Adsorption, aggregation and

phase separation in colloidal

systems

Jing Dai

KTH Royal Institute of Technology

School of Chemical Science and Engineering

Department of Chemistry

Applied Physical Chemistry

SE-100 44 Stockholm, Sweden

ii

Copyright © Jing Dai, 2017. All rights are reserved. No parts of this thesis can be

reproduced without the permission from the author.

Paper I © 2015 American Chemical Society

Paper II © 2017 American Chemical Society

TRITA CHE Report 2017:88

ISSN 1654-1081

ISNB 978-91-7729-647-8

Akademisk avhandling som med tillstånd av Kungliga Tekniska Högskolan i

Stockholm framlägges till offentlig granskning för avläggandet av teknologie

doktorsexamen fredagen den 9 feb kl 10:00 i sal F3, KTH, Lindstedtsvägen 26,

Stockholm. Avhandlingen försvaras på engelska.

Fakultetsopponent: Prof. Peter Griffiths, University of Greenwich, UK

iii

To my parents

iv

v

Abstract

The thesis presents work regarding amphiphilic molecules associated in aqueous

solution or at the liquid/solid interface. Two main topics are included: the temperature-

dependent behavior of micelles and the adsorption of dispersants on carbon nanotube

(CNT) surfaces. Various NMR methods were used to analyze those systems, such as

chemical shift detection, spectral intensity measurements, spin relaxation and, in

particular, self-diffusion experiments. Besides this, small angle X-ray scattering

(SAXS) was also applied for structural characterization.

A particular form of phase transition, core freezing, was detected as a function of

temperature in micelles composed by a single sort of Brij-type surfactants. In mixed

micelles, that phase transition still occurs accompanied by a reversible segregation of

different surfactants into distinct aggregates. Adding a hydrophobic solubilizate shifts

the core freezing point to a lower temperature. Upon lowering the temperature to the

core freezing point, the solubilizate is released. The temperature course of the release

curves with different initial solubilizate loadings is rationalized in terms of a

temperature-dependent loading capacity.

The behavior of amphiphilic dispersant molecules in aqueous dispersions of carbon

nanotubes (CNTs) has been investigated with a Pluronic-type block copolymer as

frequent model dispersant. Detailed dispersion curves were recorded and the

distribution of the dispersant among different available environments was analyzed.

The amount of dispersed CNT was shown to be defined by a complex interplay of

several factors during the dispersion process such as dispersant concentration,

sonication time, centrifugation and CNT loading. In the dispersion process, high

amphiphilic concentration is required because the pristine CNT surfaces made

available by sonication must be rapidly covered by dispersants to avoid their re-

attachment. In the prepared dispersions, the competitive adsorption of possible

dispersants was investigated that provided information about the relative strength of

the interaction of those with the nanotube surfaces. Anionic surfactants were found to

vi

have a strong tendency to replace Pluronics, which indicates a strong binding of those

surfactants.

CNTs were dispersed in an epoxy resin to prepare nanotube-polymer composites. The

molecular mobility of epoxy was investigated and the results demonstrated the

presence of loosely associated CNT aggregates within which the molecular transport

of epoxy is slow because of strong attractive intermolecular interactions between

epoxy and the CNT surface. The rheological behavior is dominated by aggregate-

aggregate jamming.

Keywords: NMR, chemical shift, spin relaxation, self-diffusion, micelle, core

freezing, segregation, solubilization, release, adsorption, binding, surfactant, carbon

nanotube, block copolymer, dispersion, competitive adsorption, nanocomposite.

vii

Sammanfattning

Avhandlingen innefattar studier av amfifila molekyler som associerar med varandra i

lösningar eller adsorberas vid gränsytor mellan vätska och fast material. Det finns två

huvudteman: det temperatur-beroende beteendet av miceller och adsorptionen av

dispergerande molekyler till ytorna av kolnanorör (CNT, carbon nanotubes). Olika

NMR spektroskopiska metoder har använts för att studera dessa system, bland annat

mätningar av kemiska skift och spektralintensitet, spinnrelaxation, och i synnerhet

självdiffusionsexperiment. Dessutom har small angle X-ray scattering (SAXS)

utnyttjats för strukturell karaktärisering.

En specifik form av fasövergångar, solidifiering av micellens inre, har undersökts vid

olika temperatur i miceller av enskilda Brij-typ surfaktanter. Denna fasövergång

uppkommer även i miceller som består av blandningar av sådana surfaktanter och

resulterar i reversibel segregering av olika molekyltyper i separata aggregat.

Hydrofoba solubilisater bidrar till att minska temperaturen för fasövergången. När

temperaturen minskar ner till övergångspunkten, släpps solubilisaten ut ur micellen till

vattenfasen på ett sätt som beror av micellens reducerade kapacitet att behålla dessa

molekyler.

Beteendet av amfilila dispersanter i vattenbaserade dispersioner av CNT har

undersökts med en Pluronic-typ block copolymer som den vanligaste dispersanten.

Detaljerade dispersionskurvor har uppmätts och distributionen av dispersanten mellan

olika molekylära miljöer analyserats. Mängden av dispergerade CNT beror av olika

faktorer i komplex samspel, såsom dispersantens koncentration, ultrasonikeringstiden,

centrifugeringen, och den initiala mängden av CNT. Hög amfifilkoncentration krävs

under dispersionsprocessen så att de nya interna ytor som bildas med hjälp av

ultrasonikering täcks av dispersanter och ytornas omedelbara slutning undviks.

Kompetitiv adsorption av olika dispersanter har undersökts för att ge information om

den relative bindningsstyrkan till CNT-ytor. Anjoniska surfaktanter som visat sig att

ha en hög kapacitet för att ersätta Pluronic binder starkt till CNT.

viii

Kolnanorör dispergerades även i epoxyresin med ändamålet att skapa polymer-nanorör

kompositer. Den uppmätta molekylära mobiliteten i epoxyn tyder på att det skapats

lösa nanorörsaggregat i dessa system. Inom dessa aggregat, starka attraktiva

intermolekylära krafter mellan epoxy och CNT leder till en långsam molekylär

transport av epoxy. De reologiska egenskaperna domineras av aggregatens ömsesidiga

obstruktion.

