KUMAR NANJUNDIAH ALL RIGHTS RESERVED

Transcription

1 2007 KUMAR NANJUNDIAH ALL RIGHTS RESERVED

2 STUDY OF CONFINEMENT AND SLIDING FRICTION OF FLUIDS USING SUM FREQUENCY GENERATION SPECTROSCOPY A Dissertation Presented to The Graduate Faculty of The University of Akron In Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Kumar Nanjundiah December, 2007

3 STUDY OF CONFINEMENT AND SLIDING FRICTION OF FLUIDS USING SUM FREQUENCY GENERATION SPECTROSCOPY Kumar Nanjundiah Dissertation Approved: Accepted: Advisor Dr. Ali Dhinojwala Department Chair Dr. Mark D. Foster Committee Member Dr. Gary R. Hamed Dean of the College Dr. Stephen Z.D. Cheng Committee Member Dr. Mark D.Foster Dean of the Graduate School Dr. George R. Newkome Committee Member Dr. Gustavo A. Carri Date Committee Member Dr. Jutta Luettmer-Strathmann ii

4 ABSTRACT Friction and wear are important technologically. Tires on wet roads, windshield wipers and human joints are examples where nanometer-thick liquids are confined between flexible-rigid contact interfaces. Fundamental understanding of the structure of these liquids can assist in the design of products such as artificial joints and lubricants for Micro-electromechanical systems [MEMS]. Prior force measurements have suggested an increase in apparent viscosity of confined liquid and sometimes solid-like responses. But, these have not given the state of molecules under confinement. In the present study, we have used a surface sensitive, non-linear optical technique (infrared-visible sum frequency generation spectroscopy [SFG]) to investigate molecular structure at hidden interfaces. SFG can identify chemical groups, concentration and orientation of molecules at an interface. A friction cell was developed to study sliding of a smooth elastomeric lens against a sapphire surface. Experiments were done with dry sliding as well as lubricated sliding in the presence of linear alkane liquids. SFG spectra at the alkane / sapphire interface revealed ordering of the confined alkane molecules. These were more ordered than alkane liquid, but less ordered than alkane crystal. Cooling of the iii

5 confined alkane below its melting temperature [T M ] led to molecular orientation that was different from that of bulk crystal next to a sapphire surface. Molecules were oriented with their symmetry axis parallel to the surface normal. In addition, the melting temperature [T conf ] under confinement for a series of linear alkanes (n =15-27) showed a surprising trend. Intermediate molecular weights showed melting point depression. The T conf values suggested that melting started at the alkane / sapphire interface. In another investigation, confinement of water between an elastomeric PDMS lens and sapphire was studied. SFG spectra at the sapphire / water / PDMS interface revealed a heterogeneous morphology. The presence of peaks related to PDMS, as well as water, suggested water puddles in the contact area and the sapphire surface had a layer of bound water. This heterogeneity picture provides insight into high friction and stick-slip behavior found in boundary lubrication. For the first time, a broadband SFG system has been coupled with a friction cell to study dynamics and molecular changes at an interface during sliding; sliding of confined alkane between sapphire and PDMS was investigated. A series of SFG spectra were taken while the confined alkane contact spot moved in and out of the laser beam. Even though the experiments were done 15 C above melting temperature, the spectra showed ordering of alkane molecules, similar to that of the confined crystal at the leading and trailing edge. The results suggest that a large portion of the resistance to sliding may come from ordering of molecules at the lens front. iv

6 ACKNOWLEDGEMENTS The journey through graduate school has been exciting and highly rewarding. I would like to show my gratitude to people who helped me along the way to successfully complete my dissertation. First, I would like to thank my advisor Prof. Ali Dhinojwala for his inspiration and guidance. I learnt a lot about solving complex problems by breaking it into smaller, simpler sections. I also learnt that it was better to do many complementary experiments before drawing conclusions. He also taught me to be the toughest critic of my work before presenting it to the scientific community. I am also grateful for his support and guidance on personal matters. I would like to thank Prof. Gray Hamed for agreeing to chair my defense committee and his suggestions in improving my dissertation. I would like to thank my other committee members, Prof. Mark Foster, Prof. Gustavo Carri and Prof. Jutta Luettmer-Strathmann for their precious time and suggestions. I would like to thank Ed Laughlin for excellent machining work and help in development of friction cell. I would like to thank Steve Roberts for helping in electronic circuits and work on lasers. I would like to thank National Science Foundation and The University of Akron for financial support. v

7 I would like to express my appreciation to my group members past and present for help in the lab. I would like to mention the help of Dr. Vikas Varshney in modeling the alkane confinement data. I would like to thank Dr. Betul Yurdumakan for showing the technique of making PDMS lens and performing JKR experiments. I would like to thank Anish Kurian for help in performing dynamic SFG measurements and Ping Hsu for some of the experiments on water confinement. I would like to thank Dr. Shishir Prasad and Sunny Sethi for help in the lab, and discussions on different topics of science during coffee breaks. I would like to thank Dr. Hasnain Rangwalla and Prof. Quinn for the Latex 2e class style files for writing my dissertation. I would like to thank my roommate, Rahul Kulkarni for support and proof reading my dissertation. I would like to thank my wife, Sushma for bearing with my busy schedule, providing inspiration and also proof reading my dissertation. I would like to thank Dr. Ramamurthy for proof reading my dissertation. I would like to express my whole-hearted gratitude to my parents, Smt. Sujayalakshmi and Sri. Nanjundiah and my brother Karthik for their love, inspiration and moral support without which this dream would not have been a reality. vi

8 TABLE OF CONTENTS Page LIST OF TABLES LIST OF FIGURES ix x CHAPTER I. INTRODUCTION II. HISTORICAL RECORD Historical perspective Friction Lubrication Simulation of confined systems Sum frequency generation spectroscopy Other optical techniques used to study confinement III. EXPERIMENTAL TECHNIQUES Development of friction cell Sum frequency generation spectroscopy Sample preparation vii

9 3.4 Measuring thickness of confined lubricant Determining thickness and critical angle using Airy formulas IV. RESULTS AND DISCUSSION Static confinement of linear alkanes Confinement of water Phase transitions under confinement in linear alkanes Characterization of molecular structure during sliding friction V. CONCLUSIONS BIBLIOGRAPHY APPENDIX viii

10 LIST OF TABLES Table Page 2.1 Vibration assignments for the hydrocarbon stretching region Ratios of A q of CH 2 (d + )/CH 3 stretches (r + ) and CH 3 asymmetric (r )/ CH 3 symmmetric (r + ) for hexadecane and pentadecane in contact with sapphire at various conditions Peak positions and their origin for water molecules next to various surfaces Values of parameters used to compute equation A q values obtained by fitting SSP spectra to equation 4.1 for confined alkanes of varying chain length in liquid and crystal states A.1 SFG fit parameters for water experiments generated by fitting the SFG spectra to equation A.2 SFG fit parameters for water experiments [contd...] generated by fitting the SFG spectra to equation ix

11 LIST OF FIGURES Figure Page 2.1 A block with mass M sliding on a flat surface. A spring (spring constant K s ) is connected to the block and the free end is pulled with velocity V s Sliding dynamics of the block in figure 2.1 for three different cases; (a) steady sliding (b) periodic stick-slip motion (c) chaotic motion [5] General kinetic phase diagram. The hashed region depicts the values of K s and V s for which stick-slip motion occurs [5] Microscopic and macroscopic parameters affecting frictional behavior. Generalized Stribeck curve [17] Proposed friction map of force versus sliding velocity showing different regimes [17] Behavior of rubber at the interface. (a) Rubber on a hard surface with long-wavelength roughness. (b) Soft rubber deforms in such a way as to completely follow the short-wavelength roughness. (c) Rubber surface dusted with small particles sliding on a hard substrate. (d) Rubber sliding on water [5] Boundary lubrication. (a) The polar head group in the fatty acid binds to the metal oxides (b) Hardy s concept of boundary lubrication [5] Schematic of surface force apparatus. Also seen are two different types of force measurement springs which are interchangeable allowing up to 8 orders of force measurement [32] x

12 2.9 Oscillatory force profiles for OMCTS liquid confined between mica surfaces [28] DLVO interaction potential energy E as a function of separation distance D between two flat surfaces interacting through an aqueous salt solution with attractive van der Waals force and repulsive electrostatic double-layer force [32] Experimental and theoretical interaction potential between mica surfaces in a 10 3 M KCl solution. As the solution is more diluted DLVO behavior is observed, but if the electrolyte is concentrated, in short range (small separations) oscillatory profile within the monotonically repulsive tail with a periodicity equal to water molecule (2-2.6 nm) is observed. The inset shows a theoretical computation for the same system [32] Variation in friction of wiper blade with various lubricants.(o) silicones; ( ) alcohols; ( ) glycerol solutions; ( ) teepol solutions [8] Grand canonical ensemble 3D model for studying confined liquids between solid walls in equilibrium with bulk liquid [46] Solvation force profiles for hexadecane molecules confined between gold surfaces. (o)rough surface, (*) flat surface [45] Schematic of the beams for SFG at an interface. (a) external geometry assuming n 2 > n 1. (b) internal reflection geometry n 1 > n Normal vibrational modes, notation and co-ordinate system for methyl and methylene groups SFG spectrum of alkane (15 carbons) next to sapphire surface. The points are the experimental data and the line is a fit to the data using equation Definition of Euler angles relating lab axes to molecular axes SFG spectra of octadecanethiol monolayers under different loading conditions. r + and r stand for the methyl symmetric and asymmetric modes [74] Schematic of the friction cell with coupled normal and shear sensors Schematic of the wheatstone bridge circuit to measure change in resistance [85] xi

13 3.3 Schematic of the bridge circuit used to amplify the output voltage of Wheatstone network [86] Image of the friction cell prototype in the laboratory Shear stress vs. time at different sliding velocities for a PDMS lens sliding against a polycarbamate film Schematic of the friction cell based on capacitance measurement Schematic of the circuit to measure current Schematic of the circuit to generate high frequency AC voltage Graph of voltage measured vs. the fixed capacitors values. Slope is determined by fitting the data points to a linear fit Graph of voltage measured vs. the known weights. Slope is determined by fitting the data points to a linear fit Graph of shear force vs. time of sliding for PDMS dry sliding against sapphire Schematic of the picosecond SF spectrometer. The meanings of the symbols are as follows: H, half wave plate; B, calcium fluoride plate beam splitter; M, mirror; DM, dichroic mirror; P, polarizer; F, Raman notch filter; C, IR chopper; PMT, photomultiplier tube detector Schematic of the total internal reflection geometry used. The top part shows the arrangement to study confinement or friction and the bottom part shows the setup to study spin coated polymer films or fluids next to sapphire surface Schematic of the Clark-MXR femto-second SFG system. The beam path is labeled and the additional optical elements in the beam path are as follows: (L) plano convex lens, (CL) plano concave lens, and (WP) waveplate Principle of interference of two beams to produce Newton rings [91] Thickness determination geometry. (a) schematic diagram of the path of light beam through the 30 0 prism. (b) the scatter of laser light in rubber [90] xii

14 3.17 Trace of variation in reflected light intensity with time for the contact zone between a rubber sphere and lubricated glass plate at constant load [8] Schematic of the thickness measurement setup Graph of output power vs. the input angle for an alkane confined between sapphire and PDMS. Setup shown in figure 3.18 was used SSP spectra of hexadecane (a) and pentadecane (c) liquid/sapphire interfaces at 295K. The SSP spectra for hexadecane (b) and pentadecane (d) crystal/sapphire interfaces were taken at 287 and 279K, respectively. The solid lines are fits using Eq The SFG spectra were offset along y-axis by an arbitrary amount and were scaled for clarity SSP spectra of hexadecane (a) and pentadecane (c) confined liquid/sapphire interfaces at 295K. The SSP spectra for confined hexadecane (b) and pentadecane (d) crystal/sapphire interfaces were taken at 287 and 279K, respectively. The dry PDMS/sapphire SFG spectrum is shown in (e). The solid lines are fits using Eq The SFG spectra were offset along y-axis by an arbitrary amount and were scaled with respect to the SFG spectrum for pentadecane crystal in Fig. 4.1(d) for comparison Predictions of A q,ssp for CH 3 symmetric, CH 3 asymmetric and CH 2 symmetric for an all-trans odd alkane rotating along the a-axis (in-plane, left) and b-axis (out-of-plane, right) with respect to the laboratory z-axis Pictorial representation of analysis of SFG data showing alkane chain layering upon confinement SFG vibrational spectra of water/vapor interface taken with (a) SSP, (b) PPP, and (c) SPS polarization combinations [119] Molecular structure of hexagonal ice crystal: (a) side view of the bulk near the (0001) surface; (b) top view of the (0001) plane. Red spheres represent O atoms (dark and light shades highlight higher and lower submonolayers in a single ice monolayer); gray and white spheres represent H atoms that are hydrogen bonded to neighboring molecules and free-dangling non bonded surface species, respectively. Dotted lines indicate hydrogen bonds [122] xiii

15 4.7 SFG spectra of silica / water interface taken in SSP polarization combination as function of ph. The spectra are offset for clarity. Spectrum of silica / ice interface is also shown for comparison [124] Possible hydrogen-bonding configuration of water molecules on hydrophilic silica surface: (a) protonated (SiOH) surface sites, low ph (below 2); (b) deprotonated (SiO ) surface sites, high ph; (c) structure of water/silica interface at low ph. Red and gray spheres represent O and H atoms of water molecules; large gray green, pink, and white spheres represent Si, O, and H atoms of SiOH groups at silica surface. Dotted lines indicate hydrogen bonds. [122] SFG spectra of water interface with solid and liquid hydrophobic surface. (a) water / OTS / silica interface. (b) water / air interface. (c) water / hexane interface [112] SFG spectra of controls taken using SSP and PPP polarization combinations. (a) sapphire / water interface. (b) sapphire / air interface. The markers present the data points and the solid line is a fit to the data using equation SFG spectra of controls taken using SSP and PPP polarization combinations. (a) PVNODC (long alkyl side chain acrylate) / D 2 O interface. Left axis represents PPP D 2 O region and SSP hydrocarbon region. First right axis represents PPP hydrocarbon region and Right (blue) axis represents SSP D 2 O region. (b) PDMS / D 2 O region in PPP combination. (c) PDMS / water interface. The markers present the data points and the solid line is a fit to the data using equation SFG spectra of bulk D 2 O next to sapphire taken using SSP and PPP polarization combinations. The markers present the data points and the solid line is a fit to the data using equation 4.1 [128] SFG spectra of confined water and D 2 O taken using SSP and PPP polarization combinations. (a) confined D 2 O water region. (b) deuterium region. (c) confined water. The markers present the data points and the solid line is a fit to the data using equation xiv

16 4.14 Pictorial representation of water structure in different environments. (a) sapphire / air interface. (b) bulk water next to sapphire. (c) water next to PDMS film. (d) D 2 O next to PDMS film. (e) confined water. The gray colored region represents the sapphire substrate, the blue hashed region is either the hydration layer or hydrogenbonded water and green region represents the PDMS film or PDMS lens Coefficient of sliding friction measured as a function of time at a constant velocity of 5 µm/sec. (Red line) sliding of PDMS lens on sapphire surface only in presence of hydration layer. (Blue, black line) sliding in presence of water SFG spectra of the PDMS / sapphire interface taken in SSP and PPP polarization combinations. The markers present the data points and the solid line is a fit to the data using equation DSC heating scans for alkanes of different chain lengths. The scans are offset and C19, C21 scans are scaled for clarity DSC cooling scans for alkanes of different chain lengths. The scans are offset and C15, C19 and C21 scans are scaled for clarity Transition temperatures measured during heating and cooling cycles using DSC. The abscissa represents the number of carbon units in the alkane chain. The error bars for the data are smaller than the symbol size Predictions of 4.3 for a series of chain length of alkanes in PDMS. Also shown is the variation of interaction parameter χ with chain length DSC heating scans for C15 and C21 alkane in natural rubber. The melting point depression predicted from the Flory calculation and the bulk T m are shown as vertical lines Scan of reflected HeNe intensity with temperature for C27 alkane confined between PDMS lens and sapphire substrate. Inset shows the experimental geometry Transition temperatures measured during heating and cooling cycles using helium neon reflectivity. The abscissa represents the number of carbon units in the alkane chain. The error bars for the data measured fall within the dimensions of the symbols used xv

17 4.24 Reflected intensity versus incident angle measured for sapphire- PDMS contact, C15 bulk crystal, C15 confined liquid and confined crystal. The intensity profile for the C15 confined liquid is fit to a three layer reflectivity model (equation 4.5) to determine the alkane thickness Reflected intensity versus incident angle measured for C21 confined liquid, confined crystal and C23 confined crystal. The intensity profile for C21 confined liquid is fit to a three layer reflectivity model (equation 4.5) to determine the alkane thickness Newton rings formed on illumination of contact spot using HeNe laser. (a) shows the rings formed before contact (thickness greater than λ) and (b) shows contact area (thickness less than λ ) SFG spectra in SSP polarization combination for C15 confined alkane at different temperatures SFG spectra taken in SSP polarization combination for confined alkanes of varying chain length in liquid and crystal states. The left panel shows the liquid spectra and the right panel the crystal spectra. The chain length increases from bottom to top. The experimental data are plotted as symbols and the solid line is a fit to equation Transition temperatures measured during heating and cooling cycles using DSC, helium neon reflectivity and SFG. The abscissa represents the number of carbon units in the alkane chain. The error bars for the data measured fall within the dimensions of the symbols Schematic representation of the alkane crystals at the PDMS / sapphire interface for different chain lengths. (a) C15, (b) C19, (c) C Experimental geometry and friction forces measured. (a) Schematic of the friction cell developed to have the capability to probe molecular structure using SFG along with force measurements. (b) Schematic of the experiment with the position of laser beams at start of sliding. The SFG probe beam contact spot is ahead of the confinement region. (c) The contact region overlaps with the probe. (d) The confined region is way passed the probe spot. (e) Shear stress measured for PDMS dry sliding and in the presence of alkane as a lubricant. (f) Thickness of the confined alkane film measured using Newton-ring technique xvi

18 4.32 Shear stress measured for dry and lubricated sliding on sapphire as a function of sliding velocity. ( ) dry PDMS sliding, ( ) sliding in presence of pentadecane liquid SFG spectra acquired during frictional sliding. (a) Waterfall plot of SFG intensity measured in the hydrocarbon region as the lens is moved in and out of the laser beam. The spectra were taken at 1 second interval with 1 second of accumulation time. (b) SF spectra taken at different positions of the lens with respect to the laser spot. The spectra are normalized and scaled for clarity. (+) SF spectrum of confined crystal next to sapphire taken during heating and cooling experiments under static conditions. ( ) SF spectrum of confined liquid next to sapphire taken under static contact. ( ) SF spectrum taken during sliding when the lens has not moved into the contact spot (bulk liquid). ( ) SF spectrum of confined liquid when the lens overlaps the laser spot. ( ) SF spectrum of alkane fluid at the leading edge of the lens as the lens enters the laser spot. (c) The variation in intensity of selected peak positions as the lens moved in and out of the laser spot. (*) 2840 cm 1, (+) 2950 cm 1 and (X) 2925 cm Coefficient of friction as well as SFG spectra measured simultaneously as a function of sliding distance. Sliding speed used was 1.5 µ/sec. The SF output, input visible and input infrared were P,P,P polarized respectively. The spectra were taken at 1 second intervals with 2 seconds accumulation time. From the waterfall plot the variation in intensity of selected peak positions as the lens moved in and out of the laser spot is plotted. (*) 2840 cm 1, (+) 2950 cm 1 and (X) 2925 cm Pictorial representation of ordering of linear chain alkane molecules at the leading and trailing edge as the lens slides across the sapphire substrate xvii

19 CHAPTER I INTRODUCTION Adhesion, friction and lubrication are phenomena we experience in our day to day activities. The basic activity of walking depends on friction between our feet and the surface. Fluids in our joints provide lubrication. With the human body as a near perfect example we have developed all mechanical objects like automobiles to be based on efficient use of friction and lubrication. The phenomenon of friction is a double-edged sword, helpful in certain instances while in other cases it becomes a problem. For example, during walking or playing a string instrument like the violin we need friction to be high, but in case of bearings in motors or human joints we need friction to be as low as possible. There are several complex problems where one type of friction has to be reduced while another increased, for example tires on roads. They should have low rolling friction but high sliding friction, i.e. automobiles should travel as fast as possible and also be able to brake in a very short distance when necessary. With the rising number of automobiles (moving parts, engines), lowering friction is directly related to reduction in fuel costs, wear and tear and overall energy savings. The latest trend in miniaturization of devices, like MEMS, also provides new challenges to develop efficiently lubricated systems. To address these and many 1

20 other related problems and investigate the underlying physical principles, the field of friction and lubrication is attracting a lot of industrial and academic effort. In this regard, the present work is focused on study of confinement, friction and lubrication of simple fluids with emphasis on determining the molecular structure at the interface using nonlinear spectroscopy. Friction involves sliding of one object relative to another at an interface. Thus the nature of an interface plays an important role in determining friction. Langmuir in 1916 first recognized the fact that the outermost layer of molecules at an interface governs the interactions between surfaces [1]. By adsorbing organic films onto high energy surfaces like metals, he lowered their surface energy. Langmuir proposed that molecules at the surface tend to order into a highly packed state and expose only certain portions of the molecule which determines the surface energy. The inner parts of the molecule contribute to the surface behavior. He also studied the wetting behavior of liquids and found that high surface energy liquids bead up when they come in contact with low energy surfaces and roll off from the surface. For example, when water comes in contact with well ordered organic molecules, water beads up and rolls off the surface. This observation is important since the interaction and friction behavior in the presence of fluids like water depend on the relative surface energy and wettability. Therefore, lower wettability can lead to lower friction when the fluid is trapped between the relative moving surfaces. 2

