Optimization of the Liquefaction Process in Bioethanol Production & Development of Method for Quantification of Nonsolubilized Starch in Mash

(1)

Department of Physics and Measurement Technology

Master’s Thesis

Optimization of the Liquefaction Process in

Bioethanol Production & Development of Method

for Quantification of Nonsolubilized Starch in Mash

Anna Aldén

LITH-IFM-EX--08/1917--SE

Department of Physics and Measurement Technology Linköpings universitet

(2)

(3)

Master’s Thesis LITH-IFM-EX--08/1917--SE

Optimization of the Liquefaction Process in

Bioethanol Production & Development of Method

for Quantification of Nonsolubilized Starch in Mash

Anna Aldén

Supervisor: Helena Stavklint,

Lantmännen Agroetanol AB

Examiner: Prof. Carl-Fredrik Mandenius,

Department of Physics, Chemistry and Biology, Linköping University, Sweden

(4)

(5)

Avdelning, Institution

Division, Department Division of Biotechnology

Department of Physics and Measurement Technology Linköpings universitet

SE-581 83 Linköping, Sweden

Datum Date 2008-02-20 Språk Language Svenska/Swedish Engelska/English Rapporttyp Report category Licentiatavhandling Examensarbete C-uppsats D-uppsats Övrig rapport

URL för elektronisk version

ISBN

—

ISRN

LITH-IFM-EX--08/1917--SE

Serietitel och serienummer

Title of series, numbering

ISSN

—

Titel

Title

Optimering av uppströmsprocessen vid bioetanolproduktion samt utveckling av metod för kvantifiering av olöst stärkelse i mäsk

Optimization of the Liquefaction Process in Bioethanol Production & Development of Method for Quantification of Nonsolubilized Starch in Mash

Författare

Author

Anna Aldén

Sammanfattning

Abstract

Ethanol production at Lantmännen Agroetanol AB in Norrköping began in De-cember 2000. The objective of this master’s thesis is to find and optimize factors affecting the yield of the liquefaction, a part of the upstream process. To measure successfulness of liquefaction it is desired that amount of non-solubilized starch is quantified, and hence a method for determination of non-solubilized starch in mash has to be developed.

Starch is a carbon reserve in plants. Starch granules are polymers of amy-lose and amylopectin which are polysaccharides of glucose. When a starch/water solution is heated the starch granules start to absorb water and swell, a process termed gelatinization. The swelling makes the granules susceptible to hydrolysis by enzymes such as α-amylase, this is called liquefaction. Eventually the granular structure is broken and the slurry contains solubilized starch which can be sac-charified to glucose by glucoamylase. In the bioethanol production process, the milled grain is mixed with water and enzymes. The slurry is heated, gelatinization and liquefaction occurs. Saccharification occurs simultaneously to fermentation. Ethanol is purified from the fermented mash during downstream processing.

Starch in the form of starch granules cannot be quantified. The adopted prin-ciple for determination of non-solubilized starch in liquefied mash is to wash away the solubilized starch, then quantitatively hydrolyze non-solubilized starch to glu-cose and quantify gluglu-cose.

To find and optimize factors significant for yield of liquefaction multiple factor experiments were conducted where eight factors were studied. pH, temperature in mixtank and temperature in liquefaction tank 1 were the most significant factors. The temperature in liquefaction tank 1 should be kept as is is at 74◦C. A small rise in pH should shorten the mean length of dextrins which is preferable. An increase of pH from 5.2 to 5.4 is therefore proposed. The temperature in mixtank should also be increased by a few degrees. The yield of the process should be carefully evaluated during the modifications.

Nyckelord

(6)

(7)

Abstract

Ethanol production at Lantmännen Agroetanol AB in Norrköping began in De-cember 2000. The objective of this master’s thesis is to find and optimize factors affecting the yield of the liquefaction, a part of the upstream process. To measure successfulness of liquefaction it is desired that amount of non-solubilized starch is quantified, and hence a method for determination of non-solubilized starch in mash has to be developed.

Starch is a carbon reserve in plants. Starch granules are polymers of amy-lose and amylopectin which are polysaccharides of glucose. When a starch/water solution is heated the starch granules start to absorb water and swell, a process termed gelatinization. The swelling makes the granules susceptible to hydrolysis by enzymes such as α-amylase, this is called liquefaction. Eventually the granular structure is broken and the slurry contains solubilized starch which can be sac-charified to glucose by glucoamylase. In the bioethanol production process, the milled grain is mixed with water and enzymes. The slurry is heated, gelatinization and liquefaction occurs. Saccharification occurs simultaneously to fermentation. Ethanol is purified from the fermented mash during downstream processing.

Starch in the form of starch granules cannot be quantified. The adopted prin-ciple for determination of non-solubilized starch in liquefied mash is to wash away the solubilized starch, then quantitatively hydrolyze non-solubilized starch to glu-cose and quantify gluglu-cose.

To find and optimize factors significant for yield of liquefaction multiple factor experiments were conducted where eight factors were studied. pH, temperature in mixtank and temperature in liquefaction tank 1 were the most significant factors. The temperature in liquefaction tank 1 should be kept as is is at 74◦C. A small rise in pH should shorten the mean length of dextrins which is preferable. An increase of pH from 5.2 to 5.4 is therefore proposed. The temperature in mixtank should also be increased by a few degrees. The yield of the process should be carefully evaluated during the modifications.

Sammanfattning

Etanolproduktionen på Lantmännen Agroetanol AB i Norrköping började i De-cember 2000. Målet med examensarbetet är att hitta och optimera faktorer som påverkar utbytet av likvifieringen i etanolproduktionen. För att studera utfallet av likvifieringen är det önskvärt att mäta hur mycket stärkelse som inte har löst sig, och därför måste en metod för att mäta olöst stärkelse i mäsk utvecklas.

(8)

Stärkelse utgör en kolreserv i växter. Stärkelsegranuler är polymerer av amylos och amylopektin, vilka i sin tur är polysackarider av glukos. När en stärkelse/vatten-blandning värms upp börjar stärkelsegranulerna att absorbera vatten och svälla, en process som kallas gelatinisering. Svällningen gör granulerna känsliga mot hydro-lys av till exempel enzymet a-amylas, vilket kallas för likvifiering. Efter tillräckligt mycket gelatinisering och likvifiering förstörs hela den granulära strukturen och stärkelsen övergår till löst form. Löst stärkelse kan försockras till glukos med enzy-met glukoamylas. I produktionen av bioetanol blandas malet spannmål med vatten och enzymer. Slurryn värms upp och gelatinisering och likvifiering sker. Försock-ring sker simultant med fermenteFörsock-ringen. Etanol renas fram från den fermenterade mäsken i nedströmsprocessen.

Stärkelse i granulform kan inte kvantifieras. Den valda metoden för mätning av olöst stärkelse i likvifierad mäsk innebär att den lösta stärkelsen tvättas bort, sedan hydrolyseras den olösta stärkelsen kvantitativt till glukos, vilken kan kvantifieras. Flerfaktorförsök gjordes för att hitta och optimera faktorer signifikanta för ut-bytet av likvifiering. Åtta olika faktorer studerades. pH, temperatur i mixtank och temperatur i likvifieringstank 1 visade sig vara de tre mest signifikanta faktorer-na. Temperaturen i likvifieringstank 1 ska bibehålla samma temperatur som idag, 74◦C. En liten höjning av pH borde förkorta medellängden av dextrinerna, vilket är fördelaktigt. En ökning av pH från 5,2 till 5,4 är föreslås därför. Temperatu-ren i mixtanken ska ökas några få grader. Utbytet av processen måste noggrant utvärderas under modifieringarna.

