SEARCH

SEARCH BY CITATION

Keywords:

  • biofuels;
  • biomass;
  • bio-oil;
  • liquefaction;
  • solvent effects

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results and Discussion
  5. Conclusions
  6. Experimental Section
  7. Acknowledgements
  8. Supporting Information

The liquefaction of lignocellulosic biomass is studied for the production of liquid (transportation) fuels. The process concept uses a product recycle as a liquefaction medium and produces a bio-oil that can be co-processed in a conventional oil refinery. This all is done at medium temperature (≈300 °C) and pressure (≈60 bar). Solvent-screening experiments showed that oxygenated solvents are preferred as they allow high oil (up to 93 % on carbon basis) and low solid yields (≈1–2 % on carbon basis) and thereby outperform the liquefaction of biomass in compressed water and biomass pyrolysis. The following solvent ranking was obtained: guaiacol>hexanoic acid≫n-undecane. The use of wet biomass results in higher oil yields than dry biomass. However, it also results in a higher operating pressure, which would make the process more expensive. Refill experiments were also performed to evaluate the possibility to recycle the oil as the liquefaction medium. The recycled oil appeared to be very effective to liquefy the biomass and even surpassed the start-up solvent guaiacol, but became increasingly heavy and more viscous after each refill and eventually showed a molecular weight distribution that resembles that of refinery vacuum residue.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results and Discussion
  5. Conclusions
  6. Experimental Section
  7. Acknowledgements
  8. Supporting Information

Mankind’s total primary energy demand is expected to increase from 536 EJ per year in 2010 to approximately 734 EJ per year in 2030 as a result of a growing world population and a higher average energy consumption per capita. Fossil (coal, gas, and oil) and nuclear sources are estimated to contribute to approximately 615 EJ per year in 2030, which means that approximately 119 EJ per year (≈15 %) has to be produced from renewable sources.1

Energy consumption can be divided into three main and comparable blocks, namely, heat, power, and transportation. For heat and power there are numerous renewable green sources available, namely, solar and geothermal for heat, biomass combustion for heat and power production, photovoltaic power, and flow energy (e.g., wind, wave, tidal, and rivers) mainly for power production. For transportation, there is currently little alternative except for green power (with expensive and heavy batteries) or biofuels. Biofuels are preferable because of their high energy densities (40–65 times higher based on weight compared to an electrical lithium ion battery) and their compatibility with existing fuel distribution infrastructure and transportation fleet. It is particularly easy if biofuels are fully blendable with existing fuels, which requires the conversion of biomass to a light, oxygen-lean, and hydrogen-rich liquid.

Technologies for the production of first-generation biofuels (e.g., ethanol and biodiesel), which are produced from edible biomass, are well established and provide the biofuels available on the market. However, concerns about competition between food and feed and about “cradle-to-cradle” energy efficiency will limit further large-scale growth of first-generation biofuel production. Second-generation biofuels, which are produced mainly from inedible lignocellulosic biomass, offer a great opportunity for biofuel production growth. Lignocellulosic biomass is currently widely available from waste, agriculture, and the wood processing industry (potentially 100–300 EJ per year)2 and part of this can be utilized in an economically viable way. However, the conversion of lignocellulosic biomass is not straightforward as it is much harder to break down the biomass into smaller building blocks especially because of its high lignin content.

The routes to convert (lignocellulosic) biomass to liquid fuels essentially contain four main process steps. Initially the biomass is pretreated, in which it is reduced in size and dried. The next step is the primary conversion step, in which biomass is depolymerized to produce smaller molecules from the cellulose, hemicellulose, and lignin. This can be achieved by various technologies such as hydrolysis, pyrolysis, liquefaction, or gasification. After this follows an upgrade (e.g., hydrotreatment, deoxygenation, and reforming) and/or synthesis [e.g., Fischer–Tropsch, methanol/dimethyl ether (DME), and fermentation] step after which the fuel is purified and optionally co-processed with refinery feeds.

