Ashley
McVeigh
,
Florent P.
Bouxin
,
Michael C.
Jarvis
and
S. David
Jackson
*
Centre for Catalysis Research, WestCHEM, School of Chemistry, University of Glasgow, Glasgow G12 8QQ, Scotland, UK. E-mail: david.jackson@glasgow.ac.uk
First published on 28th January 2016
The depolymerisation of an ammonia treated lignin to alkylphenols over a Pt/alumina catalyst was investigated under a range of process parameters including, pressure, mass of lignin, solvent and gas atmosphere. The depolymerisation was shown to be under kinetic control and orders of reaction in hydrogen and lignin were determined as 0.4 and 0 respectively. Hydrogen was shown to be necessary under our reaction conditions as when helium was used as the gas atmosphere poor conversion was obtained. A clear solvent effect was observed with 100% methanol being more effective than 100% water or any combination of the two with a yield of alkylphenols >40% with a selectivity of >40% to substituted 4-propyl-2,6-dimethoxyphenol compounds. This high yield using methanol as a solvent was thought to be due to the ability of the methanol to inhibit re-polymerisation. IPA/water was also found to be an effective solvent combination with a yield of alkylphenols of >20%. The depolymerisation reaction was also studied over Rh/alumina and Ir/alumina catalysts. The rhodium catalyst was found to be the most active on a weight basis being slightly more active than platinum, however on a molar basis the platinum was much more active.
In a previous paper3 we reported on how four lignins prepared from poplar and wheat straw (soda, organosolv, AFEX and ammonia) were found to have different S:G:H ratios and amounts of alkyl–aryl ether bonds and how this feedstream history affected lignin depolymerisation using a Pt/alumina catalyst. The results showed that the proportion of β-O-4 linkages was the crucial factor for both the yield and the nature of the monomeric products. Highly condensed lignin generated mainly non-alkylated phenolic products while less condensed lignin generated mainly phenolic products retaining a 3-carbon side-chain. In this paper we continue our investigation of lignin depolymerisation using the material designated “ammonia lignin” as characterised in the previous paper.3 The S:G:H ratio for this lignin was 0.65:0.35:0, while the percentage of β-O-4 linkages was ∼29% from thioacidolysis. Using this lignin we have investigated the optimisation of the yield and selectivity examining hydrogen pressure, solvent and catalytically active metal.
A 1 wt% Rh/alumina catalyst was prepared by incipient wetness impregnation on the same θ-alumina as the commercial Pt/alumina catalyst. Rhodium acetate was dissolved in sufficient water to achieve incipient wetness of the support and the solution added to the support. The catalyst was dried overnight at 343 K and calcined at 773 K for 4 h. XRD analysis of the calcined catalyst confirmed the alumina support was principally the theta phase. BET analysis gave a surface area of 102 m2 g−1 and a pore volume of 0.51 cm3 g−1, whilst CO chemisorption gave a rhodium dispersion of 121% due to formation of Rh(CO)2 suggesting a very small particle size.
A 1 wt% Ir/alumina catalyst was prepared using an incipient wetness technique on the same θ-alumina as the commercial catalyst. Iridium acetate was dissolved in sufficient water to achieve incipient wetness of the support and the solution added to the support. The catalyst was dried overnight at 343 K and calcined at 723 K for 3 h. The iridium dispersion, as measured by carbon monoxide chemisorption, was 13% assuming a 1:2 CO:Ir stoichiometry giving a particle size of ∼8 nm. From BET analysis the surface area of the catalyst was determined to be 104 m2 g−1 with a pore volume of 0.45 cm3 g−1.