(9)

Acknowledgments

First I would like to thank Lantmännen Agroetanol AB for giving me the op-portunity of doing this master’s thesis. Special thanks to my supervisor Helena Stavklint for interesting discussions, your support and knowledge. Anna-Karin Wingren and Caroline Lundell, you were always helpful in the laboratory. Thanks to everyone else at Lantmännen Agroetanol AB who made this time memorable.

I would also like to express my gratitude to Anders Brundin for advice con-cerning statistical analysis.

(10)

(11)

Nomenclature

DDGS Dried distillers grain with solubles

DF Degrees of freedom

DP Degree of polymerization, followed by a number from 1 to n. Fermentable starch Starch that is solubilized during liquefaction and available

as substrate to yeast. Also called solubilized starch. HPLC High performance liquid chromatography

L1 Liquefaction tank 1

L2 Liquefaction tank 2

MRP Maillard reaction products

MS Mean square (statistical expression) SS Sum of squares (statistical expression)

SSF Simultaneous saccharification and fermentation

Sugar profile Describing the distribution of lengths of solubilized starch in mash, i.e. content of DPn, DP3, maltose and glucose.

TiM Residence time in mixtank

TiL1 Residence time in liquefaction tank 1 TiL2 Residence time in liquefaction tank 2

TL1 Temperature in liquefaction tank 1

TL2 Temperature in liquefaction tank 2

TMix Temperature in mixtank

TS Total solids

(12)

(13)

Introduction

1.1 Background

Bioethanol production at Lantmännen Agroetanol AB began in December 2000. Alcohol production, described briefly, means preparing starch- or sugar-containing raw material for fermentation by yeast. The produced ethanol is then recovered and purified through distillation. The continuous work to improve the process at Agroetanol AB regarding increased ethanol yield, decreased resource consump-tion, easier management of the process and decreased production costs has been going on since the production started. To achieve improvements different process parameters such as temperatures, pH, flow rates and residence times can be var-ied, various combinations of raw materials and pre-treatments, e.g., milling, is a matter of interest, addition of enzymes, nutrients and other necessary additives can be optimized, among many other things. When improving a process it is a good idea to divide the process into subunits and study how each subunit can be optimized. However, it is important to remember that changes in one part of the process might affect other parts; therefore one must always have the overall picture in mind.

This thesis focuses on the upstream process part of the ethanol production process where the raw materials; wheat, barley and triticale are prepared for fer-mentation. The upstream process consists of milling, mixing, gelatinization, and liquefaction. During the upstream process, starch in the cereals is made available for the yeast. Heat and enzymes convert starch to fermentable carbohydrates. The more substrate available to yeast the better yield of ethanol. Starch that is not gained during liquefaction cannot be recovered no matter how good and optimized the rest of the process is and will eventually get lost. The loss of starch during gelatinization and liquefaction occurs in several ways:

• nonsolubilized starch, • retrograded starch, and

• side-reactions between sugars and other components of mash. 1

(16)

The different ways in which starch is lost represents different approaches to increase the yield of liquefaction; enhance solubility of starch, prevent retrograda-tion, and decrease side-reactions. The challenge, increase yield of liquefacretrograda-tion, is now divided into these three smaller pieces. Each one of these can be studied and considered separately even though the total yield is a combination of all of them. Lab-scale experiments are often performed to study responses when varying different process parameters. When the optimal conditions are found a pilot-plant experiment might be done and if it is successful; a full-scale trial follows. An organized, money, and time saving approach to study effects of process param-eters is to use experimental design. Several computer based tools, e.g., Modde, Umetrics; Minitab Statistical Software, Minitab, exist that can be helpful both in experimental design and in evaluation of experiments.

1.2 Thesis objectives

The objective of this master’s thesis is to find parameters significant for the yield of the liquefaction process at Lantmännen Agroetanol AB. The significant factors found should be optimized to give the best liquefaction yield. A definition of how liquefaction yield is measured should also be stated. Lantmännen Agroetanol AB especially desires that share of nonsolubilized starch after liquefaction should be measured and considered as a measure of successfulness of liquefaction. Since no method for measuring nonsolubilized starch in mash exists it has to be developed. Therefore another objective is to develop a method for determination of nonsol-ubilized starch in liquefied mash, to be used in this master’s thesis and also by Lantmännen Agroetanol AB as an analysis of the process.

1.3 Method

This master’s thesis was carried out at Lantmännen Agroetanol AB in Norrköping. Literature and articles concerning bioethanol production, starch, gelatinization, liquefaction, Maillard reactions and relevant enzymes were studied. The method for determination of nonsolubilized starch was divided into steps. For each step several experiments were carried out to develop and validate the method. Frac-tional factorial experiments at laboratory scale were conducted to find and opti-mize process parameters significant for the outcome of liquefaction.

1.4 Outline

In the next chapter the theoretical background for this master’s thesis is presented. It is followed by a separate section for development and evaluation of method for determination of nonsolubilized starch. Thereafter materials and methods used during the fractional factorial experiment session are presented followed by results and discussion; conclusions; recommendations; references and appendices.

(17)

Chapter 2

Theoretical background

2.1 Starch

Starch is the major energy storage component in plants and serves as a carbon re-serve. In the photosynthesis, energy from sunlight is used to combine six molecules of carbon dioxide and six molecules of water to a carbohydrate (D-glucose) and molecular oxygen. D-glucose (fig. 2.1) is one of the most abundant carbohydrates on earth and comprises 99.9 % of all the carbohydrates. [1] The D designation refers to the enantiomeric configuration of glucose which is either D or L. Each carbon atom is designated a number for a more comprehensive terminology. C1, also known as the anomeric carbon, is the carbon closest to the oxygen atom in clockwise direction. The cyclic glucose molecule is either in α- or β-configuration which gives information of the hydroxyl group at the anomeric carbon. The OH-group below the ring yields α-configuration and the OH-OH-group above yields β-configuration. See the α- and β-configuration of D-glucose in figure 2.1. In this figure the numbering system for the carbon atoms is also shown. [2]

Figure 2.1. The image to the left illustrates α-D-glucose with the hydroxyl group at the anomeric carbon extending below the ring structure. The right image shows β-D-glucose where the corresponding hydroxyl group extends above the ring structure.

(18)

α-D-glucose is polymerized into the polysaccharides amylose and amylopectin. Amylose and amylopectin are then packed into starch granules in plastids of leaves and storage organs in plants [3]. Hence, native starch is a polymer of amylopectin and amylose. Amylose is a linear polymer whereas amylopectin is branched [1]. The linear polymer amylose consists of α(1-4)-linked glucose subunits. α indicates that the hydroxyl group is below the apparent plane of the glucose subunit, just like before, and 1-4 denotes that the glucose subunits are linked between C1 and C4 of neighboring units. [2] This bond is called the glucosidic bond and is stabile at high pH but hydrolyzes at low pH [4]. The linear part of amylopectin also consists of α(1-4)-linked glucose residues. The branches are linked to the linear part by α(1-6)-linkages. [2] See figure 2.2 for a structural formula of amylose and amylopectin .

Figure 2.2. Structural formula of amylose (top) and amylopectin(bottom).

The branch points in amylopectin yields a large tree-like structure , see figure 2.3.

Glycogen, which is the analogue to starch in animals has a structure closely related to the structure of amylopectin. The difference is that the branch points occur approximately twice as often in glycogen as in amylopectin. [1] Cellulose is the major structural component of cell walls of land plants and brown algae [5]. It is made up of 10,000 or more D-glucose units linked together by β(1-4) glucosidic bonds. This minor difference, β-linkages instead of α has a huge impact on the properties of cellulose. The β-linkages leads to aggregation into extremely insoluble fibrils. [2] This is why cellulose is suited for production of paper and textiles [5].