One of the conversion routes is liquefaction, which produces a bio-oil with a moderate oxygen content, for example, 6–303 versus 35–40 wt % for pyrolysis oil4 and 40–50 wt % for the initial lignocellulose. The process temperature is milder than in pyrolysis (up to ≈400 °C) but the use of solvent often results in higher pressures (up to ≈250 bar if water is used as the solvent). An overview of the literature on biomass liquefaction is given in the Supporting Information in Table S1. Solvent selection is very important to obtain high liquefaction oil yields. Polar solvents usually perform better than apolar solvents. Combined oil and gas yields up to 99 wt % have been reported for liquefaction with and without the use of a homogeneous and/or heterogeneous catalyst. From a commercial point of view, the solvent cost should be minimized. Hence the solvent should either be cheap and easily recoverable, be produced within the process, and/or be co-processed with the bio-oil to the end products. From as early as the 1970s, processes have been developed to pilot scale based on two solvent types, namely, water and wood liquefaction recycle oil. Comprehensive overviews of direct liquefaction processes have been given by Behrendt et al.5 and Elliot et al.6 These processes all had drawbacks that the product was not suitable for the final upgrade to transportation fuels and/or that process conditions were too harsh to become economically viable. The processes with recycle oil resulted mainly in a product oil that was too viscous, and the use of water and/or reducing gas led to very high operating pressures (>150 bar). Therefore, processes have been developed up to pilot scale but have not reached commercialization. Hence, from a process and product quality point of view, further progress is still required. Recent efforts7 and this study have tried to gain more insight into the desired solvent properties and operating conditions for biomass liquefaction in which high oil yields are obtained with a moderate viscosity. In this article, we report our latest efforts to improve the economics of biomass liquefaction.7b Our approach was based on the idea that the process should:

• operate without a catalyst to avoid contamination of the resulting bio-oil, which inevitably leads to an elevated operating temperature,

• operate at a mild pressure by avoiding reactive gases such as CO or H2 and by using a high-boiling solvent,

• use an inexpensive solvent that does not require separation from the bio-oil, for example, by using a fraction of the bio-oil itself as solvent or by using a cheap refinery stream. Although inexpensive, water is disqualified by its high vapor pressure at elevated temperature.

Hence, our research focused on a limited number of solvents that show fairly high boiling points (≈200 °C) and are model components for bio-oil fractions or refinery streams, guaiacol, hexanoic acid, and n-undecane. Guaiacol represents lignin degradation products, hexanoic acid represents carboxylic acids that are formed through cellulose and hemicellulose decomposition, and n-undecane represents a refinery stream. As liquefaction appeared to proceed best with bio-oil model components, refill experiments were performed to simulate the recycling of bio-oil as the liquefaction medium. While this research was underway, another group appeared to follow a similar approach to biomass liquefaction.7a

Results and Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results and Discussion
  5. Conclusions
  6. Experimental Section
  7. Acknowledgements
  8. Supporting Information

Solvent screening

During liquefaction, biomass is depolymerized and broken down into smaller segments. The type of solvent used is of paramount importance for the obtainable yields of solid residue, oil, and gas. The solvent can (partly) dissolve original biomass polymers and its initial fragments, stabilize and dilute the products formed, and also act as a reactant. The effect of solvent selection is illustrated clearly in Figure 1 a, in which pine wood liquefaction results are shown as carbon distribution to solid, oil, and gas for the three different solvents, namely, guaiacol, hexanoic acid, and n-undecane. Both dry and wet (10 wt % on total feed basis) wood were used. In particular, wet biomass is of special interest as drying biomass is expensive and preferably minimized for each application. Additionally, a run with water and a run without any liquid (solvent free/pyrolysis) have been performed for comparison. The solid does not only have to be a product in the form of char/coke but can also be unconverted wood. The solvent type appears to have a very big impact on the carbon distribution of the products. Aromatic and/or oxygenated solvents generally perform better than an aliphatic hydrocarbon. According to carbon converted to oil, the solvent performance can be ranked as follows: guaiacol≈water>hexanoic acid≫n-undecane≫solvent free/pyrolysis (Figure 1 a). A quick look at common solvent parameters (Table S4) does not reveal a clear correlation between the solvent and its effectiveness in liquefaction. Clearly, a more extensive set of solvents would be needed to unravel such correlation but this was not the purpose of our study. Wet wood (50 % moisture, reported as 10 wt % wood, 10 w % water, and 80 wt % solvent) leads to a higher oil yield for every solvent used (Figure 1 a and b). As a downside, the operating pressures were significantly increased (from 32–56 bar without water to 60–121 bar with water). The addition of water was further investigated with guaiacol, and a minimum char yield was observed for intermediate water/guaiacol ratios between 1:4 and 4:1 (Figure 1 b). This shows that a mixture of solvents can lead to an added beneficial effect.