Nyckelord: NMR, kemisk skift, spinnrelaxation, självdiffusion, micell, solidifiering,

segregation, solubilisering, utsläpp, adsorption, bindning, surfaktant, kolnanorör,

block copolymer, dispersion, kompetitiv adsorption, nanokomposit.

ix

List of papers

I. Core freezing and size segregation in surfactant core-shell micelles

Beatrice Plazzotta, Jing Dai, Manja A. Behrens, Istvan Furo and Jan Skov Pedersen

J. Phys. Chem. B, 2015, 119 (33), 10798–10806

II. Release of solubilizate form micelle upon core freezing

Jing Dai, Zahra Alaei, Beatrice Plazzotta, Jan Skov Pedersen and István Furó

J. Phys. Chem. B, 2017, 121 (45), 10353–10363

III. Propofol adsorption at the air/water interface: a combined vibrational sum

frequency spectroscopy, nuclear magnetic resonance and neutron reflectometry

study

Petru Niga, Petra M. Hansson-Mille, Agne Swerin, Per M. Claesson, Joachim Schoelkopf, Patrick A.

C. Gane, Jing Dai, István Furó, Richard A. Campbell and C. Magnus Johnson

Submitted

IV. The dispersion process of carbon nanotubes sonicated in aqueous solutions of

a dispersant

Jing Dai, Ricardo M. F. Fernandes, Oren Regev, Eduardo F. Marques and István Furó

Manuscript

V. Block copolymers adsorbed on single-walled carbon nanotubes. Block

polydispersity and the modes of surface attachment

Ricardo M. F. Fernandes, Jing Dai, Oren Regev, Eduardo F. Marques and István Furó

Manuscript

VI. Assessing surfactant binding to carbon nanotubes via competitive adsorption:

binding strength and critical coverage

Ricardo M.F. Fernandes, Jing Dai, Oren Regev, Eduardo F. Marques and István Furó

Manuscript

VII. Polymer nanocomposites: insights on rheology, percolation, jamming and

molecular mobility

Roey Nadiv, Ricardo M. F. Fernandes, Guy Ochbaum, Jing Dai, Matat Buzaglo, Maxim Varenik, Ronit

Biton, Istvan Furo and Oren Regev

Submitted

x

The author contribution to the appended papers

I. Initiated the project, planned and performed all the NMR measurements and data

analysis. Participated in the writing of the manuscript.

II. Planned and performed the majority of the experimental work and data analysis.

Wrote the preliminary version of the manuscript and participated in writing the final

version.

III. Performed the NMR experimental work and data analysis. Wrote the NMR related

text.

IV. Participated in the project planning. Performed the majority of the experimental

work. Contributed to the data analysis and to writing the manuscript.

V. Participated in the project planning. Performed roughly half of the experimental

work. Contributed to the data analysis and to writing the manuscript.

VI. Performed less than half of the experimental work. Contributed to the data analysis

and to writing the manuscript.

VII. Performed roughly half of the NMR-related experimental work. Contributed to

writing the manuscript.

xi

Table of contents

1. Introduction ........................................................................... 1

1.1. SURFACTANTS............................................................................................................. 1 1.2. AMPHIPHILIC BLOCK COPOLYMERS................................................................................ 2 1.3. MICELLES AND SOLUBILIZATION IN MICELLES ................................................................. 3 1.4. CARBON NANOTUBES AND CARBON NANOTUBE DISPERSIONS ......................................... 7

2. Experimental ....................................................................... 13

2.1. SAMPLE PREPARATION ...............................................................................................13 2.1.1. Micelles with and without solubilizates ...................................................................... 13

2.1.2. Propofol with different concentrations in D2O ............................................................ 15

2.1.3. CNT dispersions ........................................................................................................... 15

2.1.4. CNT-loaded epoxy ....................................................................................................... 17

2.2. CHARACTERIZATION BY NMR SPECTROSCOPY ............................................................17 2.2.1. Principles of NMR ........................................................................................................ 18

2.2.2. Chemical shift .............................................................................................................. 21

2.2.3. Spin relaxation ............................................................................................................ 22

2.2.4. NMR diffusion experiments ......................................................................................... 26

2.3. ADDITIONAL CHARACTERIZATION TECHNIQUES ............................................................29 2.3.1. Small-angle X-ray scattering (SAXS) ............................................................................ 29

3. Summary ............................................................................. 31

3.1. AMPHIPHILIC MOLECULES ASSOCIATED IN SOLUTION ....................................................31 3.1.1. Micelle core freezing ................................................................................................... 31

3.1.2. Release of solubilizate from micelle upon core freezing ............................................. 35

3.1.3. The state of propofol in aqueous solution .................................................................. 41

3.2. BLOCK COPOLYMERS ADSORBED AT CARBON NANOTUBE SURFACES.............................42 3.2.1. The dispersion process of carbon nanotubes .............................................................. 42

3.2.2. Competitive adsorption of amphiphilic molecules at CNT surfaces ............................ 45

3.2.3. Polymer-CNT composite: structure and mobility ........................................................ 47

4. Concluding remarks ........................................................... 50

5. List of abbreviations ........................................................... 52

6. Acknowledgements ............................................................ 53

7. References ........................................................................... 55

1

1. Introduction

In this thesis, some studies regarding the structure, composition, and preparation of

some colloidal systems are presented. First, the basics regarding the systems

investigated and the methods used are introduced. Then, we summarize the most

important findings in the papers and manuscripts included.

1.1. Surfactants

The word surfactant arises from “surface-active agent” that illustrates surfactant

behavior: namely, that it can accumulate at an interface to reduce the free energy of

the boundary between any two immiscible phases. Generally, a surfactant molecule

contains two parts, one is soluble in a particular fluid, called the lyophilic part, and the

other insoluble in that fluid, called the lyophobic part. When that special fluid is water,

those two parts are called hydrophilic head and hydrophobic tail (Figure 1.1).

