A Molecular Dynamics Study on Cellulose Fiber Reinforced Nanocomposite

MISSING IMAGE

Material Information

Title:
A Molecular Dynamics Study on Cellulose Fiber Reinforced Nanocomposite
Physical Description:
1 online resource (67 p.)
Language:
english
Creator:
Li,Shi
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Mechanical Engineering, Mechanical and Aerospace Engineering
Committee Chair:
Chen, Youping
Committee Members:
Ifju, Peter
Peter, Gary F

Subjects

Subjects / Keywords:
celluloe -- hemicellulose -- md -- mfa
Mechanical and Aerospace Engineering -- Dissertations, Academic -- UF
Genre:
Mechanical Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Wood has a hierarchical structure, whose macroscopic properties emerge from their micro- and nanostructural level. In micro- or nano-scale, the mechanical properties of cell wall depend on its molecular interaction. However, due to the complexity of wood cell wall, it is difficult to study the interaction between the components within the cell wall by micro scale experimental sample. Thus, nano-scale simulation is a suitable way for us to study the mechanical properties of wood cell wall. The aim of this study is to use molecular dynamics (MD) simulation for elucidating the interactions between cellulose and hemicellulose in the secondary cell wall and microfibril angle (MFA) effect on the properties of fiber-reinforced composites. The models of cellulose microfibril bundle and hemicellulose matrices were obtained by MD software AMBER. Then topology and force field parameters of these models are transmitted to another MD software LAMMPS. The cellulose microfibril bundle is pulled away from three kinds of hemicellulose matrices. It is found that not only the area but also the shape of hemicellulose chain could affect the interaction between cellulose and hemicellulose. Furthermore, a compression loading has been applied on the systems with different MFAs. The MFA effect on the mechanical properties and the failure mechanism of the systems has also been studied.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Shi Li.
Thesis:
Thesis (M.S.)--University of Florida, 2011.
Local:
Adviser: Chen, Youping.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2011
System ID:
UFE0043441:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 A MOLECULAR DYNAMICS STUDY ON CELLULOSE FIBER REINFORCED NANOCOMPOSITE By SHI LI A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Shi Li

PAGE 3

3 To my father and mother who have always supported me in all that I have done

PAGE 4

4 ACKNOWLEDGMENTS I would first and foremost like to thank my graduate advisor Dr. Youping Chen for supporting me with a funded and sponsored project and for her valuable insights and patience with we faced our challenges. Special thanks to my graduate committee: Dr. Peter Ifju and Dr. Gary Peter for their willingnes s to review my work. Thanks also to my fellow PhD students and friends: Qian Deng, Liming Xiong, Ning Zhang, Shengfeng Yang and Yongtao Li in Atomistic and Multiscale Mechanics group at the University of Florida for their invaluable comments during my time as part of the group. Extra thanks go to Qian Deng for his guidance and comments during the research. Special thanks to Brad LaCroix for his willingness to revise my thesis. My utmost gr atitude is to my parents. Thanks for their supporting and encouraging me throughout college and graduate school.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 ABSTRACT ................................ ................................ ................................ ................... 10 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 12 1.1 Wood Components ................................ ................................ ........................... 12 1.1.1 Cellulose ................................ ................................ ................................ .. 12 1.1.2 Hemicellulose ................................ ................................ .......................... 13 1.1.3 Lignin ................................ ................................ ................................ ....... 14 1.2 Cell Wall ................................ ................................ ................................ ............ 15 1.3 Software ................................ ................................ ................................ ............ 18 1.3.1 Gaussian 0 3 ................................ ................................ ............................ 18 1.3.2 AMBER ................................ ................................ ................................ .... 18 1.3.3 LAMMPS ................................ ................................ ................................ 19 1.4 Scope of This Thesis ................................ ................................ ........................ 20 2 CELLULOSE MICROFIBRILS MODELING ................................ ............................ 21 2.1 Structure of Cellulose Microfibril ................................ ................................ ....... 21 2.2 Building Cellulose Microfibril in AMBER ................................ ............................ 24 2.2.1 Force Field GLYCAM_06 ................................ ................................ ........ 24 2.2.2 Original Model of Cellulose Microfibril ................................ ..................... 24 3 HEMICELLULOSE CHAIN MODELING ................................ ................................ .. 29 3.1 Introduction to the Structure of 4 O methylglucuronoxylans ............................. 29 3.2 Build Model with AMBER ................................ ................................ .................. 30 4 DATA TRANSFERRING FROM AMBER TO LAMMPS ................................ .......... 33 5 INTERFACIAL INTERACTIONS ................................ ................................ ............. 40 5.1 Molecular Structure of the Cellulose Hemicellulose System ............................. 40 5.2 Relaxation of the Whole System ................................ ................................ ....... 43 5.3 Applying Velocity with Water Effect ................................ ................................ ... 45 5.4 Applying Incremental Increasing Loads with Water Effect ................................ 46 5.5 Applying Incremental Increasing Loads without Water Effect ........................... 48

PAGE 6

6 5.6 Molecular Simulation Applying Loads in Two opposite Directions on Z Model ................................ ................................ ................................ ................ 49 6 MICROFIBRIL ANGLE (MFA) EFFECT ................................ ................................ .. 55 6.1 Introduction to Microfibril Angle (MFA) ................................ .............................. 55 6.2 MFA Effect ................................ ................................ ................................ ........ 55 7 CONCLUSIONS ................................ ................................ ................................ ..... 62 LIST OF REFERE NCES ................................ ................................ ............................... 64 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 67

PAGE 7

7 LIST OF TABLES Table page 4 1 Static energy analysis of a representative structure of the cellulose microfibril using both AMBER11 and LAMMPS ................................ ................. 37 6 1 Mechanical property parameter of six models with vario us MFA ........................ 60

PAGE 8

8 L IST OF FIGURES Figure page 1 1 The structure of cellulose ................................ ................................ ................... 12 1 2 Diagrammatic representation of wood secondary cell wall ................................ 16 2 1 The network of cellulose molecu les connected by hydrogen bonds ................... 21 2 2 stal structure along the c axis ................................ ......... 22 2 3 ................................ ................... 23 2 4 Cellulose unit conformation. Atomic labeling and torsion angle parameters of primary interest are denote d ................................ ................................ ............... 25 2 5 Two pairs of cellulose unit (4G B residue) in a cellulose chain ............................ 26 2 6 The model of cellulose microfibril bundle ................................ ............................ 28 3 1 A theoretical structural model of 4 O methylglucuronoxylan .............................. 30 3 2 Schematic representat ion of 4 O methylgluc uronoxylan ................................ .... 32 4 1 Temperature variation during a 150,0 00 timestep relaxation ............................. 38 4 2 Total energy variation durin g a 150,000 timestep relaxation .............................. 39 5 1 A cross sectional s ide view of a one layer z model ................................ ............ 41 5 2 A cross sectional s ide view of a one layer u model ................................ ............ 41 5 3 A cross sectional side view of a one layer r model ................................ ............. 42 5 4 Temperature variation during a 100,000 time step relaxation ............................. 44 5 5 Total potential energy variation during t he 100,000 time step relaxation ............ 44 5 6 Force versus time curves of the z model upper cellulose microfibril. .................. 45 5 7 Force versus time curves of the u mo del upper cellulose microfibri l .................. 46 5 8 Force versus time curves of the upper cellulose microfibrils with water ............. 47 5 9 Force versus time curves of the upper cellu lose microfibrils without water ........ 49 5 10 Force diagram of the positive x axial loading. ................................ ..................... 50

PAGE 9

9 5 11 Force diagram of the negative x axial loading ................................ .................... 50 5 12 Force displacement curve of the posi tive x axia l loading simulation .................. 51 5 13 Force displacement curve of the nega tive x axial loading simulation ................. 51 5 14 A cross sectional view of a one layer z model relaxed structure ........................ 52 5 15 A cross sectional view of one layer structure when the cellulose microfibril is released from hemicellulose matrices und er the positive x axial loading ........... 52 5 16 A cross sectional view of one layer structure when the cellulose microfibril is released from hemicellulose matrices und er the negative x axial loading .......... 53 6 1 Cross sectional views of one layer in six models. The MFA are 0 , 5 10 15 20 and 25 ................................ ................................ ................................ 56 6 2 Cross sectional views of the six on e layer models after relaxation ..................... 57 6 3 Compression process of the model with 0 MFA ................................ ................ 58 6 4 Compression process of the model with 5 M FA ................................ ................ 58 6 5 Compression process of the model with 10 MFA ................................ .............. 58 6 6 Compression process of the model with 15 MFA ................................ .............. 59 6 7 Compression process of the model with 20 MFA ................................ .............. 59 6 8 Compression process of the model with 25 MFA ................................ .............. 59 6 9 Stress strain curves of six models with various MF A under th e same loading rate compression ................................ ................................ ................................ 60