thumbnail image

Figure 1. a) Products distribution obtained after liquefaction of wood (10 wt %) in guaiacol, hexanoic acid, and n-undecane with and without the addition of 10 wt % of water. The product distribution of wood pyrolysis (solvent free) and wet wood are shown. b) Solid yields plotted versus water content using four different solvents. The temperature was approximately ≈300 °C with a reaction time of 30 min. Details of the experiments can be found in Table S2.

Download figure to PowerPoint

The conversions and effectiveness of the solvents observed here are generally in agreement with the data reported for similar solvents. Phenol, which is chemically similar to guaiacol, shows a very good performance for biomass liquefaction.8 Hydrocarbon solvents, such as tetralin,9 toluene,10 and “Shellsol”11 give a somewhat lower oil yield. If water is used as a co-solvent, almost no solids and little gas are obtained for guaiacol (Figure 1) and phenol,8b which shows that the addition of water as a co-solvent is generally favorable. Water as a pure solvent (e.g., the hydrothermal liquefaction of wet biomass) generally results in lower oil yields (up to 58 wt %).3b, 12

Product quality

The bio-oil produced from the liquefaction of biomass still needs further processing if transportation fuels are desired. Co-processing with fossil crude oil (fractions) seems an attractive way to upgrade it as use can be made of an already installed infrastructure and known technology. Promising results in this field have been obtained for a different route in which, on a lab scale, biomass is added to a fluid catalytic cracking (FCC) unit by pyrolysis and hydrotreatment.13 For this, however, there should be compatibility between the biocrude and fossil stream. As an initial product quality evaluation, the molecular weight of biocrudes produced in guaiacol are compared with typical refinery streams (Figure 2). Wood liquefaction oil shows a wide weight distribution, whereas the refinery streams are, as expected, cut off molecular weight distributions. If dry wood is used, the molecular weight shows a heavy fraction (>1 kDa) comparable with vacuum residue streams. If water is added, the heaviest components (>5 kDa) are typically consumed and/or not produced. However, the oil is still significantly heavier than vacuum gas oil and would need a cracking/depolymerization (and likely hydrotreatment) step for high-end fuel production. The amount of bio-oil that is heavier than guaiacol (>150 Da) was determined by gel permeation chromatography (GPC) and accounted for approximately 55 wt % of the wood intake, which confirms that most of the wood is converted into relatively heavy oil.

thumbnail image

Figure 2. GPC chromatogram of wood liquefaction that used guaiacol as a solvent with and without the addition of water plotted together with data of typical refinery streams. Refractive index (arbitrary units) versus polystyrene-calibrated molar mass.

Download figure to PowerPoint

Gas production is overall low. The major gaseous products are CO and CO2 (≈2 and 1 C %, respectively; Table S1) with marginal amounts of CH4 and higher hydrocarbons. H2 production is marginal compared to that of COx (≈2 mol % of the COx).

Process parameter screening

Process parameter screening was performed for both guaiacol and hexanoic acid as wood liquefaction solvents. As both solvents gave similar trends, we will focus the discussion on the study of guaiacol, which is the most effective liquefaction solvent. The product carbon yield distribution for the process variables, namely, a) wood loading, b) temperature, c) reaction time, and d) water content, are shown in (Figure 3).

thumbnail image

Figure 3. Carbon product yield distribution versus a) wood loading, b) temperature c) reaction time, and d) water content. Standard conditions for an experiment were: 10 wt % wood, ≈30 min, ≈300 °C, 0 wt % water unless specified otherwise by the x axis. The lines are illustrative except for solid yields versus temperature and reaction time, which are fitted to a first-order reaction.