During a typical experiment, 0.5 g of lignin was added to the autoclave along with 0.1 g of catalyst and 100 ml methanol–water mix (50/50, v/v). The reactor was purged with hydrogen and pressurised to 20 barg. The reactor was then heated at 10 deg min−1 to 573 K ± 1 K under a mechanical stirring rate of 1000 rpm and held at this temperature for 2 h. All reactions used these conditions as standard unless otherwise stated. At reaction temperature the typical pressure recorded was 145 barg. The optimisation reactions altered one or more of the parameters stated here but this will be highlighted in the text. The reaction mixture was filtered using a glass filter (po. 3) to remove the catalyst and any insoluble products. Any remaining high molecular weight material was solubilised in acetone and made up to 200 ml. This fraction will now be referred to as the ‘heavy fraction’. The methanol–water soluble fraction (specified as the ‘light fraction’) was centrifuged to isolate any finely dispersed solids and made up to 200 ml. A 15 ml aliquot of this light fraction was mixed with 0.2 ml of 1 g l−1 hexadecane (C16) internal standard and acidified to pH 3 using HCl. The products were then extracted with dichloromethane/dioxane (8/2, v/v) three times. The solvent was removed using a rotary evaporator and the remaining products were then solubilised in 2 ml dichloromethane (DCM) ready for analysis by GC–MS. To facilitate GC–MS analysis the products were derivatised by adding 10 μl aliquots of the DCM solution to 30 μl pyridine and 70 μl trimethylsilyl chloride (TMS), which was then left for at least 2 h prior to injection. Chemical composition was determined using a Shimadzu GC–MS-QP2010S coupled to a Shimadzu GC-2010 equipped with a ZB-5MS capillary column (30 m × 0.25 mm × 0.25 μm) with He as a carrier.
The amount of any given product produced was measured on the total ion chromatogram (TIC), and based on reference compounds and the C16 internal standard the quantity estimated. Due to the amount of products produced and their limited commercial availability, four standards were run in order to calculate a response factor. Varying concentrations (0.1, 0.5, 1, 5 and 10 g l−1) were prepared for each reference compound, using the C16 internal standard (10 g l−1). The reference samples were then run on the GC–MS using the same conditions described previously. The TMS was used to derivatise hydroxyl groups present on a given molecule. The reference compounds and their derivatised versions were guaiacol, 2,6-dimethoxyphenol, 2-methoxy-4-propylphenol and 4-methyl-2,6-dimethoxyphenol. The following equation was used to calculate the response factor (α):
α = (Intensity of reference/Intensity of internal standard) × (Mass of internal standard/Mass of reference). |
It was possible to obtain a linear relationship of intensity against mass for each reference compound and therefore calculate the resultant gradient which is equal to the response factor (α). This relationship was determined for each reference compound used and summarised in Table 1.
Reference compound | Type of unit | Response factor (α) |
---|---|---|
Guaiacol | Guaiacyl (G) | 0.833 |
2,6-Dimethoxyphenol | Syringyl (S) | 0.737 |
2-Methoxy-4-propylphenol | Guaiacyl (G) | 0.860 |
4-Methyl-2,6-dimethoxyphenol | Syringyl (S) | 0.771 |
Phenol | p-Hydroxyphenyl (H) | 0.95 |
From the data presented in Table 1, there was an obvious trend between the type of monomer unit, with respect to S or G, and the resultant response factor. Therefore we assumed the response factor for each of the other identified products based on their structure, with respect to H, G or S. The response factors for all other S, G and H units were deemed to be 0.75, 0.85 and 0.95 respectively.
In terms of the peaks collected through GC–MS analysis of the light fraction, 21 monomeric aromatic products were successfully identified through mass fragment data analysis for each peak based on the reference data and knowledge of the lignin structure. It was found that the reactions produced a range of alkylphenolic products with various functional groups. The absence of ring hydrogenation was confirmed through collaborative 2D NMR work,3 which showed an abundance of cross peaks in the aromatic region and no cyclohexanols were detected using GC–MS analysis. The presence of BTX molecules was also investigated but none could be found using UV-vis spectroscopy or GC-FID analysis. Selectivity, where used, is defined as S = the amount of the specified alkylphenol species/the total amount of alkylphenol detected species.