21 Amontons in the 16th century gave an empirical relation suggesting frictional force being directly proportional to the normal load [2]. The constant of proportionality called coefficient of friction is still valid and widely used in the industry. In the 19th century Coulomb suggested splitting the frictional force into two parts: 1) The force needed to start the motion (static friction) and 2) the force needed to sustain motion (kinetic friction). Bowden and Tabor further found that friction and adhesion were related and adhesion formed an important part in determining static friction [3]. At present, it is a well accepted fact that higher adhesion leads to higher friction for smooth surfaces [1]. However, the exact relationship between adhesion and friction is still not understood [2, 4, 5]. There are new ideas which suggest friction may be related to adhesion hysteresis [6,7]. The use of lubricants to reduce friction dates back to 1400 BC. To understand the role of lubricants in reducing friction, Roberts and Tabor conducted the first systematic study on lubrication between flexible-rigid interfaces. They studied silicone oil, synovial fluid and water with salts confined between natural rubber and a glass substrate [8]. They found that the viscosity of the fluid was similar to that of bulk (unconfined) up to thickness of 25 nm. However, below 25 nm they saw an abrupt increase in viscosity. They attributed this to an increase in asperity-asperity contact. This paved the way for further research using model atomically smooth mica surfaces to confine the fluid using the surface force apparatus. The force measurements using surface force apparatus [SFA] show new features like oscillatory force profiles 3

22 with gap thickness. This oscillatory force behavior is attributed to ordering of the molecules into discrete layers as the gap approaches molecular dimensions. But this molecular picture is based on models proposed using the friction data and atomistic simulations [9]. Therefore, there is a need to develop experiments to confirm the validity of these models. In recent years it has been demonstrated that infrared-visible sum frequency generation [SFG] spectroscopy can be used to determine molecular structure at solid / solid and fluid / solid contact interfaces [10 12]. SFG is a nonlinear optical technique, which can not only detect the nature of molecules at the interface but give an estimate of the density and orientation of these molecules. This is possible due to the unique nature of the SFG process, which is generated when the inversion symmetry is broken in a material. This happens within few atomic layers next to surfaces and interfaces. Therefore, SFG provides a tool, which can selectively probe the interface, without any contribution from the bulk of the material. In this study, we have developed a novel approach to couple SFG spectroscopy to probe the molecular structure at the interface along with in-situ friction measurements. Using this technique we have revisited the experiments done by Roberts and Tabor of confining fluids between flexible-rigid interfaces. We have used the well characterized model elastic network, polydimethylsiloxane [PDMS], to generate large uniform contact areas with sapphire prism as the counter face. We have picked linear 4

23 alkanes of differing chain length and water as model confining fluids, which also have useful real world applications. The historic perspective for friction and some of the basic concepts are given in Chapter 2. It also provides a brief insight into the present day understanding of the subject. Results form both experimental and theoretical investigations are presented. At the end of chapter 2, a brief introduction to the principles of non-linear optics and the theoretical framework for SFG is presented. Chapter 2 also explains how to interpret the SFG spectra qualitatively and highlights some of the major advantages of the technique. Chapter 3 explains the experimental techniques and the materials used during the course of this study. It provides details of the prototype friction cells developed which are capable of in-situ spectroscopy along with force measurements. It gives details of the SFG system used in terms of the optical layout and the sample geometry. It also gives an insight into the working of the new femto-second SFG system installed and its ability to study dynamics of molecules at interfaces. It also briefly describes the technique of thickness measurement using Newton Rings (light interference). Another approach using critical angle measurements to determine thickness is also presented. Chapter 4 gives the details of the experimental results and their interpretation using analytical models. The chapter is split into different sections. The first section deals with static confinement of linear alkanes. The second section deals with 5

24 confinement of water. The third section describes the changes in transition temperature of linear alkanes of increasing chain length under confinement. The fourth section comprises results from friction measurements and in-situ SFG experiments to determine changes in molecular structure at the interface during sliding. Finally, this dissertation concludes with a summary of the major results of the study and a discussion of possible future work. 6

25 CHAPTER II HISTORICAL RECORD 2.1 Historical perspective Tribology is the science of interacting surfaces in relative motion. It includes the study and application of the principles of friction, lubrication and wear. The word Tribology is derived from the (tribos), meaning to rub. Every device invented by man to enhance the quality of life has involved surfaces interacting in one form or another. The engineering aspect of tribology, namely friction, has had a long history. The first practical application of generating fire by rubbing stones has its roots in prehistory. More than 400,000 years ago, our ancestors in Algeria, China and Java used friction to chip stones to create stone tools. By 200,000 BC Neanderthals had mastered the art of lighting fire by friction of wood. By 4000 BC the Egyptians had perfected the science of moving heavy stones by using water as a lubricant while building the pyramids. The Sumerians and the Egyptians recognized the value of lubricants as is evident from the chariot, dated 1400 BC, which was found in the tomb of Yuaa and Tuau with traces of original lubricant. The use of grease, oil, water and mud as lubricants has been recorded as early as 2400 BC [5,13]. 7

26 The principles of friction were studied in the 1600s by scientists like Leonardo da Vinci, and Amantons. The Amontons laws are one of the oldest physical laws known. The first states that the frictional force is directly proportional to load also known as the Coulomb friction law. The second states that the tribological friction coefficient is independent of the contact area and loading force, i.e. µ is approximately a constant 1 [14,15]. 2.2 Friction Bowden and Tabor were the first to explain Amontons laws stating that µ is a constant 1 [16]. Their explanation is based on the assumption that the real surfaces are rough; therefore, the real contact area is very small and proportional to the load. If a load L is placed on a substrate, a single contact junction is formed first. The perpendicular pressure increases and plastic deformation of the junction takes place before the next junction is formed. Therefore, the real area of contact, A = L/σ c, where σ c is the penetration hardness or the largest compressive stress the material of the substrate can withstand without plastic yielding. To illustrate how plastic yield stress can be used to estimate real area of contact Persson assumed a steel cube of 10 cm side was placed on a steel table. The penetration hardness of steel is σ c 10 9 N/m 2 and since the load L=Mg 100 N this gives the area of real contact A 0.1 mm 2, i.e., only a fraction 10 5 of the apparent area. That being the case, the force F needed to shear the junctions with total area A is F = τ c A. Where τ c is the yield stress during 8

27 shear. But A = L/σ c giving F = (τ c /σ c )L, i.e., friction coefficient µ = F/L = τ c /σ c. Since τ c and σ c are of similar magnitude, µ 1. Figure 2.1: A block with mass M sliding on a flat surface. A spring (spring constant K s ) is connected to the block and the free end is pulled with velocity V s. Devices used in friction studies can be schematically represented as shown in figure 2.1. A mass M is moved with a velocity V s through a spring K s, where the spring K s can be an external elastic component or represent the elastic response of the system. Figure 2.2 shows the spring force as a function of time for three different cases. Assuming the motion starts from rest and the spring is in its undeformed state, on pulling, the force linearly increases with time while the block is still stationary. When the spring force reaches a critical value F a, i.e. the static friction force, the block starts to move. The motion can be categorized as either steady motion where the spring force equals the kinetic friction force F b, or stick-slip motion, where the motion is interrupted (Figure 2.2(b) or (c)). Most of the work done on sliding friction 9

28 Figure 2.2: Sliding dynamics of the block in figure 2.1 for three different cases; (a) steady sliding (b) periodic stick-slip motion (c) chaotic motion [5]. is to map out the region of steady sliding and stick-slip motion. The dynamic phase diagram is as shown in figure 2.3. It has been shown by experiments that stick-slip motion disappears if a stiff spring is used or the velocity is increased beyond a critical value. The phase diagram and the nature of the transition between stick-slip and smooth sliding either being continuous or discontinuous are important Stribeck curve and its limitations Tribologists have traditionally used the Stribeck curve to represent frictional behavior. Figure 2.4 shows the microscopic and macroscopic parameters involved and the generalized stribeck curve. It is based on macroscopic parameters like the bulk viscosity η b, derived from the famous Reynolds equation to describe friction regimes [3]. The curve describes the force experienced by the lubricant with changes in sliding velocity V, film thickness D, contact area A and bulk viscosity. Velocity and film thickness together are represented as the shear rate γ. This view of lubrication assumes that the changes in V, D and η b produce effects so similar that they exactly 10

29 Figure 2.3: General kinetic phase diagram. The hashed region depicts the values of K s and V s for which stick-slip motion occurs [5]. Figure 2.4: Microscopic and macroscopic parameters affecting frictional behavior. Generalized Stribeck curve [17]. 11

30 offset one another. But there are some issues with this approach. The parameters being used are bulk values and the exact values during sliding are not taken into account. It is also assumed that a doubling of sliding velocity has the same effect as reducing the thickness by half. This has not been experimentally verified. Therefore, to improve the Stribeck curve requires using parameters which are measurable during confined lubrication or sliding. One such parameter is the effective viscosity, η eff which is always greater than η b. Luengo et.al, have tried to combine the WLF representation used in polymers with that of the Stribeck parameters to come up with a modified Stribeck curve which can be established by measuring parameters in situ not only in the laboratory but under engineering conditions [17]. The proposed friction map is shown in figure 2.5. Experiments conducted by Dhinojwala and Granick [18] show that confined fluids flow smoothly provided the velocity is sufficiently low; flow becomes intermittent stick-slip when the shear rate is higher than the inverse of the natural relaxation time. The connection between stick-slip and forced motion at rates higher than the inverse relaxation time can help in predicting the onset of stick - slip motion. The physical interpretation is that when the lubricant is deformed at a speed which it cannot keep up with, elastic forces build up to a point at which catastrophic failure occurs and we see stick - slip behavior. So the friction map in figure 2.5 represents the effect of different parameters like load (L), velocity (V), distance between the surfaces (D) on friction force and indicates the various regimes. It can be seen that at zero load (L=0, D ) force increases 12

31 Figure 2.5: Proposed friction map of force versus sliding velocity showing different regimes [17]. linearly with velocity (newtonian behavior; η eff γ) and shows non-newtonian shear thinning at very high velocities. As load increases, that is D decreases, the newtonian flow in the Elastohydrodynamic [EHD] regime crosses into the boundary regime of lubrication. It can be seen that friction force reaches a maximum when the shear rate equals the inverse of relaxation time (Deborah number, D e =1). The effective viscosity is no longer newtonian but shows a power law dependence; (η eff γ n ) with shear rate where n varies from -0.5 to This regime occurs between lower and upper critical velocities. As the load increases (small D) solid-like creep behavior can be seen with solid-liquid transitions or stick-slip behavior. In this regime the 13

32 molecular structure and the surface lattice become important rather than the bulk viscosity [17]. Yoshizawa and Israelachvili studied the friction and rheological properties of molecularly thin films of hexadecane between two shearing mica surfaces under different conditions of load or applied pressure, film thickness, sliding velocity, temperature, and time or previous history [7, 19]. They also compared the results obtained from spherical molecules like octamethylcyclotetrasiloxane [OMCTS]. The main emphasis was on the onset of stick-slip sliding, velocity dependence, structure of molecules at the onset etc. The main results were: The interfacial friction of short, saturated chain molecules was not basically different from that of simple, spherical molecules. For both types of liquids, steady state sliding occurred after the molecules had become shear ordered into layers, similar to but not the same as the equilibrium layering that gives rise to the oscillatory interaction potential between two surfaces in a liquid [7]. The time to order or layer was found to be longer for chain molecules than spherical molecules, and longer chain molecules like polymers did not show the layering effect. Both molecules showed stick-slip sliding below a certain critical velocity V c. The critical velocity was found to be well described by the following two equations. V c C F/Kτ 0 (2.1) Where C is assumed to be equal to 1, F is the difference between static friction force, F s and kinetic friction force, F k, K is the spring constant and τ 0 is the characteristic nucleation time. Based on simulations Robbins and Thompson [20] proposed the 14

33 expression shown in equation 2.2 V c C F s σ/m (2.2) Where C 0.1, M is the mass of the stage and σ is the characteristic molecular distance. Equation 2.1 can be applied to films composed of chain molecules, while equation 2.2 is more applicable to spherical molecules. A new dimensionless number given by equation 2.3, was proposed by Robbins and coworkers [20]. It is similar to the Deborah Number and it was found to serve as a suitable criterion for smooth or stick-slip motion. S τ 0 /τ t (2.3) where S is called the stick-slip number and τ t is the mean transit time, i.e. the average time it takes a surface to traverse the characteristic slip-stick distance. With increase in temperature or velocity it was found that τ 0 increased as the liquid became more fluid like and τ 0 was found to be constant above a certain critical temperature or velocity. In this regime of high V and T, the nucleation time τ 0 was determined by the load L through its confining effect. Many of the effects observed could be further correlated in ways that resemble the rheological behavior of bulk polymer fluids. In this regard, the WLF equation [21] based on the principle of time-temperature superposition was used to propose a modified activation energy model. The relaxation time of the energy dissipating process was found to be applicable to frictional energy dissipation in the absence of stick-slip [7]. 15

34 To understand energy dissipation in stick-slip, Jacob Klein studied the frictional sliding of OMCTS molecules in the surface force apparatus [22]. High spatial and time resolved measurements demonstrated that energy is dissipated both by momentum transfer to the external system at the instant of stick, and by viscous heating during the slip of the shear-melted film. The analysis showed that the dissipative loss over a stick-slip cycle was primarily (90 % or more) via the viscous-heating mechanism during slip, with the residual kinetic energy transferred to the system at the point of stick accounting for the remaining part [22]. Friction between surfaces with neutral polymer brushes was found to be drastically smaller [23, 24] and hydrated ions can act as extremely efficient lubricants between sliding charged surfaces [25]. Raviv et al., combined these properties and studied brushes of charged polymers called polyelectrolytes, attached to surfaces rubbing across an aqueous medium. It was observed that effective friction coefficients with polyelectrolyte brushes in water were lower than 0.001, even at low sliding velocities and at pressures of up to several atmospheres. They attributed this low friction to the exceptional resistance to mutual interpenetration displayed by the compressed, counterion-swollen brushes together with the fluidity of the hydration layers surrounding the charged rubbing polymer segments [25,26]. Isarelachvili and coworkers studied ZnS nanorods coated with surfactant and dissolved in dodecane as lubricants to optimize tribological performance over a large range of loads and sliding speeds obtaining friction coefficients of approximately

35 without damage over a large range of sliding speeds up to pressures of 200 MPa. They proposed that the superior properties in the presence of nanorods was due to a combination of the nanoparticle geometry, nanoparticle stiffness, and nanoparticle interactions. In other words, weakly adhesive inter-nanoparticle and nanoparticle surface forces due to a synergy between the nanoparticle cores, the surfactant layers, the solvent, and the surfaces lead to lower friction [27] Rubber friction Friction when a rubber surface slides on a hard substrate has immense practical importance, for example, in the design of tires, wiper-blades, etc. Rubber friction is quite different from friction of other solids in the sense that rubber has low elastic modulus and has high internal friction or internal energy dissipation over a wide range of frequencies. The rubber friction force has two main contributions commonly called adhesive and hysteretic components, respectively. The hysteretic component is due to the internal friction or the energy dissipation in the bulk of the rubber. Consider a rubber sliding on a rough surface; as the rubber moves over the asperities, it deforms because of the low modulus and this creates oscillating forces at the rubber surface which leads to cyclic deformation of the bulk leading to dissipation via internal damping [5]. Friction coefficients show a similar temperature dependence as the elastic modulus, E, therefore, theories developed for rubber elasticity can be used to explain energy dissipation. The coefficient of friction can be expressed in terms of E and frequency, ω 0 as given by the equation 2.4: 17

36 µ C Im(E(ω 0)) E(ω 0 ) (2.4) Where C is a number of order of unity which depends on surface roughness. Therefore, on a hard rough surface the maximum coefficient of friction is of the order of unity. Using the viscoelastic behavior of rubber over a wide frequency range and the surface roughness, the behavior of the coefficient of friction with velocity can be predicted. Persson has provided an excellent review on this subject including a good example of sliding friction [5]. Considering the other component of friction, adhesion or interfacial adhesion can either increase or decrease the oscillations at the rubber surface. Figure 2.6 shows different scenarios of rubber sliding on a rough surface. For a hard rubber surface, the area of contact is smaller and hence the adhesion (friction) is lower. A soft rubber surface on the other hand, wraps along all the undulations on the surface, thereby increasing adhesive area and hence friction. If we have particulate matter on the surface and a hard rubber surface, once again the area of contact is lower and hence lowers friction. Similarly, if water or lubricant is present it fills in the asperities thereby reducing the viscoelastic deformations in the rubber and hence lower friction. 2.3 Lubrication The practical use of lubricants to reduce friction dates back to thousands of years. But the physical understanding of the subject can be traced to the 1800s. Tower con- 18

37 Figure 2.6: Behavior of rubber at the interface. (a) Rubber on a hard surface with long-wavelength roughness. (b) Soft rubber deforms in such a way as to completely follow the short-wavelength roughness. (c) Rubber surface dusted with small particles sliding on a hard substrate. (d) Rubber sliding on water [5]. 19

38 ducted experiments to determine the friction between journal bearings in The pressure generated in the oil holes of the journal bearings in the wheels railway-car was found to be sufficiently large thus capable of displacing the plugs covering the feed port. Intrigued by this phenomenon Reynolds started applying the principles of fluid mechanics to solve Tower s problem. In 1886 he published the classical theory of hydrodynamic lubrication which is in use even today for designing modern machinery. In hydrodynamic lubrication the solid walls are separated by fluids of thickness 10µm. In this case the physical problem can be solved by using the Navier-Stokes equation of hydrodynamics with appropriate boundary conditions and geometry of contact. The kinetic friction is given by the viscosity of the liquid lubricant [Reynolds 1886] [13]. Another mode of lubrication is when the thickness of the lubricant layer is a few nanometers. In 1922 an English biologist Hardy introduced the term Boundary lubrication to describe this mode of lubrication. He found that when two glass surfaces had absorbed thin layers of lubricant they slid over each other with very low friction. He concluded that under these circumstances the lubrication depended completely on the chemical composition of the fluid and not on its viscosity, and that a good lubricant was a fluid which was adsorbed by the solid surfaces. His experiment showed that fatty acids gave the lowest sliding friction in boundary lubrication experiments in comparison to paraffins and alcohols. The chemical composition of fatty acids includes a polar head and a long hydrocarbon tail. The polar head tends to bind relatively strongly to polar surfaces such as metal oxides (see Figure 2.7(a)), thereby 20

39 lending itself as a better lubricant. Hardy pictured a sliding interface as consisting of solid surfaces covered by monolayers of fatty acids with the hydrocarbon tail pointing away from the surfaces; the sliding occurs at the plane where the hydrocarbon tails meet as illustrated in Figure 2.7(b). Figure 2.7: Boundary lubrication. (a) The polar head group in the fatty acid binds to the metal oxides (b) Hardy s concept of boundary lubrication [5]. From the early 1980s there has been increased effort to understand sliding in case of thin or boundary lubricants. Ideally, during lubricated sliding 1) the adhesion is zero, 2) the friction forces are low, and 3) the surfaces are protected from being damaged by the repulsive forces. Under high loads the liquids can get squeezed out and the thin layer can also freeze becoming amorphous or solid-like [28]. Adhesion and friction in these scenarios suddenly increase leading to seizure. The concept of solid or liquid film when the thickness of the film is only a few molecules 21

40 is problematic and other effects like substrate lattice structure, the sliding velocity influence the sliding behavior observed Surface force apparatus The development of the SFA in the 1970s helped to bring atomically smooth surfaces close and study the friction of films of thickness a few angstroms [29 31]. One of the versions of SFA developed by Isarelachvili is shown in figure 2.8. Figure 2.8: Schematic of surface force apparatus. Also seen are two different types of force measurement springs which are interchangeable allowing up to 8 orders of force measurement [32]. 22

41 The SFA in figure 2.8 can be used to measure forces in controlled vapors as well as completely immersed in liquids. The distance resolution is about 0.1nm and the force sensitivity is about 10 8 N. Many versions have been developed by other scientists including Klein, Parker and Granick s groups [33,34]. The SFA contains two curved molecularly smooth surfaces of mica of radius 1cm between which forces are measured. The two surfaces are in cross cylinder configuration which is equivalent to a sphere next to a flat surface. The separation between the two surfaces from microns down to nanometers is measured by an optical technique using multiple beam interference called Fringes of Equal Chromatic Order (FECO). The mica sheets are partially silvered at the back and white light is passed through the contact area. The interference of light between the two coated mica sheets produces dark and bright fringes. From the position and shape of the fringes the distance between the two surfaces can be accurately measured down to 0.1nm. The space between the surfaces can be filled with the test liquid and the surface of mica can also be coated with various surfactants and polymers. The force is calculated by using the stiffness of the cantilever spring. Further details of the setup can be found in reference [30]. Using the SFA it was shown that when a liquid is confined between two molecularly smooth walls and the gap between the walls is reduced to molecular dimensions, an additional force (other than the ubiquitous van der Waals force) arises. This force is often called the structural, solvation, or oscillatory force. The origin of this force was understood in terms of the overlapping of the two density distribution 23

42 profiles of the liquid molecules from each wall [35,36]. The density distribution of the liquid molecules from one wall has a decaying oscillatory profile, with a periodicity close to some characteristic dimension of the confined molecule (the diameter if the molecule is spherical). Certain discrete or quantized gap widths were found to be energetically unfavorable. Figure 2.9 shows the oscillatory force data for OMCTS which is a non polar nearly spherical molecule between mica surfaces. Figure 2.9: Oscillatory force profiles for OMCTS liquid confined between mica surfaces [28]. 24

43 The curves shown are a collection of results from various groups who have studied oscillatory forces for OMCTS in SFA. The force is found to be an oscillatory function of distance varying between attraction and repulsion with the periodicity equal to the diameter of a liquid molecule. The oscillations decay rapidly with distance disappearing after 6-8 molecular diameters. The oscillations are found to be independent of temperature but dependent on the nature of surfaces and the presence of water Classical theory of forces in liquids between surfaces Forces between surfaces immersed in liquids can be classified into four main types, (1) Van der Waals forces, which are monotonically attractive and act between all molecules, (2) repulsive electrostatic double-layer forces, which are present when surfaces have a net electric charge like that in water (3) solvation (structural or hydration) forces, which arise due to ordering of liquid molecules when confined between surfaces. These forces can be attractive, repulsive or oscillatory, (4)repulsive entropic, also known as steric or fluctuation forces, which arise due to thermal motions of protruding surface groups like polymer head groups or thermal fluctuations of flexible fluid interfaces like surfactants and lipids. The van der Waals and electrostatic were the earliest forces to be considered and studied. This led to the famous DLVO theory named after Derjaguin, Landau, Verwey, and Overbeek who proposed it. According to the theory, if there was a net electric charge the force between surfaces 25

