Compositional Changes to Low Water Content Bio-oils during Aging

May 10, 2016 - Cite this:Energy Fuels 30, 6, 4825-4840 ... In this study, electrostatic precipitator (ESP) pine wood-derived bio-oil, which contains l...
0 downloads 0 Views 5MB Size
Article pubs.acs.org/EF

Compositional Changes to Low Water Content Bio-oils during Aging: An NMR, GC/MS, and LC/MS Study Jincy Joseph,† Matthew J. Rasmussen,§ James P. Fecteau,† Sally Kim,§ Hyunji Lee,§ Katelyn A. Tracy,† Bruce L. Jensen,† Brian G. Frederick,†,‡ and Elizabeth A. Stemmler*,§,‡ †

Department of Chemistry, Forest Bioproducts Research Institute, and Laboratory for Surface Science and Technology, University of Maine, Orono, Maine 04469, United States § Department of Chemistry, Bowdoin College, Brunswick, Maine 04011, United States S Supporting Information *

ABSTRACT: Bio-oil generated by the fast pyrolysis of biomass is an unstable material, undergoing chemical and physical transformations as the oil ages at room temperature. In this study, electrostatic precipitator (ESP) pine wood-derived bio-oil, which contains less water and does not undergo phase-separation upon aging, was characterized following accelerated aging. Bulk oil properties (percent water and viscosity) were found to increase in ways similar to conventional bio-oils. The unaged and aged bio-oil samples were characterized by gel permeation chromatography (GPC), solvent fractionation, solution 13C NMR, gas chromatography/mass spectrometry (GC/MS), and chip-based nanoelectrospray ionization, liquid chromatography, quadrupole time-of-flight (nanoESI-LC-Q-TOF) MS/MS. Using the formation of the silyated derivatives to extend the range of detectable compounds, GC/MS analysis was used to identify specific compounds that showed elevated reactivity, extending the understanding of reactivity characteristics beyond the known reactivity of aldehydes and some aromatics to distinguishing the reactivity of ring-conjugated aromatics and certain polyhydroxylated benzenes, specifically the 1,3-di-, 1,2,3-tri-, and 1,2,4trihydroxy substituted compounds. To explain the enhanced reactivity of these compounds, we propose acid-catalyzed formation of quinone methides as important intermediates. Additionally, we find significant changes to the composition of mono- and disaccharides, where specific monosaccharides (arabinose, xylose, and glucose) increased in concentration with aging and high reactivity was observed for certain sugars with furano-ring mass spectral characteristics. In contrast, we also found that three anhydrosugars (levoglucosan, mannosan, and galactosan) were largely stable with respect to aging. High mass resolution nanoESI-LC/MS/MS analyses of peracetylated samples permitted the analysis and chromatographic separation of both lignin and carbohydrate-derived oil components and were used for the identification of a putative formaldehyde−trihydroxybenzene dimer. This work provides further insights into chemically specific entities and the processes responsible for bio-oil aging. Diebold10 considered possibilities, including esterification and transesterification of acids; acetal and hemiacetal formation, transacetylization, homopolymerization, and hydration of aldehydes; phenol−aldehyde reactions; polymerization of furan derivatives; dimerization of organic nitrogen compounds; olefinic condensation of unsaturated aldehydes; and oxidation of alcohols to acids that could account for the increase in molecular weight, moisture content, and viscosity and the decrease in volatility that occur as the oil ages. According to an early oil stability study by Oasmaa and Kuoppala,11 where solvent fractionation was used to monitor changes in chemical composition, the major physiochemical changes to bio-oil occurred during the first 6 months of storage. To explain a significant decrease in the ether-soluble fraction of the oil and an increase in the higher molecular mass (HMM) lignin, condensation and polymerization reactions with ether-soluble compounds like aldehydes and ketones, which may also react with lignin constituents, were proposed to explain the formation of water-insoluble products.

1. INTRODUCTION The search for renewable energy sources is driven by the limited supply of nonrenewable fossil fuels and their adverse environmental effects.1−4 There are many methods by which biofuels are obtained, such as gasification, liquefaction, and enzymatic hydrolysis.1,4 Fast pyrolysis is a thermochemical conversion process that involves rapid heating of biomass (commonly woody biomass or other lignocellulosic feedstocks) in the absence of oxygen to produce bio-oil. Conventional processing conditions utilized a short residence time of feedstock in the reactor to optimize the yield of the liquid fraction.4,5 However, the oils produced are limited in their use as a fungible fuel because they are viscous, acidic, and characterized by high water and oxygen content;6−8 furthermore, the oils are unstable. Bio-oils stored at ambient temperatures undergo changes in physical and chemical properties, a process commonly described as “aging”.9 Oils show increases in viscosity, molecular weight, and water content10,11 with aging time, and approaches have been developed to evaluate bio-oil stability using elevated temperatures to accelerate the aging process and mimic ambient aging conditions.10,12,13 The low pH and great diversity of chemicals found in bio-oils offer ideal conditions for a number of possible aging reactions. © XXXX American Chemical Society

Received: February 1, 2016 Revised: May 2, 2016

A

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels Subsequent studies7,9,11,13−28 have established several consistent chemical changes that occur when bio-oils age, which include decreases in the concentration of carbonyl compounds, decreases in the concentration of certain lignin monomers, and increases in the amount of water-insoluble higher molecular mass, pyrolytic lignin-like material. Volatile acids, which are not significantly changed by the aging process, are responsible for the acidity of the oil and are thought to play a role in oil aging conversions via acid-catalyzed reactions;22,24 in contrast, no support has been found for the participation of free radicals in oil aging reactions.27,29 To better understand condensation reactions, the reactivity of model compounds have been explored.24 Microstructural characterization of the bio-oils using small-angle neutron scattering (SANS) has provided support for polymerization resulting from the aggregation of pyrolytic lignin units upon aging,30 and studies have been directed at characterizing structural features31−33 and reactivity22 of the pyrolytic lignin fraction in oil to gain insights into pyrolysis oil formation and subsequent aging reactions controlling formation of this oil component. A variety of chemically specific analytical techniques have been applied to study the process of bio-oil aging, including NMR,19−21 pyrolysis-based15,21 and conventional7,17,18,20,21 gas chromatography/mass spectrometry (GC/MS), electron paramagnetic resonance (EPR),27,29 Fourier transform infrared spectroscopy (FTIR),9,20,21,25 and high-resolution Fourier transform ion cyclotron resonance (FTICR) mass spectrometry.28,33 In this study, we explore compositional changes that occur when a low water content wood-derived bio-oil is subjected to accelerated aging, applying gel permeation chromatography, solvent fractionation, solution 13C NMR, GC/MS and chipbased nanoESI-LC/MS/MS. The low water content oils targeted by this study offer advantages for better distinguishing the relative reactivity of bio-oil components in a more consistent matrix that is not subject to phase separation, and our analytical approach, applying GC/MS with derivatization techniques and using high resolution MS coupled with chromatographic separations, has permitted the chemically specific analysis of higher molecular weight and more polar components. These studies revealed new insights regarding the relative reactivities of reactive oil components and the compositional changes that occurs when bio-oils age.

were carried out using a Brookfield DV-II Pro viscometer with a cone and plate option. The viscosity was measured at 45 °C at 0.07 rpm. The molecular weight distribution for aged Oil B samples was determined using GPC with a refractive index detector using a method adapted from Garcia-Perez et al.35 The separation was conducted with two Styragel Waters GPC columns (a WAT 044238 -Styragel HR 2, 5 μm particle size, 7.8 × 300 mm, molecular weight range 500−20 K column connected to a WAT 044232-Styragel HR 0.5, 5 μm particle size, 7.8 × 300 mm, molecular weight range 0−1 K column), connected in series. Molecular weights were calibrated to convert elution times to apparent molecular weights using EasiVial polyethylene glycol standards (Agilent Technologies) covering a molecular weight range of 106−35000 g/mol. Aged Oil B samples were subjected to a solvent fractionation procedure reported by Garcia-Perez et al.35 A 7 g portion of each aged Oil B sample was first extracted with toluene after thorough mixing using a glass rod to yield the toluene-soluble, extractives fraction. Toluene-insoluble material was then solubilized with 5 g of methanol per gram of oil. No methanol-insoluble material was obtained. The methanol solution was added dropwise to 500 mL of deionized H2O in an ice bath with stirring, forming a solid precipitate. The precipitate was filtered with Whatman No. 42 ashless filter paper. Water-soluble material was extracted with diethyl ether, yielding the ether-soluble fraction and ether-insoluble, carbohydrate fraction. Water-insoluble material was extracted with dichloromethane, yielding the dichloromethane-soluble, low molecular mass (LMM) lignin fraction and the dichloromethane-insoluble, high molecular mass (HMM) lignin fraction. The mass of each fraction was determined after solvent was removed using rotary evaporation followed by vacuum evacuation. 2.4. 13C Liquid State NMR. All liquid state NMR data were collected with a 400 MHz NMR Varian Unity Plus instrument. Bio-oil samples were dissolved in DMSO-d6 (Cambridge Isotope Laboratories) in a 1:1 mass ratio in 5 mm tubes using a broad band probe equipped for gradient shimming. 13C NMR spectra were measured with a sweep width of 25 kHz, 4.5 s pulse delay, 4000 transients, and 2.56 s acquisition time. The 13C spectra were referenced to the DMSO-d6 peak at 39.43 ppm. Baseline correction and integration of spectra based on chemical shift regions were performed using MestreNova software as described previously.36 2.5. Gas Chromatography/Mass Spectrometry (GC/MS). In preparation for analysis, samples were derivatized with N-methyl N(trimethylsilyl) trifluoroacetamide (MSTFA; Sigma-Aldrich) to dilute and rapidly derivatize hydroxyl, carboxylic acid, and aldehyde functional groups. Aged Oil B samples (50 mg) were placed in 1.5 mL amber vials and mixed with 200 μL of MSTFA. The samples were heated at 90 °C for 90 min to effect both dissolution and derivatization. The samples were cooled, the internal standard, biphenyl (99.5%; Sigma-Aldrich; 0.5 mL of a 4,000 mg/L solution in ethyl acetate), was added, and the solutions were diluted with ethyl acetate prior to analysis. GC/MS was carried out using a 5973N MSD (Agilent Technologies, Santa Clara CA) equipped with a capillary column (Restek-5Sil MS, 30 m, 0.25 mm i.d., 0.25 μm film thickness; Phenomenex, Torrance, CA). The oven temperature was held at 40 °C for 4 min, ramped at 2.5 °C/min to 80 °C, 4 °C/min to 250, and 10 °C/min to 310 °C and held at 310 °C for 10 min. The injection port was maintained at 280 °C and splitless injections of 0.2 μL were carried out using helium as the carrier gas. The mass transfer line was held at 280 °C. Data was collected using MSD ChemStation E.02.02.1431 (Agilent Technologies). Compounds were identified based upon comparison with authentic standards and mass spectral library matches. Quantitative analysis was carried out using biphenyl as the internal standard, with details provided in the Supporting Information. 2.6. HPLC-Chip-nanoESI-Q-TOF MS/MS Analysis. In preparation for analysis, samples were peracetylated with acetic anhydride/ pyridine to enhance the chromatographic and mass spectrometric instrument performance. Bio-oil samples (10 mg) were placed in a 1.5 mL amber vial with pyridine plus (Alltech, Pyridine-Plus Kit, 50 μL), acetic anhydride (Sigma-Aldrich; 100 μL), and acetonitrile (Fisher,