Figure 1.1 Surfactants structure and accumulate at interfaces

Usually the hydrophobic tail is formed by one or several alkyl chains, typically

involving around 8 to 18 methylene and methyl groups, but sometimes it can be more

complex with branched or unsaturated groups or even aromatic groups.1 The

hydrophilic head can be charged or neutral. Depending on this latter property,

surfactants can be divided into two broad classes, ionic (charged) surfactants and non-

ionic (uncharged) surfactants. Ionic surfactant can be further classified as anionic

2

surfactant, (that have anionic functional groups as head groups, such as carboxylate,

sulfonate and phosphate) or cationic surfactant (which may, for example, have amine

or ammonium groups as head group) and zwitterionic surfactant (where the head group

contains both negative and positive charges). For a non-ionic surfactant, the polar head

group often consists of polyhydroxyls or polyoxyethylene. In polyoxyethylene-based

surfactants each oxyethylene group in the head is denoted as E and, if the hydrophobic

part is an alkyl chain of m methylene/methyl units, then this surfactant can be referred

to as CmEn. For an ionic surfactant, the volume of hydrophilic head group is usually

much smaller than that of the hydrophobic tail. For polyoxyethylene-based surfactants,

the hydrophilic and the hydrophobic groups often have similar sizes and therefore it

can also be considered as a short AB block copolymer.

1.2. Amphiphilic block copolymers

Polymers are a large molecule formed with a great number of small molecules

covalently bound in a repeating structure whose properties are insensitive to adding

one or a few of those repeating units to it.2 If the polymer contains only one type of

repeating unit (A), it is called a homopolymer; if the polymer is formed by two or more

sorts of repeating units (A,B, etc.), it is a copolymer and can be classified according to

the arrangements of the constituting units, such as block copolymers, graft copolymers,

alternating copolymers, and statistical copolymers (Figure 1.2). If the different blocks

in block copolymers have different polarities, such as one of them hydrophilic and the

other one hydrophobic, then that copolymer behaves as an amphiphilic

macromolecule.

Amphiphilic block copolymers containing poly(ethylene oxide) (PEO) have been

extensively studied because the PEO segment is environmentally friendly, non-toxic,

biodegradable, has a good biocompatibility and a wide range of solubility.3-8 Among

those polymers, we focused on the Brij series and Pluronics. The Brij polymer is

composed of an alkyl chain of particular length and a PEO chain of particular length.

The variation of those hydrophilic and hydrophobic segment lengths makes their

properties, such as molecular weight, melting point, hydrophilic/lipophilic balance,

etc., different that permits different applications.9-13 Pluronic triblock copolymers are

3

also one of the most commonly used amphiphilic copolymers. They consist of PEO

blocks as hydrophilic segments and poly(propylene oxide) (PPO) blocks as the

hydrophobic part. The PEO and PPO blocks are structured as: PEOx-PPOy-PEOx

where the number of PEO and PPO blocks (also known as PEO/PPO ratio) can be

varied to change the molecular properties and, consequently, the phase behavior that

may permit a variety of applications.14-19

Figure 1.2 A homopolymer and different kinds of copolymers.

1.3. Micelles and solubilization in micelles

As described previously, the fundamental property of a surfactant is packing at the

interface to lower the interfacial tension. Actually, surfactant molecules can also be

associated with each other in bulk. This process of forming aggregates is called

surfactant self-assembly. The driving force is to prevent hydrophobic groups being in

contact with water, thereby reducing the free energy of the system.20 Different

aggregates with various structures could be formed and the self-organized structure

depends primarily on the volume and length of the hydrophobic chain and the area of

hydrophilic head group, summarized in the form of the critical packing parameter

(cpp)21, 22

4

𝑐𝑝𝑝 =𝑉ℎ𝑐

𝑎ℎ𝑔𝑙ℎ𝑐 (1.1)

where 𝑉ℎ𝑐 and 𝑙ℎ𝑐 are the volume and length of the fully extended hydrocarbon chain,

and 𝑎ℎ𝑔 is the hydrated hydrophilic head group area.

Figure 1.3 The different schematic structures of surfactant aggregates that may form in

water.

In general, if the volume of the hydrocarbon chain is equal to the volume of the

cylinder that is defined by area 𝑎ℎ𝑔 and length 𝑙ℎ𝑐, one obtains cpp=1, for which a non-

curved hydrophilic-hydrophobic interface is favored, and lamellar bilayer aggregates

are formed. If cpp>1, the shape of the molecule is similar with an inverted wedge,

bicontinuous phases or reverse micelles are favored. If cpp<1, an interface curved

toward the hydrophobic domain is favored, yielding bilayer vesicles (1/2<cpp<1),

5

cylindrical micelles (1/3<cpp<1/2) or spherical micelles (cpp<1/3). Those different

structures are illustrated in Figure 1.3.

In this thesis, the aggregates are mainly micelles. The process they are formed by is

called micellization and the concentration at which this process commences is called

the critical micelle concentration, abbreviated as cmc. Below the cmc, surfactants are

monomers in the bulk. Once the concentration reaches the cmc, they start to be

aggregated into micelles. The number of surfactants within a micelle is referred to as

the aggregation number. Normally, the cmc and aggregation number can be

determined by many techniques, such as surface tension, dynamic light scattering,

fluorescence spectroscopy, small angle X-ray scattering (SAXS), sedimentation

velocity and nuclear magnetic resonance spectroscopy (NMR).23-28 Each surfactant has

its own cmc. For example, in a PEO-based surfactant (such as the Brij family), a larger

polar head group leads to a moderate increase of cmc. The micelle is not a rigid

aggregate, but is dynamic in several ways: the molecules move within the micelles,

the chains have high flexibility, and individual molecules exchange between the free

surfactant state in bulk and the micelles.

