M. T. Holtzapple, M. K. Ross, N.-S. Chang, V. S. Chang, S. K. Adelson,

and C. Brazel

Department of Chemical Engineering, Texas A&M University,
College Station, TX 77843-3122

The MixAlco Process converts biomass into mixed alcohol fuels. The biomass is first treated with lime to render it more digestible. Then, it is fed to a mixed culture of acid-forming microorganisms that produce salts of volatile fatty acids, such as calcium acetate, propionate, and butyrate. These salts are separated from the fermentation broth and thermally converted to ketones that are subsequently hydrogenated to alcohols, such as 2-propanol, 2-butanol, 2-pentanol, and 3-pentanol. Design data are presented related to the lime treatment, fermentation, thermal conversion, and hydrogenation. A preliminary economic evaluation indicates that alcohol fuels can be produced from negative-value biomass, such as municipal solid waste and sewage sludge, for about $0.19/L ($0.72/gal).

Figure 1 shows a schematic of the MixAlco Process which converts biomass (e. g. municipal solid waste, sewage sludge, agricultural residues, energy crops) into mixed alcohol fuels. To enhance digestibility, the biomass is treated with lime. Then, using a mixed population of acid-forming microorganisms such as those found in cattle rumen, the lime-treated biomass is converted to volatile fatty acids (VFA’s) such as acetic, propionic, and butyric acids. To prevent the pH from decreasing as the acids are formed, a neutralizing agent is added to the fermentor; thus VFA salts — such as calcium acetate, propionate, and butyrate — exit the fermentor. These salts are concentrated to dryness and then are thermally converted to mixed ketones (e. g., 2-propanone, 2-butanone, 2-pentanone, 3-pentanone) that are subsequently hydrogenated to mixed alcohols (e. g., 2-propanol, 2-butanol, 2-pentanol, 3-pentanol).

In the fermentor, both lime and calcium carbonate are possible neutralizing agents. Lime is a much stronger alkali and therefore can attain the optimal rumen pH of 6.7 whereas calcium carbonate can achieve a pH of only 5.8 to 6.2. Although lower pH slows the fermentation rate, it also discourages methanogens and thus can

© 1997 American Chemical Society

increase the selectivity toward VFA’s. This advantage, plus lower cost, favors calcium carbonate as the neutralization agent.

The fermentor temperature can be adjusted for mesophilic (~40°C) or thermophilic (~55°C) microorganisms. Of the total VFA’s, mesophilic microorganisms produce 60-70% acetic acid whereas thermophilic microorganisms produce 80-90%; thus, temperature is an important parameter that affects the product distribution. Higher fermentor temperatures have an advantage of reducing the cooling water required to remove metabolic heat; however, for fuel production, the greater energy content in the higher VFA’s favors lower fermentor temperatures.

A common approach to producing alcohol fuels from biomass is to enzymatically convert treated biomass with extracellular enzymes that hydrolyze polysaccharides to soluble sugars that are fermented to ethanol which is recovered by distillation. Compared to this approach, the MixAlco Process offers the following advantages:

• It is adaptable to a wide variety of feedstocks.

• Aseptic process conditions are not required.

• Inexpensive tanks can be employed.

• Expensive extracellular enzymes are not required.

• The fermenting organisms regenerate themselves.

• Cells and enzymes can be recycled without contamination risk.

• The fermenting organisms are stable.

• The process is robust.

Elements of the MixAlco Process have been investigated for many years. In 1914, Lefranc received a patent (7) on a process to convert waste biomass into butyric acid which was neutralized with calcium carbonate. The calcium butyrate was thermally converted to ketones for high-octane "Ketol" motor fuel (2). More recently, Playne (3) has further developed this technology by incorporating various pretreatments (4) and membrane separation techniques (5). In the U. S., only ethers
and alcohols may be added to fuel, so the MixAlco Process converts the ketones into alcohols.

The future growth opportunities for lactic acid are in its use as a feedstock for potentially large-volume applications. In Table I, these applications are classified into four categories — biodegradable polymers, oxygenated chemicals, "Green” chemicals/solvents, and plant-growth regulators. The overall size of this opportunity, both in terms of mass/volume and product sales value, is substantial. For the U. S. markets, this could be approximately 6.4-8.4 x 109 lb/yr (2.9-4.0 x 106

9

tons/yr), with sales volume between approximately $4.3-6.2 x 10 /yr. The volume and selling price projections for the new products (i. e., the degradable plastics, "green" chemicals, and derivatives) are made on the basis of several published studies by Battelle and others and some internal Argonne estimates. It should be noted that the high volumes can be reached only when the prices are within the acceptable ranges (Table I) and vice versa. The list in Table I is by no means comprehensive nor would all these products (particularly the oxychemicals) be derived from lactic acid in the near future. It should be noted, however, that recently a large U. S. agriprocessing company, Cargill, has announced a potential plaht of 250 x 106 lb/yr by 1997/1998 (5), substantiating that large-volume, economical manufacturing of lactic acid may be feasible with new technologies and for new or existing products.