2. EXPERIMENTAL SECTION 2.1. Bio-oil Samples. Bio-oils were obtained through fast pyrolysis of pine (Pinus strobus) sawdust in a fluidized bed reactor, described previously,5 based on the Waterloo design.34 The pyrolysis reactor was equipped with a condenser and an electrostatic precipitator (ESP) for collecting liquid products. We used two different temperatures (90 and 80 °C) for aging two bio-oil samples (Oil A and B; generated in separate pyrolysis runs). The pyrolysis conditions for both runs included temperatures of 550 °C and feed rates of 1.96 (Oil A) and 3 g/min (Oil B). The samples were kept frozen at −5 °C while in storage and remained refrigerated throughout experimentation. 2.2. Accelerated Aging Conditions. The bio-oil, separated into five 10 g samples in glass vials, was purged with nitrogen gas. The capped samples were then heated at 90 °C for 0, 8, 16, 24, and 32 h (Oil A) or 80 °C for 0, 10, 20, 30, and 40 h (Oil B). 2.3. Water Content, Viscosity, Gel Permeation Chromatography (GPC), and Solvent Fractionation. Karl Fischer titrations were performed on Oil A and B samples for measurement of water content using a Brinkmann Karl Fischer Titrando 841 instrument and ASTM procedure E 203-01. Oils were dissolved in methanol (2:1 methanol:oil ratio) and titrated with Hydranal-Composite 2 (SigmaAldrich) in triplicate. Viscosity measurements for aged Oil B samples B

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels Optima LC/MS; 100 μL). The mixture was heated to 80 °C overnight (usually 12 h). After heating, samples were diluted with acetonitrile (750 μL). In some experiments, d6-acetic anhydride (Aldrich, 99 atom % D) was used in place of acetic anhydride. Liquid chromatographic/mass spectrometric analysis was performed using a 6530 quadrupole time-of-flight (Q-TOF) mass analyzer (Agilent Technologies, Santa Clara CA) using nanoelectrospray ionization (nanoESI). Mass spectra (MS and MS/MS) were collected in positive ion mode; the ionization voltage ranged from 1850 to 1950 V and the ion source temperature was held at 350 °C. Spectra were internally calibrated using the proton-bound dimer [CH3CN+H2O +H]+ with exact mass 60.044 and hexakis(1H, 1H, 4H-hexafluorobutyloxy)phosphazine (HP-1221; C24H18O6N3P3F36) with exact mass 1221.9906, continuously introduced and detected as [M + H]+. CIDMS/MS experiments were executed with precursor ions subjected to CID using N2 as the target gas with the collision energy adjusted based upon precursor m/z. Chromatographic separation and nanoelectrospray ionization (ESI) were performed using a 1260 Chip Cube system (Agilent Technologies) using an ProtID-chip with a 40 nL enrichment column and a 150 mm × 75 μm analytical column (Agilent Technologies). The enrichment and analytical columns were packed with 300 Å, 5 μm particles with C18 stationary phase. The mobile phases were 0.1% formic acid in H2O (A) and 0.1% formic acid, 2% water in acetonitrile (B). Samples were loaded onto the enrichment column using 90% mobile phase A. Analytes were eluted using a mobile phase gradient (Supporting Information, Table SI-1) with increasing percentages of mobile phase B. Data was collected using MassHunter B.05.00 software and analyzed with MassHunter Qualitative Analysis B.05.00 software (Agilent Technologies).

set of samples, designated Oil B, with an initial water content of 2.5%. The oil was aged at 80 °C in 10 h intervals, which corresponds to 0−10 months of effective ambient aging.12 Again, phase separation was not observed and the water content increased by approximately 0.8% by mass after aging for 40 h (Supporting Information, Figure SI-1). The rise in water content is again consistent with condensation reactions taking place at early times during aging. To further evaluate the impact of aging on bulk properties, we measured the viscosity of Oil B (Supporting Information, Figure SI-2), which increased from 0.68 to 1.97 kcp after being aged at 80 °C for 40 h. The overall increase in viscosity is consistent with observations of others.6,7,14,17,39 Our oil does, however, have much high viscosity, and our measurements show closer agreement in magnitude to values reported for another lower water content oil (filtered whole pine wood oil) subjected to aging.25 Oil B samples aged at 80 °C for up to 40 h of actual aging were subjected to GPC analysis to characterize changes in the molecular weight distribution (Figure 1). The GPC curves

3. RESULT AND DISCUSSION Accelerated aging, used in this study, involves subjecting an oil to higher than ambient temperatures for shorter periods of time to mimic the aging that typically occurs over several months when the oil is stored at room temperature.12 The accelerated aging conditions we used were designed to produce multiple samples separated in time by equal intervals leading to 16 months (using 90 °C treatment) or 10 months (using 80 °C treatment) of effective ambient aging, with effective ambient aging times estimated based upon work by Elliot et al.12 More details regarding the effective ambient aging time estimates can be found in the Supporting Information. 3.1. Water Content, Viscosity, GPC, and Solvent Fractionation Analysis. The pine-ESP fast pyrolysis oil used in this study contained 3−5% water, which is low compared to conventional pyrolysis oils (typically 15−25%).4,37 Initial aging experiments were conducted to determine if changes to the water content, which has been used in previous studies to provide evidence of condensation reactions upon aging, was impacted in similar ways for our oils. A set of samples of Oil A, which contained 5% water, were aged at 90 °C with 8 h intervals for up to 32 h, conditions that should simulate 16 months of ambient aging.38 No phase separation was observed over the course of aging, which may be attributed to the lower total water content of the oil. A roughly 1% increase in water content was measured during the first 16 h of actual aging (8 months of effective ambient aging; see Supporting Information, Figure SI-1), but no systematic change was detected after that point. Despite the lower initial water content, this increase is similar to the 1.6% increase in a 16% water content oil reported by Diebold,10 suggesting that the total water content has no dramatic effect on water-producing condensation reactions. To focus in greater detail on the initial, more rapid, changes in water content and to use the 80 °C temperature more commonly employed for oil stability testing, we aged a second

Figure 1. GPC curves for Oil B samples aged at 80 °C for 0, 20, and 40 h of actual aging, corresponding to 0, 5, and 10 months of effective ambient aging. Abundances reported as a percentage of the total detected peak area for each sample are plotted vs the log of the molecular weight.

show a significant shift toward signals characteristic of higher molecular weight species as the bio-oil ages. The weight-average molecular weight (M̅ w) increased by about 60% (from 393 to 632 Da), in reasonable agreement with previous results.9,18 The number-average molecular weights (M̅ n) also increased (from 109 to 145 Da); however, these values, which are more influenced by lower molecular mass components, were lower than other literature reports,9,18 resulting in high polydispersity index values, which increased from 3.6 to 4.4 for Oil B (data summarized in Supporting Information, Table SI-2). Finally, we applied solvent fractionation35 to aged samples of Oil B to assess compositional changes to oil components through the isolation of a toluene-soluble (extractives) fraction, a water-soluble, ether-soluble (polar aromatics) fraction, a water-soluble, ether-insoluble (carbohydrate) fraction, and two water-insoluble fractions containing low and high molecular mass (LMM and HMM) lignin. The masses (wt %) of each fraction for aged Oil B samples are summarized in Table 1. In agreement with early work on bio-oil aging,11 we also observed an increase in the water-insoluble HMM and LMM lignin fractions and a decrease in the fraction masses for extractives, polar aromatics, and carbohydrates with bio-oil aging (Table 1). In summary, our percent water, viscosity, GPC, and solvent fractionation results demonstrate that our low water-content oils respond to accelerated aging conditions in general ways C

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

The spectra were integrated to determine the percentage of each type of carbon, according to the chemical shift regions of Joseph et al.,36 as indicated in the top of the Figure 2. The NMR peak integrations vs aging time are shown for Oils A and B in Figure 3. For both Oil A (90 °C; Figure 3A) and B (80 °C;

Table 1. Mass of Dried Fractions (wt %) for Oil B Samples Following Accelerated Aging at 80 °C and Solvent Fractionation fraction mass (wt %) aging time (h) toluene-solubles (extractives) methanol-solubles

0

10

27.6 25.4 0 0 Water-Solubles ether-solubles (polar aromatics) 12.4 7.5 ether-insolubles (carbohydrates) 37.5 34.3 Water-Insolubles dichloromethane-solubles (LMM 13.9 15.6 lignin) dichloromethane-insolubles 2.5 9.1 (HMM lignin) total 93.8 91.9

20

30

40

23.9 0

22.1 0

16.2 0

7.9 35.7

7.2 35.8

7.2 29.8

14.6

15.0

20.8

13.8

13.8

13.8

96.0

93.8

87.8

Figure 3. Integration of the characteristic regions (defined in Figure 2) of the 13C spectra for (A) Oil A aged at 90 °C and (B) Oil B aged at 80 °C, presented as percentage of total area.

that are consistent with changes observed for higher watercontent oils. 3.2. Solution 13C NMR Analysis. To gain greater insight into chemical changes that occur upon aging, we measured the 13 C NMR spectra of oils subjected to aging. We initially characterized Oil A samples, which were subjected to longer aging at a higher temperature (90 °C). A stacked plot of 13C NMR spectra, normalized to the DMSO peak, is shown in Figure 2. The stacked spectra show that only a relatively small number of carbon peaks, highlighted by arrows in Figure 2, changed significantly with aging. The 13C NMR experiments were repeated for Oil B, aged at 80 °C for 40 h for 10 months of effective ambient aging (spectral data not shown).