Amphiphilic macromolecules behave somewhat similarly to simple surfactants, that is

they form micelles in solution, but their micellization is not completely the same. The

difference is related to the molecular weight of the polymer, and its hydrophobic and

hydrophilic moiety lengths, which give rise to a larger micellar core as well as a larger

overall micelle size. An additional important factor is the polydispersity of the

polymer, which also makes the aggregation more complex. Importantly, for

amphiphilic block copolymers there often is no sharp cmc even if the polymer has a

narrow molecular weight distribution.29, 30 Instead aggregation commences in a certain

range of concentration, and this concentration range may vary depending on the

different experimental techniques and conditions. Besides the cmc behavior, there is

also a strong temperature dependence in micellization for many amphiphilic block

copolymers, and the characteristic temperature is called the critical micelle

temperature. For example, for Pluronics, micelle formation is not only molecular

weight and PEO/PPO ratio modulated, but also temperature dependent.31-33 In the case

of Pluronic F127, the cmc decreases dramatically by increasing temperature, because

6

the PPO moieties are more hydrophobic at higher temperatures, which leads to easier

formation of micelles and lower aggregation numbers.16

Micelles have a hydrophobic core part, which can act as a good container for

hydrophobic molecules. In other words, the solubility of hydrophobic compounds in

aqueous solution can be increased dramatically in the presence of micelles. When the

surfactant concentration is lower than the cmc, the solubility of hydrophobic molecules

is still very low as it is in pure aqueous solution. Above the cmc, hydrophobic

molecules can be loaded in micelle core part until one reaches a solubilization

equilibrium. Here we need to distinguish this kind of solubilization (or loading) from

forming emulsions or microemulsions. First, emulsion is a suspension, which is

heterogeneous with particle size often larger than 1000 nm, and furthermore, those big

droplets are not in equilibrium and undergo coalescence until one obtains macroscopic

phase separation. On the contrary, micelles and microemulsions are part of equilibrium

phases and are homogeneous with sizes generally less than 100 nm. When a

hydrophobic compound is solubilized in micelles, the size of particles in the system

will not change significantly.34-37 Although micelles and microemulsions have similar

core-shell structures, the composition of the core part is different. As shown in Figure

1.4a, the micelle core is the hydrophobic region with the amphiphile tails mixed with

the solubilizate (Figure 1.4b), while the core part of a micromulsion consists of a small-

molecular hydrophobic liquid of oil, which is surrounded by the amphiphilic film

(Figure 1.4c). There exist also nanocapsules that are typically bigger than micelles,

and have polymeric films around the inner oil droplet.38 Solubilization is one of the

most important phenomena for micellar solutions, which can be used as detergents and

for formulation of drug carriers, where micelles can act as stimuli responsive particles

that can release molecules as defined by the different functional groups of the

constituting amphiphilic molecules.

7

Figure 1.4 Illustration of (a) a micelle, (b) a micelle with solubilizate loaded in it and (c) a

microemulsion/nanocapsule, where red represents the solubilizate or oil phase.

1.4. Carbon nanotubes and carbon nanotube dispersions

The introduction of CNTs can be traced back to the year 1991 when Iijima discovered

CNTs having a hollow cylindrical nanostructure with nm-order diameter and with

sheets of hexagonally arranged carbon as its wall.39 Due to their cylindrical structure

and small diameter, CNTs have a very high aspect ratio, that is the length/diameter

ratio in the order of 1000. In CNT walls, each carbon atom binds to three other carbons

through sp2 carbon-carbon bonds. This bond is even stronger than the sp3 bonds in

diamond, and therefore CNTs can have extreme mechanical strength. Moreover, the

p-electrons in the sheet form a π-system40, 41 and therefore the CNTs can be electrically

conductive.42

CNTs can be classified into two major types, single-walled CNTs (SWNTs) and multi-

walled CNTs (MWNTs). Single-walled CNTs are essentially single graphene layers

rolled up to a seamless cylinder while multi-walled CNTs are composed by two or

more concentric graphene cylinders. In reality, CNTs are not formed by rolling up

sheets of graphene but through other approaches.

In the picture presented above, rolling can proceed along different directions (called

the chiral angle 𝜃) within graphene layer. There are three familiar directions and

thereby resulting structures are illustrated in Fig. 1.5. for CNTs: zig-zag, armchair and

chiral. The different hypothetical rolling directions are defined by so-called chiral

8

vector, Ch=na1+ma2, shown in Figure 1.5, where the integers n and m indicate the steps

along the graphene lattice real space unit vectors (a1, a2). The structure of CNTs can

be assigned by those integers. When m=0, the chiral angle is 0o, and the CNTs are zig-

zag type; when m=n, the chiral angle is 30o, and the CNTs are armchair type. In all

other cases, the CNTs are chiral. The electronic properties of CNTs depend on the

chiral angle: if the number (n-m) is a multiple of three, the CNTs are conductive,

otherwise semiconductive. MWNTs consist of two or more rolled-up graphene sheets,

each of which can has its own chirality, and therefore the properties of MWNTs are

more complex.

Figure 1.5. Classification of the wall structure of CNTs by their chirality.

CNTs have some unusual properties, such as outstanding mechanical and electrical

performance, and can therefore be applied in a lot of areas. Moreover, CNTs have

interesting magnetic and optical properties as well, which also have been explored in

many studies.42-48 However, to exploit the full potential of CNTs in various

applications, it is often required that the CNTs exist individually, that is separated from

each other. In other words, they should not be present in form of aggregates (such as

bundles, mats and ropes), but well dispersed instead.

9

Many methods were proposed and tested to achieve this. One possible approach is to

dissolve CNTs in some suitable organic solvents. Since the surface of CNTs is

hydrophobic, using an organic hydrophobic solvent would be the first possible

approach. Unfortunately, only a few solvents can dissolve/disperse a limited amount

of CNTs, such as chloroform, N,N-dimethylformamide (DMF), o-dichlorobenzene

(ODCB) and, the best among them, N-methyl-2-pyrrolidone (NMP).45, 48-50 Another

drawback for this method is that those organic solvents are neither environmentally

benign nor biocompatible, which would limit their applications.