Download figure to PowerPoint

With increasing wood loading, the oil yield increases at the expense of the solid yield. This suggests that a higher concentration of liquefaction intermediates/products stimulate further liquefaction of biomass polymers. This effect was not observed with wood loading variation by using hexanoic acid as a solvent. However, the mixture obtained at a high wood loading is very viscous and much higher loadings would only wet the wood instead of introducing free liquid. Temperature has the strongest effect on liquefaction; as the temperature increases, more solid is converted to oil and permanent gas. The oil yields show a large increase, and the gas yields increase linearly with temperature. The solid residue still has a fibrous appearance similar to the fed wood, which indicates that the solid seems to originate directly from the wood rather than from oil degradation. Longer reaction times result in higher oil yields and lower char yields. A small increase of gas yield is observed, which also results in a higher pressure (32 bar for 30 min compared to 45 bar for 120 min; Table S1). As already seen with solvent screening, the addition of water has a beneficial effect on the amount of oil produced. Water is normally present in biomass and expensive to remove by drying. Even very high amounts of water seem to be beneficial at the cost of increasing the operating pressure. Although limited, the data allowed a preliminary kinetic analysis of the liquefaction reaction. A first-order reaction was fitted to the temperature and time profiles reported in Figure 3 b and c [Eq. (1)]:(2)

  • equation image(1)
  • equation image(2)

and in which X=(1−solid yield/100) is the conversion of the solid (the solid yield [% carbon] is taken from Figure 3 c and d), k [s−1] the first-order rate constant, t [s] the reaction time, k0 [s−1] a pre-exponential factor, Ea [kJ mol−1] the activation energy, R [kJ mol−1 K−1]=0.008314 the ideal gas constant, and T [K] the temperature. At the same time, the conversion (X) was defined as the decrease in solid residue, which assumes that the solid mainly consists of the unconverted feed rather than of solid product (char). For the temperature regime examined, this assumption holds quite well; however, at higher temperatures a significant amount of char formation is expected. A very good fit was obtained for temperature and reaction time variation (Figure 3 b and c) with an activation energy (Ea) of 101 kJ mol−1 and a pre-exponential factor (k0) of 1.18×106 s−1.

Process concept

A possible conceptual process for biomass liquefaction is illustrated in Figure 4. Here biomass is converted in recycled bio-oil, after some tailoring. Light products (gas, water, and light organic compounds) and heavy products (solid and very heavy bio-oil) are removed, and the large amount of middle-range liquid is used as recycle and product bio-oil. This concept is interesting because it is very simple and could be used for primary conversion at the biomass product site. The product bio-oil could then be transported for the final upgrade, for example, by co-refining with fossil streams. In this way, the volumetric energy density is increased significantly and an initial purification step is applied.

thumbnail image

Figure 4. Conceptual process for the liquefaction of biomass for the production of bio-oil.

Download figure to PowerPoint

In an attempt to validate such a concept, a few refill experiments were performed with guaiacol as the start-up solvent and wet wood. As proposed, the light and heavy products were removed between every refill of wood. The light products were removed by releasing the pressure at 200 °C prior to the final cooling, and the heavy products were removed by filtering the bio-oil through a 1 μm filter. With each refill experiment (see also Table S3), the amount of original guaiacol is reduced to reach approximately 38 wt % of the initial intake in the last experiment (run 5). As shown in Figure 5 and Table S3, the overall molecular weight and viscosity of the bio-oil increases with each refill. The overall molecular weight is, after the 5th run, in the same range (heavy end) as vacuum residue and would need to be upgraded significantly for the production of transportation fuels. Interestingly, the effectiveness of the recycled oil remained very high for only low amounts of solids that remained after each experiment (<1 wt % per refill; Table S3). It even surpassed the performance of the start-up solvent guaiacol to reduce the amount of solid residue. A similar effect was observed in the process parameter study in which the oil yield increased with increasing wood loading (Figure 3 a). Elemental analysis (Table S3) shows that significant deoxygenation has taken place after the fifth run as the oxygen content decreased from approximately 47 wt % in wood to approximately 32 wt % in the bio-oil. Such an oxygen decrease is consistent with the liberation of water in the lights and the CO and CO2 found in the gas products. It should be specified that the elemental composition of the oil was calculated from that of the liquid product after the contribution of water and guaiacol, which were present in the liquid product, were excluded.