Polymers of lactic acids are biodegradable thermoplastics. A fairly wide range of properties are obtainable by copolymerization with other functional monomers, such as glycolide, caprolactone, and polyether polyols. The polymers are transparent, which is important for packaging applications. They offer a good shelf life because they degrade slowly by hydrolysis, which can be controlled by adjusting the composition and molecular weight. These have potential uses in a wide variety of consumer products, such as paper coatings, films, moulded articles, foamed articles, and fibers. Some of the published information on some of the properties of lactic copolymers that approach those of large-volume, petroleum-derived polymers (such as polystyrene, flexible polyvinyl chloride [PVC], and vinylidene chloride) are summarized in the article of Lipinsky and Sinclair (7). There are numerous patents and articles on lactic acid polymers and copolymers, their properties, potential uses, and processes that date back to the early work by Carothers at DuPont. A discussion of this work is beyond the scope of this article. Several reference articles and patents (12-15) can provide the reader with a basis for further information.

Among the other new product opportunities, the use of lactate esters as "green” solvents is substantial because they are high-boiling (nonvolatile), nontoxic, and degradable compounds. With increasing consumer and political consciousness with environmentally sound products, the use of these solvents as replacements for other solvents or cleaners could be a very important expansion opportunity for lactic acid. The estimates of the volume of these (Table I) are based on typical volumes and lower prices than several intermediate volume non-volatile solvents, such as n-methyl pyrrolidone, di-basic esters and such. Low-molecular-weight polymers of L-lactic acid (degree of polymerization [dp] 2-10) have been recently discovered to stimulate plant growth in a variety of crops and fruits when applied at a low level (16-17). These findings may lead to specialized products and formulations that would incorporate L-polylactic acid as or into controlled release or degradable mulch films for large-scale agricultural applications.

The cost of the enzymes for enzymatic hydrolysis of cellulosic biomass is clearly the critical parameter from an economic point of view. Most of the industrial enzymes are produced by organisms isolated from natural sources by a labor intensive, unpredictable classical screening, strain selection, medium optimization for over production, fermentation and recovery process development. Screening of naturally occurring

microorganisms still may be the best way to obtain new strains and/or enzymes for commercial applications (60). Fundamental tasks and strategies for commercial development of an enzyme from natural sources are shown in Figure 2. Recombinant DNA technology and protein engineering have also proven as effective means of production of industrial enzymes (61). The marketing of all enzymes is subject to a

Screening for microorganisms

I

Culture selection

1

Fermentation studies (preliminary)

I

Isolation, purification and final characterization

I

Evaluation

II

Toxicology

1

Regulatory agency

Improvement of fermentation and recovery process
development

I

Product formulation

II

Marketing

Figure 2. Strategies for commercial development of an enzyme

variety of Federal laws and regulations. The generally recognized as safe (GRAS) status of an industrial enzyme depends on the source of its origin. Federal laws, regulations and policies that have an impact on industrial enzymes have been reviewed by Fordham and Block (62).

Gerhard Knothe, Robert O. Dunn, and Marvin O. Bagby

Oil Chemical Research, National Center for Agricultural Utilization
Research, Agricultural Research Service, U. S. Department of Agriculture,
1815 North University Street, Peoria, IL 61604

Vegetable oils and their derivatives (especially methyl esters), commonly referred to as “biodiesel,” are prominent candidates as alternative diesel fuels. They have advanced from being purely experimental fuels to initial stages of commercialization. They are technically competitive with or offer technical advantages compared to conventional diesel fuel. Besides being a renewable and domestic resource, biodiesel reduces most emissions while engine performance and fuel economy are nearly identical compared to conventional fuels. Several problems, however, remain, which include economics, combustion, some emissions, lube oil contamination, and low — temperature properties. An overview on all aspects of biodiesel is presented.

The use of vegetable oils in diesel engines is nearly as old as the diesel engine itself. The inventor of the diesel engine, Rudolf Diesel, reportedly used groundnut (peanut) oil as a fuel for demonstration purposes in 1900 (7). Some other work was carried out on the use of vegetable oils in diesel engines in the 1930’s and 1940’s. The fuel and energy crises of the late 1970’s and early 1980’s as well as accompanying concerns about the depletion of the world’s non-renewable resources provided the incentives to seek alternatives to conventional, petroleum-based fuels. In this context, vegetable oils as fuel for diesel engines were remembered. They now occupy a prominent position in the development of alternative fuels. Hundreds of scientific articles and various other reports from around the world dealing with vegetable oil-based alternative diesel fuels ("biodiesel") have appeared in print. They have advanced from being purely experimental fuels to initial stages of commercialization. Nevertheless, various technical and economic aspects require further improvement of these fuels.

