Radicalization and Radical Catalysis of Biomass Sugars: Insights from First-principles Studies.

Ab initio and density functional calculations are conducted to investigate the radicalization processes and radical catalysis of biomass sugars. Structural alterations due to radicalization generally focus on the radicalized sites, and radicalization affects H-bonds in D-fructofuranose more than in D-glucopyranose, potentially with outcome of new H-bonds. Performances of different functionals and basis sets are evaluated for all radicalization processes, and enthalpy changes and Gibbs free energies for these processes are presented with high accuracy, which can be referenced for subsequent experimental and theoretical studies. It shows that radicalization can be utilized for direct transformation of biomass sugars, and for each sugar, C rather than O sites are always preferred for radicalization, thus suggesting the possibility to activate C-H bonds of biomass sugars. Radical catalysis is further combined with Brønsted acids, and it clearly states that functionalization fundamentally regulates the catalytic effects of biomass sugars. In presence of explicit water molecules, functionalization significantly affects the activation barriers and reaction energies of protonation rather than dehydration steps. Tertiary butyl and phenyl groups with large steric hindrances or hydroxyl and amino groups resulting in high stabilities for protonation products drive the protonation steps to occur facilely at ambient conditions.

Petroleum resources are expected to be exhausted in the next few decades, and alternative energy sources are being sought. Owing to the abundance, sustainability and easy availability, cellulosic biomass has been regarded as a promising energy source to replace petroleum123; nonetheless, the direct transformation of cellulosic biomass to downstream products remains a grand challenge4.

Glucose, the monomer of cellulose, has often been used as protype to investigate the transformation of cellulosic biomass56789101112131415. The Brønsted-acid catalysis of glucose is driven towards the formation of humin precursors and reversion products, while under similar conditions, fructose can be activated readily and with a sequence of facile reaction steps converted to 5-(hydroxymethyl)furfural (HMF) or other platform chemicals161718. Sn-BEA zeolite and HCl solutions were skilfully combined by Nikolla et al.19, realizing the “one-pot” synthesis of HMF from glucose with relatively high conversion and selectivity. On the other hand, glucose has lower acidity than fructose20212223 and hence is less reactive under basic conditions. Recently, Liu et al.24 have demonstrated that organic amines are efficient for the isomerization of glucose to fructose. These clearly state that whether in acidic or basic environments, fructose acts as the pivotal intermediate for the transformation of cellulosic sugars.

In the past few decades, radical catalysis has emerged into a flourishing area2526, while reports relating to biomass transformations are scarce. Electron magnetic resonances were used to study the irradiation of crystalline D-glucose, D-fructose and sucrose27282930, and several C-centered radicals were assigned with help of density functional calculations303132. The radical stabilities can be expressed using the isodesmic H-transfer reactions3334,

Reaction enthalpy of Eq. 1 is also referred to as radical stabilization energy (RSE). The RSE calculations with acceptable accuracy allows for quantitative estimation of H-transfer reaction energies and optimization of the properties for new reagents35. A comprehensive understanding of C-centered sugar radicals is lacking, which was presently conducted with high-level ab initio calculations (MP2/aug-cc-pVTZ). D-glucose (monomer of cellulosic biomass) and D-fructose (pivotal intermediate for cellulose utilization), in both α- and β-anomers, have been considered. To best of our knowledge, all previous sugar conversions focused on O sites, which are preferred over C sites during protonation161718363738 and deprotonation20212223. Here a systematic study was also conducted for O-centered sugar radicals, and comparisons with the results of C-centered radicals clearly indicated that the formation of C-centered radicals is always preferential, thus suggesting the possibility to activate and convert the C-H bonds of biomass sugars.

Csonka et al.39 evaluated the performances of different density functionals and basis sets for the various β-D-glucose conformations. However, performances of different density functionals and basis sets for sugar radicals remain elusive, which will be conducted in this work. On such basis, the enthalpy changes and Gibbs free energies for all radicalization processes were presented with high accuracy, which can be referenced for subsequent experimental measurements. Finally, a variety of radical catalytic routes were designed and it clearly demonstrated that radicalization that focuses on activation of C sites can be used for the transformation of biomass sugars; in addition, we found that the presence of explicit water molecules can significantly alter the reaction paths and energies. These are helpful for the transformation of biomass sugars that can solve the global energy crisis.