thumbnail image

Figure 5. Molecular weight distribution measured by GPC of five refill experiments. The insert shows the solvent guaiacol peak. The refractive index (RI) signal is used if the y axis is plotted as arbitrary units.

Download figure to PowerPoint

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results and Discussion
  5. Conclusions
  6. Experimental Section
  7. Acknowledgements
  8. Supporting Information

This paper reports our search for a cheap biomass liquefaction process that operates at a mild pressure, a mild temperature, without catalysts, and without the need for expensive solvent/bio-oil separation. The following can be concluded:

• Polar solvents are more effective than apolar ones as the solvent effectiveness decreases in the order of guaiacol≈water> hexanoic acid≫n-undecane≫none. Polar solvents allowed deep liquefaction (>80 %) at approximately 320 °C without any catalyst.

• Wet wood leads to a higher oil yield and lighter bio-oil than dry wood for all solvents tested. Hence, there is no need to dry the biomass too well before processing. However, the addition of water increases the operating pressure significantly.

• For guaiacol, the rate of wood liquefaction appears to proceed by a first-order reaction in wood with an apparent activation energy of 101 kJ mol−1 and a pre-exponential factor of 1.18×106 S−1.

• Attempts to recycle the bio-oil through refill experiments showed high oil yields with a reduced oxygen content and low solid yields. However, successive refills led to a rapid build-up of heavy products and an increase in the viscosity of the liquid. Hence, further reduction or withdrawal of the heavy product prior to recycling seems to be required.

Based on these findings, we propose a simple, small-scale, and decentralized liquefaction process that uses the product bio-oil as the liquefaction medium after the removal of the light and heavy products. However, further improvements are still required, particularly to reduce the formation of heavy products or to remove them from the recycled bio-oil by physical separation or chemical conversion. These topics will be the subject of future publications.

Experimental Section

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results and Discussion
  5. Conclusions
  6. Experimental Section
  7. Acknowledgements
  8. Supporting Information

Chemicals

Pine wood was purchased from Rettenmaier & Söhne GmbH (Germany), smashed to a particle size below 0.5 mm and then dried at 105 °C for 24 h. The composition of the wood is listed in Table 1. The wet wood was prepared by mechanically mixing dry wood and deionized water and kept for 24 h before use. Three ways of adding wood, water, and solvent to the system were tested, which all resulted in a similar liquefaction performance, namely, i) water and wood were mixed and then the solvent was added, ii) wood and solvent were mixed after which water was added, and iii) water and solvent were mixed after which wood was added. Guaiacol (Sigma–Aldrich, >98 %), hexanoic acid (Sigma–Aldrich, 99 %), and n-undecane (Sigma–Aldrich, >99 %) were used without further treatment. Some characteristics of these solvents are given in Table S4.

Table 1. Composition of pine wood.[14][a]
BiomassComposition [wt %] (dry)MetalsComposition [mg kg−1] (dry)ElementsComposition [wt %] (daf)[b]
  1. [a] Average from the literature and our own elemental analysis. [b] Dry and ash free.

cellulose35K34C46.6
hemicellulose29Mg134H6.2
lignin28Ca768O (by difference)47.2
  total ash2600N0.04