44 was repulsive and if it were absent, then van der Waals forces dominated and the surfaces were attractive (Figure 2.10) [32]. Figure 2.10: DLVO interaction potential energy E as a function of separation distance D between two flat surfaces interacting through an aqueous salt solution with attractive van der Waals force and repulsive electrostatic double-layer force [32]. But it was found that in the case of colloids, clay sheets and lipid bilayers in aqueous conditions, the DLVO theory could be applied only up to a certain thickness ( 10 molecular diameters). As the thickness was lowered it was found that the liquid could not be treated as a structure-less continuum and the monotonic decay of the van der Waals force needed an additional solvation force with oscillatory force profile 26

45 to account for the observed experimental behavior. Figure 2.11 shows the oscillatory behavior of force with distance for aqueous Potassium Chloride solution going from attraction to repulsion with a periodicity equal to the solvent molecular diameter. Figure 2.11: Experimental and theoretical interaction potential between mica surfaces in a 10 3 M KCl solution. As the solution is more diluted DLVO behavior is observed, but if the electrolyte is concentrated, in short range (small separations) oscillatory profile within the monotonically repulsive tail with a periodicity equal to water molecule (2-2.6 nm) is observed. The inset shows a theoretical computation for the same system [32]. 27

46 The entropic repulsion force has been found in amphiphilic molecules like lipid bilayers and surfactants in aqueous solution where the oscillatory force is replaced by a monotonic repulsive hydration-fluctuation force which has a rather long range of about Angstroms. The oscillatory solvation force is found to be damped due to the thermal fluctuations of the head groups [32] Extrusion of liquids Extrusion of liquids trapped between rubber and glass as in the case of windshield wipers has drawn a lot of attention [37]. In investigations it was found that the windshield wipers carry out two separate functions for which two different characteristics are desired. On one hand, when the screen is wet the wiper needs to be soft and smooth so that it can spread the water into a thin film in situations when good wetting is desired. On the other hand when the glass is dry it has to scrape dirt and insects off the screen. A soft rubber surface will then exert an impracticably large frictional resistance. Under these conditions a hard rubber surface is desirable [8]. To gain a better understanding of this effect, Roberts and Tabor studied drainage of silicone oil, water with salts, synovial fluid and saliva between smooth polyisoprene rubber hemispheres and a glass substrate. They found the viscosity of the fluids to be the same as that of bulk liquids up to a thickness of 25 nm. And below 25 nm they observed an increase in viscosity which they attributed to asperity contacts and not due to any form of ordering in the liquid molecules (Figure 2.12). 28

47 Figure 2.12: Variation in friction of wiper blade with various lubricants.(o) silicones; ( ) alcohols; ( ) glycerol solutions; ( ) teepol solutions [8]. 29

48 Mugele and co-workers have studied the dynamics of drainage of liquids between smooth elastic walls using the Surface Force Apparatus [SFA] [38 42]. It is known that when a liquid is pressed between smooth surfaces, continuous drainage is observed till the properties follow the behavior of the bulk liquid. But, as it reaches molecular dimensions drainage occurs in discrete steps suggesting layering of the molecules. The expulsion of individual layers of molecules has been termed as a layering transition. Persson and Tossati [43] have presented a model to describe the nucleation and subsequent dynamics of layering transitions. In their model, they consider a film of thickness (number of layers) n, which is in a metastable state because of the applied normal stress being higher than the threshold pressure for that thickness. In such situations thermal fluctuations in the film create a small region with radius ρ where the liquid density is reduced. Because of the reduced density the film can no longer sustain the normal force exerted by the substrates. Hence the substrates relax elastically into the film thickness with the radius ρ. The nucleus of thickness (n-1) grows, since the energy gained due to the elastic relaxation exceeds the restoring forces that tend to close the hole. The energy of the nucleus was suggested to have three major contributions. U(ρ) = 2πρΓ + πρ 2 p o (1 ν 2 ) P 2 3D c E ρ3 (2.5) 30

49 The first term is the line energy along the circumference of the nucleus with Γ representing the line tension. The second term is the change in interface free energy with spreading pressure p o. The last term in the equation represents the elastic relaxation energy of the substrate within the area of the nucleus. The contribution that stabilizes the nucleus is given by E, where E represents the elastic modulus, ν the Poisson ratio and c a numerical constant of the order of unity. It can also be inferred that for softer substrates ρ will be smaller. The local relaxation of the substrates reduces the potential barrier for nucleation of layering transitions thus destabilizing the lubricant layer. Once a sufficiently large nucleus is formed it spreads over the whole area in a short time. Mugele and coworkers studied the effect of approach velocity and substrate elasticity experimentally using OMCTS as the model lubricant [44]. They used different thickness of mica sheet to change the substrate modulus with thinner sheets giving lower modulus due to the underlying glue used to fix the sheets in cross-cylinder geometry of SFA. For thicker mica sheets they saw no dependence on rate of loading and layering transitions in agreement with the Persson s theory. But for thinner mica (lower E) they saw no layering for lower rates up to the critical rate and above the critical rate they noticed trapping of liquid in the contact zone. They attributed this to the outerward bending of the soft substrate on application of pressure moving the maximum pressure line to the edges thus moving the liquid inward creating a droplet with 4 molecular layers in their case. With further increase in pressure and with time 31

50 the trapped liquid drained out of the contact zone and the trapped liquid also showed a layering transition. 2.4 Simulation of confined systems Gao, Luedtke and Landman studied equilibrium structures, solvation forces, and conformational dynamics of thin confined films of spherical molecules (modeled as Lenard-Jones [LJ] spheres), straight-chain alkanes of variable lengths (n-hexadecane and n-tetracosane), and a branched alkane (squalane), using a grand canonical ensemble molecular dynamics method [45, 46] for simulations of confined liquids. This method combined constant pressure simulations with a computational cell containing solid surfaces and both bulk and confined liquid regions in equilibrium with each other (figure 2.13), thereby allowing for investigations of confined films under experimental conditions similar to those in a laboratory. They studied the nature of equilibrium of confined liquids and the effects of molecular structure (shape, chain length, and branching) on the properties of such systems. For all the liquids they investigated, layered density oscillations in the confined films were found with the number of layers depending on the width of the confinement. The LJ and straight-chain alkane confined films showed enhanced layering order as well as a higher degree of in-plane molecular ordering, in comparison to those found for the squalane film, with the latter showing a high tendency for interlayer molecular interdigitation. In regard to transition between different states 32

51 Figure 2.13: Grand canonical ensemble 3D model for studying confined liquids between solid walls in equilibrium with bulk liquid [46]. (trans-guache) they found reduced conformational transition rates in the confined alkane films in comparison to its bulk [46 50]. Gao and Landman also studied the effect of surface roughness on the structure, solvation forces and the interfacial slip during sliding using molecular dynamic simulations. They used hexadecane molecules as a model lubricant and developed a gold surface as the flat surface. To create a rough gold surface, the temperature was increased beyond the melting temperature to create random fluctuations in the surface molecules and the system was then quenched below melting temperature. The solvation force calculated as a function of confined fluid thickness for flat and rough surfaces is shown in figure It was found that the flat surface showed an oscillatory force profile but as the surface became rough the oscillations disappeared showing a monotonic repulsive profile [45,49]. For the static case the reduced ordering 33

52 Figure 2.14: Solvation force profiles for hexadecane molecules confined between gold surfaces. (o)rough surface, (*) flat surface [45]. propensity of the rough-surface confined film was exhibited by significant inhibition of the development of density-layered (stratified) structures in the film with consequent strong suppression of the development of solvation forces. In its place there was a liquid-like response with higher molecular mobilities, and continuous expulsion of molecules from the confined region in response to reduction of the gap-width. The difference in the nature of the molecular films in flat and rough surface confinements was also evident in the characteristics of the free-volumes calculated for the two systems. The flat surface confinement exhibited local minima at gap widths corresponding to well-layered film configurations, while the rough surface confinement exhibited a free volume decrease with the gap width in a smooth monotonic manner. When the surfaces were sheared, in case of the flat surface, interfacial slip was seen with very small energy dissipation or shearing inside the lubricant film. But in the 34

53 case of rough surfaces, a layer of the film was found to attach itself to the surface thereby moving the slip plane inside the film instead of the interface. This increased the shear dissipation and the viscosity of the confined liquid was found be inversely proportional to the free volume. Thus, the viscosity of the fluid was lower leading to higher energy dissipation in the case of rough surfaces due to the higher free volume [45]. In another simulation they added external vibrations to the surfaces and showed that the kinetic friction could be drastically reduced [50]. Control of friction in the lubricated junction was demonstrated with a transition from a high-friction stick-slip shear dynamics of the lubricant to an ultra low kinetic friction state, termed as a super kinetic friction regime, occurring for Deborah number values D e > Sum frequency generation spectroscopy The invention of Light Amplification by Stimulated Emission of Radiation or in short LASER in the middle of the 20th century gave a new source of monochromatic high power uniform light beam [51]. This led to the discovery of phenomena like frequency doubling of light known as SHG or second harmonic generation and frequency mixing of light with possible summing and difference of frequency known as sum frequency generation (SFG) and difference frequency generation (DFG) respectively. Most of the theoretical and experimental framework was done in the 1960s by Bloembergen and was termed as an extension of classical optics [52]. These and other related phenomena are collectively called Nonlinear Optical (NLO) effects. 35

54 In 1987, Shen and coworkers showed SFG surface vibrational spectra from methanol and pentadecanoic acid on water [53] using the optical parametric amplification technique they had developed to generate pulsed tunable infrared light. At the same time, Harris et al., at Bell labs, generated vibrational spectra of Langmuir Blodgett films of cadmium arachidate on silver and octadecane thiol on gold using Raman shifting technology [53]. The first SFG spectra from polymer surfaces were published in 2000, in external geometry by Zhang et al. [54], and novel internal reflection geometry by Gautam et al. [55]. Further details of the history and refinement of the NLO phenomena, laser technology, development of SFG instrument and derivation of the relevant optical expressions presented can be found in other literature Basic principles The optical phenomena encountered are governed mainly by the principles of linear optics [56]. It is known that when a molecule is placed in a electromagnetic field, the interaction of the field with the molecules induces a dipole, µ, whose strength depends on the strength of the electric field applied µ = µ 0 + αe (2.6) where E is the electric field, α the molecular polarizability and µ 0 is the static dipole. The induced dipole is not necessarily in the direction of the electric field and α is given by a (3X3) tensor. In condensed phases the expression is modified in terms of dipole moment per unit volume, referred to as the polarization P. The molecular 36

55 polarizability gets replaced by linear susceptibility, χ (1) and the static dipole by static polarization P 0. The modified equation can be written as: P = P 0 + ǫ 0 χ 1 E; χ (1) = Nα/ǫ 0 (2.7) Where ǫ 0 is the dielectric constant. For polymers P 0 is very small and hence neglected. The linear susceptibility for an isotropic molecular material is proportional to the number of molecules, N and the molecular polarizability of individual molecules averaged over all their orientations < α >, and inversely proportional to its dielectric constant [57]. When the incident beam becomes intense, as in the case of high energy pulsed lasers, the electric fields become comparable to the fields felt by electrons in the molecule. This makes the response of the electrons to the incident field nonlinear or the electrons are no longer in harmony with the field. Thus, the need to add additional higher order terms to the polarization equation arises. The molecular polarizability and the polarization in case of condensed media can be written as: µ = µ 0 + αe + β : EE + γ : EEE +... (2.8) P = P 0 + ǫ 0 (χ (1).E + χ (2) : EE + χ (3) : EEE +...) (2.9) where β and γ are known as first and second order hyperpolarizabilities. The symbolic notation β : EE is a short form for j,k β ijke j E k, where i,j,k represent the cartesian coordinate system and β is a 3X3 tensor. χ (2) and χ (3) are second and third order susceptibility respectively and χ (2) is a third rank tensor and χ (3) is a fourth rank 37

56 tensor and so on. The additional consequence of these higher order terms is that the refractive index and absorption coefficient also becomes intensity dependant. A more interesting effect is that the frequency of light coming out of the material undergoes a change [52,57,58]. The input electric field is oscillating with time, t at a frequency ω which can be expressed by replacing E with Ecosωt. Then using linear optics it can be shown that: P (2) = χ (2) : EE = ǫ 0 χ (2) : EEcos 2 ωt (2.10) = 1 2 ǫ 0χ (2) : EE(1 + cos2ωt) The first term inside the bracket represents the static polarization and the second term represents the polarization oscillating at a frequency 2ω, resulting in the output intensity oscillating at 2ω. This is the basis of second harmonic generation or SHG [59]. Now consider the case where instead of one input frequency there are two input beams with frequencies ω 1 and ω 2. The polarization in this case can be written as: P (2) = 1 2 ǫ 0χ (2) : EE(cos(ω 1 + ω 2 )t) = 1 2 ǫ 0χ (2) : EE(cos(ω 1 ω 2 )t) (2.11) Herein lies the basis of NLO phenomena of sum frequency generation, SFG and difference frequency generation, DFG. Let us consider some of the properties of χ (2). As mentioned before, χ (2) is the third order symmetric tensor with 27 elements represented as χ (2) ijk. Therefore, if we perform an inversion operation on the tensor, then χ (2) ijk = χ(2) i j k. As a consequence χ(2) should be equal to zero for the above 38

57 equality to be true (ie., χ (2) i j k = χ(2) ijk ) in media which are invariant under inversion symmetry operation. As a result, SFG is forbidden in media which are invariant or in other words centrosymmetric [59]. Most of the bulk of material is centrosymmetric due to isotropic distribution of dipoles. But when the bulk is cut to create surfaces, the centrosymmetry is broken. Therefore, for isotropic materials the symmetry breakdown occurs perpendicular to the interface. So out of the 27 components of the second order nonlinear susceptibility, for azimuthal isotropic interfaces, only seven components can be disregarded and of those only four are independent [59,60]. In terms of cartesian coordinates: χ xxz = χ yyz, χ xzx = χ yzy, χ zxx = χ zyy and χ zzz. In case of infrared visible SFG, polarized visible beam at a fixed frequency ω 1 and Figure 2.15: Schematic of the beams for SFG at an interface. (a) external geometry assuming n 2 > n 1. (b) internal reflection geometry n 1 > n 2. polarized tunable infrared beam, ω 2 are overlapped in time and space at a surface or interface. The beams involved are shown in figure The output emission is a 39

58 coherent beam at a frequency (ω vis + ω IR = ω 3 ) in both reflection and transmission. The angles of incidence are known and hence the output direction can be determined by applying conservation of momentum. In most cases the reflected beam is used for detection and the relation governing conservation of momentum is given below: 2πω 3 c sinθ ω3 = 2πω 1 c sinθ ω1 = 2πω 2 sinθ ω2 (2.12) c where c is the speed of light. Since we are probing the surface or the interface, a thin polarized slab can be considered as the region from which SF is being generated. By choosing either s or p polarization for input beams and selecting a polarization for detection by placing a polarizer in the output beam, the four independent components of χ (2) can be probed. Here, s and p stand for light polarized perpendicular to the plane of incidence and in the plane of incidence respectively. The convention used when SSP, SPS and PPP combinations are mentioned is that the first letter corresponds to the polarization of SFG output, the second letter to the polarization of the visible beam and the last letter to the polarization of IR beam. The expressions for the relationship between the components of χ (2) probed and the combination of polarizations used was given by Zhuang and Miranda [61]. According to their convention, the susceptibility, χ (2) for a given combination of input visible, input IR and output SFG say SSP is replaced by an effective susceptibility, χ (2) eff,ssp. The relationships are given below: χ (2) eff,ssp = [T s(ω 3 )L yy (ω 3 )][T s (ω 1 )L yy (ω 1 )][T p (ω 2 )L zz (ω 2 )]sinθχ yyz (ω 2 ) (2.13) χ (2) eff,sps = [T s(ω 3 )L yy (ω 3 )][T p (ω 1 )L zz (ω 1 )][T s (ω 2 )L yy (ω 2 )]sinθχ yzy (ω 2 ) (2.14) 40

59 T s and T p are the transmission fresnel coefficients for s and p polarized light. They account for the losses due to reflection of the input and output beams before reaching the interface and emission out of the sapphire prism. L yy and L zz are the fresnel coefficients which account for the microscopic field fluctuations due to any dipole dipole interaction of molecules within the polarization sheet. The relationship for PPP becomes more complex with all the independent components of χ (2) coming into picture. In most cases, though PPP spectra are used occasionaly, SSP and SPS spectra are mainly used to compute quantitative results. Therefore, the intensity of SF signal is proportional to the χ (2) and the input electric fields. Using a simplified approach, SFG intensity, I ( ω 3 ) = (E ( ω i )) 2. Therefore, the equation for emitted intensity is given by: I ( ω 3 ) = 8Π 3 ω 2 sec 2 θ 3 c 3 n 0 (ω 3 )n 0 (ω 1 )n 0 (ω 2 ) χ(2) eff 2 I ( ω 1 )I ( ω 2 ) (2.15) Additionally, the effect of the molecules present at the polarization sheet has to be considered. The components of the nonlinear susceptibility χ (2) ijk can be written as: χ (2) eff,ijk = (χnr ijk exp(iϕ) + ν χ R ijk,ν) (2.16) The first term χ NR ijk, is called the non-resonant term and depends on the continuum dielectric properties of the material and is associated with a phase ϕ. The second term is the resonant contribution which arises due to the excitation of molecules into higher energy levels by absorption of infrared light. This absorption 41

60 occurs when the frequency ω 2 matches with the resonance frequency ν. Since the SFG signal is proportional to the square of equation 2.16, it follows that the resonant and non-resonant terms can interfere either constructively or destructively. However, the extent of interference depends on the relative phase of χ NR ijk. Substituting equation 2.16 and simplifying equation 2.15 we get χ (2) ijk,ν = (χnr ijk exp(iϕ) + ν A ijk,ν ω ν ω IR iγ ν /2 (2.17) Each vibrational mode of the surface molecule contributes to χ (2) ijk,ν, which is characterized by the resonant frequency ω ν, the damping constant Γ ν, and the line amplitude A ijk,ν. Scanning the IR frequency leads to the resonantly enhanced SFG signal proportional to χ (2) ijk 2 at ω ν with line-width (or full width at half maximum, FWHM) of Γ ν. Hence, by fitting the obtained SFG signal we can determine the parameters, ω ν, Γ ν, and A ijk,ν, for each resonant vibration. The fitting program used for our spectra analysis was developed in our laboratory by our previous group member Dr. Alexander Schwab [62] Typical SFG spectrum and how to interpret the same Most of the work involved in this dissertation has been based on using a model fluid to understand the phenomena of confinement and friction. Therefore, the fluids chosen are series of n-alkanes having methylene(ch 2 ) and methyl (CH 3 ) groups and differing only in the number of methyl groups present in the chain. The infrared region in which these groups absorb is from 2750cm cm 1. So most of the SFG spectra are 42

61 reported for this region. Before proceeding further to look at relevant SFG results, it is useful to know the vibrational assignments for the groups involved. In general for the C-H stretching modes of methyl and methylene groups the vibrational assignments are transferable from one molecule to the next in the absence of connections to inductive non-carbon hetroatoms [63]. Therefore, for majority of the methyl and methylene groups observed in these studies the vibrational frequencies established by Synder and coworkers based on extensive IR and Raman measurements of bulk alkanes and polyethylene apply [64 66]. Following the convention of Synder, methyl modes are labeled r and methylene modes are labeled d, with superscripts + and - distinguishing between symmetric and asymmetric modes, relative to the respective group symmetry axis. Table 2.1 summarizes the corresponding vibrational assignments for these groups. Figure 2.16: Normal vibrational modes, notation and co-ordinate system for methyl and methylene groups. 43

62 Table 2.1: Vibration assignments for the hydrocarbon stretching region. Mode Description of the vibration Peak position d + Symmetric CH 2 stretch 2850 r + Symmetric CH 3 stretch 2880 d FR Asymmetric CH 2 stretch Fermi resonance 2900 d Asymmetric CH 2 stretch 2916 r +FR Symmetric CH 3 stretch Fermi resonance 2940 r Asymmetric CH 3 stretch 2959 Figure 2.16 illustrates the methyl and methylene C-H vibrational modes. The r a and r b modes arise from splitting of the r mode as a result of lowering the methyl group symmetry from C 3v to C s, when the methyl group is attached to an alkyl chain [66]. But, when the methyl group becomes free to rotate around its c axis, for example when the methyl group is outside the crystalline region, the splitting disappears and the two split modes recombine and become degenerate as the lowered group symmetry is removed [67]. In addition to stretching modes, table 2.1 shows assignments for Fermi modes. Peaks due to Fermi band arise from an overlap between an energy level corresponding to resonant C-H bending overtone and an energy level corresponding to a resonant C-H stretching mode. The coincidence of these energy levels allows for sharing of energy and can result in enhancement of the overtone peak 44

63 at the expense of the stretching mode [67]. Since the vibrational energy levels are sensitive to local environment around the molecules, the Fermi resonant bands are also expected to be sensitive to molecular environment. Additionally, Synder in his studies of alkanes also found that these vibrational frequencies corresponding to methylene and methyl modes were also sensitive to chain conformation and packing [64,65]. An SFG spectrum of Pentadecane (alkane with 15 carbon atoms) next to sapphire is shown in figure The y axis is the SF intensity and the x axis is the scanned IR Figure 2.17: SFG spectrum of alkane (15 carbons) next to sapphire surface. The points are the experimental data and the line is a fit to the data using equation wavelength. The spectrum shows broad peaks related to both methylene and methyl groups at corresponding frequencies given in Table 2.1. Since we are considering alkane liquid next to sapphire, the molecules are disordered and hence low intensity 45