Numerous different vegetable oils have been tested as biodiesel. Often the

This chapter is not subject to U. S. copyright. Published 1997 American Chemical Society

vegetable oils investigated for their suitability as biodiesel are those which occur abundantly in the country of testing. Therefore, soybean oil is of primary interest as biodiesel source in the United States while many European countries are concerned with rapeseed oil, and countries with tropical climate prefer to utilize coconut oil or palm oil. Other vegetable oils, including sunflower, safflower, etc., have also been investigated. Furthermore, other sources of biodiesel studied include animal fats and used or waste cooking oils. Sources of biodiesel with some emphasis on developing countries have been discussed (2).

Several problems, however, have impaired the widespread use of biodiesel. They are related to the economics and properties of biodiesel. For example, neat vegetable oils reported to cause engine deposits. Attempting to solve these problems by using methyl esters causes operational problems at low temperatures. Furthermore, problems related to combustion and emissions remain to be solved. The problems associated with the use of biodiesel are thus very complex and no satisfactory solution has yet been achieved despite the efforts of many researchers around the world. This article will briefly discuss economics and regulatory issues as well as conventional diesel fuel (petrodiesel) and then focus on research on the use of biodiesel in a diesel engine.

Recent theoretical and experimental work has advanced our understanding of the meta­bolic pathways and elucidated the intrinsic metabolic potential and kinetic limitations of microbial conversion of glycerol to 1,3-propanediol by the major strains isolated. The maximum product yield conceivable is about 0.72 mol/mol in fermentation with glyce­rol as a sole carbon and 1.0 mol/mol in fermentation with cosubstrates. However, these high yields are normally not achieved in fermentations with high final product concen­tration which is kinetically constrained to about 65 -70 g/1. Further work is needed to optimize the distribution of reducing power released from the metabolism of glycerol and/or cosubstrate and to reduce the formation of toxic by-products for simultaneously achieving a high PD yield and concentration. This may be best achieved by re-de­signing the metabolic pathway and properly controling the cultivation conditions. Pro­gress has already been made in these directions. The use of recombinant techniques also opens new ways for producing PD from other cheap carbon sources. It will become mo­re evident that the microbial production of 1,3-propanediol is technically more flexible and economically more competitive than the chemical route.

Morris Ag-Energy operates a dry-milling ethanol plant at Morris, Minnesota. At the time of these experiments the capacity of this plant had been increased from 4.5 million to about 6 million gpa, primarily through debottlenecking and improved operation. The plant employs conventional batch fermentation and primary distillation systems and a molecular-sieve dehydration system. The fermentors are relatively small, of shallow-tank design, about 14,500 gallons (58000 1) capacity with a working depth of about 8 ft (2.5 m). They are fitted with top-drive slow — speed agitators and internal cooling coils. One fermentor was used for all the in­plant runs; it was modified for pH monitoring and sparging with air or other gasses. A bilobe rotary compressor (Roots blower) of about 60 cfm free air capacity was used for vacuum or headspace gas recirculation.

In this operation, com is ground in a hammermill, then mixed with water (48 gpm, 182 1/min) and recycled thin stillage (39 gpm, 148 1/min) to make a mash of 20 0 Brix. The starch is gelatinized and digested at 90 °С with 120 ml/min of commercial bacterial alpha amylase (IBIS) in a series of stirred tanks; the pH is controlled at 6.5 with ammonia. Further saccharification with glucoamylase (Alltech, 1 volume per 2900 volumes mash) occurs in the fermentor after cooling to 32 °С and pH adjustment. Yeast is produced continuously in a semiaerobic yeast propagator fed the same mash; all of the glucoamylase is added through the yeast propagator so that the yeast see a high initial glucose concentration. The yeast propagator operates at pH 3.5 (adjusted with sulfuric acid) with a cell count typically 0.5×10? to 1.5×10? ml"* and viability 75%. The metabolism of the yeast in the propagator is primarily fermentative but they retain the ability to quickly consume added oxygen. For these experiments the yeast suspension from the propagator was 1/8 the total fermentor charge, resulting in an initial pH of 5.6 which declines during the course of the fermentation to a limiting value of about 3.8.

After fermentation the beer is pumped to a series of beer wells where fermentation is completed and then to the distilling column. Although the normal residence time in the fermentor is 40 h, some of the experimental runs were kept at least 48 h to monitor the completion of the fermentation.

Methods

Analytical. Cell viability was determined using methylene blue (21) , a microscope, and a hemocytometer; total and viable cell counts were determined from the same data. Glucose and glycerol were determined enzymatically using prepared commercial reagents (Sigma 315-100 and 337-40A). Total glucose was determined after acid hydrolysis (0.25M H2SO4, 30 min, 100 °С); for most runs this analysis was employed only as a check of the total conversion. FAN (free alpha-amino nitrogen) was determined by the EBC ninhydrin method (22).