PAGE 10

10 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science A MOLECUL AR DYNAMICS STUDY ON CELLULOSE FIBER REINFORCED NANOCOMPOSITE By Shi Li August 2011 Chair: Youping Chen Major: Mechanical Engineering Wood has a hierarchical structure, whose macroscopic properties emerge from their micro and nanostructural level. In micro or nano scale, the mechanical properties of cell wall depend on its molecular interaction. However, due to the complexity of wood cell wall, it is difficult to study the interaction between the components within the cell wall by micro scale experime ntal sample. Thus, nano scale simulation is a suitable way for us to study the mechanical properties of wood cell wall. The aim of this study is to use molecular dynamics (MD) simulation for elucidating the interactions between cellulose and hemicellulose in the secondary cell wall and microfibril angle (MFA) ef fect on the properties of fiber reinforced composites The models of cellulose microfibril bundle and hemicellulose matrices were obtained by MD software AMBER. Then topology and force field paramete rs of these models are transmitted to another MD software LAMMPS. The cellulose microfibril bundle is pulled away from three kinds of hemicellulose matrices. It is found that not only the area but also the shape of hemicellulose chain could affect the inte raction between cellulose and hemicellulose. Furthermore, a compression loading has been

PAGE 11

11 applied on the systems with different MFAs. The MFA effect on the mechanical properties and the failure mechanism of the systems has also been studied.

PAGE 12

12 CHAPTER 1 INTRODUCTION 1.1 Wood C omponents Wood is composed of cellulose microfibrils (40% 50%), hemicelluloses (15% 25%) amorphous lignin (15% 30%) and extractives. Additionally, structural proteins (1% 5%) can be found in most plant cell walls. Besides, nu merous enzymes are also contained in the plant cell walls. Plant cell walls also contain numerous enzymes, which do not affect the mechanics of structure. 1.1.1 Cellulose Cellulose is the major chemical component of the cell wall of wood. About 40% 50% o f all plant matter is cellulose. It is a polysaccharide composed of linear chains of D glucose units. The degree of polymerization can be from several hundred to over ten thousand. These D glucose units are l 1,4 glycosidic bonds. Figure 1 1. T he structure of cellulose The formula of the organic compound of cellulose is (C 6 H 10 O 5 ) n Due to the effect of hydroxyl groups at C2, C3, and C6 positions of the D glucose unit, cellulose has a

PAGE 13

13 strong tendency to form intra and inter molecular hydrogen bo nds on the linear cellulose chains. The net of hydrogen bonds among the cellulose chains can stiffen the straight chain. Furthermore, the effects of hydrogen bonds can promote aggregation into a crystalline structure. Native cellulose, which is also called cellulose I, is known to have two distinct s from different origins. Both of these crystal allomorphs exist in various proportions in different materials. Generally, algae and bacteria are rich while higher plant life like cotton and wood, are rich crystalline allomorph with disordered chains is associated with its cryst allite surface in higher plants (Ding and Himmel 2006 ) Due to restricted access to water and chemicals, the chemical attack can be expected to occur on the crystalline cellulose like form of cellulose is one of the objects this paper concerns. Due to the activity of the hydroxyl groups, cellulose can provi de large amount s of derivatives with useful properties by reacting with various chemicals. The most important commercial materials of cellulosic derivatives are cellulose esters and cellulose ethers. 1.1.2 Hemicellulose Hemicellulose is a polysaccharide r elated to cellulose. Generally, it comprises about 20% of the mass of the plant. The hemicellulose can be any of several heteropolymers (matrix polysaccharides). It exists in almost all plant cell walls, usually along the cellulose chains. Even though, the components of cellulose and hemicellulose

PAGE 14

14 are very similar, they ha ve different properties because the structure is quite different. Cellulose is crystalline, so that it is strong and tough. However, hemicellulose is quite flexible with a random and amorp hous structure. Furthermore, cellulose always contains only anhydrous glucose, but hemicellulose can contain varying amount s of sugar monomers i ncluding glucose, xylose, mannose, galactose, rhamnose and arabinose which can also be found in heteropolyers. In addition most o f the D pentose sugar and some very small amounts of L sugars can be found in hemicellulose. Usually, xylose contains mostly hemicellulose in sugar monomers. However, mannosic and galactosic acid are also sometimes found to be comprise d of hemicellulose. In almost all the plant cell walls, hemicelluloses are embedded among cellulose microfibril bundles They are in the form of chains, which form a network of cross linked fibers to connect cellulose and pectin. Though cellulose and hemic ellulose both exist as a chain in plant cell walls, the length of the chains are quite different. In cellulose, about 10,000 to 15,000 glucose monomers comprise a polymer. But hemicellulose chain s consist of only about 200 sugar monomers. Furthermore, some branches of sugar monomers exist on the main chain of hemicellulose, but cellulose is not branched. Xylan, glucuronoxylan, arabinoxylan, glucomannan, and xyloglucan are some common hemicellulose found in plant cell wall s 1.1.3 Lignin Lignin is another ke y component of the plant cell walls. Lignin makes up about one quarter to one third of the dry mass of the wood. It fills the space between cellulose, hemicellulose and other components of the cell wall and then forms covalent bonds to different polysaccha ride. Thus, lignin also connects cellulose, hemicellulose

PAGE 15

15 and pectin. The mechanical strength of the cell wall and even that of the whole plant will be enhanced because of the effect of lignin. The molecular mass of lignin can be more than 10,000 Da (Da: unified atomic mass unit Dalton) Unlike cellulose and hemicellulose, lignin is so heterogeneous that the primary structure of this kind of biosynthesis could not be defined The degree of polymerization of lignin is quite difficult to measure. The lignin consists of various kinds of substructures that randomly bind with each other. Due to its structural complexity, there is not a well defined or standard structure for us to build the model. Therefore this study was focus ed on the fiber reinforced cellulose hemicellulose system. 1.2 Cell W all The plant cell wall is a tough and relatively rigid structure that surrounds the cell. Its thickness can be up to several micrometers. Because of its toughness and rigidity the plant cell walls provide a defined shape for the cells and protect the cell itself. None of the cell membrane can be comparable in strength to the plant cell wall. The cell wall is the main difference between animal cell s and plant cell s The c ell wall acts as a filtering mechanis m and the major function of the cell wall is to be like a pressure vessel, offering protection against mechanical stress. The composition and properties of the cell wall can be varying depending on the growth conditions and biological samples There are up to three kinds of different layers in plant cell walls: primary cell wall, secondary cell wall, and middle lamella.

PAGE 16

16 Figure 1 2. Diagrammatic representation of wood secondary cell wall The cell walls in young cells are thin and flexible which can allo w the cell to grow and expand. The first cell wall of these growing cells is called the primary cell wall. Though it should be flexible enough for growth it should also be sufficiently strong to protect the cell itself. Several interconnected polysacchari de matrices and a small amount of proteins are contained in the primary cell wall. The matrices consist of cellulose, hemicellulose, and pectins. Xyloglucans and galactoglucomannan are the common components of the hemicellulose chains, while the homogalact uronan and rhamnogalacturonan are often observed to form the pectin matrices. When the cell is fully grown, the primary cell wall may stay alive, and another new layer may be deposited by the cell. This new layer which is called the secondary cell wall is much thicker than the primary layer. Not all cells contain the secondary cell wall, but the secondary cell wall exists in almost all of the xylem. Though it is can be found anywhere in the plant, the word itself me known xyl em

PAGE 17

17 tissue. Because of the secondary cell wall, wood plants can hold themselves up against gravity. Like primary cell wall, the secondary cell wall also consists of cellulose, hemicellulose and lignin. But the ratios of these components are quite different from that in primary cell wall: instead of xyloglucans, xylan takes more weight and more cellulose exists. Besides, pectins and structural proteins can be absent from the secondary cell wall. Recently, the lignifications of the secondary cell wall was cons idered to be a key reason for the compressive strength, while the cellulose microfibrils bring the tensile strength to the cell wall, even to the plant. The structure of the secondary cell wall could be imagined as the same as reinforced concrete: cellulos e microfibrils act like the reinforcing steel bar, which give the whole structure tensile strength; and the he micellulose lignin matrices act as the concrete. Another important interest of the secondary wall is its three distinct layers S1, S2, and S3. I n those three layers, the cellulose microfibrils angles (MFA) are varying, so that the strength of S1, S2, and S3 could be different. Among these three layers, the thickness of the S2 layer is much greater than that of the S1 and S3 layers (Barnett and Bon ham, 2004) This means that for the measurement of secondary wood cell wall, the results can be approximated to the mechanical properties of S2 layer. The middle lamella, which connects the cell walls of adjacent plant cells together, is a layer full of pe ctins. The outmost layer is very similar to the primary wall, especially when the plant cell has a thick secondary wall. Thus, the compound middle lamella consists of two primary cell walls and middle lamella, and sometimes the secondary wall of two adjace nt cells. This layer also provides stability to the plant.