64 with broad peaks are noted. This reiterates the fact that intensity and width of the SFG signal is based not only on the molecules present at the surface, but also on the orientation order at the surface Determination of chain orientation using SFG spectra The A ijk,ν determined by fitting has to be related to the molecular parameters to get an idea of the molecular structure. This can be done by relating the second order molecular continuum non-linear susceptibility χ (2) to the second order non-linear hyperpolarizability β lmn. The molecular hyperpolarizability β ν,lmn of the ν th vibrational resonance is expressed as a product of the IR dipole moment tensor and the Raman polarizability vector. Therefore, for the ν th vibrational mode to be SFG active, it needs to be both IR and Raman active. As a result, vibrational modes for centrosymmetric molecules, which posses inversion symmetry are SFG inactive due to the mutual exclusion rule which states that vibrational modes may be either IR or Raman active, but not both for centrosymmetric molecules [59]. Therefore, using the relationships below, a reasonable qualitative description of the molecular vibrations at the interface can be determined using SFG. But SFG spectra may also involve effects not included in the above analysis. These include the possibility of both homogeneous and inhomogeneous broadening of line shapes as well as destructive and constructive interference of peaks due to relative phase of the peaks with respect to the non-resonant contribution [68]. 46

65 Second order non-linear hyperpolarizability β lmn is given by: χ R ijk,ν = (N [ U ijk,lmn β ν,lmn ]f(ψ, θ, φ, σ)dψdθdφ) (2.18) lmn Where N is the number of molecules at the interface. The U ijk,lmn are Euler transformation coefficients based on the Euler angles (θ, ψ, φ) shown in figure 2.18 that relate the xyz=ijk lab co-ordinate system with the abc=lmn molecule fixed co-ordinate system. These transformations allow analysis of orientation of molecules at the interface relative to the lab axes. Hirose, et al., published a tabulation of U ijk,lmn coefficients, which are used in the present work [60]. The f(ψ, θ, φ, σ) term is a function describing the probability of a molecular orientation occurring within the solid angle described by the width σ of a distribution of angles around the Euler angles (θ, ψ, φ) representing the average molecular orientation. Figure 2.18: Definition of Euler angles relating lab axes to molecular axes. The value of this function is lower when we have a wider angular distribution (larger σ) or more isotropic average orientation. This implies that the SFG signal is 47

66 lower. This suggests that SFG intensity not only depends on the number of molecules contributing to the signal but also on the orientation of these molecules. That is, if orientation distribution width σ is small we have higher SFG signal. Experiments carried out in this work mainly use linear chain alkanes of different chain lengths. The groups present in these molecules are CH 2 along the chain and CH 3 at the ends. To determine the orientation of the alkane chains it is necessary to determine the orientation of these groups with respect to the surface. Let us fix the abc axes such that the c axis always coincides with the C 3 (CH 3 ) and C 2 (CH 2 ) symmetry axes. Orientation of these groups can be considered to be azimuthally isotropic with respect to φ and ψ Euler angles. Using transformations given by Hirose et al., [60], the c axes can be made to coincide with the laboratory Z axis and the angle by which the c-axis was rotated gives the orientation of the group symmetry axis. Considering the CH 3 group and assuming the group at the surface to belong to C 3v symmetry point group, the relation between A q values fit to the spectra and the molecular hyperpolarizability averaged over φ and ψ solid angles are given by [69]; A yyz,r + = N CH3 [(β ccc,r + β aac,r +) < cosθ > < cos3 θ > 2 + β aac,r + < cosθ >] (2.19) A yzy,r + = N CH3 [(β ccc,r + β aac,r +) < cosθ > < cos3 θ > ] (2.20) 2 for symmetric (r + ) stretching and for asymmetric (r stretching as; A yyz,r = N CH3 [ β caa,r (< cosθ > < cos 3 θ >)] (2.21) A yzy,r = N CH3 [β caa,r < cos 3 θ >] (2.22) 48

67 Similarly assuming azimuthal isotropy for the CH 2 group about its symmetry axis and C 2v symmetry point group, we can get irreducible representation for symmetric (d + ) stretching; A yyz,d + =N CH2 [(β aac,d + + β bbc,d +) < cosθ > + 2 < cos3θ > < cosθ > (β aac,d + + β bbc,d + 2β ccc,d +) ] 16 A yzy,d + = N CH2 [(β aac,d + + β bbc,d + 2β ccc,d +) and for asymmetric (d ) stretching; (2.23) < cos3θ > < cosθ > ] (2.24) 16 A yyz,d = N CH2 [(β aca,d ) < cos3θ > < cosθ >] (2.25) A yzy,d = N CH2 [(β aca,d ) < cosθ > < cos3θ > < cosθ > (β aca,d ) ] (2.26) 2 8 The <> brackets represent ensemble averages over the solid angle and were obtained using the method outlined by Simpson and Rowlen [70]; π < cosθ >= κ cosθf (θ)sinθdθ (2.27) 0 π < cos3θ >= κ cos3θf (θ)sinθdθ (2.28) 0 where, π κ = [ f (θ)sinθdθ] 1 (2.29) 0 Here, κ is a normalization constant, and f (θ) is the net angular distribution function defined as: f (θ) = + n= f(2πn + θ) + f(2πn θ) (2.30) 49

68 Typically the function f(ψ, θ, φ, σ) is approximated by a gaussian distribution of width σ given by equation and the range of n used is from n = -3 to +3. f(θ) = 1 σ 2Π exp (θ θ 0) 2 2σ 2 (2.31) The procedure described above can only be applied if we know the tensor components or symmetry point groups for the molecule. But some inferences about the chain tilt can be obtained by just taking the ratio of A q s for different resonant peaks. Taking the ratio eliminates the necessity to know the number density of the groups involved. Assuming azimuthal isotropy, the only angle left to consider is θ, which gives the tilt. Therefore, taking ratio of A q,r +/A q,r can give the tilt of symmetric c axis of CH 3 group; similarly, A q,d +/A q,d can give tilt of CH 2 group with respect to the z axis. The ratio of r + /d + can tell us about the concentration of CH 3 groups relative to CH 2 groups at the surface. Therefore, fitting SFG spectra can give us valuable information about molecules present at the surface and their orientation SFG studies of friction and adhesion Phenomena of friction and adhesion are governed by what happens at the interface. Both static and dynamic properties of interfacial molecules are important to understand the behavior during various loading conditions. As discussed in the previous sections, there have been a lot of experimental and simulation based studies to understand the effect of confinement and sliding on interfacial molecules. But an in-situ molecular level probe is necessary to obtain structural information. Only recently 50

69 it has possible to study these hidden interfaces and determine the composition and orientation of molecules by SFG [71]. The SFG technique is ideal since it is forbidden in the bulk due to centrosymmetry of most organic molecules and the symmetry is broken at the surface or interface. SFG has also been employed to study other problems related to interfaces like polymer /liquid, polymer/ solid [72] and polymer / polymer interfaces. Initial work using SFG as a tool to study contact interfaces was done on ordered Langmuir Blodgett films and monolayers under pressure from 10MPa to 600MPa [72 74]. The geometry used was a smooth bronze ball coated with gold on which alkanethiols of different chain lengths were deposited. A sapphire prism was used as the counterface in order to apply pressure. Berg and Klenerman and Du and coworkers found that the signal from the surface methyl modes decreased as the pressure was increased. In other words, the well packed molecules disordered in their terminal methyl groups on applying pressure. The SFG spectra taken are shown in figure They also noted that as the pressure was removed the structure recovered and the signal was comparable to that before loading. Contrary to this, Fraenkel and coworkers found that no major changes in structure for densely packed fatty acids. They saw that the spectra before, during and after loading were similar at least qualitatively. Beattie et al. studied zinc arachidate monolayer at the interface between a sapphire prism and fused silica lens before, after and during contact with increas- 51

70 Figure 2.19: SFG spectra of octadecanethiol monolayers under different loading conditions. r + and r stand for the methyl symmetric and asymmetric modes [74]. 52

71 ing pressure [75]. They also looked at structure during sliding. They found that the monolayer was resistant to pressure and shear induced conformational order. However, frequency shifts, change in intensity ratios and drop in peak intensities were found on bringing in contact. After separation, the monolayer was found to be transferred to the silica lens. SFG was employed by Yurdumakan et al. to study confinement of polydimethylsiloxane chains between an oxidized PDMS elastomer and hydrophobic well packed OTS monolayer deposited on sapphire [76]. They found surprisingly well ordered Si-(CH 3 ) 2 groups next to the template of ordered methyl groups of OTS. This ordering was found to be reminiscent of layering observed for symmetric molecules confined between atomically smooth mica surfaces. Harp and coworkers used SFG to understand the phenomena of adhesion and adhesion hysteresis between polymer / polymer interfaces [10]. They compared SFG spectra of mechanically contacted PVNODC and PS films as well as interface annealed films and found significant restructuring of surface groups on annealing. This was found to match well with the result of higher adhesion for annealed films compared to mechanical contact. Therefore the strong adhesion observed was related to structural differences in PVNODC at the annealed interface. 2.6 Other optical techniques used to study confinement Many researchers have tried to couple force based measurements with optical techniques to further the insight into understanding material properties under confine- 53

72 ment and friction. Coupling an optical microscope to the friction instrument is one of the basic experiments conducted by several researchers studying rubber friction like Tabor, Roberts, Schallamach, Barquins, Maugis, and Johnson to name a few ( [77] and references therein). The interesting wave like pattern generated when rubber slides on smooth surface due to stick-slip was termed Schallamach waves based on the theory he proposed to explain this phenomenon. With the invention of many complex systems to measure forces as small as van der Waal force, there have been new techniques adapted to look directly at the molecules involved. With the development of SFA in the 1970s there have been attempts to attach spectroscopy capabilities to look at the 100 micron contact spot between the mica surfaces. In this regard, Salmaron and coworkers have tried to attach a high speed camera and visible light source to study drainage and lubrication in SFA [44]. Granick and coworkers have tried to couple SFA and fluorescence with dielectric and Raman spectroscopy [78,79]. They studied translational diffusion of a fluorescent dye embedded at a dilute concentration in a confined fluid at rest and during shear [80]. They found that the time scales of intensity-intensity autocorrelation functions were essentially the same during shear and at rest, except they were faster during shear by a factor of 2 to 5. The complex molecularly thin systems retained a high degree of fluidity at the molecular level even though friction experiments show that the effective shear viscosity diverges when fluids become molecularly thin [30, 81, 82]. Confocal Raman spectroscopy was coupled with SFA to investigate the confinement and shear induced changes in Ra- 54

73 man spectra of polydimethylsiloxane (PDMS) liquids confined between atomically smooth mica surfaces at thicknesses less than the unperturbed radius of gyration of the polymer [78]. They found that under static conditions the PDMS chains oriented preferentially parallel to the confining surfaces. But application of shear caused the time-averaged polymer conformations to become more nearly isotropic in the plane of shear. Changes in photoluminescence and absorption of light upon confinement of conjugated polymer MEH-PPV was studied in SFA. Unidirectional shear with amplitude 20 times the surface spacing provided preferential alignment while the solvent evaporated. Chain alignment was quantified from both photoluminescence and absorption spectra. A bimodal distribution of chain alignment was observed parallel to the shear direction in 2 of the cases but perpendicular to the shear direction in of the cases [78]. Akbulut and coworkers performed shear measurements and optical absorption spectroscopy in the surface forces apparatus to measure shear-induced phase transition of an anisotropic (dye) molecule confined between two shearing mica surfaces in aqueous solution. They found the highly anisotropic cyanine dye molecules in thin water films showed only a weak effect of molecular anisotropy on shear-induced ordering, friction forces, and the onset of shear-induced crystallization, although dramatic changes were seen when the confined molecules ultimately crystallized [83]. Cann and coworkers used IR spectroscopy in internal reflection to study hydrocarbon lubricants in Hertzian geometry [84]. Spectra sampled from the contact region showed 55

74 peak intensity changes, frequency shifts and shape changes. The analysis suggested an increase in gauche defects of alkane chains in the high pressure region implying a more globular molecule with lower volume but high energy. They also found evidence of ordering close to the metal surface when the lubricant films were less than 100 nm thick. In the next chapter, instruments designed and developed, the experimental techniques used and sample preparation methods are discussed. 56

75 CHAPTER III EXPERIMENTAL TECHNIQUES In this chapter, along with the already existing experimental techniques used during the course of research, the instruments that were built to specifically study friction with insitu spectroscopy are discussed. 3.1 Development of friction cell Most of the work carried out in the last century to understand friction has been to perform force measurements and to come up with a molecular picture to understand friction [2,3,5,9]. But, there have been no direct experiments probing the structure. In this regard, a new cell design is proposed which allows the use of nondestructive and interface sensitive optical spectroscopy to ascertain the chemical signature of molecules during sliding. Two prototype instruments were built, one having strain gauge sensors to measure force with normal and shear load measurements being coupled (model 1) and the second with capacitance sensors to measure deflection with normal and shear load being isolated (model 2). The design, calibration and testing of the instruments is summarized in the following sub sections. 57

76 3.1.1 Friction cell with coupled normal and shear force - model 1 The main focus during the development of the friction cell was to be able to do spectroscopy during sliding. To achieve this, it was decided to use an equilateral sapphire prism as the substrate. It allowed for light to pass in and out of the cell and take advantage of the critical angle for total internal reflection to probe the surface. Since sum frequency spectroscopy involves multiple beams with one of the beam having infrared frequencies, it was decided to have the sapphire prism as the stationary substrate. The sliding surface was attached to a double cantilever spring attachment, which allowed for applying normal force as well as measuring shear force during sliding. The schematic of the setup is shown in figure 3.1. Figure 3.1: Schematic of the friction cell with coupled normal and shear sensors. 58

77 The schematic shows the different parts of the friction cell. The displacement is made possible by two Picometer motors from NewFocus, Inc. They have a resolution of 30 nm per step and are computer controlled via a Picometer driver with three channels. They work on the principle of stick-slip and a screw rotates with each pulse. The maximum speed achievable is 2000 steps per second. The moving surface is mounted on a double cantilever spring made of spring steel. The force is measured using strain gauge sensors Principle of force measurement using strain gauge It is known that when a material is stretched or compressed it undergoes a change in length. The ratio of change in length to original length is termed as strain. In 1856, Lord Kelvin reported that a metallic conductor under strain showed a change in electrical resistance. This was put into practical use as a strain gauge in The metallic foil-type strain gauge used in our cell consists of a grid of wire filament (a resistor) of approximately in. (0.025 mm) thickness, bonded directly to the spring steel surface by a thin layer of epoxy resin [85]. When a load is applied to the surface, the resulting change in surface length is communicated to the resistor and the corresponding strain is measured in terms of the electrical resistance of the foil wire, which varies linearly with strain. The foil diaphragm and the adhesive bonding agent work together in transmitting the strain, while the adhesive also serves as an electrical insulator between the foil grid and the surface. A Wheatstone bridge is used to measure the change in resistance. 59

78 A Wheatstone bridge is a divided bridge circuit used for the measurement of static or dynamic electrical resistance. The schematic is shown in figure 3.2. Figure 3.2: Schematic of the wheatstone bridge circuit to measure change in resistance [85]. In Figure 3.2, if R1, R2, R3, and R4 are resistors of equal value, and a voltage, V IN, is applied between points A and C, then the output between points B and D will show no potential difference. However, if R4 is changed to some value which does not equal R1, R2, and R3, the bridge will become unbalanced and a voltage will exist at the output terminals. In the friction cell four individual gauges are attached, two 60

79 on either surface and form a part of the bridge. So when the cantilever experiences a shear force, two gauges are in tension and the other two are in compression. By connecting the gauges in alternation as shown in Figure 3.2, if a positive tensile strain occurs on gauges R2 and R3, and a negative strain is experienced by gauges R1 and R4, the total output, V OUT, is four times the resistance of a single gauge. Even after the enhancement in signal, the voltage measured is very small and hence a high impedance amplifier is attached to the output to bring the signal to a measurable scale of 0-1 Volt. The schematic of the electric circuit used is shown in figure 3.3. Figure 3.3: Schematic of the bridge circuit used to amplify the output voltage of Wheatstone network [86]. The ideal strain gauge would change resistance only due to the deformations of the surface to which the sensor is attached. However, in real applications, temperature, material properties, the adhesive that bonds the gauge to the surface and the stability of the metal, all affect the detected resistance. In the present scenario there 61

80 is another drawback which couples the normal and shear force measurement. As it can be seen from the above discussion, the strain gauge measures deformation in all directions experienced by the surface to which it is attached. Hence, on applying the normal load, the cantilever buckles deforming the strain gauge. Therefore the force measured is a sum of the normal and shear components. The image of the friction cell prototype built in the lab is shown in figure 3.4. To calibrate the force, a basket was attached to one end of the cantilever and known Figure 3.4: Image of the friction cell prototype in the laboratory. weights were placed in the basket. The corresponding voltage was measured using a Keithley 2700 multi meter. The slope was determined and the scaling factor was found to be 3958 µv/gram (3% error). The minimum force which can be accurately measured with this prototype was found to be 0.1 mg (1 mn). A typical force curve 62

81 measured is shown in figure 3.5. The y-axis is plotted in terms of shear stress, i.e. the shear force divided by contact area and x-axis is the time of sliding. The different regions of the curve represent different velocities of sliding. The results shown are for a polydimethyl siloxane lens sliding against a crystalline polycarbamate film. Figure 3.5: Shear stress vs. time at different sliding velocities for a PDMS lens sliding against a polycarbamate film. To overcome the problems associated with the coupling of normal load and shear load during measurement a new friction cell was developed. The new cell uses capacitance measurement to measure deflection of the cantilever and in turn determine the force. 63

82 3.1.2 Friction cell with separate normal and shear load sensors - model 2 Measurement of friction force is well documented in literature. There exist various designs depending on the application and mating materials for which the test is designed. But, the designs by Prof. Tabor s group over the years have mainly focused on understanding friction. The design used here is based on the model published by Tabor in 1970s [8]. The main feature of our device is the capability of in situ spectroscopy along with force measurement. The other advantage of the design is the separation of normal and shear force measurement. The schematic of the cell is shown in figure 3.6. The cell is made of aluminum with the central block moving on low friction bearings. The displacement is achieved by using the Picometer motors as in the previous prototype. The moving block is built in two layers. The bottom layer has a double cantilever platform to which capacitor plates are attached forming the shear deflection measuring setup. On top of this platform another double cantilever platform is mounted perpendicular to the base as shown in figure 3.6. Another set of capacitor plates are attached to the top platform to measure the normal deflection. The sliding surface is mounted on the top platform on a glass plate which gives an option to optically monitor the sliding geometry. The capacitance based force measurement is described in the section below Capacitance measurement It is known from theory of capacitors that the capacitance is inversely proportional to the gap between the plates and directly proportional to the area of the plates. The 64

83 Figure 3.6: Schematic of the friction cell based on capacitance measurement. relevant equations to determine force are given below. C = ǫǫ 0A d (3.1) Where C is the capacitance, ǫ is the dielectric constant, ǫ 0 is the permittivity of air, A is the area of the plates and d is the distance between the plates. The capacitance based sensor has been used in many previous and present designs in literature [33,87, 88]. They are also used in surface force apparatus. They have very high sensitivity and changes in capacitance of attofarads [10 18 F] can be measured [89]. In the present prototype, the capacitor plates are made of polished steel having a diameter of 2 cm. One plate is attached to the double cantilever spring with an insulator like glass slide in between. The other plate is attached to a stationary block with an insulator separating it electrically. The gap between the plates is kept approximately 65

84 at 500 microns. The plates are attached with very thin emulsion coated copper wire (transformer winding wire). The measurement of capacitance is indirectly done by applying a varying voltage (AC) across the plates. We know that capacitance is directly proportional to the charge on the plates [Q] and inversely proportional to the potential difference [V]. Q = CV ; I = Q t = C V t (3.2) Hence applying a varying voltage to the plates generates a current [I], capacitance being a constant. If a fixed amplitude AC voltage [50 KHz, 12 V] is applied, the current can be measured and, in turn, capacitance can be determined. The circuitry used for measuring current and AC voltage generation is shown in figures 3.7, 3.8 respectively. The current generated being very small is converted back to voltage and amplified and measured using Keithley multi meter Figure 3.7: Schematic of the circuit to measure current. 66

85 Figure 3.8: Schematic of the circuit to generate high frequency AC voltage. Knowing the capacitance, the distance between the plates can be determined, since A is a constant, ǫ =1 and ǫ 0 = 8.85x10 12 F/m. Once the change in distance is known, the force applied can be calculated using the relation: F = K d (3.3) Where K is the spring constant. To determine the spring constant we need to know the relationship between the voltage measured and the capacitance. In order to accomplish this, a set of capacitors with known capacitance are attached to the circuit and the corresponding voltages are measured. The slope is determined as shown in figure 3.9 and is found to be 0.55 ± 0.01 Volt per 1pF. On attaching a basket to the double cantilever beam via a frictionless pulley, objects of known weight are placed in the basket and the corresponding deflection is measured. The slope is determined which gives the spring constant K as shown in figure

86 Figure 3.9: Graph of voltage measured vs. the fixed capacitors values. Slope is determined by fitting the data points to a linear fit Figure 3.10: Graph of voltage measured vs. the known weights. Slope is determined by fitting the data points to a linear fit 68

87 The K values are found to be N/m and N/m for the shear and normal load cantilever beams. Using these values the forces can be calculated by measuring the deflection of the beams. A typical force curve measured is shown in figure The y axis represents shear stress, i.e. force divided by contact area and in this particular case the contact area was 0.40 mm 2. The x axis is the time of sliding and different regions of the curve represent sliding at different velocities. Figure 3.11: Graph of shear force vs. time of sliding for PDMS dry sliding against sapphire. The procedure for data acquisition was completely automated by using Lab- View for both the prototypes. More results and in-situ spectroscopy data are presented in the results and discussion chapter. Attempts were also made to carry out JKR experiments using the same apparatus. 69

88 3.2 Sum frequency generation spectroscopy The theory behind SFG spectroscopy was discussed in section 2.5. In this section the two SFG spectrometers on which data was taken is described. The source for both spectrometers are pulsed lasers. The main difference between the two is that the pulse width for the first SFG system is 1 picosecond and the second SFG system is 150 femtosecond. The picosecond system was supplied by Spectra Physics Inc. and the femtosecond system by Clark-MXR Inc. It is important to understand the terminology used in taking a SF spectrum before delving into details of the setup and instrumentation involved. As known from the previous section 2.5, SFG spectroscopy involves three laser beams of different frequency. Two of them are the input and their sum is the output. All the beams are polarized and hence while taking a spectrum different permutations of polarization are possible. The convention used in identifying the spectra is as follows. For example in a SSP spectrum, S stands for TE (transverse electric) polarization or perpendicular to the plane of incidence and P stands for TM (transverse magnetic) polarization or in the plane of incidence. The first letter corresponds to the polarization of the output SF wave, the second to the input visible wave and the last to the input infrared wave. So SSP stands for s polarized output, s polarized visible and p polarized IR. 70