Figure 3B) the relative abundance of the aromatic carbon signals rapidly decreased during early heating times, with the Oil B data showing a substantial decrease within 10 h. For Oil A, aged for a longer effective ambient aging time, the aromatic signals continued to decrease more slowly at longer aging times. If the relative areas are calculated excluding the aromatic carbons, there is little to no change in percent composition of the remaining functional groups at longer effective ambient aging times.

Figure 2. 13C NMR Spectra of Oil A following aging for 0, 8, 16, 24, and 32 h (bottom to top) at 90 °C, corresponding to a total of 16 months effective ambient aging time. Arrows indicate the peaks that decreased after aging for 8 h. D

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

groups (174.3 ppm for oleic acid; 167.3 ppm for benzoic acid) and for the ester at 169.8 ppm from 5-acetoxymethyl-2furfuraldehyde.36 Thus, while this signal appears in a region suggesting that it may result from esterification reactions, more specific support for this hypothesis was not provided by compounds included in our in-house database. 3.3. GC/MS Analysis. To provide a more detailed chemical characterization of Oil B aged for 0, 10, 20, 30, and 40 h, we analyzed samples by GC/MS. A complete listing of the >100 compounds identified and quantified by GC/MS analysis can be found in the Supporting Information (Tables SI-3 and 4); a subset of compounds, sorted into class categories, can be found in Table 2; structures appear in the Supporting Information (Figure SI-3). Because the oils contain a large number of polar components that are not detected by direct GC/MS analysis, the oils were derivatized with MSTFA to form trimethylsilyl (TMS) derivatives of hydroxyl and carboxylic acid groups.45 This has allowed us to study more polar sugars and polyhydroxylated benzenes. Furthermore, MSTFA can be used to generate MSTFA-adducts with aliphatic and aromatic aldehydes,45−47 permitting the detection of both small and aromatic aldehydes in the sample. An initial set of experiments made use of ethyl acetate as a solvent to dissolve the oil samples prior to derivatization; however, the oils were not fully solubilized and compounds were not fully derivatized following heating and derivatization. As an alternative approach, the derivatizing agent MSTFA was added directly to the oil to serve as both a derivatizing agent and solvent; following quantitative dilution and the addition of an internal standard (biphenyl), the derivatized oils were analyzed by GC/MS. This approach successfully solubilized components of the oil (from visible inspection under a ×10 microscope) and provided essentially complete derivatization of compounds with multiple functional groups, as assessed by comparing signals characteristic of fully derivatized levogluocosan (H3; present as the triTMS derivative when fully derivatized) with signals characteristic of incompletely derivatized levoglucosan (observed as mono- and diTMS forms).48 Only trace signals for the diTMS derivative of levoglucosan was detected for all analyses, providing support for our assumption that levoglucosan and other components were completely derivatized. The conditions used for derivatization also promoted formation of the TMS-derivative of the enol form of some ketones, including 2-hydroxy-2-cyclopenten-1-one; however, the derivatized keto-form was the most abundant form detected and the keto/enol ratio was consistent (12.1 ± 1.2; n = 12) for the analysis of oil samples subjected to derivatization following aging. Using this approach, we quantified components of the oil (included as Supporting Information Tables SI-3 and 4). The Supporting Information also provides details regarding the basis for compound identifications and the approaches used for absolute or relative quantification, including our use of extracted ion chromatograms (EICs) to provide more specific measures of compositional changes. It should be noted that, while our derivatization approach provides a more complete representation of polar biooil components, we are unable to detect more volatile components, including small acids, which coelute with the solvent and residual MSTFA. Representative total ionization chromatograms (TICs) for the 0 and 40 h aged sample are shown in Figure 4A and B. Because of the high abundance of levoglucosan (H3) in the sample, the scale has been adjusted to better display less

Conversion of aromatic carbons into alkyl carbons during aging under a nitrogen atmosphere is unlikely; however, oligomerization/polymerization reactions occur during aging and are thought to be responsible for the formation of the water-insoluble LMM and HMM lignin fractions of aged Oil B. These larger, solid-like materials would produce signals that would be broadened because of dipolar coupling, reducing the signal/noise ratio in the aromatic region and effectively reducing the integrated intensity. This hypothesis is supported by 13C NMR spectra of the LMM and HMM aged Oil B fractions, which showed weak (LMM lignin fraction) or negligible (HMM lignin fraction) solution NMR signals (data not shown). The reduced signals from these components of aged Oil B qualitatively accounts for the net 10% reduction of the aromatic carbon intensity in the whole bio-oil samples (Figure 3). Complementing the information provided by integration of spectral ranges, we also considered the notable decreases for specific resonances, which occurred at 33.1, 36.2, 56.0, 62.6, 100.0, 161.7, 188.5, 202.1, and 204.4 ppm for both oils and, additionally, at 22.0, 26.2, 68.1, and 75.8 ppm for Oil B (80 °C). A peak at 163.8 ppm (Oil B) increased upon aging for both oils; two peaks appearing at 25.9 and 209.7 ppm increased in Oil A. When working with such complex NMR spectra it is difficult to assign compound-specific sources to individual 13C NMR peaks; however, 13C NMR peaks can be assigned to particular functional groups with more confidence. We approached these assignments informed by previous work19,40−43 and using an in-house NMR library of over 50 compounds that may be present in pyrolysis oils measured using the conditions applied in this study.36 We first considered specific peaks that decreased in abundance in our two aging experiments. The four peaks in the range of 20−37 ppm that decreased upon aging can be assigned to alkyl carbons, while the peak at 56.1 ppm can be reliably attributed to methoxy-group carbons (OCH3) attached to an aromatic ring, a group present in many lignin-derived compounds like eugenol, guaiacol, and coniferyl alcohol. Four possibilities exist as the source of the 62.6 ppm peak. The signal could originate from CH2OH carbons from carbohydrates, with signals that appear at 61.2 to 61.5 ppm,36 from CH2OH carbons from glycolaldehyde dimers,44 from CHCH CH2OH carbons, a bonding motif present in coniferyl alcohol with the signal appearing at 61.6 ppm,36 or from CH2OH carbons in aliphatic alcohols, with signals that appear in the range of 55 to 65 ppm, depending upon substitution and branching on adjacent carbons.36 The signal at 100.0 ppm appears in a region close to signals for the anomeric carbon in the anhydrosugar levoglucosan36 and is also close to the chemical shift observed for the aromatic carbon located between two ring-substituted hydroxyl groups in 4-ethylresorcinol.36 The decreasing intensity peaks at 202.1 and 204.4 ppm are specific to aldehyde or ketone carbons that show closest agreement with the signals from aliphatic aldehydes, like citronellal (203.2 ppm) or a subset of ketones, some of which are cyclic (e.g., 3-methyl-1,2-cyclopentanedione at 202.0 ppm). This resonance does not, however, correspond well with signals for aromatic aldehydes (191.0 ppm for benzaldehyde; 192.2 for 3,4-dimethoxybenzaldehyde; 177.9−178.5 ppm for four different furfurals).36 The signal at 163.8 ppm, which is one signal that increased in abundance for both oils with aging, can be classified as an acyl type of carbonyl group; however, this signal is shifted upfield relative to signals for most carboxylic acid E

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels Table 2. Bio-oil Compounds Detected Following Derivatization with MSTFA and GC/MS Analysisa Aliphatic Aldehydes A1 formaldehyde A2 acetaldehyde A3 2-butenal A4 glycolaldehyde A5 hexanal Hydroxy Ketones and Polyols B1 1-hydroxy-2-propanone B2 ethylene glycol B3 2-hydroxy-2-cyclopenten-1one B4 2-hydroxy-3-methyl-2cyclopenten-1-one B5 2-deoxytetrono-1,4-lactone B6 1,3-dihydroxy-propan-2-one B7 glycerol B8a glycolaldehyde dimer, 1 B8b glycolaldehyde dimer, 2 B9 pinitol Aliphatic Acids and Hydroxyl Acids C1 lactic acid C2 glycolic acid C3 hexanoic acid C4 levulinic acid C5 palmitic acid C6 oleic acid Phenols P1 phenol P2 o-cresol P3 m-cresol P4 p-cresol P5 2,4-dimethylphenol Alkyl-Substituted Catechols D1 4-methylcatechol D2 3-methylcatechol D3

4-ethylcatechol

E1 E2 E3 E4 E5

Hydroxybenzenes 1,2-dihydroxybenzene 1,3-dihydroxybenzene 1,4-dihydroxybenzene 1,2,3-trihydroxybenzene 1,2,4-trihydroxybenzene Furans

F1

2-furanmethanol

F2 F3 F4

furfural 5-methyl furfural 5-hydroxymethyl furfural

G1 G2 G3 G4 G5 G6 G7 G8 G9

Guaiacols guaiacol 4-methylguaiacol 4-ethylguaiacol 4-vinylguaiacol 4-propylguaiacol eugenol isoeugenol (Z) isoeugenol (E) vanillin

G10

acetovanillone

Figure 4. Total ionization chromatograms (TICs) from the GC/MS analysis of Oil B derivatized with MSTFA. (A) Unaged oil sample. (B) Oil aged for 40 h at 80 °C. Compounds defined by peak labels are identified in Table 2; peaks h3, h6, h7, and h12 are unidentified carbohydrates; for compounds H4−H6, “α” and “β” refer to the alphaand beta-anomeric forms, respectively; “f” and “p” refer to the furanoor pyrano-ring forms, respectively. Peak labels appear in red when peaks decreased in abundance by greater than 25% following aging; peak labels appear in blue when peaks increased in abundance.