PAGE 18

18 1.3 Software 1.3.1 Gaussian 03 Gaussian is a series of electronic structure packages, which provides state of the art capabilities for predicting all the scientific and modeling features of atoms, molecules, and reactive systems. Without any artificial limitations on calculations, Gauss ian can predict the energy, molecular structures, and vibration frequencies of molecular system, by utilizing ab initio, density functional theory, semi empirical, molecular mechanics, and hybrid methods. Molecular dynamics is often used as a tool for com putational modeling for large scale biomolecular simulation. The length scale for biomolecular simulation is rather small, ranging from 10 10 m to 10 8 m. Similarly, the time scales of biomaterial motions are from 10 15 to 10 3 seconds. The length scales i mplemented in molecular dynamics can fully satisfy the requirements of biomolecular simulation; however, the state of art simulation could only do its best to get to 10 8 seconds, due to the limit of the efficiency of parallel computers. Although a longer time scale is better, the current MD techniques are sufficient to figure out the internal dynamics within biomolecules. 1.3.2 AMBER AMBER is a package of programs that can be used to run molecular dynamics simulations, particularly on bio molecules, such a s proteins, nucleic acids, polysaccharides etc. mber is also the name of one of the most popular empirical force fields. However, in this software, amber force field is combined with other force fields which can be implemented. As a molecular dynamics (MD) simulation package AMBER allows users to build their own model by using some basic residue models and manipulating molecules. Also,

PAGE 19

19 in order to build some unusual molecular, the data generated from experiences or other simulation software can be imp orted into AMBER The main program, sander, allows for NMR refinement based on NOE derived distance restraints, torsion angle restraints, and penalty functions based on chemical shifts and NOESY volumes. The molecular simulation is carried out by sander. M oreover, AMBER includes some post process programs to analyze MD trajectories. A variety of calculations can be computed by AMBER for example, root mean square deviation (RMSD) from a reference structure, time correlation functions, hydrogen bonding analy sis, solvent mediated energy of biomolecules, and so on. 1.3.3 LAMMPS LAMMPS is an acronym for Large scale Atom Molecular Massively Parallel Simulations. Unlike AMBER which concentrate s on bio molecules, this software can model most systems, like biological material, metal, and coarse grained systems. It contains almost nearly all force fields and boundary conditions. Due to its great efficiency, a simulation with millions or billions of particles can be performed by LAMMPS. Because it is a freely available open source code, users can modify the programs as necessary And users are encouraged to upload their modifications or updated programs to LAMMPS, which in turn has made LAMMPS a fr equently updated. The number of pages in the manual has increas ed from less than 300 to more than 600, which supports the notion that LAMMPS is one of the most popular MD simulation software. But LAMMPS cannot do everything. The molecule structure can not be modeled and sophisticated analyses shoul d not be carried out by the software itself. Furthermore, LAMMPS provides many force field formats, but the parameters of these

PAGE 20

20 force fields should be input by users themselves. Thus, in this thesis, the three kinds of software are used. 1.4 Scope of T his T hesis The goal of this research is to develop a model for the secondary cell wall of wood through computational method. With structural and chemical composition information obtained from existing publications, cellulose microfibrils and hemicellulose matr ices, which are the main components in wood cell wall, are obtained by using existing well tested molecular dynamics simulation codes. Based on the models of wood structural parts, a cellulose fiber reinforced nanocomposite model is developed to identify t he mechanical properties and recognize the failure mechanism of it. Chapter 2 of this thesis is devoted to the details of building the model of cellulose microfibril and hemicellulose chain. The model in Chapter 2 is built by using AMBER, which is a type o f software suitable for chemical and biological engineering. To test the mechanical properties of the nanocomposite, in Chapter 3, data obtained from AMBER is transferred to another MD code LAMMPS, which is more professional and efficient. A shear loading is applied on cellulose fiber reinforced nanocomposite to recognize the mechanism of interfacial failure between cellulose microfibrils and hemicellulose matrices in Chapter 4. Chapter 5 is devoted to the mechanical properties of the nanocomposite and the relevance of the microfibril angle (MFA).

PAGE 21

21 CHAPTER 2 CELLULOSE MICROFIBRI LS MODELING 2.1 Structure of C ellulose M icrofibril Cellulose is linear chain polymer. It is so straight that no branching or coiling happens on the molecule. The structure of cellulo se is a stiff rod like conformation, which is composed by multiple glucose residues. The hydroxyl groups on the glucose from one chain could form hydrogen bonds with the oxygen molecules on itself or on another neighbor chain. So hydrogen bonds are the maj or factor to hold the chains of cellulose firmly together. Figure 2 1 shows the conformation of the network of hydrogen bonds. Figure 2 1. The network of cellulose molecules connected by hydrogen bonds Due to the effect of the network of hydrogen bonds, several chains will aggregate together to form a cellulose elementary fibrils. The elementary fibrils were first observed by X ray scattering studies, and were cellulose microfibrils depends on various bio mater ials. Ding and Himmel proposed a

PAGE 22

22 model of cellulose microfibril that is composed of 36 chains. This model is derived from direct visualization, so we tried to make a 36 glucan chains model of cellulose microfibril. Natural cellulose microfibril is known to chain chain monoclinic unit cell. In solid state, it is observed chapter, the phase cellulose is the predominant component of higher plants, like wood By using some advanced experimental tec hnology, like X ray diffraction and NMR related researches. The unit cell constant for cellulose is: = 96.5, a = 7.784 , b = 8.201 , c = 10.380 (Nishiyama et al ., 2002). The crystal structure of cellulose is shown in Figure 2 2 and Figure 2 3. Figure 2 axis

PAGE 23

23 Figure 2 3. View of cellulose crystal structure obliquely

PAGE 24

24 2.2 Building C ellulose M icrofibril in AMBER 2.2.1 Force F ield GLYCAM_06 In molecular dynamics simulation, force field is a key factor to determine whether the simulation is reasonable In AMBER, the choice of force field becomes more important, because it may determine the difficulty in buildi ng the model of molecules. Different kinds of force fields will contain different series of basic residues. For example, the one I used for this thesis is call GLYCAM. GLYCAM comes from the initial letters of Glycosides and glycoproteins with AMBER. This force field is designed to build oligosaccharides and to simulate almost all kinds of carbohydrate s All pertinent force field terms in this force field, have been explicitly specified a nd so no default or generic parameters are employed. The GLYCAM force field parameters are developed by quantum mechanics. Free energy perturbation and direct deltaG calculations are used to study the structure or sequence of the carbohydrate. Some experim ental techniques, such as X ray diffraction, NMR spectroscopy, and mass spectrometry are employed to combine with the computational methods. In GLYCAM_06 force field, which is the most updated edition, glucose is one of the standard residues, so the model of cellulose microfibril bundle is based on the glucose residue. 2.2.2 Original M odel of C ellulose M icrofibril >4) linked D configuration rotamer, 1 and 4 means the linkage positions are C 1 and C 4 respectively. Figure 2 4 shows the atomic labeling and the torsion angles parameters of the cellulose chain conformation. C1 O4 O4

PAGE 25

25 C4 C5 C6 C2 O2 C3 O3 H3O, C6 O6 H 6 O. Figure 2 4. Cell ul ose unit conformation. Atomic labeling and torsion angle parameters of primary interest are denoted By employing the package named Leap in AMBER, one chain of D glucose can be built straight based on the parameters fi rst discovered by Nishiyama et al. (2002) Figure 2 5 shows a part of cellulose chain. I nitially, a 21 nm length chain of cellulose was built 41 residues are in the chain: the residue on the head of the chain is ROH ( OH), the one at the end of the chain is 0GB, and the other one s in the middle are 4GB. It means that the lack of one hydrogen atom is on those two atoms, so that those two atoms could connect to other residues.