(19)

2.1 Starch 5

Figure 2.3. The large bush like structure of amylopectin. [1]

The molecular masses of amylose and amylopectin vary with the source of the starch but generally amylopectin is the larger one. Amylose contains on average 500-2000 glucose residues per molecule. [2] Amylopectin can contain 100 000 glucose subunits. [1] The ratio of amylose to amylopectin content also varies with the botanical origin of the starch. Wheat starch granules contain approximately 25% amylose and 75% amylopectin while waxy maize starch contains 1% amylose and 99% amylopectin [2]. The relative amount of amylopectin and amylose affect the properties of starch which is further discussed in sec. 2.2.3 on page 7.

2.1.1 Starch granules

Naturally amylopectin and amylose form starch granules. The sizes and shapes of the granules depend on the source of starch. Wheat starch granules have a disk-pancake look and are about 10 µm thick and 20-30 µm in diameter. Starch granules from other botanical sources ranges in sizes from 0.5 to 45 mm. [1] The granules are stabilized of double helices mainly formed by the branches of amylopectin. Amylose is thought to be in a single-chain state in native starch granules since double helices of amylose are resistant to enzyme degradation (termed resistant starch) and has therefore no natural function. Amylopectin molecules separate the amylose chains. [6] The helices are stabilized by hydrogen bonds both within the same molecule and between adjacent molecules [2]. The double helices are ordered in a parallel arrangement and form either A, B or C crystalline patterns. The A-pattern structure consists of a single double helical complex surrounded by six parallel double helices packed in a hexagonal pattern with water molecules between the complexes. A-pattern structure is found in cereal grain starches such as maize, wheat and rice. The B-patterns is similar to the A-pattern, except for that the middle helical complex is replaced by water molecules. Starches from stem, tuber and fruit, like potato, sago, and banana starches show the B-patterns structure. In C-type starches, both A- and B-pattern are found. This occurs for example in pea and tapioca starches. [1]

Due to formation of highly ordered double helices by amylopectin and the single chained helices of amylose, starch granules consist of semi-crystalline regions

(20)

interspersed with amorphous regions. The semi-crystalline regions contain only amylopectin while amorphous regions also contain amylose. Consequently, starches with a relatively high content of amylose have a lesser degree of crystallinity. [1]. The structure of a starch granule is shown in fig. 2.4

Figure 2.4. The image shows the structure of a potato starch tuber. A is an electron microscopic image of the granules. B shows the semi-crystalline regions interspersed with amorphous regions and C shows the mixture of amylose and amylopectin in amorphous regions.[4].

Besides from amylopectin and amylose starch granules have, especially the surface region, a non-carbohydrate content of lipids (1-5% w/w for cereal starches [1]) and proteins [6]. Lipids complex with the hydrophobic interior of single chain amylose helices, sometimes also with branches of amylopectin if they are long enough [7]. Lipids also exist in free form in the granules [6, 7]. Most of the proteins are believed to belong to starch biosynthetic enzymes which are caught in the granule during synthesis [8].

2.2 Upstream processing in bioethanol

produc-tion

During the upstream process, the grains are prepared for fermentation by yeast. The upstream process can be separated into milling, mixing, liquefaction, gela-tinization and saccharification which are all detailed for in the following sections.

(21)

2.2 Upstream processing in bioethanol production 7

2.2.1 Milling

Milling is the first procedure in the bioethanol production process. The purpose of milling is to expose the starch granules to water in the subsequent gelatinization process. The larger the ratio of surface area to volume of the particles, the easier it is for water and enzymes to permeate the kernel. Therefore a fine grounded meal gives higher ethanol yield than coarse grounded meal. There are also disadvantages with small particle sizes. Small particles increase the load of both centrifuges and evaporators in the downstream process. Therefore the particle size has to be a compromise between yield and minimizing problems in downstream processing.[9] The hammer mill is the most common milling equipment among distillers. The hammer mill consists of a grinding chamber wherein hammers rotate at high speed. To exit the chamber the particles have to pass a sieve and therefore have to be smaller than a certain size. This allows for setting a maximum particle size in the meal. To inspect the conditions of the mill and the sieve, analysis of particle size distributions are made regularly. Roller mill is another type of milling equipment where the grains are nipped between a pair of rollers. It is suitable for small cereal grains.[9]

2.2.2 Mixing

Directly after milling the meal passes on to a mixing tank. In the mixing tank, meal, water and enzymes are mixed together during warming. Backset stillage 1 and nutrients required for fermentation might also be added here. The temperature is kept slightly above or under gelatinization temperature (see next sec. 2.2.3). Too extended gelatinization leads to high viscosity and should be avoided here due to problems with moving the slurry, though higher temperature facilitates heating in the next step. [9]

2.2.3 Gelatinization

In order to make starch available to yeast (Saccharomyces cerevisiae) it has to be hydrolyzed to fermentable sugars such as glucose, maltose (two α(1-4)-linked glucose units) and maltotriose (three α(1-4)-linked glucose units).[10] Hydrolysis of starch can occur either by acids or by the action of enzymes. Acids in combination with high temperature and pressure hydrolyze starch. Enzymes can do the same under relatively mild conditions and also give a higher yield. Since the 1960s enzymatic hydrolysis is more common and also because it is used in Agroetanol AB’s process acid hydrolysis will not be considered in the current work. [2] Before the enzymes can gain access to the starch molecules the granular structure has to be broken down. This occurs in a process called gelatinization. [9] Starch granules suspended in water swell to a limited extent and absorb water up to 30% w/w. When the suspension is heated above a certain temperature the hydrogen bonds in the starch granule rupture, the granules starts to swell irreversibly and the

1_{Backset stillage is recycled thin stillage. Thin stillage is the liquid part of the residual mash} after evaporation of the ethanol soultion in the downstream process.

(22)

granular structure is gradually lost. This temperature is called the gelatinization temperature and varies with the source of the starch. [1, 2] During swelling, leaching of granule constituents, mainly amylose occurs [11] which increase the viscosity of the surrounding slurry [4]. Further increase of the temperature leads to extended swelling of the granules and increased viscosity of the slurry and eventually breaking of the granules [4]. Large granules swell more easily than small granules due to a lesser extent of molecular bonding and have consequently a lower gelatinization temperature. High amylose content starches generally have higher gelatinization temperatures because linear amylose molecules can be more densely packed than the bush-like amylopectin molecules and hence more hydrogen bonds will form.[2] See table 2.1 for gelatinization temperatures of different cereals.

Table 2.1. Gelatinization temperatures for some cereal starches. [2]

Cereal starch source Gelatinization range [◦C]

Barley 52-59

Wheat 58-64

Rye 57-70

Corn (maize) 62-72

High amylose corn 67->80

Rice 68-77

Sorghum 68-77

2.2.4 Liquefaction

After gelatinization the swollen starch granules and the solubilized polysaccharides (leached out of the granules) are susceptible to hydrolysis by enzymes. The enzyme used here is usually a thermotolerant form of the endo-acting enzyme α-amylase [9]. α-amylases are usually isolated from bacterial sources. Hydrolysis by a-amylase converts the polysaccharides/starch to oligosaccharides of varying length. The endo-acting α-amylase hydrolyzes α(1-4)-glucosidic bonds in the inner parts of amylose and amylopectin [4]. α-amylase cannot hydrolyze α(1-6)-glucosidic bonds of amylopectin. The liberated water-soluble oligosaccharide chains are called dex-trins when linear, and α-limit dexdex-trins when branched. The products are released in α-configuration. The purpose of liquefaction is to make the mash less viscous and to prepare it for the next step, saccharification [9].