G11 guaiacylacetone G12 homovanillyl alcohol G13 vanillic acid G14 3-vanilpropanol G15 coniferyl alcohol (Z) G16 coniferyl alcohol (E) G17 coniferylaldehyde (E) Sugars and Anhydrosugars H1 galactosan H2 mannosan H3 levoglucosan H4 arabinose H5 xylose H6 glucose H7 cellobiosan H8 cellobiose Resin Acids J1 isopimaric acid J2 dehydroabeitic acid J3 abeitic acid Stilbenes K1a methoxy/hydroxy stilbene (E) K2b dimethoxy-dihydroxy stilbene (E) Syringols S1 syringol S2 4-methylsyringol S3 4-ethylsyringol S4 4-vinylsyringol S5 4-ethenylsyringol S6 4-(1-propenyl)-2,6dimethoxyphenol (Z) S7 4-(1-propenyl)-2,6dimethoxyphenol (E) S8 syringaldehyde S9 sinapaldehyde

(signifying an increase) compound labels. These compounds include low molecular weight aldehydes, ring-conjugated phenolics, specific hydroxybenzene isomers, and certain carbohydrates. Reactive Aldehydes. Aldehydes, specifically, or carbonylcontaining compounds, in general, have been found to be reactive components in bio-oils11,13,18,21,49 and bulk decreases in carbonyl content have been found to correlate with other measures of oil stability (viscosity and production of waterinsoluble content).13 Using our MSTFA-derivatized oil samples, we made use of fragmentations characteristic of MSTFA-aldehyde adducts (Supporting Information Figure SI4), which produce ions at m/z 134, 184, and 228,45,47 to identify aldehydes present in the unaged and aged oil samples; extracted ion chromatograms for the characteristic m/z 184 signals were used to selectively highlight changes in aldehyde abundance as a function of aging (Figure 5A and B, respectively). As illustrated by peaks in Figure 5, where decreases in known (labeled with a compound identifier) or unknown (labeled with an a) aldehydes are indicated by red colored labels, significant decreases in aldehydic concentrations occur upon aging; furthermore, we find that aliphatic aldehydes, including formaldehyde (A1), acetaldehyde (A2), and glycolaldehyde (A4), decrease more significantly than aromatic aldehydes, including the furfurals (F2−4), and the guiacol/syringol aldehydes: vanillin (G9), syringaldehyde (S8), coniferyl aldehyde (G17), and sinapaldehyde (S9). To compare the kinetic reactivity of these compounds, we made use of peak areas from EICs selectively targeting specific compounds and calculated the response ratio relative to the peak area for our internal standard, biphenyl (IS). This approach has allowed us to compare the relative reactivities of compounds that may coelute with other components to assess the relative reactivity of compounds for which standards are not available for quantitative measurements. The kinetic reactivity of different types of aldehydes are illustrated in Figure 6A, where relative changes in concentration are displayed as a function of aging time for representative

a

A complete listing of compounds, retention times, and bio-oil concentrations can be found in Supporting Information Tables SI-3 and 4; structures in Figure SI-3.

abundant components. As summarized in Table 2, detected components include those commonly reported in pyrolysis oils, including substituted guaiacols, syringols, alkylated phenols, hydroxyl ketones and polyols, furans, aliphatic and aromatic aldehydes, resin and fatty acids, carbohydrates, and substituted stilbenes. Compounds that have changed most significantly in abundance are indicated by red (signifying a decrease) or blue F

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 5. Extracted ion chromatograms (EICs) for the m/z 184 fragment ion characteristic of MSTFA-aldehyde adducts from the GC/ MS analysis of Oil B derivatized with MSTFA. (A) Unaged oil sample. (B) Oil aged for 40 h at 80 °C. Compounds defined by peak labels are identified in Table 2; peaks labeled with lower case “a” are unknown aldehydes; peak labels appear in red when peaks decreased by greater than 25% following aging.

aldehydes; data for all compounds can be found in Supporting Information Figure SI-5A. While formaldehyde (A1) and glycolaldehyde (A4) showed significant decreases in concentration with aging time, the aromatic aldehydes vanillin (G9) and syringaldehyde (S8) remained unchanged with aging. Very minor reactivity was observed for the furfural (F2) and 5hydroxymethyl furfural (F4), while the ring-conjugated aromatic coniferyl aldehyde (G17) and sinapaldehyde (S9) show modestly enhanced reactivity. We also note that dimeric forms of glycolaldehyde (B8a and b) were detected in our GC/ MS analysis and also decreased significantly upon aging (see Supporting Information Figure SI-5H), suggesting that aldehydic components may also be present as a reservoir of oligomeric forms under the acidic conditions of the bio-oil. Thus, our work (with a low water-content oil) demonstrates that aliphatic aldehydes, in particular, are highly reactive and change most significantly upon aging, consistent with the expected deactivation of the carbonyl carbon for aromatic aldehydes. Reduced reactivity for vanillin (G9), furfural (F2), and 5-hydroxymethyl furfural (F4) compared with formaldehyde (A1) was also noted in a recent study where model compounds were subjected to accelerated aging with pyrolytic lignin under acidic conditions.22 Our conclusions regarding the relative reactivity of aldehydic components is in agreement with observations from our 13C NMR measurements, where we also observed rapid decreases in peak intensity for specific signals that showed closest agreement with signals for aliphatic, but not furfurals or aromatic, aldehydes. Reactivity of Ring-Conjugated Aromatics. A second group of compounds that showed high reactivity upon aging had the shared feature of being able to form stabilized carbocations, commonly via a side-chain conjugated to a guaiacol/syringol ring. While most members of the large group of guaiacols and syringols found in the Oil B exhibited little to no reactivity in the oil, as shown qualitatively in Figure 4 by compounds labeled with a “G” or “S” and by the representative kinetic profile for 4-ethylguaiacol (G3) shown in Figure 6B (information on all compounds provided in Supporting Information, Figures SI-5B and C), notable exceptions were found for conjugated aromatics, most significantly 4-vinylguaiacol (G4), 4-vinylsyringol (S4) and the cis- and trans-

Figure 6. Normalized abundances (mean ± SD; n = 3) for selected TMS-derivatives of pyrolysis oil components, as measured by GC/MS plotted as a function of aging time at 80 °C. (A) Aldehydes. (B) Guaiacols and syringols. (C) Phenols and hydroxybenzenes. (D) Carbohydrates: monosaccharides. (E) Carbohydrates: anhydrosugars and disaccharides. Compounds identified in Table 2; lower case identifiers indicate an unidentified component; for compounds H4− H6, “β” refers to the beta-anomeric form and “p” refers to the pyranoring form. Full data set can be found in Supporting Information Figure SI-5.

isomers of coniferyl alcohol (G15 and G16), as shown in Figure 6B; more modest enhanced reactivity was associated with the ring-conjugated aldehydes (G17 and S9; Figure 6A), as discussed above, and for isoeugenol (G8; Figure 6B). For isoeugenol and coniferyl alcohol, which were both observed as cis- and trans-isomers, the trans-isomers (G8 and G16, respectively) showed enhanced reactivity compared with the cis-isomers (G7 and G15, respectively; see Supporting Information, Figure SI-5B for G7 results). In contrast, with compounds where the side chain was conjugated to the guaiacol/syringol ring, compounds substituted with carbonyl (G10 and G11) or carboxylic acid (G13) groups were not reactive (see Supporting Information Figure SI-5B). The reactivity of some of these compounds has been previously reported by other researchers.11,20,21 When we considered structural features that most strongly correlated with the observed reactivity, one theme arose, consistent with acid-catalyzed formation of stabilized carbocaG

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels tions. For example, the reactivity of 4-vinylguaiacol (G4) can be rationalized by the addition of an electrophile (H+, for example) to the vinyl group to produce a benzylic carbocation (see Scheme 1A), with the benzylic carbon a site for

proposed mechanism. In this case, protonation of the γhydroxyl group, followed by water loss, can be invoked as the step initiating formation of the protonated quinone methide (shown for G16 in Scheme 2A). For both isomers of coniferyl

Scheme 1

Scheme 2

nucleophilic attack. Furthermore, the resulting benzylic carbocation intermediate is stabilized by resonance via the para-phenyl hydroxyl group, producing an intermediate that can be described as a protonated quinone methide (Scheme 1). While olefinic condensations with other olefins10 have been evoked to explain the reactivity of bio-oil components with an unsaturated side chain, the protonated quinone methide is susceptible to attack by other nucleophiles, including alcohols, water, and other electron-rich aromatics present in the bio-oil. For example, acid-catalyzed additions of water and ethanol to the vinyl group of 4-vinylguaiacol are documented through the formation of 4-(1-hydroxyethyl)-2-methoxyphenol)50 and 4-(1ethoxyethyl)-guaiacol.51 Furthermore, ortho- and para-quinone methides are used synthetically for symmetric and asymmetric syntheses;52−54 notably, both 2- and 4-vinylphenols have been exploited to form reactive quinone-methides using weaker acids, including acetic acid, as the catalyst at 80 °C.55 For the 2and 4-vinylphenols, the resulting quinone methides also exhibited their characteristic enhanced reactivity toward nucleophiles at the benzylic carbon.55 When we further considered this potential mechanism in the context of other reactive and unreactive compound structures, we explored the mechanism by comparing the relative reactivity of two isomers: eugenol (G6; unreactive; see Supporting Information, Figure SI-5B) and isoeugenol (G8; moderately reactive). Eugenol, with a double bond that is not conjugated to the ring, is unable to form the protonated quinone methide (Scheme 1C), while the carbocation that would be formed from isomeric isoeugenol, with a ring-conjugated bond, is stabilized as a protonated quinone methide (Scheme 1B), supporting the relative reactivity of these two compounds. The enhanced reactivity of the cis- and trans-isomers of coniferyl alcohol (G15 and G16) can also be explained by the