89 3.2.1 Picosecond SF system The schematic layout of the optical system is shown in figure The system has two laser sources. The first is a Ti:Sapphire laser (Tsunami, Spectra Physics Inc.) at 800 nm wavelength and 1 Watt power. It has an 82 MHz repetition rate and a pulse width of 150 femtosecond. The second source is a Nd:YLF (Merlin, Spectra Physics Inc), frequency doubled laser having a nanosecond pulse width at 532 nm wavelength. It has a repetition rate of 1 KHz and power of 10 watts. Both beams are pumped into the regenerative amplifier (Spitfire). Before the femtosecond pulse enters the amplifier it is stretched to 1 picosecond and it forms the seed. The seed and the pump (532 nm) overlap on the Ti:Sapphire crystal in the Regen cavity. After a few passes through the cavity, as the pulse builds in intensity, the pockel cell in the cavity changes polarization and releases a pulse whose pulse width is further compressed in the pulse compressor. The output from Spitfire is a beam with 1 Watt power,1 ps pulse width and has a wavelength of 800 nm. This beam is passed into the Optical Parametric Amplifier [OPA], where the tunable infrared is generated from the visible pulse. The unused visible pulse along with the tunable IR beams are ejected out of the OPA. The visible at 800 nm has about 150 mw power and the IR about 1-2 mw. A small part of the IR is reflected using a calcium fluoride plate before exiting the OPA to normalize the spectra since the IR power is not constant over the wavelength range. In this system the IR can be tuned over the range of 1800 cm 1 to 3800 cm 1. The IR beam on exiting the OPA passes through a waveplate and polarizer before being focused on 71

90 Figure 3.12: Schematic of the picosecond SF spectrometer. The meanings of the symbols are as follows: H, half wave plate; B, calcium fluoride plate beam splitter; M, mirror; DM, dichroic mirror; P, polarizer; F, Raman notch filter; C, IR chopper; PMT, photomultiplier tube detector. 72

91 the sample using a 150 mm lens. The 800 nm visible beam on the other hand passes through a waveplate and delay stage before overlapping with IR at the sample. This delay stage is a motorized stage to alter the path length of the visible beam such that both IR and visible beams can be overlapped in space and time at the sample surface. The output SF beam which is spatially separated from the input beams is routed to a grating spectrometer using folding mirrors. The input beams are blocked using filters. The spectrometer improves the signal to noise by selecting the calculated SF wavelength and the light photons are measured using a Hamamatsu Photo Multiplier Tube (PMT). The photons are measured using a gated Photon counter (SR 400, Stanford Research Systems). The input IR power is measured using a pyroelectric detector attached to a lock-in amplifier (SR850, Stanford Research Systems). All the motors and the instruments are computer controlled using LabView programming interface. Typically, the SF spectra taken using this setup is done by scanning the IR wavelength in 5 cm 1 steps over the molecular resonance range. The sample cell along with the use of internal reflection geometry to further enhance the SF signal is discussed in the next section SF spectroscopy in total internal reflection (TIR) geometry To achieve the signal enhancement by TIR an equilateral prism made of sapphire is used. Sapphire has a refractive index of 1.76 which is higher than refractive indices of most organics studied. Total internal reflection can be achieved when a ray of light passes from a higher refractive index medium to a lower one. In this case 73

92 both the input beams are incident on one face of the equilateral prism as shown in figure By choosing the proper angle of incidence, the light can undergo total Figure 3.13: Schematic of the total internal reflection geometry used. The top part shows the arrangement to study confinement or friction and the bottom part shows the setup to study spin coated polymer films or fluids next to sapphire surface. internal reflection at the sapphire-organic interface. This increases the concentration of electric field at the surface, thereby enhancing the nonlinear activity. The SF signal generated can be a factor of 10 higher than that achieved using external geometry (direct incidence on the surface). This also allows the use of the sapphire surface as a stationary substrate on which a rubber substrate can slide or fluid molecules can be confined and the changes in the chemical structure can be monitored. In the experiments presented here, a rubber lens made of PDMS was used to confine 74

93 fluid molecules like n-alkanes and water. Other experiments can also be done, the surface can be brought in contact with bulk liquids filled in a steel chamber (not shown) or polymer films can be coated on the surface and the chemical signature of molecules at the interface can be studied. The other advantage of using a sapphire prism is that it does not show strong nonlinear activity from its bulk. For fluids confined between sapphire and PDMS an incident angle of 8 degrees was used. To study the sapphire air or polymer air interface, an angle of 42 degrees and for a sapphire water interface an angle of 16 degrees were employed. These angles were chosen by modeling the enhancement of SF signal from the interface as a function of the incident angle Femto-second SFG system The schematic of the femto-second SFG system is shown in figure This is a one box system with two levels of optics. It was procured from Clark-MXR Inc. In the bottom level, there is a low power telecommunication laser, which provides the seed. The main power source (the pump) is lamp based. The power generated is 10 watts of 1 KHz nano-second second pulses at 532 nm using a Nd:YAG rod and second harmonic crystal. The pump and the seed are mixed in the regen cavity at the upper level inside a Ti:Sapphire crystal. The output after amplification and pulse compression is a 150 femto-second pulse at 775 nm with 1800 µj energy. The entire setup is placed in a cast iron casing cooled with water circulation to 20 0 C. To use this laser for SFG experiments, the output beam is split into two parts with an 80:20 75

94 beam splitter. The smaller part of 350µJ is used as the visible source after passing through a lens pair to enlarge the beam profile and later passing through a notch filter. The notch selects only a small portion of the beam with center at 775 nm and pulse width of 1 nm. This increases the temporal width from 150 fs to 1-2 ps. The beam is passed through a long focal length lens to focus on the sample as a spot of 500 µm. The remaining power from the main beam is used to generate Figure 3.14: Schematic of the Clark-MXR femto-second SFG system. The beam path is labeled and the additional optical elements in the beam path are as follows: (L) plano convex lens, (CL) plano concave lens, and (WP) waveplate. tunable infrared beam using the TOPAS-C OPA. The OPA has two parts: the first part generating the signal and idler and the second part generating IR using NDFG (differential frequency mixing). The OPA uses white light generation from a sapphire 76

95 crystal, which acts as the seed to generate signal and idler. The signal and idler are passed through DFG crystal to generate an infrared beam. This is a non-collinear OPA, i.e., the signal and idler are spatially separated from the IR beam. The IR beam along with the visible beam is passed through a Barium Fluoride lens to focus on to the sample. The beams are spatially and temporally overlapped at the sample surface to generate SFG. The output is collected using a plano-convex lens and sent to a grating spectrometer via a Raman notch filter to remove the visible component and a waveplate to select the polarization. The spectrometer is a Jobin Yvon make 750M attached with Symphony CCD detector. The grating has 1200 lines per square inch and the CCD has 1024 X 256 pixels. The CCD is deep depletion type. It is cooled using liquid nitrogen and maintained at 140 Kelvin. The main advantage of this system is its capability to obtain spectra over 180 wavenumbers without scanning the IR wavelength. This is possible due to the temporally short femto-second IR pulse. This allows us to study dynamics and relaxation processes like friction which will be discussed in the next chapter. 3.3 Sample preparation In the following sub sections sample preparation procedure used is presented. PDMS lens preparation as well as friction cell cleaning procedure is presented. 77

96 3.3.1 Preparation of PDMS lens Polydimethylsiloxane lenses were prepared by using Sylgard 184 monomer supplied by Dow Corning Inc. The recipe consisted of 1 part crosslinker to 10 parts monomer. The monomer and the crosslinker were mixed and the air bubbles were removed. The mixture was taken in a syringe and lenses were formed under water. This was done to retain the shape of the lenses formed which were about 5 mm in diameter. If not done under water the monomer flows to form a flat sheet. The process of forming the lenses under water gives a smooth surface to the lenses. The lenses were cured under water at room temperature for 24 hours. The water was removed and the lenses were dried using dry nitrogen. The lenses were further cured in a vacuum oven for 4 hours at 60 0 C. The lenses, if used as is, were found to leach out the short chains to the surface and this led to lowering of friction [76]. Therefore, the lenses were soaked in toluene to remove the uncrosslinked chains. The toluene was replaced every 3 days for two weeks. The lenses were removed from toluene and dried under vacuum for 4 hours before use. The root mean square (rms) roughness determined by atomic force microscopy measurements of the surface of the lens was approximately 4-5 angstroms Preparation of samples for friction and SFG For studying the confinement of alkanes using SFG, the PDMS lenses were soaked in the alkanes, which are liquids at room temperature. If the alkanes were solid at room temperature they were melt cast on to the sapphire surface and then brought in contact with the lens. The alkanes used were of chain length 15 to 27. These were 78

97 procured from TCI America Inc with purity greater than 98%. They were used as received without further purification. The other fluid studied was water. Water was distilled and deionized before use. For bulk alkane or water on sapphire, the cell was filled with the fluid and a teflon gasket was used between the cell and sapphire prism to prevent any leaks. The cell was also equipped with an attachment for heating and cooling and temperature was measured at two locations using thermocouples. The temperature was controlled using a Lakeshore 330 temperature controller. To maintain a uniform heating and cooling rate, a block of copper with circulating water was placed below the cell. This was maintained at a constant temperature of 10 0 C. A thorough cleaning procedure was used to clean the cell and the sapphire prism before experiments. The prisms were first wiped with toluene using a soft tissue and then sonicated in toluene for 1 hour. The prisms were then washed in copious amounts of water and blow dried using dry nitrogen. They were then plasma cleaned using air plasma for a period of 5 minutes. For the friction cell the cell was washed in soap solution and sonicated in toluene for a period of 2 hours. The cell was further washed with water and blow dried using dry nitrogen. The cell was later kept in an air oven at C for 10 minutes and then plasma cleaned for a period of 5 minutes. For preparation of thin films on either glass or sapphire substrates, the substrates were cleaned using the same procedure as stated above. A 5% solution of 79

98 the desired polymer was prepared and coated on to the substrate using a Specialty Coating Systems spin coater, model number P6700 at 1700 rpm. 3.4 Measuring thickness of confined lubricant Techniques like profilometry, ellipsometry cannot be employed to measure thickness insitu during confinement. Hence optical interference was used to determine the thickness of the confined liquid between a flat surface and the rubber lens. For thickness greater than half the wavelength of light, two beam interference produces Newton rings. The principle of interference is shown in figure A monochromatic beam of light (wavelength λ) is incident on the contact region. The reflected light from the substrate and the front of the lens interfere due to different path lengths traveled. The difference in path length, if a multiple of λ or λ, can lead to constructive 2 or destructive interference respectively. In the present setup due to the curvature of the lens we see alternating bright and dark fringes suggesting that the thickness is varying radially. The setup was modified by Roberts and Tabor [8, 37, 90] to perform in-situ thickness measurements along with friction studies. Since the thickness was found to be less than half the wavelength of light the procedure was modified to measure thickness accurately down to 1 nanometer. A 30 0 prism was used as the substrate to pass light in and out of the contact spot as shown in figure Thickness above a quarter wavelength can be measured by counting the bright and dark fringes as they pass through the field of view. Below a quarter wavelength, 80

99 Figure 3.15: Principle of interference of two beams to produce Newton rings [91]. Figure 3.16: Thickness determination geometry. (a) schematic diagram of the path of light beam through the 30 0 prism. (b) the scatter of laser light in rubber [90]. 81

100 no fringes are seen. Yet, we see intensity in between a dark and bright fringe. If zero thickness gives a perfectly dark fringe and λ 4 thickness gives an intensity R 0, any gap of intermediate optical thickness n 1 h will, at normal incidence, give an intensity R = R 0 4r 2 sin 2 [2πn 1 h/λ] 1 2r 2 cos[4πn 1 h/λ] + r 4 (3.4) where n 1 is the refractive index of the confined liquid and n 0 for solids on either sides of the gap and r = n 0 n 1 n 0 + n 1 (3.5) For gaps of thickness less than λ, the bottom cosine term goes to unity. In contrast, the maximum intensity R 0 occurs when n 1 h = λ. In that case, the sine term goes to 4 unity and cosine term goes to -1. So the equation 3.4 can be simplified as R = R 0 sin 2 [ 2π λ n 1h]( 1 + r2 1 r 2)2 (3.6) for most experiments n 0 is close to 1.5 and n 1 is not less than 1.33 so r is approximately 0.1 and hence the last term in equation 3.6 varies less than 1% of unity. Hence we can write R = R 0 sin 2 [2πn 1 h/λ] (3.7) The plot of intensity variation below λ 4 is shown in figure Two assumptions are made to derive the equation above. First that there is exactly a

101 phase change at the liquid-rubber interface and second that the refractive index of the confined liquid is same as that in the bulk. Figure 3.17: Trace of variation in reflected light intensity with time for the contact zone between a rubber sphere and lubricated glass plate at constant load [8]. 3.5 Determining thickness and critical angle using Airy formulas Another optical technique which is popular is the use of Airy formulas to determine thickness. This is also used in designing dielectric coatings for photonics applications. The idea is to determine the fresnel coefficients for reflection and transmission at an interface [56]. If we consider a three layer model with a semi-infinite top and bottom layer and middle layer of certain thickness, the expressions for transmission and reflection coefficients can be derived by summing the amplitudes of successive reflections and refractions. Such a derivation was first carried out by G.B. Airy in 83

102 1833. Consider a beam of light incident at the interface. Part of the beam is reflected and the rest is transmitted. The transmitted beam is subsequently reflected back and forth at the boundaries of the middle layer. By adding the amplitudes of successive reflected and transmitted rays, the reflection and transmission coefficients can be determined. In doing so, it is important to account for the phase factor which arises due to the geometric path difference between two successive reflected and transmitted rays. This phase factor is directly proportional to the thickness of the middle layer. Considering the input beam to be p polarized (TM wave). Since we are using a prism as one media, the fluid as the second or middle layer and PDMS lens as the third layer, we need to determine the intensity of the reflected beam coming from the interface. The geometry and the beam angles used are shown in figure The fresnel coefficients for reflection of the p-wave at the first and second interface are [56]: r 12p = n 2Cosθ 1 n 1 Cosθ 2 n 2 Cosθ 1 + n 1 Cosθ 2 (3.8) the reflection coefficient is given by r 23p = n 3Cosθ 2 n 2 Cosθ 3 n 3 Cosθ 2 + n 2 Cosθ 3 (3.9) r p = r 12p + r 23p exp(2iβ) 1 + r 12p r 23p exp(2iβ) (3.10) given by the transmission coefficients for beam entering and leaving the prism are t p (prismin) = 2Cosθ a n 1 Cosθ a + Cosθ b (3.11) 84

103 Figure 3.18: Schematic of the thickness measurement setup. where t p (prismout) = 2n 1 Cosθ b Cosθ b + n 1 Cosθ a (3.12) β = 2πn 2dCosθ 2 ; θ b = Sin 1 [ Sinθ a ] (3.13) λ n 1 θ 1 = π 3 θ b; θ 2 = Sin 1 [ n 1Sinθ 1 n 2 ]; θ 3 = Sin 1 [ n 1Sinθ 1 n 3 ]; (3.14) Experimentally, a He:Ne beam is used as the monochromatic light source and the reflected output intensity is measured with respect to incident input angle. The data obtained is fit with the theoretically predicted output intensity given by I out =[t p (prismin) conj(t p (prismin))] [r p conj(r p )] [t p (prismout) conj(t p (prismout))] (3.15) 85

104 Since the beam is passing from a higher refractive index sapphire prism to a lower refractive index fluid, at the critical angle total internal reflection is achieved and the reflected intensity goes to a maximum. Hence this experiment not only allows us to determine thickness but also to determine the critical angle where the output signal is the maximum, so that the same angle can be used in SFG measurements. A typical result for an alkane between sapphire and PDMS is shown in figure The fit to the angle dependence data gives a thickness of 10 nm or lower. Figure 3.19: Graph of output power vs. the input angle for an alkane confined between sapphire and PDMS. Setup shown in figure 3.18 was used. 86

105 CHAPTER IV RESULTS AND DISCUSSION This chapter is divided into sections describing different aspects of confinement and friction. In the first section, data related to confinement of linear chain alkanes and their static structure will be presented. In the second section, the structure of ever important fluid, water under static confinement is discussed. In the third section, the impact of confinement on the transition temperature of alkane analogues and a comparison with existing theories of phase transitions is presented. In the last section, using the new capability to study dynamics in second to sub-second time scales, the effect of sliding on the structure of fluids under confinement is presented. The last section also shows results from friction force measurements during sliding. 4.1 Static confinement of linear alkanes In recent years, confinement of liquids has generated great interest in understanding friction, lubrication and wear between solid surfaces [8, 9, 39, 92 95]. Tires on roads, windshield wipers, movement of human joints are some examples where flexible-rigid contact interfaces are experienced. The intriguing question is whether the structure and viscosity of confined liquids is different from that in the bulk. No noticeable 87

106 change in liquid viscosity down to a thickness of 25 nm was suggested by Roberts and Tabor for rubber in contact with glass surfaces [8]. Due to surface roughness, Tabor s experiments were inconclusive for a thickness of less than 25 nm. Using atomically smooth mica surfaces, Israelachvili and coworkers have studied nanometer thin films and have shown that oscillatory force profiles are observed for a variety of liquids due to layering of molecules under confinement [9]. These results have also been supported by computer simulations [94, 96]. As a consequence of layering, the Newtonian liquids show solid-like response under confinement [92, 93]. However, more recent measurements using a different procedure to cleave mica show a modest increase in apparent viscosity for nanometer-thin confined liquids [97, 98]. We have revisited Tabor s experiment to address the open question on whether confinement affects the structure of liquids between an elastomer (PDMS in our measurements) and a rigid substrate (sapphire). Surface sensitive infrared-visible sum frequency generation spectroscopy (SFG) was used to directly probe the structure of confined fluids. These results provide the first spectroscopic measurement of molecular structure under confinement between surfaces of practical importance [99]. The sample geometry used was described in the previous chapter. This novel approach of using a flexible elastomeric lens which deforms against a flat solid surface to confine molecules offsets the need to have perfectly parallel surfaces as we approach separations of molecular dimensions. PDMS (Sylgard 184, Dow Corning) semi-spherical lenses of radius 3-4 mm were prepared using the procedure developed 88

107 by Chaudhury [100] (rms roughness of 0.6nm). The lenses were extracted with toluene to remove any unreacted PDMS chains. The lenses after drying were soaked in either n-hexadecane (Alfa Aesar, 99% purity, T m =291K) or n-pentadecane (TCI, 99% purity, T m = 283K). Before spectroscopic measurements were performed, the thickness of the lubricant for different normal loads was determined, using the procedure described in the previous chapter ( Ref. [37]), to be < 10 nm for the normal load of 10mN - 0.3N. Exact determination of thickness below 10 nm was not possible due to background scattering from the PDMS lens. An estimate from the shear stress of 0.1MPa, velocity of 10 µm/s and thickness of 1-10 nm, shows the effective viscosity of confined hexadecane to be Pa-s as compared to bulk viscosity of 3x10 3 Pa-s. This calculation is an approximation since it does not include influence of adhesive forces. Interestingly, this increase in viscosity is similar to that observed for hexadecane confined between mica surfaces [92]. SFG scans using SSP polarization (s-polarized SFG, s-polarized visible and p-polarized IR) in the CH range for alkanes ( cm 1 ) were taken for liquid alkane/sapphire, alkane crystal/sapphire, confined alkane/sapphire and confined crystal alkane/sapphire interfaces. The peak assignments based on bulk IR and Raman measurements for the CH stretches are as follows [66]: peaks between cm 1 are assigned to the CH 3 symmetric stretch (r + ); two peaks at 2950 cm 1 and 2965 cm 1 to the CH 3 asymmetric stretches (r ); peaks between cm 1 to the symmetric CH 2 stretch (d + ); peak between cm 1 to the asymmetric 89

108 CH 2 mode (d ); and peak in the range of cm 1 to the CH 3 Fermi resonance due to the overlap of symmetric overtone with the bending mode (r + FR ). To obtain quantitative information, the spectra were fitted using the following Lorentzian equation [55]. I(SFG) χ eff,nr + q A q ω IR ω q iγ q 2 (4.1) where A q, Γ q, and ω q are the strength, damping constant, and angular frequency of a single resonant vibration, respectively. χ eff,nr is the non-resonant part of the signal. Understanding the structure of confined liquids requires first the characterization of the structure of bulk liquids next to a sapphire surface. SFG spectra of hexadecane and pentadecane liquid in contact with sapphire are shown in Figs. 4.1a and c. For both hexadecane and pentadecane, the SFG spectra show peaks associated with CH 2 symmetric, CH 2 asymmetric, CH 3 Fermi and CH 3 asymmetric modes. The spectra also show an increase in intensity beyond 2970 cm 1, which can be attributed to the surface OH groups on sapphire. As described in the literature review, by fitting the spectra and taking the ratio of symmetric to asymmetric A q values, the orientation of the chains can be determined. The ratios of d + /r + and r /r + obtained from fitting the spectra are shown in Table 4.1. For alkane/sapphire interfaces, the high CH 2 and CH 3 asymmetric modes indicate the presence of gauche defects and the 90

109 Intensity (arb.) Hexadecane (a) liquid (b) crystal Pentadecane (c) liquid x 6 x 3 x (d) crystal x Wavenumber (cm -1 ) Figure 4.1: SSP spectra of hexadecane (a) and pentadecane (c) liquid/sapphire interfaces at 295K. The SSP spectra for hexadecane (b) and pentadecane (d) crystal/sapphire interfaces were taken at 287 and 279K, respectively. The solid lines are fits using Eq The SFG spectra were offset along y-axis by an arbitrary amount and were scaled for clarity. 91