Ammonia was determined by a modified Berthelot reaction (23). Ethanol was determined by gas chromatography using a Hayesep R column (Alltech Instruments). Ethanol values from the industrial runs were considered as relative values only and no interpretation was made of the absolute levels, due to possible handling losses. Carbon dioxide was determined by a modification of the Martin manometric method (24) using a commercial differential pressure sensor (Omega PX26-005DV) instead of a manometer. The reference side of the pressure sensor was connected to the vacuum pump via a 2 1 flask which acted as ballast. Linear calibration curves were obtained through at least 60 mM CO2 Concentration was related to partial pressure by the general approach of Schumpe (25); the effect of ethanol was specifically included based on literature data at low concentration (26). The effect of ethanol on CO2 solubility was judged to be sufficiently linear and reproducible to permit use of this approach over the limited range of ethanol concentrations encountered in these experiments. Fermentor pH was determined in situ with commercial instruments (Omega) and electrodes (Phoenix).

Laboratory Fermentations. These fed-batch runs employed a Biolafitte Fermentor at the BioProcess Institute, University of Minnesota, St. Paul. The medium is listed in Table I; this was based originally on the medium of Oura (27) with NH4CI and yeast extract added to simulate the ammonia and FAN levels prevailing in the industrial fermentation, but it was necessary to increase the yeast extract to even approach the rates and cell counts prevailing in the industrial fermentation. Concentrations were figured on the basis of a 16 1 final volume. The initial glucose concentration was 60 g/1; additional glucose was added to the fermentor as a concentrated solution during the run. The inoculum was Alltech alcohol-production yeast grown in 1 1 of the same medium in a 2 1 unsealed flask shaken at 200 rpm. This is the same yeast source employed by the plant to inoculate their yeast propagator. The fermentor was agitated vigorously and continuously sparged with a mixture of nitrogen and carbon dioxide. Gas mixtures were prepared by continuous metering through rotameters (Cole Parmer), using the manufacturer’s calibration graphs. A back pressure regulator was used to maintain constant pressure and thus constant CO2 partial pressure, calculated from the back pressure and the CO2 content of the sparge gas. Volumetric productivity of ethanol and glycerol was calculated as the increase in concentration since the previous point, divided by the intervening time. Productivity per cell was calculated by dividing the volumetric productivity by the geometric mean of the cell counts at the beginning and end of the interval. Exponential growth and death rates were calculated from those parts of the growth curve which were linear on semi-logarithmic plots. In the air — supplemented laboratory run, oxygen uptake was monitored by continuous mass spectrometry of the inlet and outlet gas streams. An absolute calibration was not performed but the data serve to estimate the relative level and percentage consumption of supplied oxygen. Total oxygen consumption was calculated from these data and the known flow rates.

Soybean oil pyrolyzed distillate, which consisted mainly of alkanes, alkenes, and carboxylic acids had a CN of 43, exceeding that of soybean oil (37.9) and the ASTM minimum value of 40 (154). The viscosity of the distillate was 10.2 cSt at 38°C, which is higher than the ASTM specification for DF2 (1.9-4.1 cSt) but considerably below that of soybean oil (32.6 cSt). Short-term engine tests were carried out on this fuel (155).

Used cottonseed oil from the frying process was decomposed with Na2C03 as catalyst at 450° to give a pyrolyzate containing mainly C8.20 alkanes (70%) besides alkenes and aromatics (156). The pyrolyzate had lower viscosity, flash point, and PP than DF and equivalent calorific values. The CN of the pyrolyzate was lower.

Rapeseed oil methyl esters were pyrolyzed at 550 to 850°С and in nitrogen dilution (157). The major products were linear 1-alkenes, straight-chain alkenes, and unsaturated methyl esters. CO, C02, and H2 were contained in the gas fraction. The C1(M4 alkenes and short-chain unsaturated esters were optimally produced at 700°.

Catalytic conversion of vegetable oils using a medium severity refinery hydroprocess yielded a product in the diesel boiling range with a CN of 75-100 (158). The main liquid product was a straight-chain alkane. Other products of the process included propane, water, and C02.

Soybean, babassu and some less common vegetable oils were hydrocracked with a NiMo/Y-Al203 catalyst sulfided in situ with elemental sulfur under hydrogen pressure (159). Various alkanes, alkylcycloalkanes, and alkylbenzenes were observed. Oxygen in the oil feed was liberated as C02, H 20, and CO. Decarboxylation was indicated by water and C02. C M formation indicated acrolein decomposition. Differences between more saturated and unsaturated oils were observed. Besides NiM0/y-Al203, an NiSi02 catalyst was studied {160) in the hydrocracking of vegetable oils at 10-200 bars hydrogen pressure and 623-673 K. The resulting product was a mixture of hydrocarbons, mainly alkanes, in the diesel fraction. Hydrogenolysis of palm oil over Ni/Si02 or over Co at 300° and 50 bar gave a nearly colorless oil, mainly C15.17 alkane {161). The same process gave soft solid with 80.4% C,7 alkanes when applied to rapeseed oil. An octadecane model compound gave 50% conversion over Co/oil catalyst to C17 alkane as the main product.

Catalytic hydrocracking (Rh-Al203 catalyst) of soybean oil at 693 К and 40 bar hydrogen pressure gave liquid products which were distilled to gasoline and gas oil boiling-range hydrocarbons {162). Decarboxylation / decarbonylation was again noted.