PAGE 26

26 Figu re 2 5. Two pairs of cellulose unit (4GB residue) in a cellulose chain During the modeling, another difficulty is to rotate the chain of cellulose to be parallel to the x axis. Although it is an original structure, it is difficult to determine the axis of the cellulose chain. When the chain is long enough though the chain become s a helix. I t means that if all the C1 and C2 atoms are picked out from one chain, the line composed by C1 will not be in the same direction of that composed by C2. T o determine th e axis of the cellulose chain, the first and the last GB residues are selected and the branches of the rings that are composed by atoms C6 and O6 are ignored The ring contains the atoms C1, C2, C3, C4, C5 and O5. After that, the average coordinates of So the line that goes across the two points is considered the axis of the cellulose chain. T he steps for rotation are as follows Suppose a vector (a, b, c) represents the axis, and our goal is to rotate it to parallel to the x axis. At first, rotate the vector about the z axis so that the vector lies in the xz plane . T he matrices are rotated axis, y a xis and z axis which is shown in Eq. 2 1, Eq.2 2, and Eq .2 3, respectively:

PAGE 27

27 (2 1) (2 2) (2 3) So the matrix to rotate the vector (a, b, c) about z axis to xz plane is Eq. 2 4 : (2 4) Then rotate the new vector R 1 *(a, b, c, 1) T in the xz plane to the x axis: (2 5) Thus, by using rotation matrix, the cellulose chain generated by AMBER can be rotated to x axis. chain monoclinic unit structure. And from synchrotron X ray and neutron fiber diffraction, the parameters of the unit cell have been determined by Nishiyama et al. (2002) In another one of his papers (Nishiyama et al. 2008), the parameters between two cellulose units are mentioned as C1 O4 C4 = 98.5 O4 C4 C5 = 142.3

PAGE 28

28 The refore, the whole st ructure of a 36 chain cellulose microfibril can be obtained by incorporating the same chain structure and arranging it according to the unit cell parameters. The model of our 36 chain cellulose microfibril bundle is shown in Figure 2 6. Figure 2 6. The model of cellulose microfibril bundle

PAGE 29

29 CHAPTER 3 HEMICELLULOSE CHAIN MODELING 3.1 Introduction to the S tructure of 4 O methylglucuronoxylans Hemicellulose can be composed of various sugar monomers, such as glucose, xylose, mannose, and so on. Tracheids and xylem fib ers are the main cell types in softwood (gymnosperms) and hardwoods (eudictyledons), respectively, and their walls are predominately secondary. The major non cellu losic polysaccharides in softwoods are (galacto) glucomannans, together with small amounts of 4 O methylglucuronoarabinoxylans. In the cell wall of hard wood, O acetyl (4 O methylglucurono) xylan, sometimes is also called as acetylated 4 O methylglucuronox ylan, is the major hemicellulose component. More precisely, the name of this kind of hemicellulose c an be written as 4 O methyl D glucuronopyranosyl uronic acid xylan. The weight ratio of 4 O methylglucuronoxylan c an be up to 20% of NaClO 2 delignified wo od. The most representative structure of this polysaccharide is a >4)Xyl p units of which a few unevenly distributed residues are substituted at C 2 by 4 O methylglucuronic acid (Moine and Krausz, 2006) ( 2007) the ratio of Xylans to methylglucuronic acid ranges from 4:1 to 16:1. The degrees of polymerization of this kind of hemicellulose range from 35 to 230. Then the relative amounts of xylans and 4 O methyl D Glycuronic acid were determined by integra tion of the corresponding anomeric protons, and the ratio of Xylans to methylglucuronic acid was subsequently calculated. Integration results gave an approximate value of 5.9:1, which is in agreement with gas chromatography (GC) analysis. On the basis of e xperimental data obtained from GC

PAGE 30

30 and NMR analyses, they proposed a theoretical structure model of 4 O methylglucuronoxylan: Figure 3 1. A theoretical structural model of 4 O methylglucuronoxylan 1 >4) linked xylop yranosyl residues are arranged linearly to compose the backbone. A 4 O methylglucuronic acid is substituted at C 2 of one of the six xylan residues, so that the ratio of Xylans to methylglucuronic acid becomes 6:1, very close to the approximate value 5.9:1 As mentioned above, the main components of 4 O methylglucuronoxylans are three residues: methyl, glucuronic acid and xylan. 3.2 Build M odel with AMBER It was found that the use of the AMBER force field in conjunction with the GLYCAM parameter set was a very effective method for modeling carbohydrates. However, in GLYCAM_06, which is the most updated force field of GLYCAM series, methyl is not included as one of the group of standard carbohydrate residues. So it is necessary to build a non standard residu e of methyl for our following molecular dynamics simulation with GLYCAM parameter set. A residue in AMBER should include the following information: topology, atom type, atom charge, bond length, bond angle, torsion angle and dihedral angle. Some of the inf ormation can be determined from experiment al data and others can be generated by various empirical functions and programs. The topology is always input by users,

PAGE 31

31 because the data comes from experiments. Once the topology is obtained, it was put into Gaussi an 03 to get a more reasonable, sometimes more stable electronic structure. Gaussian can also provide the charge distribution of each atom in the molecule. The charge distribution is the most important input when building a non standard residue, because it can determine the atom type in AMBER. For instance, two atoms may be the same elements, but they may have different orbital hybridization since the charges on the two atoms are not the same, these two atoms belong to two types of atoms, respectively. Acco rding to the atom types, bond length and bond angle was calculated automatically in LEaP, which is one important package for building the model in AMBER, by employing the force field parameter set. Torsion angle is not fixed, so this data needs to be obtai ned by using NMR spectroscopy. At last, dihedral angles ought to be provided by users. The advantage of Gaussian series of software is that all versions of Gaussian contain every scientific or modeling feature, and none impose an artificial limitation on c methyl and glucuronic acid are considered as a whole non standard residue, so Gaussian 03 was used for obtaining the electronic structure and charge distribution of the whole molecule. Another way to build a hemicellulose model is to take the methyl as a molecule, and then add it to glucuronic acid. Due to lack of information, a single methyl residue was not calculated. First, methyl is so small that we cannot determine its ch arge distribution when it connects to glucuronic acid and is affected by the electronic cloud of glucuronic acid. Second, even though the charge of methyl is determined, say +1 or 1,

PAGE 32

32 the angles of H C H may be different from the realistic situation. Furth ermore, even if the accurate structure of a methyl residue was calculated from Gaussian, it is still difficult to connect methyl to glucuronic acid. In order to connect two residues, the data of dihedral angles that is determined by four atoms from these t wo residues is necessary. Unfortunately, the dihedral angles cannot be readily determined. Thus, the best approach to determine the structure of the branch chain is to calculate the whole methylglucuronic acid, and then a reasonable structure is able to be calculated. In our case, because of the size of the structure (only 23 atoms are contained in the methylglucuronic acid), a full calculation without any approximation is applied, so that the structure of this non standard residue can be assumed accurate. After connecting the glucuronic acid residue to the fourth xylan of a backbone, which contains six xylans, a 4 O glucuronoxylan was obtained. Figure 3 2. Schematic representation of 4 O methylglucuronoxylan

PAGE 33

33 CHAPTER 4 DATA TRANSFERRING FR OM AMBER TO LAM MPS As I mentioned in the previous chapter, the cellulose microfibril and hemicellulose chains were built in AMBER, and transfer red to LAMMPS in order to apply loads on the whole model. There are multiple methods to apply loads in MD simulation: First, is to change the volume and/or shape of the simulation box; another way is to define a group of atoms, and then assign a velocity or force on the group to have a deformation on the whole structure. Moreover, some rigid wall or ball would be used to compre ss the model. In these simulations, I employed the first two methods to apply loads, which could be easily performed in LAMMPS. Although AMBER and LAMMPS both are MD software, it is not simple to calculate the mechanical properties of a material in AMBER There are several reasons why it is difficult to apply loads by using AMBER. First, the periodic box must be introduced if a large scale molecular dynamics simulation of the model is needed. Even though the concept of the periodic box also exists in AMBER, it is quite different from that in LAMMPS. In LAMMPS, the bonds, angles, dihedrals or anything across the boundaries still have relation with each other, so that multiple periodic boxes could be considered as a whole, whereas only some separated periodic boxes can be simulated in AMBER, since AMBER does not focus on large scale molecular dynamics simulation. Furthermore, AMBER also allow s a group of atoms to be defined and to assign a movement on the group in AMBER, which is so called Steered Molecular Dyn amics (SMD). However, no group of atoms happens to be a perfect plane in biomaterial, especially since the model is compress ed or stretch ed Moreover, it is not possible to employ a rigid wall in AMBER. SMD could apply the force

PAGE 34

34 onto four atoms which can define a plane, but the force could not be applied on the atoms within the plane except the four atoms. That is to say, this plane will not affect the model as a real rigid plane. In order to apply loads on our biomaterial model made by AMBER, it is necess ary to transfer the data or information from AMBER to other MD software, which is LAMMPS. As two different kinds of software, AMBER and LAMMPS will take different strategies to get the same goal, so all of the settings are not exactly the same. However, th e force field and main algorithm should be exactly the same. By using amber2lammps.py, which is provided by LAMMPS website, data can be transferred from AMBER to LAMMPS successfully. T o calculate the same model from AMBER, a proper format of force field sh ould be chosen in LAMMPS. In AMBER, all the force fields are the same format represented as Eq. 4 1 : ( 4 1) The equation above contains the bond style, angle style, dihedral style and pair style. Bond style set the formula to compute the bond interactions between pairs of atoms. The form of bond style in AMBER is called harmonic, which takes the format as: ( 4 2) Where K is a constant for bond energy and r0 is the equilibrium bond distance.