110 Table 4.1: Ratios of A q of CH 2 (d + )/CH 3 stretches (r + ) and CH 3 asymmetric (r )/ CH 3 symmmetric (r + ) for hexadecane and pentadecane in contact with sapphire at various conditions. Hexadecane Pentadecane d + /r + r /r + d + /r + r /r + Bulk Liquid Bulk Crystal Confined Liquid Confined Crystal CH 3 groups are tilted away from the surface normal. These spectra for liquid alkanes are similar to those reported for hexadecane next to glass [101]. Figs. 4.1 b and d show hexadecane and pentadecane spectra below T m. The SFG peaks are much sharper and intense in comparison to the broad peaks observed in the liquid spectra. The SFG spectrum for pentadecane crystal/sapphire interface has two dominant peaks associated with CH 3 symmetric and CH 3 Fermi. The contribution of CH 2 symmetric only appears as a weak shoulder. The presence of strong CH 3 peaks suggests that the pentadecane molecules are well ordered in an all-trans conformation. The hexadecane crystal signals are much weaker in comparison to pentadecane. Hexadecane is a symmetric molecule and is expected to be SFG 92

111 inactive when the molecule is in an all-trans conformation. The SFG data for hexadecane crystal are only presented to make qualitative comparison with the results from odd alkane and to show formation of all-trans conformation when the molecule crystallizes. The SFG spectra of confined alkane between the PDMS lens and sapphire substrates are shown in Figs. 4.2 a and c. The SFG peaks are associated with CH 3 and CH 2 groups of confined alkanes rather than methyl groups of PDMS. This can also be observed by the control experiment of dry contact between PDMS and sapphire substrate as shown in Fig. 4.2 e. The CH 3 symmetric and CH 3 asymmetric peaks of PDMS are at 2905 cm 1 and 2965 cm 1, respectively. In addition, the SFG intensity is much weaker for dry contact. The spectra of confined alkanessapphire interfaces has contribution from CH 3 symmetric mode and a broad convoluted peak due to CH 2 asymmetric, CH 3 Fermi and CH 3 asymmetric modes. The confinement-induced ordering can be inferred from the higher SFG signals (3 times higher than liquid spectra), the presence of CH 3 symmetric mode and the weak CH 2 symmetric mode. Both the confined and bulk pentadecane liquid spectra have strong CH 3 asymmetric signals, which indicate methyl groups are on average tilted with respect to the surface normal. Surprisingly, the confined liquid does not show differences in signal intensity between odd and even alkanes. This suggests that the order of confined alkane is not similar to that of crystals. To understand this further, we cooled the confined alkanes 93

112 2 Hexadecane (a) liquid x (b) crystal x 2.4 Intensity (arb.) Pentadecane (c) liquid (d) crystal (e) Dry Contact x 1.6 x 0.3 x Wavenumber (cm ) Figure 4.2: SSP spectra of hexadecane (a) and pentadecane (c) confined liquid/sapphire interfaces at 295K. The SSP spectra for confined hexadecane (b) and pentadecane (d) crystal/sapphire interfaces were taken at 287 and 279K, respectively. The dry PDMS/sapphire SFG spectrum is shown in (e). The solid lines are fits using Eq The SFG spectra were offset along y-axis by an arbitrary amount and were scaled with respect to the SFG spectrum for pentadecane crystal in Fig. 4.1(d) for comparison. 94

113 below the freezing temperature and the SFG spectra are shown in Figs. 4.2 b and d for hexadecane and pentadecane, respectively. Pentadecane confined crystal/sapphire interface is dominated by the presence of a strong CH 2 symmetric peak at 2840 cm 1 along with the CH 3 symmetric and CH 3 Fermi resonance modes. The SFG intensity for confined pentadecane crystal is higher than that of confined liquid/sapphire interface. The presence of strong CH 2 symmetric mode is surprising and suggests that confined crystal structure is very different from bulk pentadecane crystal/sapphire interface. The transition from confined ordered liquid to confined crystal is sluggish with the structure developing with time. In contrast, the bulk liquid to crystal transition shows no time dependent structural change. Also, the absence of the odd-even effect found in confined liquid reappears in the case of confined crystal because the SFG intensity for confined hexadecane crystal/sapphire interface is much lower than that of pentadecane crystal/sapphire interface. These observations confirm that the confined liquid is more ordered but does not crystallize upon confinement and crystallization requires cooling down below the bulk freezing temperature. The strong methylene intensity in confined crystal cannot be accounted for the presence of gauche defects as in the case of liquid alkane/sapphire interfaces [101]. With the help of a simple model, we postulate that the confined pentadecane chains are crystallizing with the chains lying flat next to the sapphire substrate. The molecular hyperpolarizability was derived by using the IR dipole moment and Raman tensor of a single C-H bond [102] and the normal coordinates for methyl vibrations as given by Sny- 95

114 a c A q (a.u.) x b z r - r + d + c y a b x a d + z r + r - y c b Figure 4.3: Predictions of A q,ssp for CH 3 symmetric, CH 3 asymmetric and CH 2 symmetric for an all-trans odd alkane rotating along the a-axis (in-plane, left) and b-axis (out-of-plane, right) with respect to the laboratory z-axis. 96

115 der [66]. The alkane is taken as an all-trans chain (a good assumption below T m ) with the coordinates shown in Fig We have also taken the methyl groups fixed with respect to each other and not freely rotating, which is important since the analysis depends on this assumption. The combined hyperpolarizability of an all-trans alkane can be constructed using Euler transformation of these two individual methyl groups to the new a, b and c coordinates. Finally, the combined molecule with a, b and c coordinates can be transformed to the laboratory coordinates X, Y and Z as shown in Fig The molecule is allowed to tilt by an angle θ and rotate around the a-axis (in-plane, χ is 270 o ) and around the b-axis (out-of-plane, χ is 0 o ). All φ angles are equally probable, since we do not expect any preferred direction of the alkane chains in the xy plane. A good match between experiments and the model is achieved for θ 0 (symmetry c-axis of the molecule parallel to the surface normal). The intensity of SFG signals for bulk pentadecane crystal/sapphire interfaces are comparable to that of self-assembled monolayers with methyl terminal groups (octadecyl trichlorosilane) [103]. Since the SFG intensity of the methyl is proportional to the square of the number density of methyl groups, this suggests that the SFG intensity for the alkane crystals is due to more than one layer next to the sapphire surface. This is pictorially shown in figure 4.4. Finally, we would like to comment on the structure of bulk pentadecane crystals next to sapphire substrate. Yeganeh [104] has interpreted the absence of CH 3 asymmetric modes as an indication of chains with θ 0 (Fig. 4.3). Hence in this case, there should have been a strong CH 2 symmetric 97

116 Figure 4.4: Pictorial representation of analysis of SFG data showing alkane chain layering upon confinement peak present in the spectrum (Fig. 4.1D). Allowing the rotation of the alkane chain out-of-plane would result in smaller CH 2 symmetric signals. However, this would also result in an increase in CH 3 asymmetric modes, which was not observed in the bulk pentadecane crystal spectrum. There is no solution for this case unless we assume that the symmetry is broken in the immediate vicinity of the sapphire substrate. This would imply that both the CH 3 groups of pentadecane molecule do not contribute equally to the SFG signal. This model has also been used to explain the results for odd and even crystalline alkanes next to the air interface [101]. The experimental data can be explained with the pentadecane chains oriented out-of-plane with the symmetry axis of the methyl group next to the sapphire surface oriented parallel to the surface normal. The SFG spectra for bulk and confined liquid/sapphire interfaces 98

117 can be interpreted quantitatively based on average tilt angles of methyl and methylene orientation. However, to obtain a realistic physical picture will require computer simulations. In summary, for the first time, we have shown that liquid alkanes confined between an elastomer and sapphire surface are much more ordered than liquid alkanes. The confined liquid crystallizes below the bulk freezing temperatures with the chains lying on the substrate with the symmetry axis parallel to the surface normal. This structure is very different from that of bulk alkane crystals in contact with sapphire. These results have important implications, as would be discussed later, in our understanding of friction and lubrication in confined geometry. 4.2 Confinement of water Water is the most common solvent or lubricant used in physical situations. The abundance of water on this planet makes it the most commonly used lubricant. Its use can be traced across many fields from biology to geology. Water in confined geometries plays a crucial role in many physical processes. Protein folding, catalysis, movement of our joints, tires on wet roads and windshield wipers are just a few examples involving confined water. A detailed understanding of confined water is necessary to unravel the functioning and physics behind processes where confined water is associated. Lubrication properties of water have been known from ancient times. Egyptians (2400 BC) used water to lubricate wooden sledges to transport 99

118 large stones to build the pyramids [13]. Moreover, nature has selected water as a base for the biological lubricants, which are far superior to the man-made oil-based lubricants [81]. Understanding the role of water in friction requires understanding the static structure of water under confinement. Roberts and Tabor carried out force measurements in the 1960s to understand lubrication between flexible-rigid interfaces in the presence of water, water with salts and synovial fluid (water based lubricant found in our joints) [8]. They were interested in problems related to tires on wet roads and action of windshield wipers. They found the viscosity of confined water to be similar to the bulk up to thickness of 25 nm and below which they attributed the increase in friction to asperity contacts. The development of SFA with atomically smooth surfaces has led to renewed efforts to study the ordering and viscosity of confined water [9, ]. Results from different groups have been contradictory to those suggested by Zhu and Granick [106] suggesting an increase in viscosity at low film thickness and Raviv and Klein [108] observing no appreciable increase in viscosity even at distances of 1nm. Therefore, to understand the effect of confinement, a complementary technique to study the structure of water molecules under confinement is desired. In this regard, SFG serves as a unique tool to study this phenomenon. For the current study, we have revisited Roberts and Tabors experiment of confinement between flexible-rigid interfaces and ascertain whether higher friction is due to asperity contact and also determine the structure of confined water. The sample geometry used is similar to the one used for study of alkanes under confinement. 100

119 The use of a PDMS lens as the flexible surface and a sapphire substrate as the rigid surface provides a unique scenario with one surface being hydrophilic and the other being hydrophobic. This has also been considered as a Janus interface by Granick and coworkers [109]. To study the structure of water under confinement it is important to understand the structure of bulk water next to hydrophobic and hydrophilic surfaces. First studies of water by SFG were done by Shen and coworkers in the 1990s [ ]. In recent years several research groups have used SFG to study various types of water interfaces [ ]. The first experiments were done to study water/air (vapor) interface. They found that water existed as an ordered hydrogenbonded network as expected from the high surface tension of water. The spectra taken in three different polarization combinations are shown in figure 4.5. The SSP spectrum in figure 4.5a shows three main peaks. A sharp peak at 3700 cm 1 is associated with the stretch mode of the OH dangling bonds at the water surface. The two broad peaks at 3400 and 3200 cm 1 are close in positions with the stretch modes of the bonded OH in bulk ice and water. They are termed as ice like and liquid like peaks, respectively. The spectra indicate that the water surface structure is partially ordered and partially disordered, presumably in the form of a mixed ordered and disordered hydrogen-bonding network [120]. The PPP spectrum shows more liquid like peaks along with the dangling OH peak at 3700 cm 1. The SPS spectrum is very weak and shows a peak at 3600 cm 1, which is assigned to 101

120 Figure 4.5: SFG vibrational spectra of water/vapor interface taken with (a) SSP, (b) PPP, and (c) SPS polarization combinations [119]. 102

121 the bonded OH stretch mode of surface water molecules with one bonded OH and one dangling OH. A partially disordered ice like structure with least number of broken hydrogen bonds at the surface was expected to give a surface concentration of dangling OH of 25% [121]. This was also found in experiments done with a mixture of methanol and water. Figure 4.6 shows the top view and side view of molecular structure of hexagonal ice crystal surface. It shows the tetrahedral bonding structure with surface dangling OH groups. Solid / water interfaces were also studied using SFG [111,117,123,124]. Figure 4.7 shows the SSP spectra of a silica / water interface taken at different bulk ph. A spectrum of silica / ice is also shown for comparison. It is seen that the spectra are qualitatively same as that of the air / water interface except that the dangling OH peak at 3700 cm 1 is missing. The presence of the ice like and liquid like peaks at 3200 and 3400 cm 1, respectively, indicate that the interfacial water molecules must again form partially ordered hydrogen-bonded network. Both peaks increase with increase in ph, with the ice like peak increasing more at higher ph indicating a better ordered network. The ice structure next to sapphire shows mainly 3200 cm 1 peak as expected for a tetrahedral ordered network. Figure 4.8 shows possible hydrogen-bonding configuration of water molecules next to silica surface as proposed by Shen and coworkers [122]. The model suggests that silica remains neutral [SiOH] at low ph (ph less than 2). In this case, two molecules can bind with H to O and 103

122 Figure 4.6: Molecular structure of hexagonal ice crystal: (a) side view of the bulk near the (0001) surface; (b) top view of the (0001) plane. Red spheres represent O atoms (dark and light shades highlight higher and lower submonolayers in a single ice monolayer); gray and white spheres represent H atoms that are hydrogen bonded to neighboring molecules and free-dangling non bonded surface species, respectively. Dotted lines indicate hydrogen bonds [122]. 104

123 Figure 4.7: SFG spectra of silica / water interface taken in SSP polarization combination as function of ph. The spectra are offset for clarity. Spectrum of silica / ice interface is also shown for comparison [124]. 105

124 one with O to H of SiOH. At higher ph, the surface gets completely deprotonated [SiO ]. In that case, three molecules can bind with H to O of SiO. Figure 4.8: Possible hydrogen-bonding configuration of water molecules on hydrophilic silica surface: (a) protonated (SiOH) surface sites, low ph (below 2); (b) deprotonated (SiO ) surface sites, high ph; (c) structure of water/silica interface at low ph. Red and gray spheres represent O and H atoms of water molecules; large gray green, pink, and white spheres represent Si, O, and H atoms of SiOH groups at silica surface. Dotted lines indicate hydrogen bonds. [122]. The effect of bringing a hydrophobic surface in contact with water has also been studied Du et al [112]. Since air is an ideal hydrophobic surface bringing any 106

125 other hydrophobic surface in contact with water should show a spectrum similar to air / water interface. This was indeed found to be true for a well prepared octadecyltrichlorosilane [OTS] monolayer in contact with water. The spectrum is shown in figure 4.9. It was however seen that the dangling OH at 3700 cm 1 seen for the air interface was red shifted to 3680 cm 1. This was attributed to the weak van der Waals interaction between the dangling OH and the terminal CH 3 group of OTS. The bonded OH showed more ice like response suggesting the formation of a more ordered bonding network next to the crystalline monolayer. Spectra taken for disordered liquid oil next to water showed spectra similar to the air water interface as shown in (figure 4.9) [112]. Molecular dynamics simulations gave a similar picture. The calculations for a water/decane interface showed an orientational distribution very similar to that of water/air interface [125]. Table 4.2 consolidates the peak positions and their origins for water molecules in different situations. To study the structure of water under confinement, the sample geometry used for confinement of alkanes was used. The confinement cell was kept in soap solution for a week with repeated changes of the solution to remove trace impurities of organics present. The sample cell was plasma cleaned before using. The sapphire substrate was prepared as before and plasma cleaned just before bringing it in contact. In spite of the literature evidence, it was necessary to do control experiments before confining water to determine changes in structure on confinement. The first control was to look at the surface of sapphire. The experiment was done at an incident angle of

126 Figure 4.9: SFG spectra of water interface with solid and liquid hydrophobic surface. (a) water / OTS / silica interface. (b) water / air interface. (c) water / hexane interface [112]. 108

127 Table 4.2: Peak positions and their origin for water molecules next to various surfaces. origin peak position (cm 1 ) liquid-like water network 3400 ice-like water network 3200 and 3000 dangling OH next to air 3700 dangling OH next to hydrophobic surface 3690 partially bound surface OH 3600 liquid-like D 2 O network 2510 ice-like D 2 O network 2375 dangling OD next to air 2740 PDMS CH 3 -symmetric 2910 PDMS CH 3 -asymmetric

128 degrees, which is the critical angle for sapphire / air interface. The spectra taken in PPP and SSP polarization combination are shown in figure 4.10a. The spectra taken throughout the study were fit to equation 4.1 to determine A q values. The spectra are similar to spectra taken by Shen and coworkers for the water / air interface (figure 4.5). It shows peaks related to liquid like and ice like water at 3400 and 3200 cm 1 respectively. We also see the peak due to dangling OH at 3700 cm 1 as seen in figure 4.5 for water / air interface. This suggests that the sapphire surface contains a layer of water or a hydration layer with bound water as well as dangling OH groups. Figure 4.10b also shows spectra taken when bulk water is brought in contact with sapphire. This control has been studied by many groups [117, 124]. The spectra shows broad peaks related to ice like and liquid like water at 3200 and 3400 cm 1 respectively. The spectrum is similar to the one shown in figure 4.7 for the silica / water interface. This has been attributed to the presence of a partially hydrogen-bonded network of water molecules. The peak at 3700 cm 1 is missing suggesting that all dangling OH groups are bonded to either H or O from neighboring water molecules. The ph of water used in the experiments was 7. The spectra taken using SSP polarization combination reflects the higher ph value showing higher intensity of the ice like peak. The next control considered was water next to a hydrophobic surface, in our case PDMS. Figure 4.11c shows SFG spectra taken in PPP and SSP polarization combinations. The left axis denotes the PPP intensity and the right axis the SSP 110

129 1000 a) 'PPP sapphire water' 'SSP sapphire water' 250 PPP intensity (arb.) SSP intensity (arb.) Wavenumber (cm -1 ) 3600 PPP Intensity (arb.) b) 'PPP sapphire air' 'SSP sapphire air' SSP Intensity (arb.) Wavenumber (cm -1 ) 3600 Figure 4.10: SFG spectra of controls taken using SSP and PPP polarization combinations. (a) sapphire / water interface. (b) sapphire / air interface. The markers present the data points and the solid line is a fit to the data using equation

130 PPP intensity (arb.) a PPP PVNODC D 2 o SSP PVNODC D 2 o PPP intensity (arb.) b PPP PDMS D 2 o PPP intensity (arb.) c 2800 PPP pdms water SSP pdms water SSP intensity (arb.) Wavenumber (cm -1 ) 3600 Figure 4.11: SFG spectra of controls taken using SSP and PPP polarization combinations. (a) PVNODC (long alkyl side chain acrylate) / D 2 O interface. Left axis represents PPP D 2 O region and SSP hydrocarbon region. First right axis represents PPP hydrocarbon region and Right (blue) axis represents SSP D 2 O region. (b) PDMS / D 2 O region in PPP combination. (c) PDMS / water interface. The markers present the data points and the solid line is a fit to the data using equation

131 intensity. The spectra were taken at an incident angle of 18 degrees which is the critical angle for the PDMS / water interface. The spectra show peaks at 2910 and 2965 cm 1 related to CH 3 symmetric and asymmetric respectively. In the water region, we see peaks related to ice like and liquid like water at 3200 and 3400 cm 1 respectively with PPP showing higher intensity. We also see a peak at 3600 cm 1, which has been assigned to sapphire OH (bonded OH stretch mode of surface water molecules with one bonded OH and one dangling OH) [120,122]. The peak is broad and includes contributions from dangling OH at 3700 cm 1. As seen in the figure 4.9, water next to a hydrophobic surface shows spectra similar to water / air interface. Therefore, the presence of the 3600 cm 1 peak suggests signal coming from the sapphire / PDMS interface. To confirm this, experiments were conducted by replacing water with D 2 O. D 2 O shows peaks in the cm 1 region and has similar characteristics of water. The PPP spectrum taken from PDMS / D 2 O interface is shown in figure 4.11b. In the D 2 O region, a broad peak covering 2375 and 2510 cm 1 has been assigned to ice like and liquid like D 2 O respectively. The peak at 2740 cm 1 is assigned to dangling OD similar to dangling OH in the water region [126]. Therefore, the presence of free OD peak at 2740 cm 1 along with the hydrogen-bonded D 2 O network is an indication of ordering of D 2 O molecules next to the PDMS surface. Another interesting aspect of the spectrum is the peak at 3600 cm 1, similar to one seen for the PDMS / water interface. This suggests that even though we are at the 113

132 critical angle for PDMS / D 2 O interface we are picking up a signal from the sapphire / PDMS interface. An, intensity comparison shows lower intensity for free OD peak at 2740 cm 1 with respect to the liquid like and ice like peaks, whereas water next to OTS showed a trend of comparable intensity. To investigate this further, the PDMS film was replaced with crystalline PVNODC film to create a surface similar to OTS. The spectrum in PPP and SSP combination taken for PVNODC / D 2 O interface is shown in figure 4.11a. In the D 2 O region, we find peaks related to ice like (2375 cm 1 ) and liquid like (2510 cm 1 ) D 2 O network with SSP combination showing more noticeable ice like features. Also, spectra from both SSP and PPP show peak at 2740 cm 1 related to dangling OD comparable in intensity with the bound OD peaks. Considering the hydrocarbon region, the spectra confirm the well packed structure of the alkyl side chains of PVNODC. This can be seen by the presence of CH 3 symmetric and CH 3 Fermi peaks at 2875 and 2935 cm 1 respectively. This structure has been previously reported in the literature [127]. These controls suggest that D 2 O orders similar to water next to hydrophobic surfaces. Also,the presence of dangling OH or OD peaks can be considered as possible markers to identify water (D 2 O) next to hydrophobic surfaces and the intensity of the peaks can be related to ordering (packing) found in the hydrophobic entity. The behavior of the 3600 cm 1 peak related to sapphire OH groups found in controls where a hydrophobic film was coated on sapphire needs further investigation. 114

133 PPP intensity (arb.) PPP sapphire D 2 O SSP sapphire D 2 O SSP intensity (arb.) Wavenumber (cm -1 ) Figure 4.12: SFG spectra of bulk D 2 O next to sapphire taken using SSP and PPP polarization combinations. The markers present the data points and the solid line is a fit to the data using equation 4.1 [128]. 115