To avoid this disadvantage and disperse more CNTs in water, another common method

is to chemically modify the surface of CNTs with some functional groups, which have:

1) favorable interactions with the water and 2) enhance the repulsions between CNTs

to avoid re-aggregation and keep them in individually dispersed state. Although a

considerable amount of CNTs can be dispersed on this manner in aqueous solution,

but one drawback of this method is that this kind of modification of the wall can change

the nature of the carbon-carbon bonds (functionalization should disrupt the π-system)

and thereby alter any of those advantageous properties of CNTs. Moreover, some

structural defects could also be introduced.41, 51, 52

An alternative and for some applications clearly better way to disperse CNTs is relying

on not covalently attached but merely adsorbed groups. In other words, one would use

a dispersant that spontaneously attaches itself to the outer surface of CNTs and provide

the two functions specified above. The dispersant needs to have affinity to water,

meanwhile, it needs to have some hydrophobic moieties, too, that may attach to the

CNTs without changing any chemical bonds. Both hydrophilic and hydrophobic

properties are required in the same molecule, which calls for an amphiphilic one.

Surfactants and amphiphilic polymers are prime candidates to act as the CNT

dispersants in water.

As described above, amphiphilic molecules can associate both in the bulk and at an

interface. Although both of those arrangements may permit hydrophobic moieties

avoiding contact with water, adsorption at a solid hydrophobic surface is different from

the self-association in solution. Typically, in solution, amphiphiles first adsorb at the

air-liquid interface until their concentration reaches the cmc, and then they start to self-

10

associate to form micelles in the bulk. If CNTs appear in solution, there is a stronger

attraction between the hydrophobic segments of amphiphiles and the surface of CNTs,

which induces the adsorption of amphiphilic molecules onto the CNTs surface.40, 41

Different from the micellization process, which requires a minimum critical

concentration, the adsorption of dispersants on CNTs can happen at a lower

concentration than cmc.53, 54 Normally, the initial adsorption of amphiphiles on CNTs

occurs in a random way, and at higher concentrations, aggregates, like hemimicelles

or cylindrical micelle, may form on the surface.54-57

To disperse pristine CNTs (always received as powder) in water, ultrasonication is also

needed in the amphiphilic solutions discussed above. During ultrasonication, high

shear forces are generated that, interacting with CNT bundles, may open up clefts in

the bundle. The newly exposed CNT surfaces in the cleft are hydrophobic and, if

nothing happens to them, they may quickly re-aggregate. However, if the dispersants

adsorb on those newly created CNT surfaces, re-aggregation between CNTs and re-

closing the cleft may be prevented.58, 59 A propagating cleft leads to complete

exfoliation. This process is known as the unzipping mechanism and is illustrated in

Figure 1.6.

Figure 1.6 The unzipping mechanism of exfoliation: (a) a bundle of CNTs; (b) a cleft at the

bundle end is opened by high local shear forces; (c) surfactant molecules adsorb on the new

surfaces created; (d) the cleft propagates and a CNT coated by surfactant is released and

remains dispersed in solution.

11

The main driving force for adsorption of a surfactant on CNTs is the hydrophobic

interaction. Besides that, π-π interactions can also play an important role in

adsorption.52, 60, 61 As mentioned above, electrons in the CNT wall can form a π system,

which is similar to that in aromatic molecules. Thus, a strong and specific interaction

can arise between CNTs and molecules containing aromatic moieties.

Similarly to surfactants, polymers can also adsorb on CNTs, with their hydrophobic

segments attached to the wall and hydrophilic segments interacting with water.59, 62

But there is some difference between surfactants and polymers. Polymers have high

molecular weight and different possible arrangement of the repeating units, which

likely will affect the interactions with nanotubes. Thus, polymers may not only loosely

adsorb on nanotube surface but can in some cases also wrap the tube by coiling around

it.63-65 The polymers in the wrapping mode of adsorption are clearly less dynamic

compared to those with loose adsorption. Hence, polymers with loose adsorption mode

can exchange between being attached on the CNTs surface and being free in bulk

solution, and achieve an equilibrium with a dynamic balance between those states.

Besides the structure of dispersants, their concentration is important as well for

dispersing CNTs. Initially, researchers prepared CNTs dispersions using surfactant

with concentrations higher than the cmc, but later it was found that it does not matter

whether the concentration is above cmc or not.66 In other words, the formation of

micelles is not a requirement for dispersibility of CNTs. Typically, the concentration

of dispersed CNTs increases as a function of dispersant concentration until a plateau

is obtained. When the dispersant concentration increased to an even higher value, the

dispersed CNT concentration may decrease.

Due to their unique mechanical,67-73 electrical74-78 and thermal73 properties, CNTs are

excellent fillers for polymers when creating composite material systems. Dispersing

CNTs in polymer melts is separate chapter regarding dispersibility. One connection to

aqueous dispersibility is that it might be advantageous, in order to achieve thin bundles

or individual CNTs, to pre-disperse CNTs in aqueous phase first, dry them, then

12

disperse this material instead of pristine CNTs in polymers melts or precursors. This

was the route chosen for the study described in Paper VII in this thesis.

13

2. Experimental

2.1. Sample preparation

To make experiments reproducible, the first step is to make the samples reproducible.

In this subsection, we focus on micelles containing solubilizates and CNT dispersions.

A number of features caught our attention and lead to a lot of work and delays even if

they did not make into the final publication - such as the choice of a suitable

solubilizate to be loaded in micelles or the sonication strength and time to disperse

CNTs. Some of those issues are going to be discussed in this section.

2.1.1. Micelles with and without solubilizates

The formation of micelles is simple and commences upon mixing and dissolution of

the surfactant/amphiphilic polymer in water at a concentration higher than its cmc. In

our research, neat Brij S20 (C18E20) or neat Brij S100 (C18E100) were dissolved in

heavy water D2O at a concentration of 1 wt%. A mixed micellar system of Brij S20

and Brij S100 was obtained by mixing the 1 wt% solution of the pure surfactants at a

ratio of 1:1. Among those three micellar solutions, only the one formed by Brij S20

(that exhibited the sharpest core freezing behavior) was used to solubilize small

hydrophobic (and a few amphiphilic) molecules.