Computational Details

In line with previous works8910111213161718, the lowest-energy conformers of α,β-D-glucopyranose and α,β-D-fructofuranose were the choice for studies, which were respectively referred to as αG, βG and αF, βF (Fig. 1). The various O and C sites potential to be radicalized were identified by atomic numbering; e.g., radicals corresponding to α-D-glucopyranose at O1 site and β-D-fructofuranose at C4 site were designated to be αGrO and βFrC, respectively. Different electronic states were considered4041, and radicals in doublet states that show obviously superior stabilities will be discussed unless otherwise specified.

Figure 1

Structures of β-D-glucopyranose (βG), α-D-glucopyranose (αG), β-D-fructofuranose (βF) and α-D-fructofuranose (αF).

All calculations were conducted with Gaussian09 suite of programs42. Structural optimizations of sugars and their radicals corresponding to the various O/C sites were performed at MP2/aug-cc-pVTZ (denoted as bs4) level39. The radicalization process and radical generation energies (ΔEr) are shown as,

Performances of other functionals and basis sets were then evaluated, using MP2/bs4 to be benchmark as suggested elsewhere39: 1) Hartree-Fock (HF)43 as well as B3LYP4445, BP8646, PBE1PBE47 and M06L4849 density functionals, in combination with the bs4 basis set; 2) Because of relatively fine agreement with MP2 results, B3LYP was further employed with 6-31G(d), 6-31 + G(d, p) and 6-311++G(d, p) (referred to bs1, bs2, and bs3, respectively) to address the effect of basis sets; 3) Single-point energies were run at MP2/bs4 level, on basis of B3LYP optimized structures. The ΔEr values of 1)~3) may deviate from those directly from MP2/bs4, and such deviations were defined to be δΔE(i) (i being the atomic numbering of O/C site). Performances of these functionals and basis sets can be assessed by the average of δΔE(i) at the various O/C sites (<δΔE)>) and standard deviations of δΔE(i) (S.D.).

The enthalpy changes (ΔH) and Gibbs free energies (ΔG) for the radicalization processes can be given as20,

where S, T and R stand for entropy (in terms of tranlational, rotational and vibrational contributions), temperature and gas constant, respectively.

The thermodynamic parameters were calculated at B3LYP/bs2 level and then corrected by δΔE(i), which have been testified by a large number of cases to achieve comparable results as those directly from MP2/bs4 method (Tables S1–S4 and Figures S1 and S2),

The solvent effects were accounted for by the self-consistent isodensity polarizable continuum model (SCI-PCM) of self-consistent reaction field (SCRF)50, in combination with B3LYP/bs2 method. The default dielectric constant (ε = 78.4) was used for water solvent.

For catalytic systems, the two-layer ONIOM methodology (MP2/bs4//M06L/bs3)5152 that has been validated sufficiently (Tables S1, S5 and Figures S3 and S4) was employed for energy calculations, on basis of B3LYP/bs2 optimized structures. The radicalized C/O sites and neighbouring C/O groups as well as adsorbents (proton and water molecules) were defined as the high-level regions, while the rest were treated as the low-level regions. This methodology can be applied for the accurate energy calculations of other sugar and larger catalytic systems.

Results and Discussion

Radicalization of D-glucopyranose

As indicated in Fig. 1, D-glucopyranose conformers (βG and αG) are inclined to construct successive H-bonds (Oi+1Hi+1…Oi type; e.g., O4H4…O3: 2.407 Å, O3H3…O2: 2.436 and O2H2…O1: 2.485 Å in βG,)2053. Figures 2 and 3 show that no proton migration or ring opening occurs for the radicalization of D-glucopyranose conformers, whether at O or at C sites. Structural alterations are generally small and restricted at the radicalization sites, and two C-centered radicals previously observed during the irradiation of crystalline α-D-glucopyranose show agreement with present results272832.