2.2.5 Saccharification

During saccharification individual glucose molecules are released by the action of an exo-acting enzyme. A commonly used enzyme is glucoamylase/amyloglucosidase. The enzyme successively hydrolyzes the terminal α(1-4)-glucosidic bond from

(23)

2.3 Factors contributing to lower starch yield 9

non-reducing2 ends of dextrins. Glucoamylase is also able to hydrolyze 6)-glucosidic bonds, but at a much slower rate, 20 to 30 times slower, than α(1-4)-glucosidic bonds. Hence, both dextrins and α-limit dextrins are degraded by glucoamylase. The product released is α-D-glucose. [9]

The saccharification step can occur both pre and simultaneous to fermenta-tion, the latter called SSF (Simultaneous Saccharification and Fermentation). If a separate saccharification step is used, the optimal conditions (e.g. pH, tempera-ture) for the saccharification enzyme can be used. Glucoamylases are usually not thermotolerant due to their fungal origin and often do not tolerate temperatures above 60◦C. The pH optimum is usually between pH 4.0 and 4.5. A problem with these pH and temperature conditions is that they also favor growth of bacterial contaminants like Lactobacillus. [9] The benefit of pre-saccharification is that high glucose concentration is achieved before fermentation which ensure that glucose is not limiting [2].

Using SSF the yeast will be spoon-fed with glucose. Too high glucose lev-els leads to osmotic stress of the yeast cells [12].The lower glucose concentration and also the competition from yeast make it harder for bacterial contaminants to survive. Low glucose concentration also prevents repolymerization of glucose to isomaltose, which gives a lower glucose yield. [9]

2.3 Factors contributing to lower starch yield

2.3.1 Retrogradation

Retrogradation of starch is the physical event when starch molecules (amylose or amylopectin) in solution reassociate via intermolecular hydrogen bonds, forming helixes and double helixes, and precipitate from the solution. [1, 13, 14]. The straight amylose chains have a greater possibility to line up in solution and form intermolecular hydrogen bonds than the branched amylopectin. Therefore, amy-lose has a much lower solubility in aqueous solutions than amylopectin.[1] Due to different ratios of amylopectin to amylose content the rate of retrogradation de-pends on the botanical origin of the starch. [14] The retrograded starch, especially retrograded amylose, is resistant to digestion by α-amylase [6, 13]. Retrograda-tion occurs after heating a starch soluRetrograda-tion in temperatures above gelatinizaRetrograda-tion temperature and then recooling it. Lower cooling temperature and greater differ-ence between heating and cooling temperature increases the rate of retrogradation. Retrogradation has been observed in samples heated to 90◦C and then cooled to 30◦C.[14]

2.3.2 Maillard Reactions

The Maillard reaction, or non-enzymatic browning involves reactions of reducing sugars, such as glucose, maltose, fructose, and ribose, and amino groups found in amino acids or proteins [15]. Since Maillard reactions can take place between

2_{Reducing ends are the ends where the anomeric carbon is free, the other ends are} non-reducing ends.

(24)

different sugars and different amino acids (or proteins) a wide variety of Mail-lard reaction products (MRP) are formed [16]. Many of the MRP have aroma and taste properties, which make them desirable for instance in bread baking [17]. The formation of MRP causes browning whose intensity can be measured spectropho-tometrically at 420 nm [16, 18]. MRP are formed in many different ways due to conditions and which sugar and amino acids that are involved. There are though some main reaction ways and the MRP are often grouped according to them, like Heyns products and Amadori products. In the formation of Amadori products an unstable Shiff base is formed in the first step. The Shiff base then undergoes the so called Amadori rearrangement to form the Amadori compound. The Amadori compound can then undergo a wide variety of further reactions to form MRP.[15] Conjugation of starch with amino acids through the Maillard reaction leads to lower swelling and solubility of starch in water compared to original starches. Also the susceptibility to hydrolysis by α-amylase is lowered and the thermal stability is increased. [19] Therefore Maillard reactions are not desirable in bioethanol production. Tauer et. al. [15] also concludes that MRP have an inhibitory effect on yeast, which leads to lower ethanol production. This inhibitory effect is however highly dependent on pH during fermentation and was weakly significant for pH 5 and highly significant for pH 7 and 8 in his study.

Factors affecting the occurrence of Maillard reactions are pH, temperature, time, sugar component and amino component among others [15, 16, 18, 19]. Heat strongly accelerates the Maillard reaction [19] and is probably the most significant factor. A prolonged heating time form more MRP, but the pH during the reaction has a significant impact on the amount of MRP formed. According to Tauer et. al [15] the same heating time but different pH (pH 5 and 7) showed lesser formation of MRP at lower pH. Of the amino acids lysine seems to be the one contributing the most to MRP formation [15, 16]. One possible explanation to this is that lysine contains two amino groups. Tauer et. al [15] who studied inhibitory effects of MRP on fermentation found none or weakly significant difference between sugar components (fructose, glucose, maltose and ribose) involved in the Maillard reaction. Kwak et. al. [16] saw that the degree of contribution to browning intensity was in the order of xylose > arabinose > glucose > maltose > fructose. Xylose and arabinose are pentoses and more reactive than the others who are hexoses. Glucose and maltose contribute more to browning than fructose due to different reaction ways. Fructose forms mainly Heyns products while glucose and maltose favor formation of Amadori products. Browning due to Amadori products is faster than browning due to Heyns products [16].

2.3.3 Nonsolubilized starch

In processes where large amounts of mash are gelatinized and liquefied some starch granules might remain intact, termed ghosts, and hence the starch stays unavail-able for fermentation. Debet and Gidley [6] have studied formation of ghosts. Their suggested mechanism for ghost formation includes cross-linking between glu-cans in the granule. During swelling of the granule, as in gelatinization, amylose molecules previously separated by amylopectin can move freely within the granule

(25)

2.4 Quantification of starch 11

and find another amylose molecule to form a double helix with. As previously mentioned in 2.3.1 on page 9 double helices formed by two amylose molecules are tightly associated. They have a melting point at 130◦C. If the swelling is rela-tively slow, as in wheat, and the cross-linking is rapid, enough glucose residues have time to participate in cross-linking which prevent the granule from bursting. The cross-linking seem to be mostly located near the surface of ghosts, which form a mechanical skin preventing further swelling and rupture of the granule. Surface proteins and lipids are involved in preventing swelling and therefore promote for-mation of ghosts. Ghost forfor-mation have though been observed when surface lipids and proteins have been removed, however in a more swollen form.[6]

2.4 Quantification of starch

Starch granules can be detected through addition of triiodide. Iodide complexes in the hydrophobic interior of the helixes formed by amylose and amylopectin. The amylose complex results in a deep-blue colored product while the iodide-amylopectin complex gives a wine-red or violet product. [1] However, this method is only qualitative and not quantitative.

Quantification of starch usually follows the general principle: quantitative hy-drolysis of starch to glucose and then quantification of glucose. Free glucose can be measured by different methods, e.g., HPLC, glucose-oxidase and peroxidase reagents, or polarimetrically. The hydrolysis is done either by acids or enzymes [20]. In AACC Method 76.13 (AOAC Method 996.11) starch is completely solubi-lized by cooking the sample in presence of thermostable α-amylase. Then dextrins are hydrolyzed quantitatively to free glucose by the action of amyloglucosidase. Glucose is measured spectrophotometrically at 510 nm after addition of glucose-oxidase/peroxidase buffer.

(26)

(27)

Chapter 3

Bioethanol production at

Lantmännen Agroetanol AB

3.1 An overview of the production process at

Lant-männen Agroetanol AB

The continuous production process at Lantmännen Agroetanol AB consists of several steps, for a schematic picture see fig. 3.1.