alcohol, the resulting quinone methide has two sites that may be the target of nucleophilic attack. The behavior of coniferyl alcohol can be contrasted with that of 3-vanilpropanol (G14), which has no ring-conjugated bond (see Scheme 2B). For this structure, which exhibited low reactivity upon aging, formation of the protonated quinone methide is not possible. Furthermore, we can explain the elevated reactivity of coniferyl aldehyde (G17) and sinapaldehyde (S9) relative to vanillin (G9) and syringaldehyde (S8) by the presence of the conjugated double bond, leading to isoeugenol-like levels of reactivity. Finally, furfuryl alcohol (F1) provides another example of a compound whose high reactivity (see Figure 6B) may be explained by a similar mechanism. In this case, water loss from protonated F1 (Scheme 2C) produces a carbocation stabilized by the electron-rich furan ring. The potential reactivity of this bio-oil component has been recognized in model compound studies.24 In contrast, 5hydroxymethyl furfural (F4) shows only modest reactivity (see Figure 6A). In this latter case, the electron-withdrawing aldehyde group would be the preferred site for protonation, precluding water loss and formation of the reactive carbocation. With reference to our 13C NMR results, which provided evidence for attenuated signals from aromatic and ring-methoxy carbons, and in the context of the results of other investigators, we can point to the possibility that the structural features associated with these specific reactive guaiacols and syringols make the most significant contributions to reactions producing the insoluble, pyrolytic lignin fraction of the oil. Furthermore, in connection with our results with aldehydic components, we find support to suggest that this class of specific ring-conjugated aromatics are most likely to be reacting with nonaromatic aldehydes, not furfurals or other aromatic aldehydes. In the acidic oil, protonated small aldehydes can also play a role as electrophilic species, replacing H+ when reacting across the double bond, which would also result in the production of activated, protonated quinone methide structures analogous to H

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

ring hydroxyl group. The reduced reactivity can be attributed to this lack of ring activation. When considering the structures and substituent effects for other unreactive, but structurally related, ring-substituted compounds, including 4-methylguaiacol (G1) and 4-methylcatechol (D1), shown in Scheme 3B, weaker ring activation is also evident (see Scheme 3B), consistent with our results showing that these compounds exhibited little to no reactivity upon aging in the oil. A mechanism that can be invoked to describe the reactivity of hydroxybenzenes is shown in Scheme 4 for 1,2,3-trihydroxybenzene (E4). Under acid-catalysis, ring-activated E4 attacks protonated formaldehyde; tautomerization and dehydration yields a resonance-stabilized quinone methide (Scheme 4A), which may also exist in other tautomeric forms (not shown). Formation of quinone methide intermediates has been documented in the synthesis of 1,3-dihydroxybenzene aerogels.62 Attack by a nucleophile, potentially another ringactivated, electron-rich aromatic, would yield a condensation product that, notably, is still highly activated with respect to subsequent reaction through additional reaction cycles (Scheme 4B). Furthermore, reactions with more complex aldehydes (glycolaldehyde or other hydroxyl-substituted aldehydes) offer possibilities for other reactions and functionalization of the oligomeric product that may catalyze the production of other higher molecular weight products. Additionally, quinone methides are also known to participate in [4 + 2] cycloaddition reactions with alkenes,52 and such reactions may also play a role in the observed reactivity of these compounds. Compositional Changes to Carbohydrates. Finally, we found evidence for significant changes in the composition of carbohydrates or components identified by fragmentation patterns as being carbohydrate-derived (possible mono- or disaccharides). Monosaccharides present particular challenges for identification and quantification by GC/MS because, in the presence of water, a sugar can exist and interconvert between multiple forms, including the α- or β-forms of five- or sixmembered ring structures (furano- or pyrano-ring, respectively) or open-ring (ketol or aldol) forms. Upon derivatization, a single monosaccharide can produce multiple isomeric products reflecting the distribution of isomers in the sample.63−65 Furthermore, the EI mass spectra of the TMS derivatives are characterized by extensive fragmentation to common, lower m/ z peaks, although some structurally diagnostic peaks are observed. For example, a dominant peak at m/z 217 correlates with fragmentation of the furano-ring form; an m/z 204 peak with pyrano-ring sugars.64,65 In contrast with monosaccharides, anhydrosugars are not subject to mutarotation because the glycosidic carbon is included in the ring formed by dehydration; reflecting this, anhydrosugars are found in a single form and produce a single peak following derivatization. In addition to the anhydrosugars (H1−H3), cellobiosan (H7), and cellobiose (H8) found in the unaged Oil B sample, shown in Figure 4A, over 20 additional peaks were assigned as being carbohydrate-derived sugars, with sugars of unknown identity labeled with a lower case “h” in Figure 4. While levoglucosan (H1) and cellobiosan (H7) showed no change in concentration with aging and modest decreases in concentration were observed for two anhydrosugars, galactosan (H1) and mannosan (H2), very signficant changes were observed for other carbohydrate-derived compounds. Because many of these putative sugars coeluted with other oil components, EICs for m/z 217 (furano-ring sugars) and m/z

that shown in Scheme 1A. The resulting activated structure has the potential to participate in additional chemical transformations with nucleophilic species in the oil (alcohols or other ring-activated lignin-derived species, for example). If the aldehyde is more complex and contains a nucleophilic site, internal cyclization or oligomerizations catalyzed by another electrophile become additional pathways to form higher molecular weight products. Selective Hydroxybenzene Reactivity. Specific hydroxybenzenes were a third class of compounds that decreased significantly upon aging. While most mono- and dihydroxybenzenes, including phenol and cresols (P1−P5) and alkylated catechols (D1−D3), showed minor to no decreases in concentration with aging (see Figures 4 and 6C; information for all compounds in Supporting Information Figure SI-5D), three specific hydroxybenzenes, 1,3-dihydroxybenzene (E2), 1,2,3-trihydroxybenzene (E4) and 1,2,4-trihydroxybenzenes (E5), exhibited very high reactivity in Oil B subjected to accelerated aging (see Figure 6C). The trihydroxybenzene isomers are products of cellulose pyrolysis;56−58 however, they have not been commonly reported as components of bio-oils. We have found that these compounds are detected by GC/MS only when derivatized. We considered the specific reactivity of the three hydroxybenzenes (E2, E4, and E5) in the context of acidcatalyzed phenol/formaldehyde reactions with formaldehyde, where phenols react with formaldehyde and other aldehydes to form −CH2− or −CH2−O−CH2− linked polymers.10,59,60 In these reactions, ortho and para positions are the active sites for electrophilic addition of protonated formaldehyde to the aromatic ring, with preference for para substitution under acid-catalysis.60 When considering the three reactive hydroxybenzenes (E2, E4, and E5) identified in the oil, the higher number of activating hydroxyl substituents and their location on the aromatic ring should result in high reactivity at ortho or para sites with respect to electrophilic attack (Scheme 3A). All Scheme 3

three reactive hydroxybenzene have three available ortho/paraactivated sites, with some sites activated by the augmenting effects of two hydroxyl groups (indicated by double arrows in Scheme 3A). The reactivity of 1,3-dihydroxybenzene with formaldehyde59 and furfural61 using acids (including acetic acid) as catalysts has been exploited for the synthesis of aerogels. In contrast with the reactive hydroxybenzenes, phenol (P1) and 1,2- and 1,4-dihydroxybenzene (E1 and E3) have available ortho/para sites; however, each site is only activated by one I

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels Scheme 4

204 (pyrano-ring sugars) were used to more specifically target changes in abundance that occurred upon aging (Figure 7). Fragment ions from the anhydrosugars (H1−H3), which appear at both m/z 217 and 204, were responsible for the most abundant signals; however, this analysis revealed five unknown sugars, characterized by a dominant m/z 217 ion and classified as furano-ring, that decreased in abundance as the oil aged (peaks in red shown in Figure 7A). We also observed four peaks that increased in abundance with aging, identified as the α- and β-forms of furano-ring arabinose (H4(α,f) and (β,f)) and xylose (H5(α,f) and (β,f); peaks shown in blue in Figure 7A). The EICs for the m/z 204 peak characteristic of pyranoring sugars reveals peaks for six sugars identified as the α- and β-forms of pyranose-ring arabinose (H4(α,p) and (β,p)), xylose (H5(α,p) and (β,p)), and glucose (H6(α,p) and (β,p)), which increased in abundance for the aged oil sample (see peaks in blue shown in Figure 7D). Using EICs for chromatographic peaks that could be reliably resolved for integration, a kinetic analysis of abundance changes with time revealed six sugars (h1, h3, h5-h7, and h12) that decreased to less than 50% with aging, all of which were classified as furano-ring sugars based upon their low m/z 204 to 217 ion ratios (see Supporting Information Figure SI-5E and F). As shown by representative kinetic profiles in Figure 6D, putative carbohydrates h3, h6, and h7 decreased to less than 50% within the first 10 h of aging time. Ten peaks, all attributed to isomers of arabinose (H4), xylose (H5), and glucose (H6) rapidly increased in abundance (see Figure 6D and Supporting Information Figure SI-5E). Between the α- or β-anomers of each monosaccharide (H4-H6), similar kinetic profiles were observed. Four unidentified carbohydrates, including h4 and h8 (see Figure 6D), remained unchanged by accelerated aging; both levoglucosan (H3) and cellobiosan (H7) showed no evidence for reaction (Figure 6E), while galactosan (H1) and mannosan (H2) showed slow decreases in abundance (Figure 6E). In contrast, cellobiose (H8) increased in abundance with aging time (Figure 6E). Carbohydrates are a significant, and frequently problematic, component of bio-oils and remain a challenging oil component to characterize in a structurally specific and reliable way.37,66,67 Previous aging studies have used solvent fractionation to quantify the carbohydrate component of the oil, which was found to decrease as a mass fraction during the first six months