Crude and partially hydrogenated soybean oil were decomposed by passage over solid acidic A1203 or basic MgO {163). The degree of unsaturation of the oil influenced product formation. Partially hydrogenated soybean oil yielded more hydrocarbons while crude soybean oil yielded a mixture of oxygenated products and hydrocarbons of lower mean molecular weight. The products derived from MgO cracking showed more unsaturates and aromatics than those from A1203 decomposition.

Kolbe electrolysis of the potassium salts of coconut fatty acids and acetic acid reportedly gave a liquid with good DF properties {164) and the products resembled those from pyrolytic procedures. This product contained 83% alkanes, mainly even-numbered compounds from Сю.24, with C12_18 being the most abundant.

The yeast Saccharomyces cerevisiae is not able to use xylose or xylitol as a carbon source for growth or fermentation {54). Hallbom et al. (55) obtained efficient conversion of xylose to xylitol by transforming S. cerevisiae with the gene encoding the xylose reductase (XR) gene {XYL1) of Pichia stipitis. Due to lack of xylitol dehydrogenase (XDH), the recombinant S. cerevisiae needs a co-carbon substrate to regenerate the cofactors and to gain maintenance energy. Hallbom et al. (56) studied the influence of cosubstrate and aeration on xylitol formation by the recombinant S. cerevisiae. With glucose and ethanol, the conversion yields were close to 1 g xylitol/ g consumed xylose. Decreased aeration increased the xylitol yield based on consumed cosubstrate, while the rate of xylitol formation decreased. Xylitol yields close to 100% could be obtained from a medium with a total xylose concentration corresponding to that of an industrial hemicellulose hydrolyzate by fed-batch cultivation of recombinant XYL1 expressing S. cerevisiae using ethanol as co-substrate (57). Recently, Roca et al. {58) investigated the effect of hydraulic residence time (1.3-11.3 h), substrate/cosubstrate ratio (0.5 and 1), recycling ratio (0.5 and 10), and aeration (anaerobic and oxygen limited conditions) on xylitol production by immobilized recombinant S. cerevisiae in a continuous packed-bed bioreactor.

For more than 10 years, batch dilute acid pretreatment techniques have been extensively evaluated at NREL on biomass feedstocks, including several species of hardwoods, herbaceous crops, and agricultural residues (39-42). The pretreatment objectives have focused largely on a prehydrolysis approach, in which the hemicellulose component of biomass is hydrolyzed via dilute acid catalysis, leaving the cellulose fraction in an insoluble form for subsequent enzymatic hydrolysis by cellulase enzymes. A detailed process economic analysis of a base-case bioethanol production process, utilizing continuous cocurrent dilute acid prehydrolysis technology, has been conducted at NREL (43). Pretreatment conditions and performance parameters used in this evaluation were based on bench-scale batch, dilute acid pretreatment experimental data for hardwood poplar species. Since that time, pilot-scale batch (44) and cocurrent pretreatment reactors have been installed and operated to collect larger-scale reactor data for the pretreatment performance of selected feedstocks. For the most part, pretreatment performance data collected in these larger reactors have been comparable to results obtained in bench-scale batch dilute acid pretreatments (44).

Numerous of sensitivity analyses were performed as a part of this base case process engineering and economic analysis. The most significant parameter in improving the overall economics of bioethanol production was maximizing the yield of ethanol from a unit of biomass. Clearly, improving the yields of each individual unit operation will contribute to increasing the overall yield of biomass. The pretreatment step is very significant in this respect, as an improved pretreatment process would result in higher yields of soluble sugars from the hemicellulose fraction and a more digestible cellulose fraction, resulting in greater conversion of cellulose via enzymatic hydrolysis by cellulase. Because bench — and pilot-scale batch and cocurrent dilute acid pretreatment methods show definite limitations in the yields of soluble sugars from hemicellulose (about 80%-85% of theoretical) and in the yields of ethanol from cellulose in pretreated biomass solids from a simultaneous saccharification and fermentation (SSF) process (about 70%-75% of theoretical), new pretreatment approaches aimed at achieving higher yields are being developed and evaluated.

During the past few years, dilute acid prehydrolysis R&D at NREL has focused on the developing processes that exploit the biphasic kinetics of xylan hydrolysis from the hemicellulose fraction of biomass. In most species being considered for bioethanol conversion, xylan is the overwhelming majority of the hemicellulose component. The kinetics of xylan hydrolysis have been widely reported to be biphasic, with an "easy-to — remove" fraction that can be removed under relatively mild conditions and a "hard-to — remove" fraction that requires more severe conditions (45-47). The hydrolysis reaction mechanism for fast — and slow-reacting xylan fractions assumes conversion first to soluble xylose oligomers, followed by conversion to monomeric xylose and finally, to xylose degradation products. Three prehydrolysis reactor configurations (batch, cocurrent, and countercurrent) have been evaluated in a model of xylan hydrolysis kinetics to determine the best configuration for maximizing xylan hydrolysis yields, while minimizing degradation product formation (48). This work showed that a countercurrent configuration is the best design for achieving high recovery of xylose equivalents, especially at high conversion levels of xylan, which is critical to the economic viability of any bioethanol conversion process. This conclusion is somewhat intuitive, as in a countercurrent configuration, the residence time of the xylose from the "easy-to-remove" xylan fraction is short compared to a batch or cocurrent configuration, resulting in less degradation product formation.