The grain is stored in four silos. The desired mixture of grain; wheat, barley and triticale is grinded in two hammer mills. The milled grain proceeds to the mixing tank where it is added to water, urea and enzymes. The enzymes added are α-amylase, β-glucanase and xylanase. The two latter possess activity towards non-starch carbohydrates, the structural material in plants [21, 22] and are added mainly for viscosity reduction. Residence time in mixing tank is approximately one hour and temperature is about 58◦C. pH is kept at 5.2. The slurry carries on to liquefaction tank 1 (L1) where gelatinization and liquefaction start. Temperature is increased to 73◦C, residence time is two hours. The hot mash then continues to liquefaction tank 2 (L2) where the temperature is increased to 89◦C. Also here residence time is two hours. The warm mash is then cooled in a mash cooler, to 31.5◦C and pH is lowered to 3.5. Thin stillage and a protease are also added here. Fermentation is the next step and occurs progressively in five fermenters. The liq-uefied mash is added both to fermenter 1 and 2. Yeast from the yeast propagation tank is added to fermenter 1 and 2. Total fermentation time is approximately 55 hours. The ethanol content in the last fermenter is about 9 %(w/w).

Downstream processing begins with vaporizing the ethanol solution from the mash in two parallel mash colons. An aldehyde stripper is used to remove heads, i.e., acetaldehyde and ethylacetate, from the ethanol solution. In the subsequent rectification tower higher alcohols, fusels (sw. "finkel") are removed. A molecular sieve absorbs residual water and the purity of the ethanol is now 99.8% (w/w).

The non-vaporized mash continues to a stillage tank. The stillage is separated into solids and liquids by decanters. The solids are further treated and eventually

(28)

Figure 3.1. Schematic picture of the production process at Lantmännen Agroetanol AB.

(29)

3.2 Properties of enzymes used in the upstream process 15

Figure 3.2. Temperature and pH profile of β-glucanase/xylanase [22]

becomes DDGS (Dried Distillers Grain with Solubles) which is sold as animal feed. The liquid part, thin stillage, is both recycled in the process and evaporated to syrup. Syrup is involved in the production of DDGS and also sold as animal feed as it is.

3.2 Properties of enzymes used in the upstream

process

3.2.1 α-amylase

The α-amylase used is thermostable and the recommended conditions according to the manufacturer is a temperature of 83-89◦C and pH of 5.4-5.8. The recommended dosage is 0.04 - 0.06 %(w/w) as a starting point. [23]

3.2.2 β-glucanase

The β-glucanase used has multiple activity towards cellulose, hemicellulose, arabi-noxylans and β-D-glucans. The pH and temperature profiles are presented in fig. 3.2

3.2.3 β-glucanase/xylanase

β-glucanase/xylanase modify and digest non-starch carbohydrates. It is mainly added for viscosity reduction. The pH and temperature profiles are shown in fig. 3.3.

(30)

(31)

Chapter 4

Method for determination of

nonsolubilized starch in

mash

As previously mentioned, loss of starch occurs during the process, for example non-solubilization of starch. It is of interest to study how much starch is lost and to what, as an evaluation of the process. To evaluate the yield of liquefaction, mash samples can be taken just after the liquefaction step, before fermentation.

4.1 General principle of method

To my knowledge, no method for direct quantification of starch in granular form ex-its. Therefore the nonsolubilized starch has to be completely hydrolyzed to glucose and then the glucose content can be quantified. Liquefied mash, in which we want to determine nonsolubilized starch content, already contains high concentrations of solubilized starch. If the nonsolubilized starch in liquefied mash is hydrolyzed to glucose, the solubilized starch will be hydrolyzed too, and the glucose content would resemble content of total solubilized starch in mash. There are two possible solutions; determine content of fermentable starch (i.e. solubilized starch) first and then hydrolyze nonsolubilized starch and quantify the total solubilized starch con-tent. The nonsolubilized starch content is calculated by subtracting fermentable starch from total solubilized starch. The other alternative is to wash away the fermentable starch, then hydrolyze the nonsolubilized starch and quantify only the nonsolubilized starch (in solubilized form). Both approaches have advantages and disadvantages. In general, when subtracting two figures, both with a certain inaccuracy the inaccuracy of the result is even larger. On the other hand the sec-ond alternative must contain a washing step to get rid of the fermentable starch, this extra procedure might also lead to more inaccuracy of the result. However the second alternative was chosen since it was believed that it would give the best accuracy when directly quantifying the glucose-concentration of interest.

(32)

4.2 Sample preparation

The sample preparation includes washing solubilized starch away from the mash followed by drying of the mash. The drying step is necessary for achieving a known water content of mash (i.e. 0%). In glucose determination methods the glucose content is specified as % (w/w) which of course is dependent of water content.

The sample preparation begins with saccharification of the solubilized starch in the liquefied mash. This procedure was mainly included because it was de-sirable to measure content of fermentable starch in liquefied mash. It was later discovered that the washing procedure used, only was effective when washing away shorter dextrins which made the saccharification step even more important. After saccharification the mash is filtered and the filtrand is dried.

4.2.1 Saccharification of solubilized starch

To determine fermentable starch, an existing method at Lantmännen Agroetanol AB for determination of remaining fermentable carbohydrates in fermented mash was used as a starting-point. In this method 20 g mash from fermenter five is diluted with deionized water to 120 g. The slurry is heated in 60◦C for 10 min before 220 µl glucoamylase is added. The slurry is then incubated at 55◦C for 3 hours where complete saccharification of solubilized starch occurs. The glucose content, corresponding to fermentable starch, is then measured by HPLC. The chromatogram also yields information about maltose, DP31 and DPn2content.

To find optimal conditions for saccharification of liquefied mash, which has a higher content of free sugars than fermented mash, different incubation times, dilutions and enzyme doses were evaluated to reach as low DPn content as possible. When DPn content approaches zero all solubilized starch is hydrolyzed to glucose which is the purpose with this procedure.

To compensate for higher concentration of solubilized starch, experiments were conducted where the enzyme dose was amplified several times. A shorter incubation-time than three hours was desirable which also was an argument for increasing the enzyme dose. HPLC-analysis were run at 1 and 3 hours of incubation time. Inter-estingly the content of DPn seemed to increase from 1 to 3 hours of incubation. Probably other compounds are included in the DPn-peak which affect the result. The difference in enzyme dose did not seem to have any effect on remaining DPn. This was probably because the lowest enzymatic dose in this experiment already was excessive. To lower the DPn content another experiment was conducted where the dilution factor was increased from six to ten, i.e. 20 g mash was diluted to 200 g. To compensate for any effects of greater volume, E-flasks with a slightly higher enzymatic dose, 350 µl, was included in the experiment. HPLC-analysis were run at 0.5, 1 and 3 hours of incubation. The DPn content seemed to be the same between 0.5 and 1 hour of incubation time and then increased between 1 and 3 hours as in the previous experiment. The DPn content was generally lower in flasks with dilution factor 10 compared to flask with dilution factor 6. The

1_{DP: Degree of Polymerization, DP3 is hence maltotriose, three glucose molecules} 2_{n is an integer, always >3 in this report.}

(33)

4.2 Sample preparation 19

glucose content did not seem to be affected much. The enzymatic dosage, 220 or 350 µl, did not seem to affect the DPn content. The higher DPn content in the lower dilution factor experiment was probably due to inhibition of the enzymatic activity by the high glucose concentration.

It was decided that an enzymatic dosage of 220 µl/20 g mash, an incubation time of 1 hour and a dilution factor of ~20 should be used in the method to effectively hydrolyze the solubilized starch to glucose.