PAGE 35

35 Angle style helps the users to calculate the angle interactions between triplets of atoms. As the same as the bond style, the angle style is also called harmonic, and the potential could be written as: ( 4 3) W 0 is the equilibrium value of the angle. Dihedral style is used for computing the dihedral interactions between quadruplets of atoms. In AM BER, C harmm is the dihedral potential. It can be expressed as: ( 4 4) W n is the multiplicity of the func tion, and d is the phase shift. Pair style describes the terms for non bonded pairs of atoms including the electrostatic interactions and van der Waals. So two terms are contained in the expression of pair style lj/cut/coul/cut: ( 4 5) W interactional potential goes to zero, r is the distance between atoms; C is the energy conversion constant, q i and q j are the charges constant. There are two cutoff distances in this non bonded expression: one is for Lennard Jones potential and the other one is for the Coulombic pair wise interaction. The cutoff distances are set for decreasing the computational expense, because the contributions from non bonded interactions are rather small at long distances. One more

PAGE 36

36 thing should be mentioned is that the Lennard Jones is different from that in the whole format of the AMBER force field. But these tw o forms are actually the same. ( 4 6) W 12 6 1/6 2 /(4A). The last one is the improper style which describes how to calculate the im proper interactions between quadruplets of atoms. However, in our simulation, we do not have to calculate this term, so I do not introduce this expression in this part. After choosing the right format of force fields, we transfer the coordinate and paramet er topology files to a data file, which is used for LAMMPS. Before applying the same algorithm of AMBER, it is necessary to modify some parameters. In AMBER, the dielectric constant is always fixed, and the value is 1. The pair wise interaction, which is a lso the non bonded part, is treated differently in various types of force fields. The pair wise interaction c an be divided into three kinds of interactions: 1 2, 1 3 and 1 4. 1 2 interaction is the interaction between a pair of atoms that are connected by a linear bond. 1 3 interaction refers to the interaction between atoms of 1 2 pairs and a bonded third atom. For example, atom A is bonded to atom B, and atom B is bonded to atom C, then the 1 3 interaction here is the interaction between atom A and C. 1 4 interaction is the one that between pairs connected by a set of two bonds. For instance, atom A is bonded to atom B, and atom B is bonded to atom C, and atom C is bonded to atom D, then 1 4 interaction refers to the interaction between atom A and D. In or der to calculate the weighted pair wise interactions, the list of bonds, angles or dihedrals will not define the 1 2, 1 3, and 1 4 interactions. And in different force fields, these three interactions have different weights. In AMBER force field, the three coefficients in front of the three

PAGE 37

37 part s are set to 0.0, 0.0, 0.5 for Lennard Jones interactions (Van der Waals) and to 0.0, 0.0, 0.8333 for Coulombic interactions (electrostatic interactions). What is different from other kinds of material is that the SH AKE algorithm is applied on the biomaterial. By applying the SHAKE algorithm, the specified bonds and angles are reset to their equilibrium lengths or angular values on each time step. This is done by applying an additional constraint force so that the new positions preserve the desired atom separations. The equatio ns for the additional force are solved via an iterative method that typically converges to an accurate solution in a few iterations. In AMBER, bond lengths and angles involving all hydrogen atoms are constrained with the SHAKE algorithm, so that the model could be stable. In the molecular simulation by using LAMMPS, the model obtained by AMBER was used as the initial model with SHAKE applied for bonds involving hydrogen atoms and a time step of 2 femtosecond (fs) In order to figure out whether the data i s transferred from AMBER to LAMMPS successfully, I compared each energy term in the initial time step of AMBER against that of LAMMPS. The comparison of each energy term is shown in Table 4 1: Table 4 1. Static energy analysis of a representative structur e of the cellulose microfibril, using both AMBER11 and LAMMPS AMBER (kcal/mol) LAMMPS (kcal/mol) bond energy 4624.9973 4624.9973 Angle energy 17319.9779 17319.9779 Dihedral energy 26877.7002 26877.7002 Total van der Waals 46506.7695 46506.7696 Total electrostatic 261864.3796 261873.4431 Total energy 264180.2856 264189.3489 Since the same force field was chosen for these two kinds of software, the bond energy, angle energy and dihedral energy are exactly the same. There are very small

PAGE 38

38 differences in computing the van der Waals and electrostatic energy, which lead to the difference in the total energy. But it is acceptable for the result, because the internal architecture of the AMBER code is different from that of LAMMPS and these numbe rs are the sum of several hundred individual interactions. Therefore, the table above shows that AMBER and LAMMPS can calculate nearly identical energies by giving the same original structure, force field, and algorithm. It is also quite important to check whether the model works after transferring data from AMBER to LAMMPS. In MD simulation, the stability of the model is the most important point to check whether the model works, because the stability of the model determined whether the result converges suc cessfully. If the result diverges as the program runs, the numbers in the model will blow up and never reach a solution. So the cellulose microfibril is relaxed under NPT ensemble for 150,000 time steps, and each time step is 2 fs. In this model, the relax ation time lasted 0.3 ns. To check whether a model is stable, the temperature and the total energy of the model are checked. Figure 4 1. Temperature variation during a 150,000 timestep relaxation From Figure 4 1, the temperature is very stable after abou t 50,000 time steps around 300K, which is room temperature. At the beginning, the temperature should be 0 50 100 150 200 250 300 350 0 50000 100000 150000 temperature (K) time step

PAGE 39

39 zero, because the initial structure is totally static. Not only is the temperature stable, but the curve of total energy is also correspondingly flat af ter 50,000 time steps. Figure 4 2. Total energy variation during a 150,000 timestep relaxation 40000 50000 60000 70000 80000 90000 100000 110000 120000 0 50000 100000 150000 total energy (kcal/mol) time step

PAGE 40

40 CHAPTER 5 INTERFACIAL INTERACTIONS As mentioned before, the hemicellulose lignin matrices act as the concrete. The reinforcement of hemicellulose could enhance the stiffness of the wood cell wall T he polymer properties could lead to the mechanism of cell wall deformation. However, it is difficult to analyze the polymer properties by experiment in nano scale. Thus, the understanding of the interface betwe en cellulose microfibrils and hemicellulose chains is very important and could be used to optimize the properties of the whole structure. It is believed that the interface plays an important role to improve the stiffness and strength of the polymer materia l. So the interfacial interaction is the focus when the upper cellulose microfibril is pulled away from hemicelluloses matrices in the axial direction. In traditional polymer matrix composites, the push out and pull out test are the particular ways to eval uate the shear strength of the interface. A critical axial load would be applied to debond the fiber from the surrounding polymer. The critical axial load can be used to estimate the shear stress. However, the shear stress in simulation can be much more co mpliant. Three models were proposed to simulate the effect of the structure of hemicellulose chains. In the simulation, similar to the experiment, the upper cellulose microfibril was applied on an axial load, and the lower microfibril was fixed. The hemice llulose chains, including the ordered long hemicellulose chains and the randomly short ones, were set free. 5.1 Molecular S tructure of the C ellulose H emicellulose S ystem Due to the shapes of the ordered hemicellulose chains, I named these three models: u model, z model and r model. All of the three models contain some random hemicellulose chains. Besides that, the cellulose microfibrils are surrounded by only

PAGE 41

41 random hemice llulose chains in the r model. The topologies of these three models are shown in Figure 5 1, Figure 5 2, and Figure 5 3. Figure 5 1. A cross sectional side view of a one layer z model Figure 5 2. A cross sectional side view of a one layer u model

PAGE 42

42 Fi gure 5 3. A cross sectional side view of a one layer r model The dimensions of each of the three systems are 23.4 74 12 6 9.0 nm All three of these models contain 89112 atoms, and the cellulose microfibrils are all the same. The bonds on the edges of c ellulose microfibrils can go across the box boundary, and the models are placed in a three dimensional periodic box, so the fiber is infinite long. Although the shapes of hemicellulose chains are different, the amounts of the hemicellulose are the same. T he entire pulling process was carried out by MD simulation. An x axial load was applied on the upper cellulose microfibril to pull it away from the hemicellulose chains around it. The whole simulation contains over 200,000 MD time steps of 2 fs, depending the load variation. The applied force was increased gradually during the whole loading time.

