-

cal Synthesis—A Systems Approach (SusChemSys) is acknowl-

edged. SusChemSys is co-financed by the European Regional

Development Fund (ERDF) and the state of North Rhine-West-

phalia, Germany, under the Operational Program Regional

Competitiveness and Employment 2007–2013.

Conclusion

We calculated the preferred mechanism for the uncatalyzed

and metal-catalyzed glucose–fructose isomerization in water

by using DFT. This allowed us to rationalize many of the avail-

able experimental results and suggest criteria for an effective

catalyst.

Keywords: biomass · computer chemistry · Lewis acids ·

reaction mechanisms · sustainable chemistry

Metal cations with moderate Lewis acidity and moderate

Brønsted acidity (pK_a=4–6) are good catalysts. Ideally, they

have only water ligands attached, or ligands further up the

spectrochemical series that are weak s donors and can accept

p back donation (e.g. CN^Àand CO). The ligands should be

moderately strong p acceptors at most; p-acceptor ligands will

increase the ligand field separation and, thus, may hinder ac-

ceptance of electron density during the hydrogen-transfer step

of the reaction. Bad ligands are strong s- and p-donors, such

as halogenides. Low-lying unoccupied orbitals that can accept

electron density, for example, unoccupied metal d orbitals are

helpful. It is also an advantage if the metal complex allows co-

ordination of glucose as a ligand because this lowers the

energy relative to catalysts with hydrogen bonding to glucose.

Aluminum(III) and chromium(III) fulfill these criteria, hence,

they yield more HMF than iron(III), copper(II), and magnesiu-

m(II). Dimeric catalysts are less effective because the energy for

complex formation is high for aluminum(III) and chromium(III)

and it is entropically less favorable for glucose to meet two

metal cations rather than one. The high temperature of the ex-

periments is necessary to overcome the computed barriers for

the different catalysts investigated (20–35 kcalmol^À1).

[1] G.-H. Wang, J. Hilgert, F. H. Richter, F. Wang, H.-J. Bongard, B. Spliethoff,

C. Weidenthaler, F. Schꢀth, Nat. Mater. 2014 , 13 , 29 3 –3 00.

[2] A. A. Rosatella, S. P. Simeonov, R. F. M. Frade, C. A. M. Afonso, Green

Chem. 2011 , 13 , 75 4 –7 93.

[3] a) T. Stꢄhlberg, W. Fu, J. M. Woodley, A. Riisager, ChemSusChem 2011 , 4 ,

45 1 –4 58; b) G. Yong, Y. Zhang, J. Y. Ying, Angew. Chem. 2008 , 120 ,

948 5 –9 488; Angew. Chem. Int. Ed. 2008 , 47 , 934 5 –9 348; c) S. Hu, Z.

Zhang, J. Song, Y. Zhou, B. Han, Green Chem. 2009 , 11 , 174 6 –1 74 9;

d) H. Zhao, J. E. Holladay, H. Brown, Z. C. Zhang, Science 2007 , 316 ,

159 7 –1 600.

[4] J. B. Binder, R. T. Raines, J. Am. Chem. Soc. 2009 , 131 , 197 9 –1 985.

[5] a) T. Wang, Y. J. Pagꢅn-Torres, E. J. Combs, J. A. Dumesic, B. H. Shanks,

Top. Catal. 2012 , 55 , 65 7 –6 62; b) X. Qian, M. R. Nimlos, M. Davis, D. K.

Johnson, M. E. Himmel, Carbohydr. Res. 2005 , 340 , 231 9 –2 327; c) T.

Stꢄhlberg, S. Rodriguez-Rodriguez, P. Fristrup, A. Riisager, Chem. Eur. J.

2011 , 17 , 145 6 –1 464.

[6] a) Y. Roman-Leshkov, J. N. Chheda, J. A. Dumesic, Science 2006 , 312 ,

193 3 –1 937; b) Y.-N. Li, J.-Q. Wang, L.-N. He, Z.-Z. Yang, A.-H. Liu, B. Yu,

C.-R. Luan, Green Chem. 2012 , 14 , 275 2 –2 75 8; c) S. Caratzoulas, D. G.