Two-Stage Countercurrent Pretreatment. A large number of pretreatment data have been collected at NREL using hardwood yellow poplar sawdust in a system of percolation reactors operated in such a way as to simulate a two-stage countercurrent dilute acid process. The selection of the reaction conditions and the details of the operation of what is now referred to as the reverse-flow two-stage system have been described in detail (49). A factorial experimental design and subsequent reaction condition optimization was conducted in this study. The conditions selected as a result of this optimization were a 10-min residence time of the solids in each stage, an acid concentration of 0.07% (w/w) sulfuric acid, a first-stage solids temperature of 174°C, and a second-stage solids temperature of 204°C. A recovery of 97.0% of xylose equivalents in the liquor stream from the xylan content of the feedstock was achieved, with about 60% in the form of soluble oligomeric xylose and 40% as monomeric xylose. Less than 3% of the xylan degraded to furfural. About 10% of the glucan was solubilized, but with much lower levels of oligomeric glucose relative to monomeric glucose. About 35% of the lignin was solubilized. The structure of the soluble lignin compounds was not determined. The degree of lignin solubilization is significantly higher than with batch or cocurrent dilute acid pretreatments. It is believed that the reverse-flow mode of operation, in which the liquor is separated from the residual solids while temperatures are still high, prevents re-precipitation and/or re-condensation of solubilized lignin back on the pretreated solids particles.

As the reverse-flow procedure includes a hot water washing of the residual solids, with the wash liquor combined with the actual preydrolysis liquor, the resulting residual solids are free of soluble compounds, including sugars, oligomers, and soluble inhibitory compounds from lignin breakdown or sugar degradation. Thus, the residual solids can proceed to the SSF conversion process without any further washing or other detoxification procedure. Standard SSF conversion studies were performed using the conditions described by Torget and co-workers (49), including a cellulase loading of 25 FPU/g cellulose. For the reverse-flow two-stage optimization run described above, an ethanol yield from cellulose of 91% of theoretical was achieved in 55 h. Both the yield and the rate are substantially higher than those reported for similar feedstocks subjected to batch dilute acid pretreatments. This represents a substantial potential improvement to the overall process economics.

Because of the near quantitative yield of xylose equivalents, the partial solubilization of glucan to glucose equivalents, and the complete displacement of soluble compounds from the residual solids during the washing step, a significant fraction of the initial carbohydrates found in the feedstock end up in the prehydrolyzate liquor. Based on the initial composition of yellow poplar sawdust and the pretreatment performance achieved in the optimization run, more than 40% of the total available carbohydrates are found in this liquor. In the past, the evaluation of pretreatment performance has been based on the enzymatic digestibility and/or SSF production of ethanol from washed, pretreated solids only. It has largely been assumed that the soluble carbohydrates in the liquid fraction of the pretreated slurry or in a separate prehydrolyzate liquor, could be converted to ethanol at assumed yields and rates and would not require any type of post-treatment prior to either a separate or cofermentation.

Hydrolyzate Detoxification. The base-case process economic evaluation described above did not include any prehydrolyzate detoxification step. It has become recognized that some type of detoxification of at least the liquid fraction will likely be required. This is even more likely in flow-through, dilute acid pretreatments compared to batch or cocurrent pretreatments because of the higher lignin solubilization. Certain compounds that result from the solubilization of lignin, including certain organic acids, higher alcohols, and phenolic-based compounds, are known to be inhibitory to fermentative microorganisms. Also, feedstocks that have highly acetylated hemicellulose in their structure are likely to release inhibitory levels of acetic acid upon hemicellulose hydrolysis. Significant efforts are under way to develop cost-effective detoxification processes that can selectively remove or convert such compounds into non-toxic forms. At the same time, modified prehydrolysis conditions that result in similar pretreatment performance levels, but that produce fewer toxic compounds, are being evaluated.

Challenges to Sugar Yield. As stated above, the analysis of the carbohydrates in the optimized prehydrolyzate liquor shows that about 60% of the xylose equivalents in the liquor are in the form of oligomeric xylose. Previous process economic analyses have assumed that solubilized carbohydrates from pretreatment would be fermentable, i. e. monomeric sugars. Options for converting oligomeric xylose to monomeric xylose have been reported (48). Two options were investigated in work performed in the laboratory of Y. Y. Lee at Auburn University (48). The first was a mild-temperature hold of the prehydrolysis liquor at the pH (~2.2) of the liquor as it leaves the pretreatment reactor. A range of hold times and temperatures was examined. The best conditions were a temperature of 130°C and a hold time of 11 h, where the monomeric xylose content increases to 97% of total xylose equivalents and the oligomeric xylose content decreases to virtually zero. The furfural content increased only marginally, indicating little increased sugar degradation under these relatively mild conditions.