There are two common methods employed to prepare micelles containing solubilizates

in them.35-37 The first one is direct dissolution where the solubilizate is added to a

micellar solution and partitions rapidly and spontaneously into the core part of

micelles. The second and more involved method is film casting where the

surfactant/amphiphilic polymer and solubilizate are both dissolved in an organic

solvent, which solvent is then evaporated leaving a film in the container wall. Water

is then added to dissolve the mixed film of surfactant and solubilizate. In our

experiments, the first and simpler method was sufficient and was thereby chosen to

prepare the samples.

A more complex point was to choose a good solubilizate for this model study (that is,

on contrast to applications where the solubilizate is given). Since the main

14

characterization method is 1H NMR, one important requirement was that the

solubilizate exhibited an easily distinguishable peak in the NMR spectrum with no

overlap with the micellar signal. Besides that, several physical properties had to be

considered. First, the density of solubilizate could not be too close to that of heavy

water, because in that case an excess oil phase did not phase separate sufficiently.

Secondly, the solubilizate had to be liquid in the experimental temperature range of 0-

20 °C to avoid signal loss by freezing. Moreover, the solubilizate could not be volatile

in the same temperature range that, in practice, meant that the boiling point had to be

far higher than 20 °C. Finally and importantly, the solubilizate was not permitted to

completely suppress core freezing since it was exactly the influence of that

phenomenon on solubilizate content that was the subject of our study.

A number of liquids were tested (such as toluene or hexamethylsilane) and among

those it was hexamethyldisiloxane (HMDSO) that fulfilled best all requirements and

was thereby selected. It presents a single 1H NMR peak at approx. 0 ppm chemical

shift and with no overlap with the micellar signals at approx. 1 ppm or above, a density

of 0.764g/mL that is far below that of heavy water, a boiling point at 100 °C and a

melting point at -59°C. While core freezing was shifted by it, it was not completely

suppressed.

A series of samples was prepared by adding different amount (from 0.5 µL to 80 µL)

of HMDSO to 1.5 mL micellar solution (Brij S20 1 wt%). After vortex mixing for 5

min, the sample was subsequently centrifuged for 30 min at 6500 rpm. This procedure

yielded a two-phase sample, where the upper part was the excess amount of HMDSO.

The transparent lower part, around 1 mL, was taken out and moved to a 5 mm NMR

tube yielding >7 cm sample column. Subsequently, a second centrifugation step was

applied to the sample in the NMR tube for 30 min at 1250 rpm in order to make sure

that any non-solubilized HMDSO that remained was moved to the sample top and

thereby out of the NMR detection range. All the samples were prepared at room

temperature and used in following NMR experiments.

15

2.1.2. Propofol with different concentrations in D2O

Propofol stock solution (0.89 mM) was received from a collaborating group. This stock

solution was then diluted with heavy water to obtain a series of propofol solutions with

different concentrations (from 0.02 mM to 0.89 mM). Propofol was also one of the

liquids we tested as solubilizate in Brij S20, but at high loading it has completely

suppressed core freezing.

2.1.3. CNT dispersions

In general, CNT dispersions were prepared by ultrasonicating the CNT powders mixed

into a dispersant solution. The role of the dispersant has already been discussed above

(see Fig. 1.6). Subsequent centrifugation removed the non-exfoliated CNT particles

and other impurities from the dispersions whose CNT content was evaluated with a

combination of thermogravimetric analysis (TGA) and Ultraviolet-visible (UV-vis)

spectroscopy.62, 79

During sonication, the cavitation bubbles can be created and also collapsed. The rapid

collapse of those bubbles generates very fast local solvent streams and pressure

gradient that exerts high shear forces on the CNTs bundles. Those forces can overcome

the van der Waals forces that keep the CNTs together and create temporary clefts.80, 81

The number and size of cavitation bubbles are dependent on sonication frequency and

power. Lower frequency ultrasound tends to create larger cavitation bubbles and the

larger the cavitation bubbles, the greater the forces generated upon collapse. Normally,

tip sonication is operated at a lower frequency than bath sonication, which is one

reason we chose tip sonication for exfoliation. In addition, tip sonication can also

deliver higher acoustic power that yields more bubbles. If high power is used for long

enough time, one can create dispersion with significant CNT content.82 Unfortunately,

sonication also leads to breaking the dispersed CNTs,83-86 and therefore the used time

and power is typically (in our case, to yield a total delivered acoustic energy density

of 1100 J/mL) a compromise between sufficiently high content of sufficiently intact

CNTs. In addition to CNT exfoliation and breakage, the other effect of ultrasonication

is to increase the temperature of the processed liquid mixture. This was shown to lead

to inferior and scattered results and therefore the sonication vial was immersed into a

16

home-made cooling bath that kept the temperature low enough during the whole

process. Other details, such as an exact and reproducible volume of the sonicated liquid,

an exact and reproducible position of the sonication tip relative to the vial were

important to ensure the reproducibility of the samples. Sonication also strongly

depended on the shape of the vials and therefore the exactly the same type of vials has

to be used for a given study.

After the sonication step, one has a suspension-dispersion containing individual CNTs,

thin CNT bundles and non-exfoliated CNT particles (and, if the starting pristine CNT

is not pure, impurities). These objects exhibit different sedimentation velocities and

therefore centrifugation is a good method to separate them. Hence, a mild

centrifugation (typically, 30 min at 4000 g) was sufficient to remove non-exfoliated

particles and thereby obtain as supernatant the desired dispersions of individual CNTs

and thin CNT bundles.83, 87-89

When the CNTs dispersions are done, their CNTs concentration has to be quantified.

The protocol for this was described in detail elsewhere.62, 79 Since CNTs absorb visible

light with a high absorption at the wavelength λ=660 nm and their apparent absorption

follows the Lambert-Beer law, their obtained absorbance A is proportional to their

concentration c:

𝐴 = 휀𝑐𝑙 (2.1)

where 휀 is the extinction coefficient, and 𝑙 is the length of light path.