PAGE 43

43 5.2 Relaxation of t he W hole S ystem Actually, the hemicellulose matrices are not connected to the microfibrils in the initial structures. Therefore, several stages of relaxation were taken for the full conne ction between cellulose microfibrils and hemicellulose matrices. In the first stage of relaxation, the system was placed under NVE ensemble. All the polymers were set as rigid bodies except the random hemicellulo se chains. The cellulose microfibrils and ordered hemicellulose chains are treated as rigid bodies because the rigid bodies can be moved to a lower total potential energy during this 10,000 MD time steps relaxation This prevents any extremely high interac tion in the system. In the second stage, cellulose microfibrils were still set as rigid bodies, but the hemicellulose matrices were released. Therefore, the hemicellulose matrices can change freely to a less constrained shape and place for MD simulation. T h e aim of this relaxation is to fully connect the hemicellulose matrices to the cellulose microfibri l bundles due to the electrostatic interactions and Van der Waals forces This stage also took 10,000 MD time steps. At last, the whole system was released so that the cellulose microfibrils and hemicellulose matrices could fully contact with each other. The ensemble for this stage of relaxation is NVT, so that the temperature of the system could be improved to around 300K, which is approximately room tempe rature. A 100,000 MD time steps are necessary in this stage to fully relax the whole system. As mentioned in the previous chapter, checking the stability of temperature and total energy during the simulation is one of the most important methods to determin e whether a system is stable enough. The figures below show the temperature and total

PAGE 44

44 energy of the z model during the third stage of relaxation. Similarly, the performance of the u model and r model are almost the same. Figure 5 4. Temperature variation during a 100,000 time step relaxation Figure 5 5. Total potential energy variation during the 100,000 time step relaxation 200 250 300 350 0 20000 40000 60000 80000 100000 Temperature (K) Time Steps Temperature 200000 250000 300000 350000 400000 0 20000 40000 60000 80000 100000 Total Energy (Kcal/mol) Time Steps Total Potential Energy

PAGE 45

45 5.3 A pplying V elocity w ith W ater E ffect In MD simulation, there are two methods of applying loads to model the pull out process: one is to set velocities on groups of atoms and the other one is to give an additional push to atoms in the simulation. At first, velocities were assigned on the uppe r cellulose microfibril. The systems in this part of simulation were also put into a box full of water. The molecule of water chosen was TIP3P, which is one of the most widely used model for chemical computation. The force of water and hemicellulose matric es on the moving cellulose microfibril was measured. Although the cellulose microfibril moved away from the hemicellulose matrices with various velocities, the force is not realistic. Figure 5 6. Force versus time curves of the z model upper cellulose mi crofibril 0 1000 2000 3000 4000 5000 6000 7000 8000 0 200 400 600 800 1000 force(kcal/mol/A) time(2fs) z_leftwards 2.00E-03 5.00E-03 1.00E-02 2.00E-02

PAGE 46

46 Figure 5 7. Force versus time curves of the u model upper cellulose microfibril In Figure 5 6 and Figure 5 7, different curves represent different loading velocities of the upper cellulose microfibril. It seems that the property of the interfa ce depends on the loading rate. The bigger the velocity assigned on the fiber, the larger the force between fiber and hemicellulose matrices is. However, the different models have almost the same curves under the same loading rate. The results may not be r easonable because: a) applying loads on fibers by assigning velocity is not accurate in this case and b) TIP3P water model is not suitable for the simulation of the mechanical properties of this material. 5.4 A pplying I ncremental I ncreasing L oads with W ate r E ffect The other way to apply loads on the same system was then attempted. The advantage of applying increasing force on the material is that both the elastic and plastic deformation stage can be observed, while the elastic deformation stage is very 0 1000 2000 3000 4000 5000 6000 7000 8000 0 200 400 600 800 1000 force(kcal/mol/A) time(2fs) u_leftwards 2.00E-03 5.00E-03 1.00E-02 2.00E-02

PAGE 47

47 shor t if the velocity is assigned on the group of atoms. On each time step of MD simulation, an atom will be affected by the atoms around it. If the velocity of the atom is too small, the atom might be pulled back. If the velocity is big enough to outweigh the deformation stages are so short in Figure 5 6 and Figure 5 7. Assuming the TIP3P water model works fine in MD simulation, the same increasing forces were applied on the fibers of the three systems. As in the previous velocity loading case, the forces of water and hemicellulose matrices on the moving cellulose microfibrils were measured. The force versus displacement curves of the three models are shown in Figure 5 8, Figure 5 8. Force versus time curves of the upper cellulose microfibrils with water Although the amounts of hemicellulose matrices are exactly the same in those three models, the contact areas between cellulose microfibrils and hemicellulose 0 500 1000 1500 2000 2500 3000 0 200 400 600 800 1000 1200 1400 force (Kcal/mol/ ) displacement ( ) Comparison zmodel umodel rmodel

PAGE 48

48 matrices are quite different because of the structures. The connection between two cellulose microfibrils in the r model is smaller than the other two, so the resulting curve should be the lowest. From the figure above, the maximum force of these three models are a lmost the same, however, the energies for failure could be depicted by the areas under the force displacement curves are different. Assuming the contact areas between fibers and hemicellulose matrices of z model, u model and r model are Az, Au, and Ar, res pectively. Since Az is the largest and Ar is the smallest, the sequence of energies for material failure is the same as that of the contact area. However, the peaks of these three curves are almost the same, which means there is rarely a difference between the largest forces to pull the fiber out from three different polymer environments. This result is reasonable, but it includes the effects of water. 5.5 A pplying I ncremental I ncreasing L oads w ithout W ater E ffect The three systems are changed in this part. All the molecules of water were deleted from the periodic boxes. In our simulation, the system of unit is called real so the unit of distance is and the unit of force is Kcal/(mol* ). After the 120,000 time steps relaxation, 0.01 Kcal/(mol* ) would be applied on each atom of the upper fiber for every 5,000 time steps. Because the total number of the atoms of one fiber is 33,264, the increasing force applied on the upper cellulose microfibril for every 5,000 time steps is 2.3*10 11 N. And the force is i ncreased 10 times, so the whole pull process is a 50,000 MD simulation time steps. The force displacement curves of three models are shown in Figure 5 9,

PAGE 49

49 Figure 5 9. Force versus displacement curves of the upper cellulose microfibrils without water From the atomic trajectory of the systems, the hemicellulose matrix is released when the f orce displacement curve began to drop. It is obvious that the curves of three models are quite different. Two possible reasons for this could be: a) the contact area s between cellulose microfibril and hemicellulose matrices of three systems are different, b) the shapes of the hemicellulose matrices in three models are different. 5.6 Molecular S imulation A pplying L oads i n T wo o pposite D irections o n Z M odel Two forces i n two opposite directions were applied on the z model to figure out the effect of the shape of the ordered hemicellulose matrices. Applying loads on the same model can ensure that the contact areas of the interface are always the same. Furthermore, two opp osite forces, which are along the positive x axial and the negative x axial direction respectively, could lead to an equivalent state of changing the shape of 0 100 200 300 400 500 600 700 800 0 200 400 600 800 1000 1200 1400 Force (Kcal/mol/) Displacement () z u r

PAGE 50

50 hemicellulose matrices. That means the structure of the interface was rotated by 180 about y axi s. As in the previous section, the same increasing force is applied on the upper cellulose microfibril bundles. Figure 5 10. Force diagram of the positive x axial loading Figure 5 11. Force diagram of the negative x axial loading