The other option evaluated for xylose oligomer hydrolysis was to determine whether sufficient xylanase activity exists in cellulase preparations to allow for enzymatically catalyzed hydrolysis of xylan oligomers to monomers under typical SSF conditions. In tests where Genencor Cytolase CL was added to prehydrolyzate liquor in amounts equivalent to 10 and 25 FPU/g cellulose as if 8% (w/v) cellulose was present, significant conversion to xylose monomer was achieved (Elander, R., unpublished results, 1996). Starting from an initial monomer level of 41% of the total xylose equivalents, the monomeric xylose level increased to 87% of total xylose equivalents with a 10 FPU/g assumed cellulose loading at 61 h, and 95% of total xylose equivalents with a 25 FPU/g assumed cellulose loading at 61 h. The 61-h time is similar to the 55-h SSF time required to achieve a 91% ethanol yield in SSF, as reported above for the reverse-flow, two-stage pretreated solids. The unknown of these experiments is whether these levels of conversion would be maintained when process concentrations of cellulose and lignin are present in the system.

In addition, the work of Elander and co-workers (48) included an analysis of potential process improvements that could result in higher concentrations of carbohydrates feeding into the fermentation steps, which would presumably result in a higher ethanol concentration exiting the fermentation step(s) and lower ethanol recovery costs. Using the volume of liquor generated from pretreatment, which was 2.2 reactor void volumes per biomass-packed reactor (49), and an assumed washed pretreated solids concentration of 40%, process analysis indicated that in a simultaneous saccharification and cofermentation (SSCF) process, in which the washed solids are combined with detoxified liquor and yields of ethanol from cellulose (49) and from xylose are known (43), an ethanol concentration exiting SSF of 3.05% (w/v) could be expected. This is somewhat lower that the base case value of -4.5% (w/v) assumed by Hinman and co-workers (43), and indicates that the volumes of liquor generated could increase ethanol recovery costs. Also, the high liquor volumes result in increased fermentation tankage requirements and would require higher steam demands in pretreatment, which would be exacerbated by the higher temperature requirements that have been used in reverse-flow prehydrolysis. This situation will certainly have an impact on the energy balance of the process and will affect overall process economics.

Reactor Design. The ultimate ability to reduce prehydrolyzate liquor volumes, but still maintain prehydrolysis sugar yields and pretreated solids enzymatic digestibility, becomes a reactor design issue, not only at the bench-scale level but, more importantly, in an engineering-scale system that is representative of a commercial-scale countercurrent prehydrolysis reactor system. The ability to achieve the contacting and movement of liquid and solids effectively in a large-scale device is a major challenge and will likely determine the ultimate commercial success of this technology. Efforts are currently under way to work with equipment suppliers to design, develop, and test engineering — scale countercurrent dilute acid prehydrolysis reactor systems. Although biomass particle sizes, reaction conditions, and the ultimate goals of the reaction are somewhat different, an extensive body of knowledge in reactor design issues related to biomass solubilization exists within the pulp and paper industry. Collaborations with pulping equipment manufacturers are being investigated as a means to more rapidly develop commercial — scale countercurrent prehydrolysis reactor systems.

New Directions at NREL. To date, the only feedstock extensively investigated using the dilute acid countercurrent process is yellow poplar sawdust. This particular feedstock is representative of a near-term waste biomass source from sawmill operations, but also is similar to potential hardwood energy crop species; thus, it provides useful information for near — and long-term applications. Ultimately, feedstock options, including other waste biomass materials, such as agricultural residues and industrial lignocellulosic wastes, and energy crop feedstocks, such as herbaceous grasses and fast-growing hardwoods, will be investigated.

In addition to dilute acid, the feasibility of a pressurized hot water countercurrent prehydrolysis process that requires no added acid catalyst is being investigated on appropriate feedstocks. Eliminating the acid catalyst requirement has a number of potentially beneficial impacts on the process, including reduced chemicals costs, potentially less expensive metals for pretreatment reactor construction because of less corrosive conditions, and reduced formation of insoluble compounds, such as gypsum, upon neutralization. Such a process is also more environmentally benign than those that utilize harsh acids, bases, or solvents. This work is being conducted using a flow-through percolation system at NREL. In addition, the investigation of a neutral pH-controlled process via a NREL-sponsored subcontract at Purdue University is currently being pursued. Other workers (50-52) have recently reported promising findings with pressurized hot water prehydrolysis approaches, but the economic impacts of some reported process conditions, such as very small biomass particle size and relatively large volumes of pressurized hot water, have not been thoroughly evaluated.