Figure 7. Extracted ion chromatograms (EICs) from the GC/MS analysis of Oil B derivatized with MSTFA. (A) m/z 217, a fragment ion characteristic of furano-ring carbohydrates; unaged oil sample. (B) m/z 217 for aged Oil B. (C) m/z 204, a fragment ion characteristic of pyrano-ring carbohydrates; unaged oil sample. (D) m/z 204 for Oil B aged for 40 h at 80 °C. Compounds defined by peak labels are identified in Table 2; for compounds H4−H6, “α” and “β” refer to the alpha- and beta-anomeric forms, respectively; “f” and “p” refer to the furano- or pyrano-ring forms, respectively; peaks labeled with lower case “h” are unknown carbohydrates; peak labels appear in red when peaks decreased by greater than 25% following aging; peak labels appear in blue for peaks that increased.

of aging.11 Based upon correlation with increases in the high molecular weight lignin fraction, condensation/polymerizations J

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Our quantitative analysis of oil components (Supporting Information Table SI-3) accounts for 21.1% of the mass of the oil before aging and 17.6% after aging. Thus, we have accounted for a mass decrease of around 16.5% during aging for 40 h at 80 °C. 3.4. Chip-Based nanoESI-QTOF-LC/MS Analysis. To extend our ability to detect more polar and higher molecular weight components of the bio-oil samples, we analyzed the samples using a chip-based nanoESI-Q-TOF-LC/MS/MS system. LC/MS techniques have been applied to the analysis of pyrolysis oil compounds using negative ion mode, hydroxidedoped ESI.70 For our study, we used an HPLC chip71 that integrates an enrichment column, a 150 mm × 75 μm reversedphase analytical column, and a nanoelectrospray ionization (ESI) tip for the separation and high sensitivity detection of compounds amenable to ionization by ESI. Because of the wide gradient needed for the analysis of complex bio-oil samples, our analyses on this chip-based system were limited to positive ion detection, an ionization mode that is not best suited for the analysis of compounds, including most bio-oil components, which lack a basic functional group. Furthermore, highly polar compounds, like carbohydrates, are not sufficiently retained to be separated by our reversed-phase column. To address these deficiencies, we peracetylated the bio-oil samples prior to analysis, based upon finding that the ionization response increased relative to underivatized components and that carbohydrates were retained and well-resolved chromatographically. To demonstrate that our peracetylation approach resulted in complete derivatization, we derivatized carbohydrate standards alone and when spiked into bio-oil samples and found no evidence for partially acetylated derivatives. As an internal quality control check to assess the peracetylation reaction for the unaged and aged Oil B samples, we looked for signals for partially derivatized pinitol (B9), which forms a penta-acetylated derivative, to determine if the derivatization reaction was incomplete; only completely derivatized product was observed. Using this analytical approach, peracetylated unaged and aged (40 h) samples of Oil B were analyzed using the chipbased nanoESI-Q-TOF-MS/MS system (see Figure 8A), and the response of the samples was compared with a method blank. Each chromatogram results from the injection of ∼750 ng of the underivatized bio-oil. A chromatogram for a reference mixture of a few higher molecular mass standards (carbohydrates, H3 and H6−H8, and two dilignols, pinoresinol (L1) and guaiacylglycerol-β-guaiacyl ether (L2)) is shown in Figure 8B. The TICs for unaged and aged oils were highly complex (Figure 8A); however, the profiles were very similar and did not reveal dramatic changes following aging. We first carried out an analysis to determine how our LC/MS results, targeting changes in compound signals, compared with those observed by GC/MS. In conducting this analysis, compound identifications were based upon accurate mass measurements, evaluation of MS/MS spectra, detection of expected mass shifts upon peracetylation with d6-acetic anhydride and, whenever available, comparisons of putative compounds with the mass spectra characteristics and retention times for authentic standards. We first examined the signals for compounds that were found to be unreactive by GC/MS. As illustrated by the exact mass EICs for 4-propylguaiacol (G5), 3-vanilpropanol (G14), and eugenol (G6) shown in Figure 9A−C, no significant changes in

that potentially reduced solubility and resulted in the transfer of carbohydrate-like components to the water-insoluble fraction were reported.11 Levoglucosan, specifically, has been found to increase in concentration during some aging experiments,10,11,18 explained via repolymerization of glycolaldehyde10,18 or acid hydrolysis of sugars;11 in other investigations, levoglucosan-derived signals were found to decrease, driven, it was suggested, by the decomposition to furfural or other small molecules.19 High-resolution Fourier transform MS analysis of bio-oil aging found negligible changes in sugar-derived products upon aging.28 Our solvent fractionation results, where we also quantified the water-soluble, ether-insoluble carbohydrate fraction (Table 1) and found a 21% decrease in this fraction, agree with prior work;11 however, our semiquantitative analysis of the detected carbohydrates only showed a 1% decrease in mass (see Supporting Information Table SI-4J). We did, however, observe compositional changes resulting from significant increases in the concentrations of isomers of arabinose, xylose, and glucose (H4−H6) and decreases to the concentrations of unknown furano-ring sugars. Our data provided no support for acidcatalyzed isomerization of glucose to fructose or dehydrations to produce 5-hydroxymethyl furfural (H4) or levulinic acid (C4) with aging, as the concentrations of these species decreased slightly or showed no significant changes, respectively, and fructose was not detected by our analysis. Based upon the identities of the sugars found to increase in abundances (arabinose, xylose, and glucose) we can hypothesize that these sugars originate from hemicellulose, presumably arabinoxylans. Hemicelluloses have been found to show lower thermal stability compared with cellulose68,69 and would not be expected to survive the pyrolysis process; however, the analysis of pyrolytic lignins has provided support for oligomeric structures containing differing amounts of hemicellulose connected to the lignin as phenyl glycosides or ferulate esters.32 Through aging, these pyrolytic lignin-associated hemicelluloses may be released, potentially changing the solubility characteristics of the pyrolytic lignin or activating sites for additional condensation reactions. Reactivity of Other Compound Classes. We also studied the behavior of representative carboxylic acids, including three small hydroxyl aliphatic acids (lactic, glycolic, levulinic, and hexanoic acids; C1−C4), two longer chain acids (palmitic and oleic acids; C5 and C6), vanillic acid (G13), and three resin acids (J1−J3). For all acids examined, the signals were more variable (see Supporting Information Figure SI-5G and I), and provided little support for the contribution of esterification reactions to oil aging chemistry. We did, however, observe changes to specific hydroxyl ketones and polyols. Alcohols may play a role in aging reactions, acting as nucleophiles in the conversion of aldehydes to hemiacetals and acetals, for example, and in other addition reactions, which may include additions to ring-activated aromatics. While none of these compounds exhibited the rapid reactivity apparent for other components, 1hydroxy-2-cyclopenten-1-one (B3), 1,3-dihydroxy-propan-2one (B6) and glycerol (B7) decreased with oil aging (see Supporting Information Figure SI-5H). As described above, the glycolaldehyde dimers (B8a and b) decreased in more significant ways with aging. Finally, we observed some other examples of enhanced reactivity for some unidentified higher molecular weight components, and for the cis/trans isomers of a putative hydroxyl-methoxy-substituted stilbene (K1a and b; see Supporting Information Figure SI-5I). K

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Na]+ ion that provides the most carbohydrate-specific signals, only weak signals were detected at the retention time characteristic of glucose for the unaged Oil B sample (Figure 9J); however, other components that yielded mass spectra information (exact masses, MS/MS spectra) characteristic of peracetylated C6H12O6 carbohydrates were detected at earlier retention times. When we examined EICs for the aged Oil B sample (Figure 9J), signals characteristic of glucose at the correct retention time were now detected; furthermore, the exact mass EIC showed decreasing intensity for earlier eluting peracetylated C6H12O6 carbohydrates (Figure 9J). This behavior agrees with our GC/MS data, where the α- and βforms of TMS-derivatized glucopyranose (H6) were detected in the aged, but not unaged, Oil B sample. More dramatic changes in chromatographic profiles were observed for signals characteristic of peracetylated xylose/ arabinose (C5H12O5) sugars (Figure 9K). We found that signals for earlier-eluting sugars with this molecular formula decrease in abundance, while later-eluting components increase in signal strength (Figure 9K) for the aged oil sample. Again, this result agrees with our GC/MS data, where the α- and β-forms of TMS-derivatized arabino- and xylopyranoses (H4 and H5) were detected in the aged, but not unaged, Oil B sample. Finally, while the signal for peracetylated cellobiose (H8) showed a modest increase between the unaged and aged samples (Figure 9L), other peaks attributed to cellobioserelated C12H22O11 carbohydrates either were unchanged or decreased in abundance. Thus, our nanoESI-LC/MS analysis provides further support for our conclusions regarding compositional changes to carbohydrate components. nanoESI-LC/MS/MS for the Characterization of Oil-Aging Products. The LC/MS/MS data for the bio-oil samples are rich sources of chemically specific information that can be probed to characterize higher molecular mass aging products and further establish structural motifs that play a role in aging chemistry. For example, when we subjected the data sets to a Find by Molecular Feature (FMF; Agilent Mass Hunter software component) analysis to generate extracted compound chromatograms (ECCs) for products detected above a defined threshold, in the time range of 5−45 min over 1000 compounds were identified as distinct chromatographic peaks (data not shown). Many later eluting components yield MS/ MS spectra showing signatures characteristic of coniferyl alcohol and other substituted aromatics. While a detailed exploration of higher molecular mass components of these samples is beyond the scope of this publication, we include one specific and one broad conclusion that can be drawn from the data. First, in support of our observation that trihydroxybenzenes are highly reactive components of the oil, activated toward potential dimerization reactions with formaldehyde, we report on the first, to our knowledge, identification of a bio-oil derived trihydroxybenzene oligomer (see Figure 10). Our identification is supported by exact mass measurements, shifts in mass upon peracetylation with d6-acetic anhydride that provide evidence for the presence of six hydroxyl groups (Supporting Information, Table SI-5) and an MS/MS spectrum showing evidence for a methylene bridge connecting two polyhydroxylated aromatic rings (Figure 10) accompanied by the expected losses from the added acetyl groups. This putative dimer is detected in both unaged and aged oil samples; however, the abundance of this component is lower in the aged oil sample,