PAGE 51

51 Figure 5 10 and Fig ure 5 11 show how the forces are applied. After applying the same loading forces on both of the models, the force displacement curves were generated, which are shown in Figure 5 12 and Figure 5 13. Figure 5 12. Force displacement curve of the positive x axial loading simulation Figure 5 13. Force displacement curve of the negative x axial loading simulation 0 200 400 600 800 1000 0 200 400 600 800 1000 Force (Kcal/mol/ ) Displacement () 0 200 400 600 800 1000 0 200 400 600 800 1000 Force (Kcal/mol/) Displacement ()

PAGE 52

52 It is observed that these two curves are almost the same except the peak area. When the force is applied along the positive x direction, the interfa ce allows much more shear force. Since the contact area of these two systems is the same, the shape or the structure should be the main factor to cause the difference. T hus, the trajectories of the atomic structure of the systems were observed Figure 5 14. A cross sectional view of a one layer z model relaxed structure Figure 5 15. A cross sectional view of one layer structure when the cellulose microfibril is released from hemicellulose matrices under the positive x axial loading

PAGE 53

53 Figure 5 16. A c ross sectional view of one layer structure when the cellulose microfibril is released from hemicellulose matrices under the negative x axial loading There are two kinds of groups in one single hemicellulose chain: the upper and lower parts which contact wi th the cellulose microfibril, and the middle part, which just connect the upper and lower parts. After observing the movement of the atomic structure of these two systems, the adhesive parts of hemicellulose matrices in both systems is released from the ce llulose microfibrils when the displacement is about 50 nm showing in Figure 5 15 and Figure 5 16. When the displacement of cellulose microfibril is between 20 nm and 50 nm, the middle parts of the hemicellulose chain had different performances in the two s ystems resulting in the peak and plateau area in Figure 5 12 and Figure 5 13 respectively In the x positive loading case, the middle parts were stretched. However, in the x negative loading case, the middle parts were rotated because of the movement of t he cellulose bundles. The different performance of the middle parts is the key reason why the two curves in Figure 5 12 and Figure 5 13 are so different. If the chain is rotated, the moving process of the chain is approximately like a rigid body rotation, which would not need too much force or energy. However, if the chain is stretched, the situation is far more complicated. Because the force field

PAGE 54

54 employed is a four body potential, if an atom is moved away from its equilibrium position, the bond, angle, an d dihedral which involves this atom would be changed. The a flat section in the curve of x negative loading case. Therefore, the conclusion of this series of simulat ions is that the interface strength is affected by the contact area and the structure of the hemicellulose matrices. The more contact area between cellulose microfibril and hemicellulose matrices, the stronger the interface is. Furthermore, unlike the cell ulose microfibril bundles, which have a semi crystalline structure, hemicellulose chain is very flexible. This has been shown in the previous chapter. The hemicellulose chain performs similar to a string with a force applied. If the force is compressive, t he string will buckle easily, but if a tensile force is applied, it can be strong despite its flexibility.

PAGE 55

55 CHAPTER 6 MICROFIBRIL ANGLE (M FA) EFFECT 6.1 Introduction to M icrofibril A ngle (MFA) The term microfibril angle (MFA) in wood science refers to the angle between the direction of the helical windings of cellulose microfibrils in the secondary cell wall of fibers and tracheids and the long axis of cell (Barnett and Bonham, 2004) Density h as long been considered the main factor of wood quality. However, as Evans and I lic (2001) mentioned, the influence of MFA on wood stiffness is significantly greater than that of density. They also claim that MFA together with density accounted for more th an 95% of the variation in longitudinal modulus of elasticity and that MFA alone accounted for more than 85% of the variation. As the tree grows, the trunk has to be stiffer, or the increasing weight of the tree will break itself. That means mature wood is stiffer than juvenile wood. It is found that MFA is consistently smaller in latewood cell walls than in those of earlywood in E. nitens (Stuart and Evans, 1994) Since the wood with a larger MFA will gra de use. As the forest industry develops, more timber, pulp and wood products are increasingly needed, so that trees are not grown to fully maturity before being harvested. This is why interest is increasing for possible methods to improve wood quality by m odifying MFA. In the following section, simulations are run on various models with different microfibril angles to determine the effect of MFA at the atomic scale. 6.2 MFA E ffect All the models in this part are based on the model with a 0 microfibril ang le. The initial structure of the system is shown in Figure 6 1:

PAGE 56

56 Figure 6 1. Cross sectional views of one layer in six models. The MFA are 0 , 5 10 15 20 and 25 The model in the upper left corner is the initial one, which contains four cellulose microfibrils and four groups of z shape hemicellulose matrices. All these models are put into a periodic box. The bonds, bond angles, and dihedral angles of the models could be across the boundaries of the box, so that the system is actually inf initely huge. Then three stages of relaxation are applied on the six systems. In the first 20,000 step relaxation stage, NVE ensemble was introduced to average the residual stress throughout the system. However, the NVE ensemble could not eliminate the res idual stress, so that another 50,000 step relaxation under NPT ensemble was applied to optimize the structure. In addition, the relaxation under NPT ensemble is also a temperature control process. The temperature of the system could be heated up and kept t o 300 K. In NPT ensemble, a thermostat and a barostat could control the

PAGE 57

57 temperature and pressure of the system, respectively. Since a compression would be applied on the model, the pressure should not be a constant value. Another 20,000 step relaxation und er NVT ensemble was employed for the final optimization. The relaxed structures are shown in Figure 6 2: Figure 6 2. Cross sectional views of the six one layer models after relaxation After the relaxation, the compression loadin g is applied on the models. The way we apply the loads is to change the size of the periodic box smoothly. The change in each box is 10% in the y direction, which is along the axial direction of the cellulose microfibrils in the 0 MFA. The change in the a tomic structure is shown in Figures 6 3, 6 4, 6 5, 6 6, 6 7, and 6 8.

PAGE 58

58 Figure 6 3. Compression process of the model with 0 MFA Figure 6 4. Compression process of the model with 5 MFA Figure 6 5. Compression process of the model with 10 MFA

PAGE 59

59 Figure 6 6. Compression process of the model with 15 MFA Figure 6 7. Compression process of the model with 20 MFA Figure 6 8. Compression process of the model with 25 MFA Every figure abov e represents a compression process of a model: the left figures are the initial structure after relaxation, the middle ones represent the elastic deformation of that model, and the right ones show the failure of these six systems. The figures on the right show the buckling occurring o n cellulose microfibril. Also, the stress strain curve of each model is plotted in Figure 6 9.

PAGE 60

60 Figure 6 9. Stress strain curves of six models with various MFA under the same loading rate compression From the curves in Figure 6 was obtained. Table 6 1. Mechanical property parameter of six models with various MFA MFA ( ) 0 5 10 15 20 25 Modulus (GPa) 58. 974 52.962 49.665 43.841 41.190 32.834 Yield Strength (GPa) 1.76 1.60 1.46 1.31 1.24 1.07 Toughness (MPa) 59.262 55.255 50.857 48.265 47.616 50.454 From Figure 6 9 and Table 6 modulus and strength of the system decreases with the increasing of MFA. And the toughness of the system seems to decrease as MFA increases. All of the models failed 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 0.01 0.02 0.03 0.04 Stress (GPa) Strain Stress strain curves MFA=0 MFA=5 MFA=10 MFA=15 MFA=20 MFA=25

PAGE 61

61 due to the o ccurrence of buckling, and buckling can happen on any part of cellulose microfibril, whether the hemicellulose chains are adhered or not. Therefore, the buckling of cellulose microfibril bundle is the main reason why the systems fail. As the results mentio ned in Chapter 5, when the connection between cellulose microfibril bundle and hemicellulose matrices is broken, the deformation of the system is much larger than that of buckling. This means that if the interaction between the cellulose and hemicellulose is broken, the buckling of cellulose microfibril bundle already occurred.