Vlachos, Carbohydr. Res. 2011 , 346 , 66 4 –6 72; d) R. S. Assary, P. C. Red-

fern, J. Greeley, L. A. Curtiss, J. Phys. Chem. B 2011 , 115 , 434 1 –4 349;

e) R. S. Assary, T. Kim, J. J. Low, J. Greeley, L. A. Curtiss, Phys. Chem.

Chem. Phys. 2012 , 14 , 1660 3 –1 6611; f) M. J. Antal, W. S. L. Mok, G. N. Ri-

chards, Carbohydr. Res. 1990 , 199 , 9 1 –1 09; g) V. Choudhary, S. H. Mush-

rif, C. Ho, A. Anderko, V. Nikolakis, N. S. Marinkovic, A. I. Frenkel, S. I.

Sandler, D. G. Vlachos, J. Am. Chem. Soc. 2013 , 135 , 399 7 –4 006.

[7] J. Guan, Q. Cao, X. Guo, X. Mu, Comput. Theor. Chem. 2011 , 963 , 45 3 –

462.

[8] a) E. A. Pidko, V. Degirmenci, R. A. van Santen, E. J. M. Hensen, Angew.

Chem. 2010 , 122 , 258 4 –2 588; Angew. Chem. Int. Ed. 2010 , 49 , 253 0 –

2534; b) E. A. Pidko, V. Degirmenci, E. J. M. Hensen, ChemCatChem 2012 ,

4 , 126 3 –1 271.

Experimental Section

[9] G. Yang, E. A. Pidko, E. J. M. Hensen, ChemSusChem 2013 , 6 , 168 8 –1 696.

[10] G. Yang, E. A. Pidko, E. J. M. Hensen, J. Catal. 2012 , 295 , 12 2 –1 32.

[11] C. Adamo, V. Barone, J. Chem. Phys. 1999 , 110 , 615 8 –6 169.

[12] Gaussian 09, Revision B.1, M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E.

Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, B. Men-

nucci, G. A. Petersson, H. Nakatsuji, M. Caricato, X. Li, H. P. Hratchian,

A. F. Izmaylov, J. Bloino, G. Zheng, J. L. Sonnenberg, M. Hada, M. Ehara,

K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O.

Kitao, H. Nakai, T. Vreven, J. A. Montgomery, Jr., J. E. Peralta, F. Ogliaro,

The reactions were carried out as follows: glucose (0.616 mmol)

and catalyst (0.070 mmol) were dissolved in distilled water (3 mL).

The reaction vessel was placed in an autoclave (22 mL) with a glass

inlet. The autoclave was closed and filled with nitrogen (60 bar),

then depressurized to expel oxygen. The autoclave was filled again

with nitrogen (60 bar) and heated to 413 K with stirring. The reac-

tion mixture was stirred at 413 K for 1 h. The reaction was stopped

by letting the autoclave cool down rapidly in an ice bath. The ex-

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

11

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

These are not the final page numbers! ÞÞ

Full Paper

M. Bearpark, J. J. Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov, R. Ko-

bayashi, J. Normand, K. Raghavachari, A. Rendell, J. C. Burant, S. S. Iyen-

gar, J. Tomasi, M. Cossi, N. Rega, J. M. Millam, M. Klene, J. E. Knox, J. B.

Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann,

O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, R. L. Martin,

K. Morokuma, V. G. Zakrzewski, G. A. Voth, P. Salvador, J. J. Dannenberg,

S. Dapprich, A. D. Daniels, ꢆ. Farkas, J. B. Foresman, J. V. Ortiz, J. Cio-

slowski„ D. J. Fox, Gaussian, Inc., Wallingford CT, 2009.

[13] G. I. Csonka, A. D. French, G. P. Johnson, C. A. Stortz, J. Chem. Theory

Comput. 2009 , 5 , 67 9 –6 92.

[14] S. F. Sousa, P. A. Fernandes, M. J. Ramos, J. Phys. Chem. A 2007 , 111 ,

1043 9 –1 0452.

[15] a) J. M. del Campo, J. L. Gꢅzquez, S. B. Trickey, A. Vela, J. Chem. Phys.

2012 , 136 , 104108; b) Y. Zhao, D. G. Truhlar, Acc. Chem. Res. 2008 , 41 ,

15 7 –1 67; c) Y. Zhao, D. G. Truhlar, J. Chem. Theory Comput. 2009 , 5 ,

32 4 –3 33.