Abbreviation | R1/R2/R3 group | Product name |
---|---|---|
H0 | R1R2R3H | Phenol |
H1 | R1R2H; R3CH3 | 4-Methylphenol (p-cresol) |
H2 | R1R2H; R3CH2CH3 | 4-Ethylphenol |
H3 | R1R2H; R3CH2CH2CH3 | 4-Propylphenol |
G0 | R1R3H; R2OCH3 | 2-Methoxyphenol (guaiacol) |
G(OH)0 | R1R3H; R2OH | 1,2-Dihydroxybenzene (catechol) |
G1 | R1H; R2OCH3; R3CH3 | 2-Methoxy-4-methylphenol |
G2 | R1H; R2OCH3; R3CH2CH3 | 4-Ethyl-2-methoxyphenol |
G(OH)2 | R1H; R2OH; R3CH2CH3 | 4-Ethylbenzene-1,2-diol |
G3 | R1H; R2OCH3; R3CH2CH2CH3 | 2-Methoxy-4-propylphenol |
G3(i) | R1H; R2OCH3; R3CHCHCH3 | 2-Methoxy-4-propenylphenol |
G3(OMe) | R1H; R2OCH3; R3CH2CH2CH2OCH3 | 2-Methoxy-4-(3-methoxypropyl)phenol |
G3(OH) | R1H; R2OCH3; R3CH2CH2CH2OH | 4-(3-Hydroxypropyl)-2-methoxyphenol |
S0 | R1R2OCH3; R3H | 2,6-Dimethoxyphenol (syringol) |
S1 | R1R2OCH3; R3CH3 | 2,6-Dimethoxy-4-methylphenol |
S2 | R1R2OCH3; R3CH2CH3 | 4-Ethyl-2,6-dimethoxyphenol |
S3 | R1R2OCH3; R3CH2CH2CH3 | 2,6-Dimethoxy-4-propylphenol |
S3(i) | R1R2OCH3; R3CHCHCH3 | 2,6-Dimethoxy-4-propenylphenol |
S3(OMe) | R1R2OCH3; R3CH2CH2CH2OCH3 | 2,6-dimethoxy-4-(3-methoxypropyl)phenol |
S3(OH) | R1R2OCH3; R3CH2CH2CH2OH | 4-(3-Hydroxypropyl)-2,6-dimethoxyphenol |
S(OH)3(OH) | R1OH; R2OCH3; R3CH2CH2CH2OH | 5-(3-Hydroxypropyl)-3-methoxybenzene-1,2-diol |
GPC profiles of the starting ammonia lignin and those obtained after reaction, with and without the Pt/alumina catalyst are shown in Fig. 1. It should be remembered that equal volumes of both the light and heavy fractions were mixed together in order to give an overall representation of the product weight distribution. It is evident from these plots that there was a clear shift to lower molecular weights from ∼11.5 min to ∼14 min. Analysis of the profiles reveals that the molecular weight of the catalytic products was 1042 Da, which is 26.8% of that calculated for the ammonia lignin (3884 Da). This is in comparison to the run without catalyst, which produced products with a molecular weight of 1304 Da. Moreover, the polydispersity of the ammonia lignin decreased from 2.45 to 1.96 and 1.75 for the tests without and with catalyst respectively. This indicates that the lignin fragment size was much more uniform than the starting lignin after both reactions. Therefore although the thermal reaction does initiate depolymerisation, the catalytic reaction is more effective.