4.2.2 Filtration and drying of saccharified mash

To facilitate the drying and washing step the liquefied and saccharified mash was filtrated and rinsed with ~100 ml of deionized water with vacuum aid using a Büchner funnel together with a Büchner flask. The free glucose is solved in water and most of it is therefore found in the filtrate. A similar result could have been achieved by centrifuging the sample, discard the supernatant, resuspend the pellet with water and centrifuge again. However no centrifuge for such big samples (200 g) was available at the laboratory. The filtrand, which contains the nonsolubilized starch, and some residual free glucose, was dried at 40-45◦C until complete dryness.

4.2.3 Washing

Total starch analysis kit (Megazyme International, Bray, Ireland)(AOAC Method 996.11/AACC Method 76.13/ICC Standard Method No. 168) was used to hy-drolyze and quantify the nonsolubilized starch in the dried filtrand. Since it was believed that some residual free glucose remained in the sample, a washing pro-cedure used for samples containing free glucose and maltosaccharides described in the method, was decided to be used. In this washing procedure ~100 mg of the sample is suspended in 5 ml aqueous ethanol (80% v/v) and incubated at 80◦C for 5 minutes. Another 5 ml aqueous ethanol is added and the sample is centrifuged for 10 minutes at 3000 rpm. The supernatant is carefully poured off. Thereafter the pellet is resuspended in aqueous ethanol and centrifuged once again, the su-pernatant is discarded. Centrifugation was in this application changed from 10 minutes and 3000 rpm to 3 minutes and 4500 rpm. The much shorter centrifuga-tion time was possible due to absence of light weight particles in the samples. The result of the washing procedure was examined by determination of free glucose in the washed samples using Total Starch analysis kit with the exception that wa-ter was added instead of enzymes, to avoid solubilization of further starch. The results for unwashed samples and samples washed once, twice and three times re-spectively are shown in table 4.1. As can be seen from the table, three washes are needed to reach acceptable low levels of free sugars in both sample A and B. It was decided that the washing procedure described in Megazyme Total starch analysis kit should be used and repeated three times in the method. Controls made during analysis of the fractional factorial experiments yielded even lower levels of free sugars in washed samples.

(34)

Table 4.1. The results for sample A and B washed none, once, twice and three times respectively. The unit is % (w/w) starch in the sample.

Sample Washes A B 0 1.658 0.704 1 0.198 0.088 2 0.110 0.068 3 0.049 0.055

4.3 Hydrolysis of nonsolubilized starch

Megazyme Total Starch analysis kit was used to quantify the remaining nonsolu-bilized starch in the washed samples. A starch control was included in the kit and since the method is validated no further controls or validations seemed necessary. Calculations were made that confirmed the plausibility of the results.

4.4 Evaluation of method accuracy

To evaluate the accuracy of the method three samples, A, B and C were with-drawn from the same mash sample (liquefied mash from the process). A, B and C were saccharified, filtered and dried separately. From each of A, B and C two samples two samples of the dried filtrands were taken and washed and analyzed for nonsolubilized starch according to the developed method. The values of the two samples for each of the three samples are presented in tab. 4.2, as % (w/w) nonsolubilized starch in the sample, together with mean values for each sample A, B and C. The total mean value is calculated together with the sample standard deviation, S. The sample standard deviation is based on the tree mean values for A, B and C respectively. The dispersion can be used as a measure of accuracy and is calculated as mean_S which in this case yields an accuracy of 3.3% which must be considered satisfying. However, for the B sample the two results are quite distinct from each other and the sample standard deviation for sample B is large, while it is good for sample A and C. It was decided that double samples should be run for all the reactors in the future experiments to avoid errors.

(35)

4.4 Evaluation of method accuracy 21

Table 4.2. Measurement values of nonsolubilized starch (as %(w/w)) for three samples of the same mash.

Sample A B C Sample 1: 2.19 2.40 2.43 Sample 2: 2.29 2.07 2.30 Mean: 2.24 2.24 2.37 Total mean: 2.28 Sample std deviation: 0.075 Dispersion: 3.3%

(36)

(37)

Chapter 5

Experimental details

5.1 Multiple factor experiments

To find and optimize factors important for yield of liquefaction multiple factor experiments were conducted where the mixtank, liquefaction tank 1 and 2 were simulated. A two-level eight-factor fractional factorial design is adopted in this study. The eight factors being studied are; temperature in mixtank, L1 and L2 (abbr. TMix, TL1 resp. TL2); residence times in mixtank, L1 and L2 (abbr. TiM, TiL1 resp. TiL2; pH (abbr. pH) of the process and addition of urea (abbr. U) or not. The eight factors and their levels are presented in table 5.1. The experimental design is found in appendix A. Sugar profile (glucose, maltose, DP3 and DPn concentrations in mash) fermentable starch, nonsolubilized starch and browning intensity are used as response factors.

Table 5.1. The investigated factors and their levels.

Mixtank L1 L2 Temp.[◦C] 55-65 74-94 74-94

Time [h] 0.5-1 1-3 1-3

pH 4.6-5.6

Urea [g] 0-1.86

The experiments were conducted in 2.5 l glass reactors placed in temperate water bath. Water was added to the reactors according to the recipe (table 5.2). Urea (Yara, Köping, Sweden) and meal were added during constant agitation. pH was adjusted according to the experimental plan using concentrated sulphuric acid (Merck, New Jersey, USA). β-glucanase (Genencor International, Pattensen, Ger-many), β-glucanase/xylanase (Genencor International, Pattensen, Germany) and α-amylase (Genencor International, Pattensen, Germany) were added according to the recipe. Immediately after addition of enzymes countdown for incubation time in mixtank started. Countdown for incubation times in L1 and L2 started

(38)

instantly after the previous incubation time was ended, even though the tempera-ture for the next incubation was not reached. Boiling water was added to speed up warming of the water bath. When incubation in L2 was finished pH was adjusted to 3,5 with concentrated sulphuric acid to lower the activity of α-amylase. Samples were taken for determination of sugar profile, absorbance at 420 nm, fermentable and nonsolubilized starch.

Table 5.2. Mash recipe

Component Qty Water 1035 g Meal 465 g Urea 0, 2, or 4 g/kg meal β-glucanase/xylanase 60 µl /kg meal β-glucanase 50 µl/kg meal α-amylase 150 µl /kg meal

5.2 Analytical methods

5.2.1 Analysis of raw material

Starch content in meal was determined using Total Starch analysis kit (Megazyme International, Bray, Ireland) according to the standard assay procedure.

Moisture content in meal was determined using Halogen Moisture analyzers HG63 and HR73 (Mettler Toledo GmbH, Greifensee, Switzerland) at 130◦C.

5.2.2 Sugar profile

Sugar profile, i.e., content of DPn, DP3, maltose, and glucose in liquefied mash was determined using a HPLC 1100 Series (Agilent Technologies, Waldbronn, Ger-many) with a 100*7,8 mm column (BioRad Laboratories Inc., CA, USA) and 0,005 M sulphuric acid as mobile phase.

5.2.3 Maillard reaction products

To assess the formation of brown colored products descending from Maillard reac-tions, the absorbance of the supernatant from mash centrifuged at 5000 rpm for 10 minutes was measured spectrophotometrically at 420 nm against deionized water. When necessary, appropriate dilutions were made to assure an optical density less than 1.5.

(39)

5.2 Analytical methods 25

5.2.4 Quantitative analysis of fermentable and

nonsolubi-lized starch

The quantitative analysis of fermentable (i.e. solubilized starch) and nonsolubi-lized starch is divided into five steps;

Saccharification 10 g of liquefied mash was added to autoclaved 250 ml

Erlenmeyer-flask (E-Erlenmeyer-flask) and diluted with pre-heated, to avoid retrogradation of starch, deionized water to approximately 200 g. The dilution factor was calculated. A blank was also prepared using only deionized water. The E-flasks were incubated for 10 minutes in water bath at 60◦C. Glucoamylase (Novozymes A/S, Bagsvaerd, Denmark) was added and the E-flasks were incubated at shaking table in water bath for one hour at 55◦C. Samples were analyzed using HPLC, as described above, for determination of fermentable starch.