With A and l measured/known, the last required parameter to obtain the CNT

concentration is the extinction coefficient 휀. This quantity is not a priori known and

has to be calibrated against CNT mass in a dispersion. The calibration method was the

following one.62 A certain volume 𝑉𝑠 of the CNTs dispersion was first freeze dried, the

mass 𝑀𝑠 was determined by weighing and then TGA was applied to get the mass

fractions of CNT and dispersant. Finally, the concentration of CNT can be calculated

as:

17

𝑐𝐶𝑁𝑇 = [𝑀𝑠 × (1 −𝜙𝑠

𝜙𝑑)] ×

1

𝑉𝑠 (2.2)

where 𝜙𝑠 is the mass fraction of dispersant in supernatant and 𝜙𝑑 is the dispersant

decomposition fraction (which was obtained in a separate TGA experiment). The

advantage of this combined procedure is that the CNT concentration can be obtained

by rapid spectroscopic experiments on readily accessible UV-vis spectrometers.

2.1.4. CNT-loaded epoxy

As described above, CNTs need to be pre-dispersed and then mixed with epoxy resin

to form a composite. The CNT pre-dispersions were prepared by sonication with

Pluronic F127 as dispersant. The process was as above, with multi-walled CNTs added

at 2 mg/mL to an aqueous F127 solution (1.5 mg/mL), followed by tip sonication with

approx. 310 J/mL delivered acoustic energy density. During this process, the

temperature was kept at 0 oC. After sonication, the dispersion was centrifuged for 20

min at 1000 g. After centrifugation, the supernatant was collected and freeze-dried to

remove the water. The final yield of cotton-like CNT-F127 powder was subsequently

mixed with epoxy resin by a planetary mixer for 10 min to obtain the composite.

2.2. Characterization by NMR spectroscopy

NMR spectroscopy was used in this thesis as the main tool to get the information about

the structure, dynamics and association of molecules. Well-established NMR

relaxation and diffusion methods were used for analyzing the state of

surfactants/amphiphilic polymers in aqueous solutions or at the surface of CNTs.

Besides NMR, the structural method of SAXS also played an important role in our

work.

18

2.2.1. Principles of NMR

NMR spectroscopy is based on the nuclear property of nuclear spin which makes that

nuclei can interact with an external magnetic field. Some nuclei lack this property and

they cannot thereby be used for NMR spectroscopy.90

The spin is defined as the intrinsic angular momentum and magnetic moment that exist

in all the elementary particles such as protons, neutrons and electrons. It is

characterized by spin quantum number I that is I=1/2 for protons, neutrons and

electrons. Nuclei with odd mass number have a spin I that is odd multiples of 1/2 (1/2,

3/2, 5/2…), while nuclei with even mass number but with odd numbers of protons and

neutrons have integer spin I (1, 3, 4, 5, 7, by chance there is no I=2 nucleus). Nuclei

with even mass number and even atomic number such as 12C and 16O have no spin.

A nuclear spin I can have 2I+1 possible states, that are indexed by the quantum number

m, m=−I, −I+1, −I+2, …, I. Hence, the component of angular momentum 𝜤 along any

spatial direction z can be quantized as:

𝛪𝑧 = 𝑚ℏ (2.3)

where ℏ is the Planck’s constant divided by 2𝜋. The simplest case is spin I=1/2, such

as for the 1H, 13C, 31P nuclei, where there are two possible states with m=1/2 and

m=−1/2.

The magnetic moment 𝝁 of a nucleus is connected to its angular momentum 𝜤 by

simple proportionality:

𝝁 = γ𝜤 (2.4)

where γ is the gyromagnetic ratio, specific for each nucleus. It is a scalar that can be

either positive or negative and therefore the nuclear magnetic moment can be parallel

or antiparallel, respectively, to its associated angular momentum.

In the absence of an external magnetic field, all the possible 2I+1 possible nuclear spin

states have the same energy. In a magnetic field B0 whose direction is assigned as z,

the nuclear magnetic moment interacts with the field as:

19

Em=−𝜇𝑧𝐵0 (2.5)

and the energy levels split as given by:

𝜇𝑧 = 𝑚ℏ𝛾 (2.6)

For the simplest case of spin I=1/2, there are two energy levels associated with m=1/2

and m=−1/2, with the energy difference between those two states given as:

Δ𝐸 = ℏ𝛾𝐵0 (2.7)

The two energy levels are populated by spins aligned either along or against the

external magnetic field. In thermal equilibrium, the Boltzmann distribution prevails

yielding:

𝑁ℎ𝑖𝑔ℎ

𝑁𝑙𝑜𝑤= 𝑒−Δ𝐸 𝑘𝐵𝑇⁄ (2.8)

where 𝑘𝐵 is the Boltzmann constant and T is the absolute temperature. while the

number of nuclei at the two energy levels are denoted as Nlow (at the lower energy

level) and Nhigh (at the higher energy level).

Without an external magnetic field, the nuclear spins have no preferred orientations,

and therefore the total nuclear magnetization, the vector sum of the individual nuclear

magnetic moments, is zero. In an external magnetic field, the number of spins aligned

along and against the magnetic field are different and a net nuclear magnetization

arises parallel to the direction z of the applied magnetic field. This net magnetization

is proportional to the population difference between different spin states, Δ𝑁 =

𝑁𝑙𝑜𝑤 − 𝑁ℎ𝑖𝑔ℎ . This population difference and thereby the equilibrium nuclear

magnetization is very small,91 Δ𝑁/𝑁 being in the order 10-4-10-8, depending on the

field strength and the gyromagnetic ratio. As one consequence, a large amount of

sample is required for NMR spectroscopy, performed with concentrations typically at

or above the mM range.

In NMR experiments, radiofrequency (rf) pulses, delivered by a coil arranged around

the sample, are applied to excite this net magnetization out of its equilibrium state.

20

This rf field can be seen as a time-dependent magnetic field that is aligned orthogonally

to 𝐵0. To be able to excite nuclear spin transitions, its frequency has to match the

energy difference between two different spin states, that is ℎ𝜈 = Δ𝐸 = ℏ𝛾𝐵0. Hence

arises the resonance condition:

𝜈0 = 𝛾𝐵0/2𝜋 (2.9)

This resonant frequency is called the Larmor frequency, sometimes also given in terms

of angular frequency 𝜔0 = 𝛾𝐵0.