Figure 2

Optimized structures for the radicalization of β-D-glucopyranose (βG) at the various O/C sites.

Figure 3

Optimized structures for the radicalization of α-D-glucopyranose (αG) at the various O/C sites.

Radicalization at O sites reinforces associated C-O bonds; e.g., the C3-O3 bonds are equal to 1.420 and 1.378 Å in βG and βGrO, respectively. The electronic densities on radicalized O atoms decline and thereby proximate H-bonds are impaired; e.g., the Mulliken charges of O3 in βG and βGrO amount to −0.834 and −0.802, causing the elongation of O4H4…O3 H-bond from 2.407 to 2.429 Å. Radicalization at C sites strengthens the associated C-O and C-C bonds, where the pyranyl C-C/C-O bonds are apparently less affected than dangling C-O bonds; e.g., the C2-C3, C3-C4 and C3-O3 bonds are 1.511, 1.513 and 1.420 Å in βG and 1.486, 1.486 and 1.363 Å in βGrC, respectively. The C3-O3 bond has been corroborated by the delocalized π-electron system developed by the lone-pair electrons of O3 and half-empty orbital of C3, as evidenced by electron transfers from O3 to C3. Whether in α- or in β-anomer, the hydroxyl (-OH) groups are alternatively up and down with respect to the pyranyl ring that are not beneficial for the formation of Oi+1Hi+1…Oi type H-bonds, whereas such H-bonds can be substantialized due to radicalization at C sites; e.g., the C2C4C3O3 dihedrals are optimized at 122.32° and 143.62° in βG and βGrC, and the larger dihedrals are in favour of O3H3…O2 and O4H4…O3 H-bonds that are 2.436 and 2.407 Å in βG while 2.225 and 2.367 Å in βGrC, respectively. C2 site of αG is an exception, where the −O2H2 and −O1H1 groups fall at the same side of pyranyl ring, and radicalization elongates O2H2…O1 H-bond distance from 2.210 to 2.315 Å.

Radicalization of D-fructofuranose

Radical structures of D-fructofuranose conformers (αF and βF) are given in Figs 4 and 5. The structural alterations due to radicalization are generally local and focus on radicalization sites, which is in line with the condition of D-glucopyranose conformers54. Proton migration/ring opening occurs only at O2 site of αF, where the furanyl ring opens by breaking C2-C3 bond with production of the C2-O2 and C3-O3 delocalized π-electron system.

Figure 4

Optimized structures for the radicalization of α-D-fructofuranose (αF) at the various O/C sites.

Figure 5

Optimized structures for the radicalization of β-D-fructofuranose (βF) at the various O/C sites.

Radicalization at O sites reinforces dangling C-O bonds and at C sites also strengthens associated furanyl C-C/C-O bonds; meanwhile, proximate H-bonds are elongated, which are consistent with the results of D-glucopyranose conformers; e.g., O2H2…O1 H-bond are 2.211 Å in αFrO and 2.191 Å in αF. However, the changing trends of H-bonds become elusive for radicalization at C sites as a result of inconsistent roles played by interlaced H-bonds: 1) Shortening as in D-glucopyranose; e.g., O1H1…O5 H-bond: 2.057 Å in βFrC vs. 2.602 Å in βF; 2) Apparent elongation; e.g., O2H2…O1 H-bond: 2.704 Å in αFrC vs. 2.191 Å in αF; 3) Construction of new H-bond; e.g., the O3H3…O4 H-bond: 2.449 Å in αFrC. Radicalization causes associated hydroxyl (-OH) groups to approach the furanyl plane; e.g., the C2C4C3O3 dihedrals are 118.98° and −141.97° in αF and αFrC, respectively, and consequently, HO3H3…O6 H-bond is damaged by radicalization of αF at C3 site, and the −O3H3 group shifts to the other side of furanyl ring and creates O3H3…O4 H-bond. That is, radicalization at C sites in D-fructofuranose rather than D-glucopyranose conformers result in more pronounced effects on proximate H-bonds.