Filtration The saccharified sample was filtrated through a quantitative

filterpa-per, grade 454 (VWR International AB, Stockholm, Sweden) with vacuum aid using a Büchner funnel together with a Büchner flask. The residue was rinsed with approximately 100 ml deionized water.

Drying The residue from the filtration was dried until completely dry in 40-45◦C

Washing ~100 mg of the dried sample was weighed into test tubes. 5 ml of

ethanol was added and the contents were mixed on a vortex stirrer. After incubation in waterbath at 80◦C for 5 minutes and mixing, 5 ml additional ethanol was added and the tubes were centrifuged at 4500 rpm for 3 minutes. The supernatant were carefully poured off. The samples were rinsed with 10 ml ethanol by mixing and centrifugation as above. This procedure was repeated three times to assure that free sugars were washed off.

Hydrolysis The remaining starch in nonsolubilized condition in the washed

sam-ple were hydrolyzed and quantified using Total Starch analysis kit. The general principle for the kit is described in section 2.4 on page 11.

5.2.5 Statistical analysis of multiple factor experiments

The results from the multiple factor experiment were analyzed in Modde 7.0.0.1 (Umetrics, Umeå, Sweden). At first, non-significant factors were removed from the additive regression model:

Yresponse = β0+ β1T M ix + β2T iM + β3T L1 + β4T iL1 +

+β5T L2 + β6T iL2 + β7U + β8pH

where the βs are the regression coefficients and TMix, TiM, TL1, TiL1. TL2, TiL2, U and pH represent the coded ((-1) - 1) values of the factors. For some of the responses the experimental design were complemented with additional experiments to explore non-linear relations between factors and responses. A regression model

(40)

containing all the significant main effects, square effects and two factor interactions were then made. R2 and Q2 were calculated for each model. R2 and Q2 provide information about how well the model fits the data respective how well the model predicts new data. To check the adequacy of the models, examination of the residuals was done. The normal probability plot of the residuals was used to check the normality assumption. The residuals were also plotted against predicted value to discover any nonconstant variance. The statistical significance of the model was established by comparing the calculated F-value for the model with the listed F-value. As a rule of thumb the calculated F-value for the model should be at least 3-5 times greater than the listed F-value to be statistically significant [24]. Contour plots were then helpful in finding optimal settings for the significant factors.

(41)

Chapter 6

Results and discussion

6.1 Calculations

The results gained from the analysis of nonsolubilized starch are expressed in % (w/w) of the washed and dried mash. It is desirable to instead express the results in % of total starch. The example below show how this is done for the performed experiments.

Starch content in meal

(as determ. by total starch analysis kit): 66.9 % (d.w.)

TS in meal: 88.44%

Added meal: 465 g

Mash (at the end of exp.): 1498.8 g Fermentable starch (as glucose): 20.02 %(w/w) Non-sol. starch in sample: 3.25 %(w/w)

At first the amount of solubilized starch, i.e. fermentable starch that are washed off have to be calculated:

20.02% ·162

180 = 18.02% 1498.8g ·18.02

100 = 270.1g

18.02%(w/w) of the mash in the reactor at the end of the experiment is fer-mentable starch (as starch). The fraction 162₁₈₀ is the weight ratio of starch to glucose. Approximately 270.1 g of the added meal is washed off as fermentable starch. The equation below calculates how much meal is left as a washed and dried sample:

465g ·88.44

100 − 270.1 = 141.2g

This is the meal that are analyzed for nonsolubilized starch. The result was in this case 3.25 %(w/w) nonsolubilized starch in the sample.

(42)

141.2g ·3.25 100 = 4.6g

i.e. 4.6 g starch existed in nonsolubilized form in the reactor. Of the added starch, 4.6 g constitutes

4.6g 465g ·88.44₁₀₀ ·66.9

100

= 1.7% of the total amount of starch added to the reactor.

The percentage of solubilized/fermentable starch of total starch is 98.2%. 270.1g

465g ·88.44₁₀₀ · 66.9 100

= 98.2%

Together, solubilized and nonsolubilized starch constitutes 1.7 + 98.2 = 99.9% of the total amount of starch added. However, many of the figures used in the calculations are measured and have therefore a certain inaccuracy. The more calculations done, the larger the inaccuracy. These calculations are made for all the experiments performed, where nonsolubilized starch was determined, and are presented in appendix C. As can be seen, the two values sometimes add up to more than 100% which is a result of the approximations.

6.2 Selection of factors

The most of the chosen factors are believed to have both positive and negative effects of the process and are therefore interesting to study.

Temperatures: It was believed that higher temperatures during the upstream

process would have positive effects for the solubilization of starch. The more energy provided, the more hydrogen bonds interrupted. However, higher temperature should also lead to formation of more MRP which lowers the yield of liquefaction.

Residence times: Residence times should have effects similar to temperatures.

Longer times should provide more energy and better solubilization of starch, but also time for formation of more MRP.

pH: The pH of the liquefaction process has not been studied at Lantmännen

Agroetanol AB before. pH affect the activity of the enzymes. The pH optimum of the enzymes used is 5.4-5.8 for α-amylase, 5-6 for β-glucanase and about 4 for β-glucanase/xylanase. pH is also affecting the formation of MRP. A more acidic pH seems to lead to a lesser extent of MRP formation according to literature.

Urea addition Urea contains two amino groups which can react with reducing

sugars and form MRP. Urea is added as a nutrient to yeast and does not need to be added until after the liquefaction if there are obvious negative effects.

(43)

6.3 Statistical analysis of multiple factor experiment 29

Other possible factors to study are enzymatic doses. However, the enzymes have probably only positive effects and the larger dose the better result, until a certain point of course. Therefore the optimization potential is low. Another factor being interesting to study is particle sizes in meal. This would possibly yield positive effects for liquefaction yield. The negative effects however, occurring during recovery of ethanol would not have been seen in this study. There are certainly lots of other factors to study, however the time was limiting and more factors would require more experiments.

6.3 Statistical analysis of multiple factor

experi-ment

In the first stage, nineteen experiments were conducted, including three center point experiments, according to the experimental design generated in Modde (Umetrics, Umeå, Sweden). Separate models were constructed for each one of the response variables; DPn, DP3, maltose, glucose, fermentable starch, nonsolubilized starch and Maillard reaction products. The models were fitted with multiple linear regression (MLR). If nothing else is stated, a confidence level of 0.95 is used. The original design was in a later stage complemented with face centers for selected factors to support a non-linear model in those factors, i.e. screening fractional factorial design to responses surface modeling (RSM). Due to time limitations and for easier visualization of results, the three most important factors were chosen, TMix, TL1 and pH. See section 6.3.8 on page 43 for more information on why these three factors were chosen. In the following sections the responses are detailed for, one by one. For some of them, the extended design is also accounted for. The best models, concerning statistical significance, correlation coefficient (R2) and Q2were requested. Thereafter an overview of factors and responses follows where different parameter settings are tried out to achieve an optimal liquefaction process. The results for all responses are shown in appendix B.