Estimation of the amount of monomers produced by GC–MS analysis showed significant differences when the reaction was performed with and without a catalyst as shown in Fig. 2. The reaction without catalyst showed a lower overall yield of 6.8% and generated mainly phenol (H0), guaiacol (G0) and syringol (S0) with selectivities of 45%, 7% and 24% respectively. In the presence of the Pt/alumina catalyst, the yield was increased to 16.4% and, unlike the non-catalytic reaction, which promoted dealkylation, the products obtained in the catalytic test were able to maintain the alkyl chain with various functional groups attached to the terminal γ-carbon. Although there was no obvious selectivity towards one particular product, phenol (14%), syringol (11%), propylsyringol (S3) (16%) and propenylsyringol (S3i) (10%) were the main products obtained, showing a selectivity towards S-units. Indeed by implementing the use of a catalyst, the selectivity towards S-units increased from 37% to 60%, and phenol decreased from 45% to 14%. The S:G ratio for the products is 0.7:0.3, which is close to the S:G ratio of the starting lignin as measured by thioacidolysis. Hence these results are in keeping with the conclusions of our earlier paper,3 where it was shown that the β-O-4 linkages were the ones most likely to be broken in the catalytic reaction. Nevertheless other linkages are also broken catalytically as well as thermally; the production of phenol when no H form was detected indicates that demethoxylation can take place. This is in agreement with recent publications on lignin pyrolysis5,6 that reported phenol as a major product at temperatures as low as 523 K.5
Fig. 2 Ammonia lignin monomer yield after reaction in the presence and absence of 1 wt% Pt/alumina (yields are estimated by GC–MS analysis). |
In high temperature water environments alumina supports may not be stable and can hydrate. Indeed a recent study investigated the stability of γ-alumina under aqueous phase reforming conditions using hot water and found that γ-alumina converted to hydrated boehmite at 473 K and above.7 The authors did however note that the presence of metal particles resulted in a significant decrease in the rate of boehmite formation. Furthermore, research has shown that the presence of oxygenates can enhance the stability of the support material by blocking the surface of the support with carbonaceous material thus preventing hydrolysis of the alumina.8,9 The type of oxygen functionality did not influence the stabilisation of the support and it was suggested that this meant that the different oxygenated molecules eventually formed the same surface oxygen species. It was also put forward that hydrolysis was prevented because the oxygen functionalities coordinated with the unsaturated alumina sites thus preventing the water molecules from accessing the sites. Additionally, the use of ethanol was proven to slow down the formation of boehmite crystals to some extent as well.8,9 Therefore although the alumina used in this study was θ-alumina, which may be expected to be more stable, after reaction the alumina support was analysed by XRD and BET. The diffraction pattern obtained of the support post-reaction did not differ from that obtained prior to reaction and there was no significant change in the support surface area after reaction. No boehmite reflections were observed and it was confirmed that the support had remained in the θ-phase, hence hydration had not taken place during the reaction.
Fig. 4 Reactions with hydrogen and with helium gas atmospheres (yields are estimated by GC–MS analysis). |
Fig. 5 Effect of changing initial hydrogen pressure, all other conditions identical (yields are estimated by GC–MS analysis). |
Fig. 6 and 7 show the effect on the overall yield and the specific alkylphenol yield of changing the solvent from 100% water to 100% methanol. The use of water was observed to be less effective than the methanol–water co-solvent for higher overall product yields. It is worth noting that the maximum pressure during the reaction with 100% water was 187 barg so much less than is required to produce scH2O, whereas with 100% methanol as the solvent the maximum pressure was 160 barg significantly in excess of that required to produce scCH3OH. As the methanol concentration was increased from 0 to 100%, the product yields increased from 11.2% to 43.5%.