134 It can be recalled that when the sapphire / water interface was investigated we saw no peak at 3700 cm 1 and only peaks at 3200 and 3400 cm 1 respectively were present (figure 4.10). These were related to ice like and liquid like hydrogenbonded network and it is suggested in the literature that the hydrogen in the water molecule bond with the surface OH groups and hence we do not see separate peaks at 3600 and 3700 cm 1. To test whether similar behavior is seen in case of D 2 O, SFG spectra were taken for D 2 O next to sapphire surface. Figure 4.12 shows the spectra taken in SSP and PPP polarization combinations. In the deuterium region, we see peaks related to hydrogen-bonded D 2 O network, but the peak at 3600 cm 1 related to partially bound dangling surface OH groups is still present. This suggests that the hydration layer on the sapphire surface cannot be easily removed or exchanged and has to be taken into consideration. With these controls, let us consider what happens to water molecules when they are confined between hydrophilic sapphire surface and hydrophobic PDMS. Figure 4.13 shows spectra taken in SSP and PPP polarization combinations for confined water c and D 2 O a at an input angle of 8 degrees (critical angle for sapphire / PDMS interface). Peaks at 2910 and 2965 cm 1 assigned to CH 3 symmetric and asymmetric modes of PDMS chains respectively can be seen in both confined water and D 2 O spectra. The confined water spectrum also shows peaks assigned to ice like and liquid like hydrogen-bonded network similar to the sapphire / water control interface spectra in figure A Peak assigned to dangling OH group (3690 cm 1 ) next 116

135 PPP intensity (arb.) a b 'PPP conf d2o' 'SSP conf d2o' SSP intensity (arb.) Wavenumber (cm -1 ) 3600 conf water PPP conf water SSP PPP intensity (arb.) c SSP intensity (arb.) Wavenumber (cm -1 ) 3600 Figure 4.13: SFG spectra of confined water and D 2 O taken using SSP and PPP polarization combinations. (a) confined D 2 O water region. (b) deuterium region. (c) confined water. The markers present the data points and the solid line is a fit to the data using equation

136 to hydrophobic surface can also been seen with intensity comparable to the bonded OH peaks. This trend is similar to spectra taken for water next to well ordered OTS or PVNODC surface (Figure 4.9, 4.11). Therefore, the spectra from confined water can be interpreted as a mixture of spectra of bulk water next to sapphire and water / air interfaces. The presence of PDMS peaks points to a heterogeneous landscape with puddles of water and PDMS in contact with sapphire. Confinement of D 2 O also features peaks in the water region suggesting the presence of hydration layer of sapphire. Along with the hydration layer,the presence of D 2 O is confirmed by the presence of ice like and liquid like features of D 2 O network at 2375 and 2510 cm 1 (figure 4.13b). But the overall intensity in the D 2 O region is lower compared to the water region suggesting lower concentrations of D 2 O. This lower intensity of the spectra makes it harder to ascertain the presence of a free OD peak (2740 cm 1 ) in comparison to free OH from the hydration layer. The presence of hydrogen-bonded water (D 2 O) or bulk water D 2 O under confinement can also be observed from thickness measurements. Using the Newton ring interference technique as explained in the experimental section, the thickness of water under confinement was found to be 40 ± 5 nm. Using spectroscopic information and thickness values, a heterogeneous landscape at the sapphire surface can be inferred. A molecular picture in different situations is shown in figure Figure 4.14a, for the sapphire surface shows the presence of a water hydration layer at the surface with dangling OH groups. When bulk water is added and the 118

137 Figure 4.14: Pictorial representation of water structure in different environments. (a) sapphire / air interface. (b) bulk water next to sapphire. (c) water next to PDMS film. (d) D 2 O next to PDMS film. (e) confined water. The gray colored region represents the sapphire substrate, the blue hashed region is either the hydration layer or hydrogen-bonded water and green region represents the PDMS film or PDMS lens. 119

138 sapphire / water interface is probed we see that water molecules hydrogen bond with the surface OH groups and create a hydrogen-bonded network (figure 4.14b). If a hydrophobic film like PDMS or PVNODC is coated on sapphire, we see that the hydration layer on the sapphire surface still exists and that bulk water next to the hydrophobic surface orders like bulk water next to air. The extent of order indicated by an ice like peak increases with the commensurability of the water network with the surface (hexagonal packing of surface). Dangling OH groups as seen next to air are also found in spectra of water next to a hydrophobic surface (figure 4.14c). Similar behavior is shown for D 2 O next to a hydrophobic surface (figure 4.14d). The elastic nature of the PDMS network leads to the formation of puddles on confinement of water (D 2 O) molecules. A heterogeneous landscape with PDMS contact areas and puddles of bonded water can be inferred (figure 4.14e). Regions where PDMS comes in direct contact with sapphire show dangling OH groups from the hydration layer of sapphire. It is interesting to examine whether the sliding behavior of elastomer in the presence of water can be explained from this picture. In this regard, figure 4.15 shows the coefficient of friction measured at a constant velocity of sliding of 5 µm/sec. It can be seen from the figure that in the presence of only a hydration layer,the sliding of the PDMS lens shows a higher coefficient of friction compared to that observed in the presence of water. From figure 4.15, it can be seen that the coefficient of friction in the presence of water is still greater than 1. This can be explained by the heterogeneous landscape. 120

139 Coeffecient of Friction Velocity - 5 µm/sec PDMS dry sliding with water 1 with water Time (sec) Figure 4.15: Coefficient of sliding friction measured as a function of time at a constant velocity of 5 µm/sec. (Red line) sliding of PDMS lens on sapphire surface only in presence of hydration layer. (Blue, black line) sliding in presence of water. 121

140 The direct PDMS contact asperities leads to an increase in frictional force and,on extrapolating it can be argued that one of the reasons for the stick-slip behavior may be due to PDMS (asperity) contact [8]. When a uniform layer of water is formed, it leads to low friction and smooth sliding in accordance with the no dramatic change in viscosity has been observed for water of 1 nm thickness between smooth surfaces [107,108]. The next question is why does the PDMS dry contact show such a high coefficient of friction when the above experiments suggest the presence of a water hydration layer on sapphire, which is difficult to remove. To study this aspect, SFG spectra of the PDMS static contact with sapphire in the water region were collected. Figure 4.16 shows the spectra taken in SSP and PPP polarization combination of the PDMS / sapphire interface. The incident angle used was 8 degrees which is the critical angle. The spectra show features at 2910 and 2965 cm 1 relating to PDMS CH 3 symmetric and asymmetric modes. We also see a high intensity peak at 3690 cm 1 corresponding to the dangling OH group of sapphire hydration layer next to a hydrophobic surface. Apart from the concentration, we know that the signal intensity is a function of orientation in SFG spectra. Therefore, it can be inferred from the spectra that the PDMS network under load is interacting with the OH groups on the sapphire surface. This may also explain the high adhesion hysteresis seen for PDMS on glass [76]. 122

141 PPP intensity (arb.) SSP sapphire pdms PPP sapphire pdms SSP intensity (arb.) Wavenumber(cm -1 ) 3600 Figure 4.16: SFG spectra of the PDMS / sapphire interface taken in SSP and PPP polarization combinations. The markers present the data points and the solid line is a fit to the data using equation

142 In summary, for the first time the molecular structure of water under confinement in between an elastomer and a rigid surface has been successfully spectroscopically probed. It is seen that the sapphire surface is covered by a hydration layer with dangling OH groups which are difficult to remove or replace. Confinement of water leads to trapping of water molecules in puddles, suggesting a heterogeneous landscape with PDMS contact regions mixed with water regions. Coefficients of sliding friction measured show higher values for dry sliding (bound water present) compared to those measured in the presence of bulk water. Probable reasons for higher friction coefficients are suggested based on the molecular picture from the static spectroscopy measurements. Experiments are underway to capture changes in SFG spectra during sliding by performing in-situ measurements. The ability to take snap shots of molecular structure in time scale of seconds can leads to a better understanding of molecular changes during frictional sliding. This can also help in answering questions related to effect on the heterogeneous landscape during sliding and whether water finally forms a uniform layer removing PDMS contacts. 4.3 Phase transitions under confinement in linear alkanes The fundamental interest in this area is to understand the physical effects of surface forces and reduced dimensionality. Examples of such systems include materials confined in narrow pores, reverse micelles and slit geometries like lubrication in hard disks. On reduction in size of the confined space approaching the range of inter- 124

143 molecular forces, we can expect changes in phase transitions like freezing and even the formation of new phases. As seen in wetting and capillary condensation, physical behavior of the confined molecules can be anticipated to depend on the interplay between molecule-boundary and molecule-molecule interactions. The nature of the phases formed, their structure and properties (diffusion rates, shear properties) as well as the effect of pore size and morphology are of considerable interest and are at present not well understood. Therefore extending the work done in the previous section, it was decided to study the effect of chain length on structure and on the melting - freezing transition temperature under confinement at the flexile elastomer - rigid substrate interface. Let us first review the literature in the field of phase transitions under confinement. Most of the experimental work has been carried out in porous media either ordered or disordered. Examples for ordered media include zeolites, porous silicon, carbon nanotubes, MCM-41 and SBA-15. Most of these are in the form of cylindrical pores. Some of the disordered media include porous glass, silica aerogel, carbon aerogel, activated carbon fibers etc. These do not have regular shapes and can have a network of pores, hence making characterization of surface and pore properties difficult. In most cases, the dimensions of the pore are of the order of 1nm to tens of nm [129]. Another geometry that can be used is one where the gap width can be controlled with a slit or a two dimensional slab using the surface force apparatus. The SFA also provides atomically smooth surfaces and flexibility to alter the surface 125

144 properties [130]. The fluids studied under confinement range from simple fluids like linear and branched alkanes, water, benzene, carbon tetrachloride [34, ] etc. to complex fluids like ionic fluids [134, 135], polymers [136, 137], colloids [ ], liquid crystals [ ] etc. The experimental techniques used to study the phenomena vary from thermal measurements to spectroscopy. Some of the techniques used are DSC thermal analysis, adiabatic calorimetery, nuclear magnetic resonance, SFA, neutron quasi-elastic or inelastic scattering, Dielectric relaxation spectroscopy, and light scattering measurements. There is also a body of literature focusing on molecular simulations using Monte Carlo as well as molecular dynamics approaches. There are also many theories developed like liquid state theory and crystal state theory based on dislocation-mediated melting and theories on two dimensional crystal formation and melting [144,145]. Excellent reviews by Gelb et al, [146] on experiments in pores, Christenson [130] on experiments in slit geometry and Simionesco et al. [129] on simulations and theoretical models give an idea of the current understanding of the subject. Some experimental evidence presented in the literature shows a depression in the freezing temperature of confined fluids where the extent of the depression increases with a decrease in pore size [146]. Experiments in controlled slits using SFA have also shown an elevation of the freezing temperature [130].A qualitative analysis of the results points to the nature of surface (interface) playing the main role in deciding whether there is a depression or elevation of the freezing temperature. Regarding the 126

145 structure of the confined fluids,the crystals formed under confinment have been found to be more defective than bulk crystals and in some cases an absence of crystallization and a formation of an amorphous core have been noted [147]. To explain these results the most commonly used simple thermodynamic treatment has been based on the Gibbs-Thomson equation that is obtained either by equating the free energies of the confined liquid and solid [148] or by determining the temperature at which the chemical potential of the confined solid equals that of the bulk reservoir [149]. The shift in freezing temperature, T f is related to the slit size, d by equation 4.2: T f = T f,conf T f,bulk = 4(σ ws σ wl )T f,bulk d H f ρ s (4.2) Where σ ws and σ wl are the wall-solid and wall-fluid surface tensions, H f is the bulk latent heat of melting and ρ s is the density of the solid. In equation 4.2, the sign of the shift in freezing temperature is given by the difference of the surface tensions. Thus, equation 4.2 predicts that the freezing temperature will be decreased or increased compared with the bulk value when the pore wall prefers the liquid phase to the solid phase or prefers the solid phase to the liquid respectively. [] Considering the present study, one of the walls of the confining geometry is an elastomer. Therefore, the porosity of the elastomer and the affinity of the fluid to enter the matrix need to be considered. The linear alkanes used as the confined fluid swells the PDMS elastomer. This aspect of rubber swelling has been previously studied in depth. Researchers as early as the 1960s have looked into swelling of natural rubber with fluids like benzene [150]. They reported an unexpectedly large freezing-point depression of benzene in 127

146 crosslinked natural rubber. They hypothesized that the polymer network constituted a mesh which restricted the size of the frozen solvent crystals and resulted in 5-15 K depression, which could not be predicted by solution thermodynamics alone. But the use of crystal size or Gibbs-Thomson equation cannot account for the large depression observed, as seen by measurement of the crystal size using X-ray line broadening experiments [151]. Considering the swollen elastomer network as a polymer solution, the theory of mixing presented by Flory [152] can be applied to explain the melting and freezing point depression observed. According to the theory, a depression in transition temperature is expected due to the lowering of the chemical potential of solvent molecules trapped in the elastomer matrix. A theoretical curve for melting point depression based on the Flory-Huggins lattice model is obtained by equating the chemical potential of the solvent crystal with that of the surrounding solution. The resulting equation is : ( 1 ) ( 1 ) = R [ln(1 ν T m Tm 0 2 ) + [1 (1/x)]ν 2 + χν H 2] 2 (4.3) f Where, H f is the bulk heat of fusion of the solvent, ν 2 is the polymer volume fraction, x is the degree of polymerization of the polymer and χ is the polymer-solvent interaction parameter. For elastomeric networks, x is assumed to be very large so that [1 (1/x)] is equal to 1. The equation fits data for uncrosslinked and lightly crosslinked networks but fails for highly crosslinked networks. The experimental values of depression in freezing point are greater than for depression in the melting.while the phenomena of freezing depend on under cooling and nucleation, the Flory calcu- 128

147 lation is only strictly applicable to equilibrium melting. Taking this into account, the problem was revisited by McKenna and workers [153,154] to study melting behavior in benzene / natural rubber systems. They conclude that a sizable portion of the melting point depression in benzene / natural rubber systems can be accounted for by the lowering of the thermodynamic potential of solvent molecules and the remaining portion can be attributed to various physical effects of restriction of crystal size and difficulty in nucleation. New theories combining Flory s calculation and the Gibbs- Thomson theory of depression due to crystal size have been developed [155, 156]. Additional terms describing network elasticity have been added [155] to better fit the experimental results. The advances in instrumentation and new theories have led use of melting point depression to characterize rubber networks [157,158]. Continuing the work in the previous section, experiments were carried out to study the effect of chain length on transition temperatures under confinement. Three separate tools were employed to study the effect. First; differential scanning calorimetery of PDMS lenses swollen with alkanes of varying chain lengths were conducted. DSC experiments were conducted to study the effect of concentration of alkanes in bulk rubber on transition temperature. Next; helium neon laser source was used to look at transition temperature under confinement using the same geometry as used in previous section. Last, SFG spectra were taken at different temperatures to study the changes in structure with temperature and thereby determine phase transitions. The DSC experiments can be thought of as probing the macroscopic phenomena, HeNe 129

148 reflectivity probing micron distances from the sapphire substrate and SFG probing the molecular processes involved. DSC thermal analysis was performed using sealable liquid pans in the TA universal 2000 system. The ramp rate was 0.5 K /min. The samples were equilibrated above the melting temperature for 1 hour before a temperature scan. This was done to give sufficient time for the PDMS lens to swell because of the alkanes. The concentration of alkane in the pan was determined from the equilibrium concentration measured by swelling the lens in a vial of alkane and measuring the weight of the lens before and after swelling. The heating curves for 4 different chain lengths measured are shown in figure 4.17 and the cooling curves are shown in figure Repeat scans were carried out to improve statistics. Considering the heating curves, at first glance we see that all chain lengths show peaks at the crystal to rotator transition which match the bulk value, since this transition is not concentration dependent. However, in the case of the rotator to melt transition we see differences between short and long chains. In the case of C15, the concentration of alkane used was 15%, and we see no transition at the bulk T m indicated by vertical (solid) line. We see a broad peak starting from 6 0 C. This suggests that all the alkane is inside the rubber and the broad transition indicates the presence of crystals of varying size (network pore size distribution). The cooling curve for C15 shows crystallization beginning at 2.5 C suggesting existence of hysteresis between crystallization and melting. In the case of longer chain alkanes, 130

149 0.3 C C C x 2 Heat Flow (Watt/g) C21 C27 Tm T conf 30.9 C C x C C C Temperature ( 0 C) Figure 4.17: DSC heating scans for alkanes of different chain lengths. The scans are offset and C19, C21 scans are scaled for clarity. a concentration of 10% was used even though the equilibrium swelling concentration was lower. The DSC heating curves reflect this with the scans showing rotator to melt transition at bulk T m. The cooling curves in figure 4.18 show lower hysteresis and T f close to T m for C19 - C27 suggesting the presence of alkane crystallizing outside the lens. Figure 4.19 shows the compiled results of variation of bulk T m, confined T m, T f, confined T f and lower limit of confined T m with chain length. To analyze this trend, let us consider a combination of a Flory calculation and Gibbs-Thomson theory discussed in the material on natural rubber [153]. 131

150 0.4 C15 X C Heat Flow (W/g) C19 C21 X C X C C C Temperature ( 0 C) Figure 4.18: DSC cooling scans for alkanes of different chain lengths. The scans are offset and C15, C19 and C21 scans are scaled for clarity. Using the Flory-Huggins lattice model in equation 4.3 as applied by McKenna and coworkers [153], the melting point depression was calculated for alkanes with increasing chain lengths confined in PDMS elastomer. The interaction parameter χ used in the equation 4.3 was independently calculated using solubility parameters at the critical temperature of the bulk crystal melting transition [152]. Table 4.3 gives the values of parameters used in equation 4.3. The equation used to calculate the solubility parameters for alkanes is given as a footnote in table 4.3. The value of the solubility parameter for PDMS is taken from the literature. 132

151 60 50 Transition Temperature ( 0 C) DSC_PDMS_heat DSC_PDMS_cool DSC_heat DSC_PDMS_lowlimit Chain length Figure 4.19: Transition temperatures measured during heating and cooling cycles using DSC. The abscissa represents the number of carbon units in the alkane chain. The error bars for the data are smaller than the symbol size. 133

152 Table 4.3: Values of parameters used to compute equation 4.3. chain length (n) ν alkane Bulk T m (K) H f (J/mol) [159]V (cm 3 /mol) χ a a The interaction parameter χ was determined using solubility parameters [152]. The equation is given by where δ alkane = MPa 0.5 [161]. χ = V (δ alkane δ pdms ) 2 RT (4.4) n n [160]. The value of δ pdms is taken as

153 Figure 4.20 shows the calculated χ as well as the melting point depression variation with chain length. It can be seen from the figure that χ increases with chain length suggesting lower miscibility for longer chain alkanes. Similarly melting point depression also shows an increase with chain length. Comparing figures 4.19 Depression in Melting point ( 0 K) chi 'delta T' Chain Length Figure 4.20: Predictions of 4.3 for a series of chain length of alkanes in PDMS. Also shown is the variation of interaction parameter χ with chain length. and 4.20, it is clear that the predicted melting point depression using equation 4.3 is higher compared to that seen in our experiments. The only chain length which shows measurable depression is C15. Among the temperatures plotted in figure 4.19 the 135

154 trend which comes close to the depression predicted by theory is the lower limit of T m. But according to the Flory calculation, the predicted depression should match the higher limit of T m corresponding to the overall concentration of alkane present inside the rubber. To explain this anomaly, let us consider the effect of χ in equation 4.3. The predicted increase in χ with chain length cannot be reversed, since entropy increases with increase in chain length and free energy of mixing favors phase separation compared to miscibility. The only other parameter is the volume fraction of the alkane. A decrease in volume fraction leads to higher depression and an increase in volume fraction leads to more alkane present outside the rubber as seen in figure But, before coming to the conclusion that the model does not capture reality (experimental data), it is important to ascertain the accuracy of experimental procedure adapted. To do this, experiments were carried out using natural rubber as is discussed in the reference [153]. Figure 4.21 shows DSC heating scans for C15 and C21 alkane confined in natural rubber. Concentrations of 15% and 10% were used for C15 and C21 respectively. Using the solubility parameter technique as before, the values of melting point depression predicted by the Flory calculation are also shown along with the bulk T m. It can be seen from the figure that the values predicted by the Flory calculation are close to the maximum T m values obtained from the experiments. Therefore, the experimental procedure is not flawed, but the underlying physical process in case of PDMS is not captured by Flory calculation. 136

155 Heat flow (W/g) Heat flow (W/g) C21 in NR 10 C15 in NR FH theory = C C Temperature ( 0 C) 30 FH theory = C C 40 T m Figure 4.21: DSC heating scans for C15 and C21 alkane in natural rubber. The melting point depression predicted from the Flory calculation and the bulk T m are shown as vertical lines. 137

156 Considering that the motivation at the start of this study was to look at phase transitions of alkanes under confinement between an elastomer and a rigid surface, helium neon [HeNe] reflectivity measurements were performed to look at phase transition close (microns) to the sapphire surface. Figure 4.22 shows the experimental geometry and a scan of reflected helium neon intensity with temperature for C27 alkane. The alkane swollen PDMS lens was brought in contact with a sapphire prism mounted on a steel cell to confine the alkane molecules at the interface. The pressure was sufficient to deform the lens and create a uniform contact surface of 2 mm. A HeNe laser beam with a chopper and photo detector attached to a SR 850 lock-in amplifier was used to observe the reflected intensity from the contact area. Since the experiments were started from the melt state, to get maximum reflection, an incident angle close to the critical angle for the liquid (8 degrees) was used. The steel cell was equipped with a rod heater and was mounted on a copper block maintained at a constant temperature of 10 C. The temperature of the cell was varied at a rate of 0.2 K /min using a Lakeshore 331 temperature controller attached to the rod heater. The evanescent field from the laser probes changes in refractive index close to the sapphire surface (micron thickness). The confined alkane was held at a temperature 5 C above the bulk T m before starting the scan. Figure 4.22 shows a drop in reflected intensity when the alkane in the contact region crystallizes. The drop in intensity is due to a change in the refractive index from 1.45 to 1.55 on crystallization. When we start heating the crystals, we see a transition when the crystals melt and we recover 138

157 Figure 4.22: Scan of reflected HeNe intensity with temperature for C27 alkane confined between PDMS lens and sapphire substrate. Inset shows the experimental geometry. 139