6.3.1 DPn

A problem concerning statistical evaluation of DPn content is that the same mea-surement value might represent different things since dextrins of polymerization grade four and higher is included in the value. Some experimental conditions might favor formation of shorter dextrins while other conditions favor formation of dextrins of higher polymerization grade, without any change in the total content of DPn. Therefore the DPn content of two separate experiments might seem equal when they really are not. This is probably the reason why an appropriate model could not be accomplished for the DPn yield. The only significant effect found was pH. The coefficient was negative which means that a lower pH favors formation of DPn. A plausible explanation is that the pH optimum of the α-amylase used is 5.4-5.8. At more acidic pH, like 4.6 used in this study, α-amylase works slower and less formation of shorter dextrins like maltotriose, maltose and glucose occurs. A DPn content with dextrins of higher polymerization grade represents more work to

(44)

amyloglucosidase in the next step of the process while shorter dextrins represent less work.

A confidence level of 0.9 (instead of 0.95) revealed temperature in mixtank and temperature in L1 as statistically significant main effects besides from pH. The design was complemented with additional experiments at the face centers for TMix, TL1 and pH. With the additional experiments the three main effects TMix, TL1 and pH together with the interaction effect TMix*pH were significant at confidence level 0.95. Experiment C24, which had an abnormally low value was excluded from the analysis. The coefficient plot is shown in fig. 6.1.

Figure 6.1. Coefficient plot for DPn.

The regression model of DPn yield on the significant factor is: YDP n= β0+ β1T M ix + β3T L1 + β8pH + β18T M ix∗pH

where the β’s are the regression coefficients with following values

β0= 9.3686 β1= −0.5254 β3= 0.5823 β8= −0.9756 β18= 0.4940

and TMix, TL1 and pH represent coded values of the factors.

In table 6.1 the analysis of variance (ANOVA) for DPn is shown. According to the F-test the model cannot be considered statistically significant. The calculated F-value, 10.27 is only three times the listed F-value, F4,19 = 2.90 (95%). The R2

(45)

Table 6.1. Analysis of variance for DPn. SS = Sum of Squares, MS = Mean Square, DF = Degrees of Freedom

Source of variation SS MS DF F-value

Regression 31.12 7.78 4 10.27 Residual 14.39 0.76 19 Total 45.51 1.98 23 R2 _0.68 Q2 _0.42 F listed value F4,19= 2.90(95%)

predicted value (plots not shown) did not imply any violations of the normality or equal variance assumptions. The model is though poor and not useful for prediction of new data, which can be seen on the poor value of Q2. Though the coefficients imply that lower temperature in mixtank, higher temperature in L1 and lower pH give the highest concentration of DPn in mash.

The interaction plot for temperature in mixtank and pH (fig. 6.2) shows that at pH 4.6 the temperature in mixtank influence the formation of DPn while the production of DPn at pH 5.6 only has slightly variations over the temperature range studied. However, at pH 5.6 the DPn content always underpass the DPn content at pH 4.6.

(46)

6.3.2 DP3

For the DP3 response four main effects were clearly significant; TMix, TiM, TL1, and pH. Coefficients are positive for TMix and TiM and negative for TL1 and pH which means that higher temperature and longer time in mixtank together with lower temperature in L1 and more acidic pH give the highest concentration of DP3 in liquefied mash. At significance level 0.90 TL2 is also significant.

The experimental design was complemented with additional experiments at the face centers for TMix, TL1 and pH. Except for the four significant main effects previously mentioned the main effect TL2 and two two-interaction effects were significant with the new data included; TMix*pH and TiM*pH. The coefficient plot is shown in fig. 6.3.

Figure 6.3. Coefficient plot for DP3.

The regression model obtained for yield of DP3 as a function of the significant variables is as follows:

YDP 3 = β0+ β1T M ix + β2T iM + β3T L1 + β5T L2 +

+β8pH + β18T M ix∗pH + β28T iM∗pH

where the β’s are the regression coefficients with following values β0= 2.1034 β1= 0.1318 β2= −0.1113 β3= −0.1337

(47)

and TMix, Tim, TL1, TL2 and pH represent the coded values of the factors. Table 6.2 depicts the analysis of variance for DP3. The values of R2 and Q2 are intermediate. According to the F-test the model is statistically significant.

Table 6.2. Analysis of variance for DP3.

Source of variation SS MS DF F-value

Regression 1.68 0.24 7 17.82 Residual 0.23 0.01 17 Total 1.91 0.08 24 R2 _0.88 Q2 _0.71 F listed value F7,17= 2.61(95%)

In fig. 6.4 the normal probability plot of the residuals is shown. The plot is satisfying. The plot of residuals versus predicted value does not show any particular pattern and cause no concerns (plot not shown).

Figure 6.4. Normal probability plot of the residuals for the model of YDP 3.

6.3.3 Maltose

The proceeding for the maltose response was the same as for DP3 and DPn. The insignificant effects in an additive model including all the eight main effects were excluded. The same factors as in the DP3 response turned out to be sig-nificant, TMix, TiM, TL1 and pH. TiM was just about to be significant with p-value 0.048 (<0.05). This design was also complemented with the additional face centered points for TMix, TL1 and pH. The main effect TiL2, the interaction

(48)

term TMix*pH, the square terms TMix*TMix and pH*pH were revealed as sig-nificant effects with the additional data included as shown in the coefficient plot for maltose, fig. 6.5

Figure 6.5. Coefficient plot for maltose.

The regression model obtained for yield of maltose as a function of the signifi-cant variables is as follows:

YM altose = β0+ β1T M ix + β2T iM + β3T L1 + β6T iL2 +

+β8pH + β11T M ix2+ β88pH2+ β18T M ix∗pH

where

β0= 10.1107 β1= 0.4567 β2= 0.3807 β3= −0.4282

β6= 0.2844 β8= 1.3215 β11= 0.6384 β88= −1.0945 β18= −0.4019

and TMix, Tim, TL1 and pH represent coded values of the factors.

The ANOVA table for maltose is shown in table 6.3. According to the F-test the model can be considered statistically significant, for 95% of confidence since the calculated F-value, 32.31 is several times greater than the listed value, F8,16 = 2.59. The values of R2and Q2 are acceptable.

The normal probability plot of the residuals is shown in fig. 6.6. It is satisfac-tory though experiment 1 and 7 is slightly deviating from the rest.

Optimization of the Liquefaction Process in Bioethanol Production & Development of Method for Quantification of Nonsolubilized Starch in Mash

Department of Physics and Measurement Technology

Master’s Thesis

Optimization of the Liquefaction Process in

Bioethanol Production & Development of Method

for Quantification of Nonsolubilized Starch in Mash

Anna Aldén

Optimization of the Liquefaction Process in

Bioethanol Production & Development of Method

for Quantification of Nonsolubilized Starch in Mash

Anna Aldén

Abstract

Sammanfattning

Acknowledgments

Nomenclature

Contents

Chapter 1

Introduction

1.1

Background

1.2

Thesis objectives

1.3

Method

1.4

Outline

Chapter 2

Theoretical background

2.1

Starch

2.1.1

Starch granules

2.2

Upstream processing in bioethanol

produc-tion

2.2.1

Milling

2.2.2

Mixing

2.2.3

Gelatinization

2.2.4

Liquefaction

2.2.5

Saccharification

2.3

Factors contributing to lower starch yield

2.3.1

Retrogradation

2.3.2

Maillard Reactions

2.3.3

Nonsolubilized starch

2.4

Quantification of starch

Chapter 3

Bioethanol production at

Lantmännen Agroetanol AB

3.1

An overview of the production process at

Lant-männen Agroetanol AB

3.2

Properties of enzymes used in the upstream

process

3.2.1

α-amylase

3.2.2

β-glucanase

3.2.3

β-glucanase/xylanase

Chapter 4

Method for determination of

nonsolubilized starch in

mash

4.1

General principle of method

4.2

Sample preparation

4.2.1

Saccharification of solubilized starch