PAGE 62

62 CHAPTER 7 CONCLUSIONS This paper mainly researches the interactions between cellulose and hemicellulose in the secondary cell wall and the microfibril angle (MFA) effect on the properties of fiber reinforced nanocomposites. The structural model of cellulose microfibril bundle is generated by using the geometric parameters from experiments. The structure is stable through energy equilibrium. As the periodic boundary condit ion is applied, the cellulose microfibril bundle has an infinite length. We also built a model of hemicellulose chain. 4 O methylglucuronoxylan is the main hemicellulose found in the secondary cell wall of many hardwoods. Xylan and methylglucuronic acid a re two main components of this hemicellulose. A unit chain of hemicellulose is built, which consists of six xylans and one methylglucuronic acid. Various fiber reinforced cellulose hemicellulose systems based on the models of cellulose microfibril bundle and hemicellulose chain can be designed. In order to explore the interactions between cellulose and hemicellulose, three different hemicellulose matrices are designed. From the results of MD simulation on those three systems, it is found that the more area between cellulose microfibrils and hemicellulose matrices the stronger interactions exhibit. However, contact area is not the only factor affecting the interactions. The shape of hemicellulose chain could also affect the interfacial interactions. During t he deformation of the system, if some part of the chain is stretched, more force is necessary to pull the cellulose microfibril bundle away from the hemicellulose matrices. However, if some part of the chain is compressed, this part will

PAGE 63

63 be rotated. Only a small amount of force can depart the cellulose microfibril bundle and hemicellulose matrices. Furthermore, a study of the microfibril angle (MFA) effect on the properties of fiber reinforced composites based on MD simulation shows the similar result as th e and the yield strength is not as high as the one with a smaller MFA. The failure of all the systems is buckling. It is also observed that the buckling could be displayed on a ny part of cellulose microfibril bundle, no matter the adhesive hemicellulose chain lies on it or not. Therefore, we conclude that the buckling of the cellulose bundle is the crucial factor of the failure in the systems. This is related to the structures of the two components: cellulose microfibril bundle has a semi crystalline structure, while hemicellulose chain is much more flexible. The deformation is very small when buckling appears. However, hemicellulose chain can move some distance with cellulose m icrofibril bundle before the interaction between them is broken. In other words, if the interaction between cellulose and hemicellulose was broken, the buckling of cellulose microfibril bundle had occurred and the system had collapsed.

PAGE 64

64 LIST OF REFER ENCES [1] T. H. Giddings, D. L. Brower Jr., and L. A. Staehelin, J. Cell. Biol. 84, 327 (1980). [2] M. C. McCann, B. Wells and K. Roberts, J. Cell Sci. 96,323 (1990). [3] J. P. Vincken, W. S. York, G. Beldman, and A. G. J. Voragen, Plant Physiol. 114, 9 (1997) [4] T. Imai and J. Sugiyama, Macromolecules 31, 6275 (1998). [5] V. L. Finkenstadt and R. P. Millane, M acromolecules 1998, 31, 7776 (1998). [6] J. F. V. Vincent, J. Exp. Biol. 202, 3263 (1999). [7] A. Teleman, J. Lundqvist, F. Tjerneld, H. Stlbrand, and O. Dahlman, Carbohydr. Res. 329, 807 (2000). [8] C. Picard, J. Gruza, C. Derouet, C. M. G.C. Renard, K. Mazeau, J. Koca, A. Imberty, and C. H. D. Penhoat, Biopolymers, 54, 11 (2000). [9] D. J. Cosgrove, Plant Physiol. 125, 131 (2001). [10 ] R. Evans and J. Ilic, For. Prod. 51, 53 (2001). [11] E. N. Landis, S. Vasic, W. G. Davids, and P. Parrod, Exp. Mech. 42, 389 (2002). [12] Y. Nishiyama, P. Langan, and H. Chanzy, J. Am. Chem. Soc. 124, 9074 (2002). [13] D. Kretschmann, Nat. Mater. 2, 7 75 (2003). [14] J. Fahlen, L. Salmen, J. Mater. Sci. 38 119 (2003). [15] J. Keckes, I. Burgert, K. Fruhmann, M. Muller, K. Kolln, M. Hamilton, M. Burghammer, S. V. Roth, S. Stanzl Tschegg, and P. Fratzl, Nat. Mater. 2, 810 (2003). [16] S. J. V. Frankland and V. M. Harik, Surf Sci 505, L103 (2003). [17] C. Grindon, S. Harris, T. Evans, K. Novik, P. Coveney, and C. Laughton, Phil. Trans. R. Soc. Lond. A 362, 1373 (2004). [18] J. R. Barnett and V. A. Bonham, Biol. Rev. 79, 461 (2004). [19] P. Fra tzl, I. Burgert and H. S. Gupta, Phys. Chem. Chem. Phys. 6, 5575 (2004). [20] S. Wan, P. Coveney, and D. R. Flower, J. Comput. Chem. 25, 1803 (2004).

PAGE 65

65 [21] W. Chen, G. Lickfield, and C. Q. Yang, Polymer 45, 1063 (2004). [22] Z. Xia and W. A. Curtin, Phys. Rev. B 69, 233408 (2004). [23] D. J. Cosgrove, Nat. Rev. Mol. Cell Biol. 6, 850 (2005). [24] L. J. Gibson, J. Biomech. 38, 377 (2005). [25] A. C. Balazs, T. Emrick, and T. P. Russell, Science 314, 1107 (2006). [26] C. Gonzlez J. Llorca, Acta Mater. 54 4171 ( 2006 ). [27] I. Burgert, Am. J. Bot. 93, 1391 (2006). [28] J. F. Matthews, C. E. Skopec, P. E. Mason, P. Zuccato, R. W. Torget, J. Sugiyama, M. E. Himmel and J. W. Brady, Carbohydr. Res. 341, 138 (2006). [29] J. Hanus and K. Mazeau, Biopolymers 82, 59 (2006). [30] S. Ding and M. E. Himmel, J. Agric. Food Chem. 54, 597 (2006). [31] S. E. Stanzl Tschegg, Int. J. Fract. 139, 495 (2006). [32] S. R. Alam, J. S. Vetter, P. K. Agarwal, and A. Geist, Proc. ACM SIGPLAN Symp. Principles and Practice of Parallel Programming (PPoPP), ACM Press 59 (2006) [33] T. C. Clancy and T. S. Gates, Polymer 47, 5990 (2006). [34] T. Yui, S. Nishimura, S. Akiba, and S. Hayashi, S. Carbohydr. Res. 341, 2521 (2006). [35] C. J. Kennedy, G. J. Cameron, A. D. C. Apperley, C. Altaner, T. J. Wess, and M. C. Jarvis, Cellulose 14, 235 (2007). [36] C. Moine, P. Krausz, V. Chaleix, O. Sainte Catherine, M. Kraemer, and V. Gloaguen, J. Nat. Prod. 70, 60 (2007). [37] L. Donaldson, Wood Sci. Technol. 41, 443 (2007). [38] L. S. Schadler, L. C. Brinson, and W. G. Sawyer, JOM 59, 53 (2007). [39] M. Bergenstrhle, L. A. Berglund, and K. Mazeau, J. Phys. Chem. B 111, 9138 (2007). [40] M. E. Himmel, S. Ding, D. K. Johnson, W. S. Adney, M. R. Nimlos, J. W. Brady, and T. D. Foust, Science 315, 804 (2007).

PAGE 66

66 [41] B. Clair, J. Gril, F. Di Renzo, H. Yamamoto, and F. Quignard, Biomacromolecules 9, 494 (2008). [42] C. M. Altaner and M. C. Jarvis, J. Theor. Biol. 253, 434 (2008). [43] K. N. Kirschner, A. B. Yongye, S. M. Tschampel, J. Gonzalez Outeririno, C. R. Daniels, B. L. Foley and R. J. Woods, J. Comput. Chem. 29, 622 (2008). [44] M. S. Shell, R. Ritterson and K. A. Dill, J. Phys. Chem. B. 112, 6878 (2008). [45] S. J. Heo, L. Jang, P. R. Barry, S. R. Phillpot, S. S. Perry, W. G. Sawyer, and S. B. Sinnott, J. Appl. Phys. 103, 083502 (2008). [46] X. Hao, Q. Han and X. H. Yao, Compos. Sci. Technol. 68, 1809 (2008). [47] Y.Nishiyama, G. P. Johnson, A. D. French, V. T. Forsyth, and P. Langan, Biomacromolecules 9, 3133 (2008). [48] N. Terashima, K. Kitano, M. Kojima, M. Yoshida, H. Yamamoto, and U. Westermark, J Wood Sci 55, 409 (2009). [49] T. Yui and S. Hayashi, Cellulose 16, 151 (2009). [50] Y. Nishiyama, J. Wood Sci. 55, 241 (2009). [51] D. B. Wells, and A. Aksimentiev, Biophys. J. 99, 629 (2010).

PAGE 67

67 BIOGRAPHICAL SKETCH Shi Li was born and raised in Wuhan, China He graduate d magna cum laude from Huazhong University of Science and Technology in 2009 with a Bachelor of Science in Mechanical Engineering Shi c ame to Gainesville, Florida in 2009 and was admitted to be a graduate department. He worked with Dr. Youping Chen on his master degree in summer 2010. Shi enjoys reading and exerci sing.