Thermodynamic Calculations

The ΔEr and RSE values of sugars (βG, αG, βF and αF) corresponding to the various O/C sites are calculated at MP2/bs4 level (Tables 1 and S4). Eqs 1 and 3 show that they can be correlated with each other,

Table 1

MP2/bs4 calculated radical generation energies (ΔEr) for D-glucopyranose and D-fructofuranose conformers at the various O/C sitesa,b.

	βG		αG		βF		αF
	O	C	O	C	O	C	O	C
C1/O1	495.5	427.3	500.5	441.1	496.7	429.2	494.3	440.1
C2/O2	505.9	438.0	503.8	443.9	516.6		436.4
C3/O3	506.0	429.2	506.6	431.5	505.7	430.0	500.4	440.7
C4/O4	506.5	427.7	510.7	431.1	497.4	428.0	504.6	432.4
C5/O5		439.8		443.1		435.1		434.1
C6/O6	495.8	428.6	503.1	425.3	502.4	422.4	494.5	434.7
Average	501.9	431.8	504.9	436.0	503.8	428.9	486.0	436.4

aEnergy units in kJ/mol.

bIn each case, the lowest radical generation energy has been highlighted in bold.

Accordingly, ΔEr and RSE have exactly the same changing trends. The ΔEr data of D-glucopyranose and D-fructofuranose can be comparable, and for each sugar (βG, αG, βF and αF), the ΔEr values corresponding to the various O/C sites are close to each other, and C rather than O sites are preferred for radicalization, which is consistent with the experimental observations272829303132. In addition, the ΔEr data of two anomers (α/β) can differ substantially, probably as a result of anomeric effects and H-bonding diversities2053.

As indicated in Figure S1 and Table S1, density functional methods are apparently superior to HF for treatment of radical sugars and obtain close results as MP2. Then B3LYP, currently one of the most used density functionals, is combined with different basis sets to demonstrate the effect of basis sets. It shows that bs2 and bs3 achieve comparable results with bs4 while bs1 deviates significantly (Figure S2 and Table S1), suggesting that diffuse functions are necessary for treating sugar radicals.

Balancing the computional cost and accuracy, B3LYP/bs2 is selected for further studies, and a number of cases in Tables S2 and S3 demonstrate that the ΔH and ΔG data calculated at B3LYP/bs2 level deviate remarkably from those of MP2/bs4, while those corrected by energy deviations (δΔE(i)) are also reproducible. G4MP2 is probably the most familiar composite method for energy calculations55, and other composite methods with zero-point vibrational energies being computed at lower theoretical levels have also been reported5657. The ΔH and ΔG data for radicalization of sugars at the various O/C sites are calculated this way and collected in Table 2. The lowest ΔHr and ΔGr values for O sites are 466.1 (O1) and 381.2 (O1) kJ/mol for βG, 470.0 (O1) and 440.8 (O1) kJ/mol for αG, 466.4 (O1) and 437.2 (O1) kJ/mol for βF and 407.5 (O2) and 377.1 (O2) kJ/mol for αF while for C sites are 401.4 (C4) and 351.3 (C3) kJ/mol for βG, 398.2 (C6) and 368.7 (C6) kJ/mol for αG, 394.9 (C6) and 365.1 (C6) kJ/mol for βF and 405.3 (C4) and 372.2 (C4) kJ/mol for αF, respectively. The radical structure of αG at C6 site has been identified before32, and this coincides well with the present results that αGrC has the lowest ΔHr and ΔGr values. The comprehensive thermodynamic parameters given in Table 2 can be referenced for subsequent experimental measurements. Solvent effects are further included for the calculations of ΔH and ΔG, and it shows that these thermodynamic parameters are affected slightly by addition of water solvent (Tables 2 and S6).

Table 2

Enthalpy changes (ΔHr) and Gibbs free energies (ΔGr) for the radicalization of D-glucopyranose and D-fructofuranose conformersa,b.