A number of alternate pretreatment methods have also recently been investigated through a series of NREL-sponsored subcontracts. These include the use of various acid catalysts in a dilute acid prehydrolysis mode, such as nitric acid, phosphoric acid, and phosphoric acid/organosolv; alkaline-based methods, such as lime pretreatment and ammonia recycle percolation; and physicochemical methods, such as hydrogen peroxide extrusion and ammonia freeze explosion. Many of these methods have been widely discussed in the literature, but it is difficult to compare the relative effectiveness of these various approaches because of variation in feedstocks tested, analytical methods, and pretreatment performance measurement and reporting protocols. A common-basis evaluation of these approaches that will determine the relative effectiveness and process economic implications of these various pretreatment approaches is currently being conducted. This can serve as a basis for making rational choices in pretreatment technologies for given feedstock and process applications. Promising pretreatment technologies will likely be further developed from additional bench-scale work and initial integration with subsequent detoxification and fermentation steps, potentially leading to engineering scale equipment design and testing in the future. Another key goal of this evaluation is to determine which pretreatments are best suited to various feedstocks.

Complete Hydrolysis Studies. Very recently, the general concept of dilute acid countercurrent prehydrolysis developed at NREL has been extended to investigate the possibility of a full hydrolysis of both hemicellulose and cellulose to soluble sugars. This concept has been motivated by the fact that thus far, costs of commercial cellulase preparations from industrial enzyme suppliers have been prohibitive for use in bioethanol processes (43у 53). Although ongoing process development and research plans at NREL show promise in substantially reducing cellulase production costs, the ultimate costs of cellulase production are still unclear. Thus, a full hydrolysis process that substantially reduces (or even eliminates) cellulase requirements is an option worthy of investigation.

The use of dilute sulfuric acid to totally hydrolyze the carbohydrates in lignocellulosic feedstocks for ethanol production was widely abandoned by 1990 due to the relatively low yields of glucose (50%-60%) as compared to potentially high yields (80%-95%) from enzymatic hydrolysis of cellulose (54-56). Even though kinetic modeling exercises conducted at NREL (57) suggested that yields as high as 88% could be obtained using a countercurrent reactor configuration, laboratory studies could only demonstrate total sugar yields (C5 and C6 sugars) of about 60% (56). However, several observations were made to explain the low yields. The material of construction of the experimental reactor (Carpenter Cb20-3), when subjected to 1% acid at elevated temperatures, leached chromium ions at a concentration that leads to degradation of glucose formed in the reactor. Additionally, as the biomass was hydrolyzed, the packed bed collapsed, which led to non-ideal fluid dynamics that caused increased degradation of the glucose.

The renewed interest in total hydrolysis dilute acid processes has been motivated by modeling exercises, which indicate that with a hydrolysis liquor residence time about one-seventh of the solid residence time, yields of glucose could approach those obtained with an enzymatic process (57). Such a process could result in high volumes of hydrolyzate liquor relative to countercurrent dilute acid prehydrolysis. The impact of producing liquor volumes even greater than those for countercurrent dilute acid prehydrolysis is to diminish the overall process efficiency. Although overall cost savings may be realized through reduced cellulase requirement, process improvements to reduce the quantity of the resulting liquor volumes is required. Again, innovative pretreatment reactor design aimed at construction feasibility and reducing required liquor volumes will play a key role in determining the viability of a dilute acid total hydrolysis process.

Numerous treatments have been developed to enhance the enzymatic digestibility of lignocellulosic biomass including: physical (e. g., ball milling, two-roll milling), chemical (e. g., dilute-acid hydrolysis, alkali), physico-chemical (e. g., steam explosion, Ammonia Fiber Explosion), and biological (e. g., white-rot fungi) (6). For the MixAlco Process, alkaline treatment is selected because the acids produced in the fermentor will neutralize the alkali, thus allowing recovery of the treatment agent. Of the various alkalis that are effective (e. g., sodium hydroxide, ammonia), lime was selected because of its low cost and compatibility with other process steps.

Compared to other alkalis, the literature on lime treatments is relatively sparse. Most of the studies have been performed by animal scientists seeking simple, room-temperature treatments to enhance ruminant digestibility. Because the treatment temperature was low, their results were poor; the general consensus is that lime is not as effective as other alkalis. However, by optimizing the reaction temperature and other conditions, lime is a very effective treatment agent.

Table I summarizes the results of some recent lime treatment studies using extracellular enzymes to hydrolyze the biomass. Compared to untreated biomass, lime-treated biomass has an enzymatic digestibility roughly ten fold larger. Because of its low lignin content, herbaceous biomass requires only lime treatment. However, because of its high lignin content, woody biomass requires the addition of oxygen to partially oxidize the lignin and remove it from the biomass. In addition, woody biomass requires more severe time and temperature.

Table П shows that lime treatment roughly doubles the ruminant digestibility of biomass. Comparing the digestibilities reported in Table I versus Table П, the digestibility within the ruminant is greater than that achieved with extracellular enzymes. This suggests that an industrial process based on a mixed culture of microorganisms may have advantages over one based upon extracellular enzymes.