Figure 8. Chip-nanoESI-Q-TOF MS analysis of peracetylated bio-oil samples and standards. (A) TICs of peracetylated Oil B unaged and aged for 40 h at 80 °C and a method blank; compounds identified in Table 2. (B) Overlay of EICs for standards of levoglucosan (H3), glucose (H6), cellobiosan (H7), cellobiose (H8), pinoresinol (L2), and guaiacylglycerol-β-guaiacyl ether (L3); all detected as peracetylated derivatives. Mobile phase composition changed from 98% mobile phase A (0.1% formic acid in water) to 100% mobile phase B (0.1% formic acid and 2% water in acetonitrile.

signal intensity were observed. For these examples and for >20 additional unreactive compounds examined, we found excellent agreement with the GC/MS reactivity classifications. In contrast with the signal consistency for eugenol (G6; Figure 9C), we found that the intensity for isoeugenol (G8) decreased for the aged Oil B sample (Figure 9C). When we examined signals for other compounds found to decrease in the aged oil sample, including the trihydoxybenzenes (E4 and E5), 4vinylguaiacol (G4), and coniferyl alcohol (G16), shown in Figures 9D, E and F, we again found agreement with our GC/ MS results, with all four compounds showing significantly reduced signal intensities in the aged oil sample. Finally, we probed the data for signals originating from carbohydratederived components of the oil, where the LC/MS should provide superior capabilities for the detection of higher molecular weight components and reduced fragmentation, relative to GC/MS, for determination of carbohydrate molecular mass. Our analysis of the unaged and aged oil samples for carbohydrate-derived sugars resulted in the detection of levoglucosan (H3), cellobiosan (H7), and cellobiose (H8) in both samples; for larger carbohydrates, only a very low abundance signal characteristic of cellotriosan was detected. EICs comparing the response for levoglucosan (H3) and cellobiosan (H7) in the unaged and aged oil samples are shown in Figure 9G and H. In agreement with our GC/MS results, we found no significant changes in signal intensity between the unaged and aged Oil B samples. We also found highly consistent EIC profiles for signals assigned to pinitol (B9; Figure 9I) and a deoxyglucose (data not shown). In contrast, we found evidence for significant changes to the carbohydrate composition, based upon the unaged and aged EIC profiles for signals characteristic of other sugars. For example, when we generated an exact mass EIC for peracetylated glucose, using the exact mass for the [M + L

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 9. Exact mass EICS from the chip-nanoESI-Q-TOF MS analysis of aged (40 h at 80 °C) and unaged peracetylated Oil B samples: (A) 4propylguaiacol (G5) at m/z 209.117, (B) 3-vanilpropanol (G14) at m/z 267.123, (C) eugenol (G6) and isoeugenol (G8) at m/z 207.102, (D) 1,2,3and 1,2,4-trihydroxybenzene (E4 and E5) at m/z 253.071, (E) 4-vinylguaiacol (G4) at m/z 193.086, (F) coniferyl alcohol (G16) at m/z 205.085, (G) levoglucosan (H3) at m/z 289.209, (H) cellobiosan (H7) at m/z 577.176, (I) pinitol (B9) at m/z 405.139, (J) glucose (H6) and other pentaacetylated C6H12O6 carbohydrates at m/z 413.105, (K) arabinose/xylose tetraacetylated C5H10O5 carbohydrates at m/z 341.084, (L) cellobiose (H8) and other octaaacetylated C12H22O11 carbohydrates at m/z 619.187. EICs were generated using a 20 ppm window.

suggesting that it is formed very rapidly and undergoes subsequent reactions as the oil ages. On a broad scale, we were interested in determining if our peracetylation chip-nanoESI-Q-TOF MS/MS approach provided support for shifts to higher molecular weight products as the oil aged. Such changes are evident in the molecular weight shifts observed by GPC in past studies and in our GCP analysis of unaged and aged Oil B samples (see Figure 1). Furthermore, mass shifts to higher m/z signals have been reported for the direct (no LC) analysis of aged and unaged samples by atomospheric pressure photoionization (APPI) and ESI-FTICR MS. Although our peracetylation LC/MS approach cannot be compared with GPC in absolute terms (because of mass shifts resulting from peracetylation of the samples, in-source fragmentation and clustering during ionization, variations in

ionization efficiency, and mass discrimination effects) we felt that, by direct comparisons of samples analyzed under similar conditions, we could obtain useful comparative information about how molecular mass changes varied between the two samples. To answer this question, we integrated extracted ion chromatograms for 100 Da mass ranges starting at m/z = 100 and ending at 3000. Our results show a maximum ion intensity at an m/z value of ∼350 (Figure 11A). Signal intensity decreased significantly for m/z values above 1,500. When we calculated the weightaverage molecular weight for the nanoESI-LC/MS analysis of these two samples, we found a mass shift from 491 to 529 Da for the unaged to aged samples. Our LC/MS values for the mass shift were lower than those determined by GPC, where a M

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

shift from 393 to 632 g/mol was observed, but both show a shift to higher molecular mass for the aged samples. To better focus on changes to the molecular weight distributions, we calculated the ratio of signals for the aged to unaged samples; this ratio showed a continual increase with m/ z value for the aged Oil B sample relative to the sample that was not subjected to accelerated aging (Figure 11B). This result again indicates that our nanoESI-LC/MS approach provides evidence for the conversion of lower molecular mass components to higher mass components as the oil ages. Furthermore, we compared our LC/MS data with that measured using GPC (Figure 1); while a comparison of the results in terms of the absolute ratios is not appropriate for reasons listed above, we can compare the general trend where both types of analyses show a decrease in lower mass components and a shift toward higher molecular weight components with oil aging. This suggests that the higher mass products detected by our nanoESI-LC/MS approach can be targeted for structural characterization. Figure 10. Mechanism of formation and CID-MS/MS spectrum for the m/z = 517.134, [M + H]+ ion from a putative trihydroxybenzene dimer with formaldehyde, detected at 39.9 min in peracetylated Oil B, analyzed by chip-nanoESI-Q-TOF MS/MS. (A) Mechanism proposed for dimer formation, shown for 1,2,3-trihydroxybenzene (E4) reacting with protonated formaldehyde. (B) CID-MS/MS spectrum with ion isolation was carried out using a 1.3 Da window; collision energy was 22 eV; nitrogen was used as the collision gas.

4. CONCLUSIONS Our analysis of bio-oil aging, using a combination of 13C NMR, GPC, GC/MS, and nanoESI-LC/MS reinforces findings that have appeared in the literature and adds new information regarding the identification of highly reactive oil components. Our 0.8% increase in water content observed after aging is consistent with condensation reactions that form water as a byproduct. Our GC/MS analyses of derivatized oil samples extended the range of compounds that have been reported and provided more discriminating analyses by which to compare relative compound reactivity in the oil upon aging. We identified structural features of reactive compounds that correlate with reactivity and report on the presence and high reactivity of certain (1,3-di-, 1,2,3-tri-, and 1,2,4-tri-) polyhydroxylated benzenes and propose the formation of quinone methides as important intermediates that account for the high reactivity of selective compounds. We also find that, while anhydrosugars are largely stable with respect to aging, significant compositional changes occur to glucose, xylose, arabinose, and other putative carbohydrate-like components; specifically, we find evidence for the disappearance of certain furano-ring sugars and the appearance of possibly hemicellulose-derived sugars as the oil ages. We hypothesize that some of the sugars are produced by aging-induced release of lignin-carbohydrate oligomers. Our derivatization GC/MS analysis accounted for approximately 16% of the mass of the compounds that decrease upon aging. Our nanoESI-LC/MS examination of peracetylated bio-oil samples has supported conclusions based upon GC/MS and also shows agreement with GPC data that has documented shifts in molecular weight upon aging. Using LC/MS analyses, we found the first compound-specific evidence for the formation of a trihydroxybenezene−formaldehyde dimer, presumably generated from one of the two trihydroxybenenes that decrease in abundance with oil aging.



Figure 11. (A) Integrated areas for 100 Da ranges for time = 5−50 min, each point corrected for signals from the method blank for chipnanoESI-Q-TOF MS analysis of peracetylated Oil B unaged and aged for 40 h at 80 °C. (B) Ratio of signals from the aged/unaged sample from part A. (C) Ratio of signals for GPC data shown in Figure 1.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.energyfuels.6b00238. N