	βG		αG		βF		αF
	ΔHr	ΔGr	ΔHr	ΔGr	ΔHr	ΔGr	ΔHr	ΔGr
O1	466.1	381.2	470.0	440.8	466.4	437.2	464.0	435.1
O2	477.5	399.3	473.5	443.8	486.7	455.7	407.5	377.1
O3	476.3	399.8	476.1	446.9	477.4	449.4	471.5	442.2
O4	476.8	401.7	482.2	453.7	469.7	439.1	475.6	442.9
O6	465.5	392.7	472.4	441.4	471.0	440.0	463.5	434.9
C1	401.8	353.0	413.8	382.5	401.7	370.4	412.1	380.1
C2	411.6	360.3	416.2	381.8
C3	403.1	351.3	405.1	374.1	403.6	373.4	413.6	378.2
C4	401.4	355.7	404.8	373.9	401.3	371.7	405.3	372.2
C5	413.9	358.7	415.6	380.6	408.1	373.7	407.4	374.6
C6	401.9	359.0	398.2	368.7	394.9	365.1	407.0	377.0

aEnergy units in kJ/mol.

bIn each case, the lowest ΔGr value has been highlighted in bold.

Radical Catalysis

As discussed earlier, for each sugar (βG, αG, βF and αF), the most preferential site for radicalization is always at C sites; that is, radicalization is particular by offering a route to activate the C-H bonds in sugars, in contrast to previous attempts focusing on catalysis of O sites. In addition, the ΔEr data of C sites in D-glucopyranose and D-fructofuranose conformers are close to each other suggesting a comparable reactivity, and this provides insightful clues to the direct catalytic transformation of D-glucose instead of using D-fructose as the immediate product as widely recommended27891011121319.

Thermodynamic calculations in Section 3.3 require structural optimizations at MP2/bs4 level that is computationally costly and seems not appropriate for catalytic studies. Here an alternative method is used: two-layer ONIOM(MP2/bs4//M06L/bs3) energies are calculated on basis of B3LYP/bs2 optimized structures. Tables S1, S5 and Figures S3 and S4 show that the alternative obtains comparable ΔEr and RSE data as the method in Section 2; in addition, this methodology can be safely applied for the accurate energy calculations of other sugar and larger catalytic systems. Figure 6 gives an example for the direct radical transformation of sugars. The conversion from βGrC to βGrC through H-radical (H•) transfer requires an overwhelmingly large activation barrier (ΔEa = 199.5 kJ/mol) that is difficult to proceed at ambient conditions; however, the direct radical transformation of βGrC is facile, and the H•-deprivation reaction (βGrC + CH3• → βGH + CH4) is barrierless, which is further driven by the high exothermicity (ΔEr = −299.7 kJ/mol). In βGH, the pyranyl ring is opened with emergence of two functional groups (−C1H = O and −C5 = O) that are ready for the future chemical synthesis.

Figure 6

Illustration of catalytic transformation of C-centered sugar radicals.

Energy calculations (kJ/mol) are reported at MP2/bs4//B3LYP/bs2 level. Deprivation of H radical (H•) is realized with the reaction of βGrC + CH3• → βGH + CH4.

A variety of functional groups can be implanted by radical catalysis2658, see Table 3. The reaction energies of R1~R8 with βGrC (ΔEr) are calculated at ONIOM(MP2/bs4//M06L/bs3)//B3LYP/bs2 level and affected pronouncedly by the radical stabilities: ΔEr generally has the opposite trend as RSE (Table 3). Functionalized sugars are then combined with Brønsted acids to drive cellulosic sugars to downstream products. For the Brønsted-acid catalysis of D-glucopyranose, the initial protonation and dehydration steps are considered to be rate-determining1659, and these two steps are depicted in Fig. 7. In line with the previous work16, the Brønsted acid is modeled as Zundel complex H5O2+ (an approximation of proton in water), and hence the protonation energy (ΔEpro) is defined as,

Table 3

ONIOM(MP2/bs4//M06L/bs3)//B3LYP/bs2 calculated energies for the radical catalysis of βGrC1a,b.