[23] NBO Version 3.1, E. D. Glendening, A. E. Reed, J. E. Carpenter, F. Wein-

hold, University of Wisconsin, Madison, 1998.

[24] Y. Romꢅn-Leshkov, M. Moliner, J. A. Labinger, M. E. Davis, Angew. Chem.

2010 , 122 , 913 8 –9 141; Angew. Chem. Int. Ed. 2010 , 49 , 895 4 –8 957.

[25] TS₁, PT₂, TS₃, and PT₃were not analyzed because they are generally

lower in energy or of the same energy as PT₁and TS₂. The energy for

separation of the open-chain fructose–catalyst complex was not deter-

mined because the ring closure (TS₃) requires more activation for the

uncatalyzed than for the catalyzed reaction. The ring has to be estab-

lished if HMF is to be formed later in the reaction. The deprotonated

and protonated species involved in the proton transfers are not taken

as reference points because they are not stable minima in the reaction;

their energies were generated with a thermodynamic cycle.

[26] S. Bogatko, P. Geerlings, Phys. Chem. Chem. Phys. 2012 , 14 , 805 8 –8 066.

[27] Entry of Merck Millipore for CAS 10125–13–0, copper(II) chloride dihy-

drate.

[16] E. A. Pidko, V. Degirmenci, R. A. van Santen, E. J. M. Hensen, Inorg. Chem.

2010 , 49 , 1008 1 –1 0091.

[28] L. Nagy, H. Ohtaki, T. Yamaguchi, M. Nomura, Inorg. Chim. Acta 1989 ,

159 , 20 1 –2 07.

[17] a) C. Gonzalez, H. B. Schlegel, J. Chem. Phys. 1989 , 90 , 215 4 –2 161; b) C.

Gonzalez, H. B. Schlegel, J. Phys. Chem. 1990 , 94 , 552 3 –5 527.

[18] V. Barone, M. Cossi, J. Phys. Chem. A 1998 , 102 , 199 5 –2 001.

[19] Y. Takano, K. N. Houk, J. Chem. Theory Comput. 2005 , 1 , 7 0 –7 7.

[20] A. V. Marenich, C. J. Cramer, D. G. Truhlar, J. Phys. Chem. B 2009 , 113 ,

637 8 –6 396.

[29] J. Briggs, P. Finch, M. C. Matulewicz, H. Weigel, Carbohydr. Res. 1981 , 97 ,

18 1 –1 88.

[30] T. Kurahashi, T. Mizutani, J. Yoshida, J. Chem. Soc. Perkin Trans. 1 1999 ,

46 5 –4 74.

[21] S. A. Bogatko, E. J. Bylaska, J. H. Weare, J. Phys. Chem. A 2010 , 114 ,

218 9 –2 200.

Received: March 1, 2014

[22] A. F. Hollemann, N. Wiberg, Lehrbuch der Anorganischen Chemie,

Vol. 101, Walter de Gruyter, Berlin, New York, 1995.

Published online on && &&, 0000

&

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

12

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

ÝÝ These are not the final page numbers!

Full Paper

FULL PAPER

&

Transition Metals

C. Loerbroks, J. van Rijn, M.-P. Ruby,

Q. Tong, F. Schꢀth, W. Thiel*

&& – &&

The mechanism of glucose–fructose

conversion was investigated for differ-

ent transition-metal catalysts in water

(see scheme). The mechanism is influ-

enced by the Lewis and Brønsted acidity

of the metal catalysts, which are gov-

erned by their electronic structure for

a given set of ligands.

Reactivity of Metal Catalysts in

Glucose–Fructose Conversion

Chem. Eur. J. 2014, 20, 1 – 13

www.chemeurj.org

13

ꢁ 2014 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

&

These are not the final page numbers! ÞÞ

Article Doi

Reactivity of metal catalysts in glucose-fructose conversion

DOI: 10.1002/chem.201402437

Source and publish data:

Authors:

Article abstract of DOI:10.1002/chem.201402437

Full text of DOI:10.1002/chem.201402437

Products guided by the article

R&D Labs maybe for 123-76-2

Relevant to this article

Hot Product