Fig. 7 Effect of MeOH–water solvent composition on alkyl phenol yield. The yields for the S3+ monomers are shown at the top of the graph (yields are estimated by GC–MS analysis). |
The yield of 43.5%, which was obtained at a concentration of 100% methanol, also gave the highest selectivity towards one particular product, namely 4-(3-hydroxypropyl)-2,6-dimethoxyphenol (38%, S3(OH)), suggesting that some control over selectivity could be achieved dependent upon the nature of the starting lignin. A more detailed figure (Fig. 8) is shown separating the components of the G3+ and S3+ species. What is immediately obvious is that alkylphenols 4-(3-hydroxypropyl)-2-methoxyphenol (G3(OH)), 2,6-dimethoxy-4-propylphenol (S3) and 4-(3-hydroxypropyl)-2,6-dimethoxyphenol (S3(OH)) are highly favoured. These results indicate that by using 100% methanol, or a methanol–water mixture, condensation of the lignin was inhibited allowing greater depolymerisation. Indeed for the experiment using 100% methanol the molecular weight of the residual polymer was reduced to 918 Da, a reduction of 76% from that of the starting lignin. The reason that methanol is so effective may be due to its ability to stabilise free radicals generated during the depolymerisation by acting as a capping-agent.18 Indeed in a recent study Hensen and co-workers21 showed that supercritical ethanol acted as a capping agent resulting in a high depolymerisation yield from lignin of over 80% with no char formation. The sc-ethanol also acted as a formaldehyde scavenger, which helped minimise lignin recombination, a key aspect in minimising char.21
Fig. 8 Expanded breakdown of S and G alkylphenol products obtained with 25/75% water/methanol and 100% methanol as solvents (yields are estimated by GC–MS analysis). |
However a simpler explanation may also be relevant, Fig. 6 also shows a measure of the solubility of lignin in the solvent mix and it can be seen that activity and solubility follow similar shaped lines. Therefore it is possible that much of the improved yield found as the methanol concentration increases is due to increased solubility of the lignin22 allowing more effective interaction with the catalyst. Indeed a similar effect was found by Minami and Saka23 when examining the effect of scMeOH on woody biomass. In that study a high water content in the scMeOH resulted in low solubility of lignin-derived products causing a reduction in yield. Pure alcohol solvents have also been examined by Song et al.24 over a nickel catalyst. In that study the solubility of the lignin in the solvents was determined to be a significant factor in their efficacy, although other factors such as hydrogen transfer were also proposed.24 Methanol has also been used as a hydrogen transfer agent.25 In a recent study a copper catalyst was used to generate hydrogen via in situ methanol decomposition and this hydrogen (and carbon monoxide) facilitated lignin depolymerisation and hydrogenation.25
Isopropanol (IPA) had been used, in conjunction with formic acid, for solvent liquefaction of lignin, in air via solvolysis.26 It was found that the highest amount of phenolics was obtained after treatment with formic acid and methanol or isopropanol, which suggested that IPA would be a suitable solvent in our process. Fig. 9 compares the product distribution obtained using methanol–water and IPA–water (50:50 v/v). The experiment using IPA–water yielded 24.3% of monomeric products in comparison to 16.4% obtained using the standard MeOH–water mix. Although the IPA–water solvent resulted in a general improvement in yield it principally favoured the longer chain alkyl products for both G and S motifs.
Fig. 9 Comparison between methanol and IPA as a co-solvent with water in a 50/50 mix (yields are estimated by GC–MS analysis). |
Given that isopropanol has a slightly lower critical point (508 K, 54 bar) compared to methanol (513 K, 78.5 bar) IPA will have reached its critical point (maximum pressure during reaction 130 barg) thus aiding the depolymerisation reaction in a manner similar to methanol. Nevertheless IPA being a more effective solvent than methanol was slightly surprising given that it is less polar and hence would have a lower ability to dissolve lignin and solubilise water. However it is possible that some IPA underwent catalytic dehydrogenation to acetone under our standard reaction conditions to give an IPA–acetone–water mix during the reaction. Acetone is capable of dissolving the lignin and its heavier products, so such a mix would promote dissolution of the lignin and its intermediates. Moreover, IPA is also known to donate hydrogen during its dehydrogenation to acetone.
Hydrogen donor solvents (HDSs) were first used in coal liquefaction to stabilise free radicals that would otherwise form char27 so it is possible that any IPA/acetone solvent mix reduced re-polymerisation of the lignin by stabilising reactive phenols or reaction intermediates. Work comparing IPA against tetralin28 found that the IPA system showed higher selectivity to alkylated products than the tetralin system due to the higher hydrogen donor capability of IPA. The authors concluded that, in agreement with our results, HDSs were effective in converting lignin.28 However at this stage we have no evidence that transfer hydrogenation is occurring.
Fig. 10 Effect of changing catalytically active metal on the yield of alkylphenols (yields are estimated by GC–MS analysis). |
This journal is © The Royal Society of Chemistry 2016 |