No.	Chemical formula	RSE	ΔEr	ΔEpro	ΔEdeh
				−13.3	75.2
R1	H3C•	0	−381.0	−52.9	51.7
R2	(CH3)3C•	−55.4	−302.6	−89.4	53.6
R3	CH2 = CHCH2•	−84.2	−295.7	−53.5	47.0
R4	C6H5• (phenyl)	10.9	−399.5	−85.8	46.7
R5	CH3C• = O (acetyl)	−77.8	−325.9	−44.6	91.8
R6	OH•	35.3	−439.2	−46.4	31.2
R7	Cl•	−24.0	−364.2	10.8	52.4
R8	H2N•	−1.7	−388.2	−140.0	47.7

aEnergy units in kJ/mol.

bData of the first line are for βG.

Figure 7

Regulation of catalytic effects of sugars by radical functionalization.

Energy calculations (kJ/mol) are reported at ONIOM(MP2/bs4:M06L/bs3)//B3LYP/bs2 level. High-level regions are displayed in ball and stick while low-level regions in stick.

With regard to βG, the protonation (ΔEpro) and dehydration (ΔEdeh) energies are equal to −13.3 and 75.2 kJ/mol, respectively. As indicated in Table 3 and Fig. 7, both of ΔEpro and ΔEdeh, especially the former, are affected significantly by functionalization; e.g., the protonation step becomes very favourable for R8 (H2N•, ΔEpro = −140.0 kJ/mol), while R7 (Cl•) causes this step to be unfavourable (ΔEpro = 10.8 kJ/mol). That is, the catalytic effects of cellulosic sugars can be fundamentally regulated by functionalization.

Figure 8 shows the protonation (ΔEpro) and dehydration (ΔEdeh) processes of βG in presence of two explicit water molecules. Protonation of βG at O1 site causes the formation of one additional water molecule, and this step is thermodynamically unfavorable (ΔEpro = 90.2 kJ/mol) with a moderate activation barrier (ΔE‡pro = 97.8 kJ/mol). The protonation and dehydration processes of βG with different functional groups (R1~R8) are also calculated and the results are given in Table 4 and Figures S5–S12. It indicates that functionalization affects significantly the activation barriers and reaction energies of the protonation steps. For R5, O1H constructs strong H-bonding with the acetyl O atom (1.772 Å, see Figure S9), and this adds the difficulty of protonation resulting in a considerably large activation barrier (ΔE‡pro = 131.3 kJ/mol). The protonation thermodynamics can be significantly driven by the functional groups with more steric hindrances (e.g., R2, R4 > R1, R3 > H• in βG) or by the high stabilities of protonation products (R6 and R8). In the case of R6, the protonation product is stabilized pronouncedly by two strong H-bonds (O2H2…O1: 2.062 Å and O7H7…O1: 1.469 Å, see Figure S10). In contrast, the reaction energies of dehydration seem not to be much affected by different functional groups, which are generally around 90.0 kJ/mol except R6 corresponding to the particularly stable protonation product (ΔEdeh = 117.9 kJ/mol).

Figure 8

Protonation and dehydration of the O1 site in β-D-glucopyranose (βG) in presence of explicit water molecules.

Energy calculations (kJ/mol) are reported at ONIOM(MP2/bs4:M06L/bs3)//B3LYP/bs2 level. High-level regions are displayed in ball and stick while low-level regions in stick.

Table 4

ONIOM(MP2/bs4//M06L/bs3)//B3LYP/bs2 calculated energies for the radical catalysis of βGrC1 in presence of two water moleculesa,b.

No.	chemical formula	ΔE‡pro	ΔEpro	ΔEdeh
		97.8	90.2	93.8
R1	H3C•	80.0	36.2	92.0
R2	(CH3)3C•	35.5	15.5	84.7
R3	CH2 = CHCH2•	84.4	46.5	85.3
R4	C6H5• (phenyl)	88.8	13.2	81.9
R5	CH3C• = O (acetyl)	131.3	90.3	83.9
R6	OH•	74.1	0.9	117.9
R7	Cl•	111.9	84.6	93.0
R8	H2N•	30.9	−55.7	97.6

aEnergy units in kJ/mol.

bData of the first line are for βG.

Additional Information

How to cite this article: Yang, G. et al. Radicalization and Radical Catalysis of Biomass Sugars: Insights from First-principles Studies. Sci. Rep. 6, 29711; doi: 10.1038/srep29711 (2016).