158 the reflected intensity for the melt. The temperature corresponding to 50% change in intensity was noted as the transition temperature. Heating and cooling cycles were repeated to get reproducible values. Similar experiments were carried out for the series of odd alkanes from 15 to 27 carbon units. The transition temperatures measured as function of chain length are complied in figure Transition Temperature ( 0 C) He_Ne_heat He_Ne_cool DSC_heat Chain length Figure 4.23: Transition temperatures measured during heating and cooling cycles using helium neon reflectivity. The abscissa represents the number of carbon units in the alkane chain. The error bars for the data measured fall within the dimensions of the symbols used. 140

159 The maxmimum melting point depression is observed for C19 and the depression decreases with increase in chain length. But, surprisingly, C15 and C17 show negligible depression with melting point close to bulk T m. Before considering an explanation for the results it is important to know whether the reflected intensity drop we are considering is due to a phase change to the crystal state. Lower transition temperatures can be obtained if the reflected intensity change is due to an intermediate state. To confirm this would require determination of refractive index in liquid and crystal states under confinement and in bulk. This was done by measuring reflected intensity as a function of incident angle at a fixed temperature. Then, using a three layer reflectivity model described in the experimental section, the critical angle measurement data were fitted to get the refractive index and thickness in the liquid and crystal state. Figure 4.24 shows reflected intensity data measured with respect to incident angle for sapphire - PDMS dry contact, C15 confined alkane liquid, crystal and bulk alkane crystal next to sapphire. The intensity profile for the confined liquid was fit to the three layer model given by equation 4.5 I out =[t p (prismin) conj(t p (prismin))] [r p conj(r p )] [t p (prismout) conj(t p (prismout))] (4.5) The refractive indices from the fit match the values found in the literature. The best fit is obtained for a thickness of 10 nm. But it is not possible to pin point the exact value as seen from the other fits using 100 nm and 250 nm thickness. The intensity profiles for sapphire - PDMS contact, bulk C15 crystal and confined crystal show a 141

160 0.8 n sapphire = 1.76 n pdms = 1.41 n c15 liq = 1.45 He-Ne reflected intenstiy (I / Io) n c15 cry =1.52 c15_conf_liquid 'fit_c15_liq_ 250 nm' 'fit_c15_liq_10 nm' 'fit_c15_liq_100 nm' pdms_sapphire c15_bulk_crystal c15_conf_crystal Angle (deg) Figure 4.24: Reflected intensity versus incident angle measured for sapphire-pdms contact, C15 bulk crystal, C15 confined liquid and confined crystal. The intensity profile for the C15 confined liquid is fit to a three layer reflectivity model (equation 4.5) to determine the alkane thickness. 142

161 shallow drop, which cannot be predicted using the model. Since, the model assumes sharp and smooth interfaces, the trend seen may be attributed to the presence of surface roughness or air pockets. Experiments repeated for C21 and C23 alkane also show similar behavior. Figure 4.25 shows the reflected intensity variation with incident angle measured for confined alkanes of 21 and 23 carbon units. He-Ne reflected intensity (I / Io) n c23 cry =1.56 n c21 cry =1.558 n c15 cry =1.52 n c21 liq =1.45, d =10 nm n sapphire =1.76 n PDMS =1.413 c21_conf_cry c21_conf_liq c23_conf_cry fit_c21_liq c15_conf_cry Angle (deg) Figure 4.25: Reflected intensity versus incident angle measured for C21 confined liquid, confined crystal and C23 confined crystal. The intensity profile for C21 confined liquid is fit to a three layer reflectivity model (equation 4.5) to determine the alkane thickness. 143

162 The refractive indices calculated by fitting the initial drop for confined crystal and bulk crystal show similar values. Also, in the case of confined crystals, varying the refractive index and the thickness of the middle layer cannot fit the critical angle. A good fit is obtained only by changing the refractive index of the third layer to match with that of the crystal, suggesting the presence of either a heterogeneous structure with alkane and PDMS chains intertwined or a thick layer of crystal (microns). Since the critical angle measurement did not give an accurate value for the thickness of an alkane confined film, a Newton ring based light interference technique was used. This technique was adopted by Roberts and Tabor in the 1960s to determine thickness and study confinement of fluids [8]. The details of the technique are described in the experimental section. Figure 4.26 shows Newton rings formed when C15 alkane is confined between PDMS and sapphire. Figure 4.26a shows rings when the thickness of the liquid is greater than the wavelength of light used. Figure 4.26b shows the contact area when the thickness is less than 1 4th λ. The intensity at the center of the contact area is measured as the thickness is decreased and using equation 4.6 the thickness of the confined alkane is determined. R = R 0 sin 2 [2πn 1 h/λ] (4.6) For C15 confined liquid, the thickness was found to be 10 nm. A thickness below 10 nm is difficult to determine due to back scattering from PDMS elastomer. SFG spectra taken for the series of alkanes give an insight into what happens close to the sapphire surface. SFG spectra were taken using the same geometry to ob- 144

163 Figure 4.26: Newton rings formed on illumination of contact spot using HeNe laser. (a) shows the rings formed before contact (thickness greater than λ) and (b) shows contact area (thickness less than λ 4 ). Intensity (arb.) C C C C C C Intensity (arb.) Wavenumber (cm -1 ) Figure 4.27: SFG spectra in SSP polarization combination for C15 confined alkane at different temperatures. 145

164 serve the molecular structure changes next to the sapphire surface during transitions. Figure 4.27 shows SFG spectra in SSP polarization combination for C15 confined alkane at different temperatures. The confined alkane was first cooled to 5.5 C. The spectra were taken during the heating cycle. Since, the spectra were taken during the heating cycle an incident angle of 2 degrees was used corresponding to the critical angle for a confined crystal. The temperature was increased at a rate of 0.2 K /min and a 10 minute wait time was given at every temperature step before the spectrum was taken. The spectrum at 5.5 C is similar to one shown for a confined alkane crystal in figure 4.2d. It should be noted that the spectrum was interpreted to mean that the alkane chains on crystallization lie flat next to the sapphire surface with their symmetry axis parallel to the surface normal and show layering. The spectra show an increase in intensity as the temperature is increased, but the spectral features remain the same. At 9.6 C, the alkane melts and we recover the structure of the confined alkane shown in figure 4.2c. This suggests that the transition occurs close to 9.6 C, which is close to the bulk melting temperature. Similarly, SFG spectra for the remaining odd alkanes were measured. The spectra are shown in figure A q values obtained by fitting SSP spectra to equation 4.1 are compiled in table 4.4. First let us consider the confined liquid spectra (Left panel). The spectra and A q values for C15 confined alkane show more CH 3 (r + ) groups at the surface and lower CH 2 (d + ) concentration, which keeps increasing with chain length. The A q values for CH 3Fr (r + Fr ) follows r+ and decreases with chain length. The spectral 146

165 'c15-liq' 'c19-liq' 'c21-liq' 'c27-liq' 'c15 -cry' 'c19-cry' 'c21-cry' 'c27-cry' Wavenumber (cm -1 ) Wavenumber (cm -1 ) 3100 Figure 4.28: SFG spectra taken in SSP polarization combination for confined alkanes of varying chain length in liquid and crystal states. The left panel shows the liquid spectra and the right panel the crystal spectra. The chain length increases from bottom to top. The experimental data are plotted as symbols and the solid line is a fit to equation

166 Table 4.4: A q values obtained by fitting SSP spectra to equation 4.1 for confined alkanes of varying chain length in liquid and crystal states. d + r + r fr r c15 confined liquid c19 confined liquid c21 confined liquid c27 confined liquid c15 confined crystal c19 confined crystal c21 confined crystal c27 confined crystal

167 intensity also decreases with increase in chain length. Thus, the confined liquid spectra suggest lower orientational order and a concentration similar to the bulk liquid as chain length is increased, suggesting a lower effect of confinement as chain length increases. A similar trend is seen in the case of confined alkane crystal next to sapphire (right panel). The higher CH 2 (d + ) concentration and spectral intensity seen for C15 decreases with chain length, with C27 showing spectrum similar to bulk alkane crystal. The transition temperatures measured using SFG are compiled along with He-Ne and DSC values in figure It can be seen from figure 4.29 that SFG measurements show a trend similar to HeNe data with the depression being more pronounced compared to reflectivity measurements. The melting point depression is maximum for C19 and decreases with increase in chain length. But, surprisingly, C15 and C17 show negligible depression with melting point close to bulk T m. The trend is reproducible with both HeNe (probing micron length scale) and SFG (probing molecular length scale) confirming the result as seen in figure Combining all the above results a qualitative analysis is presented to interpret the surprising trend in melting point depression with chain length of the confined alkane. For convenience, let us split the chain lengths into two parts. The first part containing C15 and C17, which are liquids at room temperature (sample was prepared by soaking the lenses in the alkane liquid). The other set consisting C19, C21, C23 and C27 which are solids at room temperature (sample was prepared by casting thick 149

168 60 50 Transition Temperature ( 0 C) DSC_PDMS_heat DSC_PDMS_cool He_Ne_heat He_Ne_cool SFG_heat SFG_cool DSC_heat DSC_PDMS_lowlimit Chain length Figure 4.29: Transition temperatures measured during heating and cooling cycles using DSC, helium neon reflectivity and SFG. The abscissa represents the number of carbon units in the alkane chain. The error bars for the data measured fall within the dimensions of the symbols. 150

169 films on the sapphire surface and then pressing with the elastomer in the solid state and heating to above T m for lens to soak in the alkane). The second set shows considerable melting point depression with C19 showing the maximum depression. This can be interpreted as follows. The lens absorbs 50% less C19 alkane when compared to C15 as seen from equilibrium swelling and Flory-Huggins interaction parameter. As we approach the freezing temperature, most of the alkane outside gets frozen first and, as seen from DSC, more alkane is drawn out to the surface. Finally, when the confined alkane freezes, the concentration of alkane present in the rubber is low and heterogeneous contact of alkane and rubber at the interface can lead to smaller crystal sizes. Therefore, the lower concentration and lower size can lead to high melting point depression. The presence of lower crystal size can also be inferred from the HeNe data, which measures the change in refractive index due to the phase transition. Since a change in crystal structure requires much higher pressures than applied by an elastomer, the reason for a phase change can be attributed to smaller crystal size. The decrease in depression as we move towards C21, C23 and C27 can be attributed to lower confinement effects with chain length as seen from SFG (figure 4.28). DSC results as well as Flory-Huggins interaction parameters suggest lower miscibility with increase in chain length. The structure of a C27 confined crystal next to sapphire surface being similar to that of bulk alkane next to sapphire points towards transition temperature being close to T m. 151

170 Considering the first set, we see that C15 and C17 show higher miscibility with PDMS elastomer as seen from DSC and Flory calculation. Both HeNe and SFG suggest transition temperatures close to bulk T m (Figure 4.29). As mentioned above, HeNe measures changes in refractive index and, with an evanescent wave probing close to micron from the surface suggests the presence of larger crystal sizes. DSC also points towards more alkane being present inside PDMS and the smaller size of the molecule favors higher diffusion rates. Therefore, the larger crystal size can lead to transition temperature being close to bulk T m. Figure 4.30 shows a schematic representation of the alkane crystals at the PDMS / sapphire interface for different chain lengths. To get a more clear understanding of this surprising trend, where C19 shows highest depression, attempts were made to make the elastomeric surface non-porous so that the effect of PDMS swelling in the alkane could be avoided. The elastomer was coated with metals like gold and aluminum and also a thin film of mica was floated onto the lens surface. But in all cases the films cracked on application of pressure showing similar results as obtained by using plain elastomer. It may be useful to replace the PDMS elastomer with a fluorinated elastomer having a low surface energy. But, in such a case it becomes thermodynamically unfavorable for the alkane to stay at the interface. In summary, a systematic study was carried out to determine and understand the effect of chain length on phase transition under confinement. Three different tools 152

171 Figure 4.30: Schematic representation of the alkane crystals at the PDMS / sapphire interface for different chain lengths. (a) C15, (b) C19, (c) C

172 probing different length scales were adopted. A surprising trend with some chain lengths showing melting point depression was observed. A qualitative analysis was presented to interpret the observations. To further understand this complex problem with many facets, further experiments and computer simulations are necessary. 4.4 Characterization of molecular structure during sliding friction Sliding of two solid surfaces can be dramatically different in the presence of nanometerthin confined liquids. This region of boundary lubrication is very different from that associated with hydrodynamic lubrication, where the sliding resistance is governed by the bulk viscosity. It has been suggested based on force measurements that the effective viscosity of these nanometer-thin lubricants can be many orders of magnitude higher than in the bulk and even exhibit solid-like response [31,93,162]. Simulations have also shown that the confined liquid undergoes a solid to liquid phase transition upon sliding [94, 162]. Although there have been many shear measurements there is no direct evidence of the structure of the confined liquids during sliding. This has now become a matter of utmost importance with the recent findings of contaminants in the liquid or on the surfaces dramatically influencing shear and normal forces [98]. The structure of confined liquids during sliding has relevance in many fields from tribology, adhesion and wear, micro-fluidics, contact lithography, earthquakes to biological systems. We have combined broadband surface sensitive infrared-visible sum frequency generation spectroscopy (SFG) with friction experiments to measure the changes in 154

173 structure of confined alkane (pentadecane) between an elastomer and a solid surface during sliding. We observe a dramatic ordering of pentadecane molecules at the lens leading edge, with a structure similar to that of crystal state, at temperatures 13 C above the bulk melting temperature. These results indicate that the resistance to sliding is dominated by the fluid resistance at the leading and trailing edges rather than the averaged value over the contact area. This has important implications in the interpretation of the average viscosity of boundary lubricants determined by assuming uniform resistance across the contact region. The friction cell used is described in the experimental section. Friction cell - prototype 2 was employed. The novel approach of using a flexible elastomeric lens which deforms against a flat solid surface to confine molecules offsets the need to have perfectly parallel surfaces as we approach separations of molecular dimensions. The spot probed using SFG is spatially fixed and the lens is made to slide in and out of the contact spot (figure 4.31b,c,d). The shear stress measured for dry sliding is found to be higher compared to lubricated sliding in the presence of an alkane (figure 4.31e). The curves show the presence of initial stiction followed by stick-slip sliding. The thickness of the confined alkane is found to be 10 nm from the Newton-ring technique (figure 4.31f) [8]. This suggests that we are in the boundary lubrication regime and away from hydrodynamic lubrication. Friction experiments without the optical probe were conducted to map out the velocity dependence. The range of velocities accessible using the picometer motors was 30 nm /sec to 60 µ/sec. Sliding experiments were carried out for dry 155

174 Figure 4.31: Experimental geometry and friction forces measured. (a) Schematic of the friction cell developed to have the capability to probe molecular structure using SFG along with force measurements. (b) Schematic of the experiment with the position of laser beams at start of sliding. The SFG probe beam contact spot is ahead of the confinement region. (c) The contact region overlaps with the probe. (d) The confined region is way passed the probe spot. (e) Shear stress measured for PDMS dry sliding and in the presence of alkane as a lubricant. (f) Thickness of the confined alkane film measured using Newton-ring technique. 156

175 PDMS in contact with sapphire as well as in the presence of alkane at four different velocities (figure 4.32). It can be seen from the figure that dry sliding shows higher friction compared to lubricated sliding. The shear stress does not show a dependence on velocity of sliding in the range of velocities tested. The trend for dry sliding is similar to the study of PDMS sliding on glass [163] Shear Stress (MPa) Extracted PDMS Pentadecane soaked PDMS Velocity ( m/s) Figure 4.32: Shear stress measured for dry and lubricated sliding on sapphire as a function of sliding velocity. ( ) dry PDMS sliding, ( ) sliding in presence of pentadecane liquid. For the first time, coupling of a broadband femtosecond SFG system with internal reflection geometry has allowed us to collect snapshots of SFG spectra at intervals of 1 second with signals accumulated over 1 second periods. The layout of 157

176 our SFG system has been described in the experimental techniques section. Sum frequency [SF] spectra of alkane molecules were measured in the hydrocarbon stretching region. The peak assignments for linear alkane chain based on bulk IR and Raman measurements for the CH stretches are as follows [66]: peaks between cm 1 are assigned to the CH 3 symmetric stretch (r + ); two peaks at 2950 cm 1 and 2965 cm 1 to the CH 3 asymmetric stretches (r ); peaks between cm 1 to the symmetric CH 2 stretch (d + ); peak between cm 1 to the asymmetric CH 2 mode (d ); and peak in the range of cm 1 to the CH 3 Fermi resonance due to the overlap of symmetric overtone with the bending mode (r + FR ). Figure 4.33a shows a waterfall plot of SFG spectra measured as the lens slides in and out of the contact spot. The sliding speed was 7.5 µ/sec. At first glance we can see the spectral features change as the lens moves in and out of the spot. To see this more clearly, the variation in intensity of certain peaks as the lens slides is plotted in figure 4.33b. The CH 2 symmetric peak (2840 cm 1 ) shows a huge increase in signal intensity as the lens enters the laser spot and goes down as the lens overlaps with the laser spot and again shows a peak as the lens exits the laser spot. This suggests changes at the leading and trailing edge of the lens. The CH 3 symmetric (2880 cm 1 ) peak shows a similar trend but with the intensity not dropping as low in the overlapping region. The combination of CH 3 asymmetric and Fermi (2950 cm 1 ) increases in intensity as the lens moves into the laser spot remains steady and decreases as the lens moves out of the spot. The CH 2 asymmetric 158

177 Z a) b) Wavenumber (cm -1 ) c) 2840 cm -1 Entry Y Exit 2925 cm cm Distance (µm) Wavenumber (cm -1 ) Figure 4.33: SFG spectra acquired during frictional sliding. (a) Waterfall plot of SFG intensity measured in the hydrocarbon region as the lens is moved in and out of the laser beam. The spectra were taken at 1 second interval with 1 second of accumulation time. (b) SF spectra taken at different positions of the lens with respect to the laser spot. The spectra are normalized and scaled for clarity. (+) SF spectrum of confined crystal next to sapphire taken during heating and cooling experiments under static conditions. ( ) SF spectrum of confined liquid next to sapphire taken under static contact. ( ) SF spectrum taken during sliding when the lens has not moved into the contact spot (bulk liquid). ( ) SF spectrum of confined liquid when the lens overlaps the laser spot. ( ) SF spectrum of alkane fluid at the leading edge of the lens as the lens enters the laser spot. (c) The variation in intensity of selected peak positions as the lens moved in and out of the laser spot. (*) 2840 cm 1, (+) 2950 cm 1 and (X) 2925 cm

178 (2920 cm 1 )peak shows the opposite trend with intensity, dropping as the lens moves into the laser spot and increasing as the lens moves out. To relate these changes in peak intensity to molecular structure changes let us consider snapshots taken at different spatial position of the lens with respect to the laser spot. Figure 4.33c shows spectra at different spatial positions of the lens with respect to the laser spot as well as a spectrum of confined liquid taken under static contact. A spectrum of the confined crystal measured on freezing the confined liquid is also shown for comparison. The spectrum of the confined crystal, showing a high CH 2 -symmetric peak was interpreted to suggest ordering and layering of alkane chains with symmetry axis parallel to the surface normal [99]. The SFG spectrum for bulk alkane next to sapphire ( ), before the lens has moved into the laser spot is similar to one published previously [101]. The presence of the CH 2 symmetric (2850 cm 1 ) and asymmetric (2920 cm 1 ) peaks have been attributed to gauche defects and the disordered nature of the liquid next to sapphire surface. The SFG spectrum of confined liquid taken during static confinement ( ) and the spectrum measured when the lens overlaps the laser spot ( ) are similar. The absence of CH 2 peaks and presence of the CH 3 Fermi and asymmetric peaks suggests ordering of alkane molecules [99]. The spectrum of confined crystal (+) next to sapphire and the spectrum taken when the leading edge moves into the laser spot during sliding ( ) are also similar. It is important to note that the confined crystal spectrum was measured by cooling and freezing the alkane whereas sliding was carried out at room temperature (23 C). 160

179 start of sliding coefficient friction 2840 cm cm cm Distance (µm) 1000 Figure 4.34: Coefficient of friction as well as SFG spectra measured simultaneously as a function of sliding distance. Sliding speed used was 1.5 µ/sec. The SF output, input visible and input infrared were P,P,P polarized respectively. The spectra were taken at 1 second intervals with 2 seconds accumulation time. From the waterfall plot the variation in intensity of selected peak positions as the lens moved in and out of the laser spot is plotted. (*) 2840 cm 1, (+) 2950 cm 1 and (X) 2925 cm

180 To ascertain if the initial stress build up (stiction) at the start of sliding was causing the alkane molecules to order like a crystal, SFG measurements and force measurements were carried out simultaneously. Figure 4.34 shows the coefficient of friction and the SFG peak intensity variation as a function of sliding distance. In this case, sliding was started by bringing the lens in static contact overlapping the laser spot. Therefore, we are observing the changes occurring at the start of sliding. The coefficient of friction shows an initial build up related to stiction and stick-slip sliding there after. It is interesting to note that the CH 2 symmetric (2840 cm 1 ) intensity does not follow the stress build up and shows a maximum as the lens exits the contact spot. The CH 2 asymmetric (2925 cm 1 ) shows an increase in intensity as the lens moves out of the contact spot, suggesting that the left over alkane fluid relaxes to a bulk liquid state. The increase in the CH 2 symmetric peak at the leading and trailing edge of the lens as well as the similarity to the spectrum of the confined alkane crystal suggests ordering of alkane molecules at the edges. The ordering of alkane molecules as the sliding lens enters and leaves the laser spot can be pictorially represented as shown in (figure 4.35). It has been previously shown by optical microscopy that liquids get trapped when an elastomer is brought in contact with a rigid surface [8,39]. This can also be inferred from the present SFG measurements where ordering of alkane molecules is observed in the overlap of the contact area with the laser spot, but not the crystalline order as seen at the edges. This crystalline order at the edges can have important implications. In 162

181 Figure 4.35: Pictorial representation of ordering of linear chain alkane molecules at the leading and trailing edge as the lens slides across the sapphire substrate. case of rigid boundaries, alkanes may show crystalline order throughout the contact area explaining the solid like responses seen in force measurements. The present experiments give the first direct experimental evidence of such a transition taking place at the interface upon confinement. 163