Reaction procedure

An autoclave (≈45 mL, Inconel 825) with a mechanical stirrer was employed as the reactor, and all the experiments were conducted in batch mode. High heating and cooling rates were obtained by using a hot sand bath and a water bath in combination with a thin reactor wall (4 mm). The set-up was placed in a bunker and could be operated remotely to ensure safety. The temperature profile of the reactor is shown in Figure 6 a. In a typical run, wood (3 g) and guaiacol/hexanoic acid/n-undecane (27 g) were introduced into the reactor, and the slurry was stirred (≈1000 rpm) to mix it. After flushing twice with N2 and an air-tightness test, the reactor was pressurized with N2 to approximately 5.5 bar, put into a sand fluidized oven (e.g., 300 °C), and kept there for 30–120 min. After the reaction, the reactor was quenched in a water bath (≈20 °C) to cool. The reaction temperature was defined as the end temperature of an experiment, and the reaction time was defined as the stay period of reactor inside the hot sand bath. Detailed data of the experiments are reported in Table S2.

thumbnail image

Figure 6. a) Temperature profile of the reactor and b) overall procedure for product collection.

Download figure to PowerPoint

Analytical methods and product definition

A gas sample was obtained by using a syringe and quantification was performed using the end pressure at RT with the ideal gas law PV=nRT, in which V represents the gas free space of the reactor. The slurry was filtered through a glass filter paper with pore size of 1.6 μm to separate the liquid from the solid, and the solid was subsequently dried at 105 °C for 24 h. The overall procedure for product collection is shown in Figure 6 b. The oil itself was not further separated from the liquefaction solvent and acetone. Gas samples were analyzed by using a Varian Micro GC CP-4900 with two analytical columns: Molsieve 5A (10 m) C and PPQ (10 m). He was used as the carrier gas. The total amount of the gas sample could be calculated from the pressure and volume of the collection vessel. The elemental composition of the solids was determined by using a Fisons Instruments 1108 CHNS-O apparatus. GPC (Agilent Technologies 1200 series) of the oil+solvent samples was performed by using three PLgel3 μm MIXED-E columns placed in series to determine the molecular weight distribution of the liquid samples. Other operating conditions were 1 mL min−1 THF eluent, refractive index detector, 40 °C, standard 162 to 30 230 Da calibrated with polystyrene. GC–MS (Agilent Technologies, GC 7890A/MS 5975C) analysis was performed by using a Agilent HP-%MS HP19091S-443 column by dissolving the sample in acetone and filtering it through a 0.2 μm filter. The operating conditions were temperature program 45 °C (4 min), increased at 3 °C min−1 to 280 °C (20 min), carrier gas He, injector temperature 250 °C, and a split ratio of 20:1. The water concentration in the oil was measured by Karl Fischer titration by using a Metrohm 787 KF Titrino. The titrant used was Hydranal Composite 5.

Mass balances of approximately 95 wt % on the total intake (inclusive solvent) were obtained. Data are reported on carbon yield basis, by using the carbon content of the wood intake, the gas, and the solid as shown below. Accordingly the missing carbon is attributed to the liquid product [Eqs. (2)–(4)]:(3), (4)

  • equation image(2)
  • equation image(3)
  • equation image(4)

Refill experiments

Refill experiments were performed in which guaiacol was used as the start-up solvent. As a result of the nature of the refill experiments, no mass and carbon balance were measured. However, the oils were analyzed by GPC and the guaiacol content was determined by GC–MS. The autoclave was loaded with wet pine wood and heated. After the reaction, instead of cooling to RT, the autoclave was cooled to 200 °C and opened to vent off the gas and remove compounds with a low boiling point, which were condensed and quantified (lights). The autoclave was then cooled to RT and the product oil in solution was obtained by filtration (Figure 6 b) and removing the used acetone for rinsing. For the refill experiments, this oil was then used as the solvent medium. A total of five experiments (four refills) were performed. The data of the refill experiments are shown in Table S3.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results and Discussion
  5. Conclusions
  6. Experimental Section
  7. Acknowledgements
  8. Supporting Information

The authors would like to thank Shell Global Solutions International B.V. for funding this research and Benno Knaken for technical support.

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results and Discussion
  5. Conclusions
  6. Experimental Section
  7. Acknowledgements
  8. Supporting Information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
cssc_201300297_sm_miscellaneous_information.pdf253Kmiscellaneous_information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.