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels



(20) Kim, T.-S.; Kim, J.-Y.; Kim, K.-H.; Lee, S.; Choi, D.; Choi, I.-G.; Choi, J. W. J. Anal. Appl. Pyrolysis 2012, 95, 118−125. (21) Alsbou, E.; Helleur, B. Energy Fuels 2014, 28, 3224−3235. (22) Meng, J.; Moore, A.; Tilotta, D.; Kelley, S.; Park, S. ACS Sustainable Chem. Eng. 2014, 2, 2011−2018. (23) Venderbosch, R. H.; Ardiyanti, A. R.; Wildschut, J.; Oasmaa, A.; Heeres, H. J. J. Chem. Technol. Biotechnol. 2010, 85, 674−686. (24) Hu, X.; Wang, Y.; Mourant, D.; Gunawan, R.; Lievens, C.; Chaiwat, W.; Gholizadeh, M.; Wu, L.; Li, X.; Li, C.-Z. AIChE J. 2013, 59, 888−900. (25) Naske, C. D.; Polk, P.; Wynne, P. Z.; Speed, J.; Holmes, W. E.; Walters, K. B. Energy Fuels 2012, 26, 1284−1297. (26) Moens, L.; Black, S. K.; Myers, M. D.; Czernik, S. Energy Fuels 2009, 23, 2695−2699. (27) Kim, K. H.; Bai, X.; Cady, S.; Gable, P.; Brown, R. C. ChemSusChem 2015, 8, 894−900. (28) Smith, E. A.; Thompson, C.; Lee, Y. J. Bull. Korean Chem. Soc. 2014, 35, 811−814. (29) Meng, J.; Smirnova, T. I.; Song, X.; Moore, A.; Ren, X.; Kelley, S.; Park, S.; Tilotta, D. RSC Adv. 2014, 4, 29840−29846. (30) Fratini, E.; Bonini, M.; Oasmaa, A.; Solantausta, Y.; Teixeira, J.; Baglioni, P. Langmuir 2006, 22, 306−312. (31) Bayerbach, R.; Meier, D. J. Anal. Appl. Pyrolysis 2009, 85, 98− 107. (32) Fortin, M.; Mohadjer Beromi, M.; Lai, A.; Tarves, P. C.; Mullen, C. A.; Boateng, A. A.; West, N. M. Energy Fuels 2015, 29, 8017−8026. (33) Bai, X.; Kim, K. H.; Brown, R. C.; Dalluge, E.; Hutchinson, C.; Lee, Y. J.; Dalluge, D. Fuel 2014, 128, 170−179. (34) Scott, D. S.; Piskorz, J. Can. J. Chem. Eng. 1982, 60, 666−674. (35) Garcia-Perez, M.; Chaala, A.; Pakdel, H.; Kretschmer, D.; Roy, C. Biomass Bioenergy 2007, 31, 222−242. (36) Joseph, J.; Baker, C.; Mukkamala, S.; Beis, S. H.; Wheeler, M. C.; DeSisto, S. W. J.; Jensen, B. L.; Frederick, B. G. Energy Fuels 2010, 24, 5153−5162. (37) Oasmaa, A.; Meier, D. J. Anal. Appl. Pyrolysis 2005, 73, 323− 334. (38) Brown, R. C.; Meyer, T.; Fox, R.; Submramaniam, S.; Shanks, B.; Smith, R. G. A Systems Approach to Bio-Oil Stabilization−Final Technical Report, DOE/GO182052011, 2011. (39) Hilten, R. N.; Das, K. C. Fuel 2010, 89, 2741−2749. (40) Mullen, C. A.; Strahan, G. D.; Boateng, A. A. Energy Fuels 2009, 23, 2707−2718. (41) Strahan, G. D.; Mullen, C. A.; Boateng, A. A. Energy Fuels 2011, 25, 5452−5461. (42) Scholze, B.; Hanser, C.; Meier, D. J. Anal. Appl. Pyrolysis 2001, 58, 387−400. (43) Ben, H.; Ragauskas, A. J. Bioresour. Technol. 2013, 147, 577− 584. (44) Lesueur, D.; Castola, V.; Bighelli, A.; Conti, L.; Casanova, J. Spectrosc. Lett. 2007, 40, 591−602. (45) Little, J. L. J. Chromatogr. A 1999, 844, 1−22. (46) Ende, M.; Luftmann, H. Tetrahedron 1984, 40, 5167−5170. (47) Spiteller, G.; Kern, W.; Spiteller, P. J. Chromatogr. A 1999, 843, 29−98. (48) Fabbri, D.; Chiavari, G.; Prati, S.; Vassura, I.; Vangelista, M. Rapid Commun. Mass Spectrom. 2002, 16, 2349−2355. (49) Hu, X.; Gunawan, R.; Mourant, D.; Lievens, C.; Li, X.; Zhang, S.; Chaiwat, W.; Li, C.-Z. Fuel 2012, 97, 512−522. (50) Vanbeneden, N.; Saison, D.; Delvaux, F.; Delvaux, F. R. J. Agric. Food Chem. 2008, 56, 11983−11988. (51) Dugelay, I.; Baumes, R.; Gunata, Z.; Razungles, A.; Bayonove, C. Sci. Aliments 1995, 15, 423−433. (52) Toteva, M. M.; Richard, J. P. Adv. Phys. Org. Chem. 2011, 45, 39−91. (53) Pathak, T. P.; Sigman, M. S. J. Org. Chem. 2011, 76, 9210−9215. (54) Wang, Z.; Wong, Y. F.; Sun, J. Angew. Chem., Int. Ed. 2015, 54, 13711−13714. (55) Fleischer, I.; Pospech, J. RSC Adv. 2015, 5, 493−496. (56) Fabbri, D.; Chiavari, G. Anal. Chim. Acta 2001, 449, 271−280.

Experimental details for quantitative measurements; basis for relating aging times to effective ambient aging; water content; viscosity measurements; compound structures; EI mass spectra of MSTFA aldehyde adducts; normalized abundances vs time for compounds; gradient conditions for LC/MS; GPC measurements; results of quantitative analysis of Oil B before and after aging; summary of exact mass measurements for a putative trihydroxybenzene dimer (PDF)

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Author Contributions ‡

B.G.F. and E.A.S. contributed equally to the work.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We thank the Department of Energy, under Grant DE-FG0207ER46373, and the National Science Foundation, under Grant CHE-1126657, for financial support of this work. We thank Karl Bishop and David Labreque for assistance with NMR, Diane Smith and Yurai Zen for assistance with GPC, Saikrishna Mukkamala for assistance with titrations, William J. DeSisto, M. Clayton Wheeler, Nathan Hill, and Saikrishna Mukkamala for providing the oil samples, and William J. DeSisto, M. Clayton Wheeler, Barbara Cole, Paige Case, and Paige Speight for helpful discussions.



REFERENCES

(1) Bridgwater, A. V. Chem. Eng. J. 2003, 91, 87−102. (2) Mohan, D.; Pittman, C. U.; Steele, P. H. Energy Fuels 2006, 20, 848−889. (3) Czernik, S.; Bridgwater, A. V. Energy Fuels 2004, 18, 590−598. (4) Bridgwater, A. V. Biomass Bioenergy 2012, 38, 68−94. (5) DeSisto, W. J.; Hill, N.; Beis, S. H.; Mukkamala, S.; Joseph, J.; Baker, C.; Ong, T. H.; Stemmler, E. A.; Wheeler, M. C.; Frederick, B. G.; van Heiningen, A. Energy Fuels 2010, 24, 2642−2651. (6) Chaala, A.; Ba, T.; Garcia-Perez, M.; Roy, C. Energy Fuels 2004, 18, 1535−1542. (7) Diebold, J. P.; Czernik, S. Energy Fuels 1997, 11, 1081−1091. (8) Chen, D.; Zhou, J.; Zhang, Q.; Zhu, X. Renewable Sustainable Energy Rev. 2014, 40, 69−79. (9) Czernik, S.; Johnson, D. K.; Black, S. Biomass Bioenergy 1994, 7, 187−192. (10) Diebold, J. P. A Review of the Chemical and Physical Mechanisms of the Storage Stability of Fast Pyrolysis Bio-Oils, NREL/SR-570-27613; National Renewal Energy Laboratory: Golden, CO, 2000. (11) Oasmaa, A.; Kuoppala, E. Energy Fuels 2003, 17, 1075−1084. (12) Elliott, D. C.; Oasmaa, A.; Meier, D.; Preto, F.; Bridgwater, A. V. Energy Fuels 2012, 26, 7362−7366. (13) Oasmaa, A.; Korhonen, J.; Kuoppala, E. Energy Fuels 2011, 25, 3307−3313. (14) Boucher, M. E.; Chaala, A.; Pakdel, H.; Roy, C. Biomass Bioenergy 2000, 19, 351−361. (15) Garcia-Perez, M.; Chaala, A.; Pakdel, H.; Kretschmer, D.; Rodrigue, D.; Roy, C. Energy Fuels 2006, 20, 786−795. (16) Xu, F.; Xu, Y.; Lu, R.; Sheng, G.-P.; Yu, H.-Q. J. Agric. Food Chem. 2011, 59, 9243−9249. (17) Ortega, J. V.; Renehan, A. M.; Liberatore, M. W.; Herring, A. M. J. Anal. Appl. Pyrolysis 2011, 91, 190−198. (18) Fahmi, R.; Bridgwater, A. V.; Donnison, I.; Yates, N.; Jones, J. M. Fuel 2008, 87, 1230−1240. (19) Ben, H.; Ragauskas, A. J. ChemSusChem 2012, 5, 1687−1693. O

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels (57) Ponder, G. R.; Qiu, H.; Richards, G. N. Appl. Biochem. Biotechnol. 1990, 24−25, 41−47. (58) Richards, G. N.; Shafizadeh, F.; Stevenson, T. T. Carbohydr. Res. 1983, 117, 322−327. (59) Mulik, S.; Sotiriou-Leventis, C. Aerogels Handb. 2011, 215−234. (60) Uglea, C. V.; Negulescu, I. I. Synthesis and Characterization of Oligomers; Taylor & Francis, 1991. (61) Tian, H. Y.; Buckley, C. E.; Paskevicius, M.; Sheppard, D. A. Int. J. Hydrogen Energy 2011, 36, 671−679. (62) Mulik, S.; Sotiriou-Leventis, C.; Leventis, N. Chem. Mater. 2007, 19, 6138−6144. (63) Paez, M.; Martinez-Castro, I.; Sanz, J.; Olano, A.; Garcia-Raso, A.; Saura-Calixto, F. Chromatographia 1987, 23, 43−46. (64) Medeiros, P. M.; Simoneit, B. R. T. J. Chromatogr. A 2007, 1141, 271−278. (65) DeJongh, D. C.; Radford, T.; Hribar, J. D.; Hanessian, S.; Bieber, M.; Dawson, G.; Sweeley, C. C. J. Am. Chem. Soc. 1969, 91, 1728−1740. (66) Oasmaa, A.; Kuoppala, E. Energy Fuels 2008, 22, 4245−4248. (67) Tessini, C.; Vega, M.; Mueller, N.; Bustamante, L.; von Baer, D.; Berg, A.; Mardones, C. J. Chromatogr. A 2011, 1218, 3811−3815. (68) Alen, R.; Kuoppala, E.; Oesch, P. J. Anal. Appl. Pyrolysis 1996, 36, 137−148. (69) Patwardhan, P. R.; Brown, R. C.; Shanks, B. H. ChemSusChem 2011, 4, 636−643. (70) Owen, B. C.; Haupert, L. J.; Jarrell, T. M.; Marcum, C. L.; Parsell, T. H.; Abu-Omar, M. M.; Bozell, J. J.; Black, S. K.; Kenttamaa, H. I. Anal. Chem. 2012, 84, 6000−6007. (71) Yin, H.; Killeen, K. J. Sep. Sci. 2007, 30, 1427−1434.

P

DOI: 10.1021/acs.energyfuels.6b00238 Energy Fuels XXXX, XXX, XXX−XXX