In the simplest NMR experiment, shown in Figure 2.1, there is one rf pulse applied.

This pulse has a specific length and rf field strength so that its effect is to turn the

nuclear magnetization from its equilibrium state along the z axis to the perpendicular

xy-plane. Once in that plane, the magnetization precesses around the direction of 𝐵0

with its Larmor frequency 𝜔0. This motion changes the magnetic flux in the detection

coil (typically the same one as that for excitation) surrounding the sample and the

changing flux induces a time-dependent voltage in the coil. This is called the free

induction decay (FID), which is the primary time-dependent NMR signal. After being

Fourier transformed, it yields the NMR spectrum as a function of frequency.

Figure 2.1 A 90° rf pulse and the time-domain FID signal excited by it.

21

2.2.2. Chemical shift

The Larmor frequency as defined above is set by the external magnetic field. Yet, this

frequency is not exactly the same for the same nuclei either in different molecules or

at different positions in the same molecule. Namely, nuclei in atoms and molecules are

surrounded by electrons and one effect of 𝐵0 is to perturb slightly the electronic states.

This perturbation can be represented as if a small additional magnetic field 𝐵′ was

added to an external magnetic field 𝐵0, typically in the opposite direction. Hence, the

nuclei at different sites reside in a magnetic field that varies from site to site on a

manner that depends on the local electronic environment. This phenomenon is called

shielding and is represented as:

𝐵 = 𝐵0 − 𝐵′ = 𝐵0(1 − 𝜎) (2.10)

where 𝜎 is the shielding constant. Hence, the resonant frequency of a nucleus at a

particular site becomes:

𝜈 = 𝛾𝐵0(1 − 𝜎)/2𝜋 (2.11)

Instead of detecting the shielding constant, a more convenient way to represent it is by

the chemical shift 𝛿, that is defined as the difference in resonant frequencies between

a nucleus at a site of interest and a nucleus in a reference compound:

𝛿 = 106𝜈 − 𝜈𝑟𝑒𝑓

𝜈𝑟𝑒𝑓 (2.12)

typically expressed in units of ppm. Defined on this manner, 𝛿 is independent on the

external magnetic field. The chemical shift depends not only on the intramolecular but

also, to a lesser extent, on the intermolecular surrounding.27, 92-94

Because of the chemical shift, the NMR spectra typically contain many peaks. Separate

peaks with their respective individual chemical shifts are detected if there is no

exchange of the atoms between those sites or the exchange is slow relative to the

inverse of the frequency difference between the peaks belonging to those sites. If the

exchange is faster than that, one can only detect a single peak at the frequency given

by the average chemical shift.

22

Besides the chemical shift, there are additional spin interactions that may influence

NMR spectra. In liquids, the other relevant interaction is the spin-spin or J-coupling.

The effect of that is the various signals are split, depending of the arrangement of other

nuclei around a site in a molecule. This interaction also depends on the electronic

structure but the role of the electrons is to transfer, through their delocalization over

covalent bonds, the magnetic effect of a nuclear spin in a given atom to the spin

residing in another atom. There are additional and in magnitude much larger spin

interactions such as the dipole-dipole coupling. However, in a liquid their spectral

effect is averaged out by fast isotropic molecular motions and they only influence spin

relaxation, discussed in the next sub-section.

2.2.3. Spin relaxation

In thermal equilibrium state, the net nuclear magnetization is oriented along the z-axis

and there is no magnetization in the xy-plane. After a 90° rf pulse as in Figure 2.1, there

is net magnetization in the xy-plane, but the magnetization along the z-axis is zero. In

other words, rf excitation has put the sample spins into a non-equilibrium state. Hence,

like in any other non-equilibrium state, the system relaxes back to the equilibrium one.

For nuclear spins, this process is called spin relaxation that, for the simplest case of

spin I=1/2, includes two independent processes: the longitudinal relaxation and the

transverse relaxation, characterized by their respective relaxation time constants T1 and

T2. Longitudinal relaxation, also called spin-lattice relaxation, characterizes the return

of the longitudinal (that is, along the z-axis) magnetization to its equilibrium value.

Transverse relaxation, also called spin-spin relaxation, refers to the decay of transverse

magnetization to zero.

The driving force of spin relaxation is the fluctuation of the local field (or, more

generally, the fluctuations of the various nuclear spin couplings such as the dipolar

coupling, electric quadrupole, chemical shift anisotropy95) that arise from molecular

motions. The quantity that is often used to characterize the temporal properties of those

motions is the motional correlation time 𝜏𝑐. In itself, using a single parameter is a

simplification; in reality, the proper tool to analyze molecular motions is their auto-

and cross-correlation functions. Yet, in simple liquids the decay of those correlation

23

functions is well represented by a single 𝜏𝑐 which is then used to indicate the time

needed for a molecule or moiety to significantly change its orientation with respect to

the external magnetic field. When presenting a qualitative model of spin relaxation,

one can disregard the actual spin coupling that is modulated and can instead assume

that the correlation time represents the average time that is needed to experience a

significant change in a randomly fluctuating magnetic field.

Within the random local field model, the fluctuating field has components in all

different directions x, y and z. Of these components, it is only fluctuations in the x and

y directions that cause transitions among the involved energy levels, for the same

reason that it is only an rf field in the xy-plane can excite the nuclear spin system.

Because the longitudinal relaxation influences the total energy of the spin system, it

requires such transitions between energy levels. On the other hand, fluctuations in the

z-direction cause no transitions, but only change the precession speed of the spins

relative to each other which leads to a spread of spins in the xy-plane and the loss of

their sum, the total precessing nuclear magnetization. For this reason, both longitudinal

and transverse relaxation are influenced by the field fluctuations in the xy-direction,

while fluctuations in the z-direction only affect the latter process. Besides the direction,

the time scale of the fluctuations has also an effect. Namely, to be able to cause

transitions, the fluctuating field must possess temporal variation that is significant in

the range of the Larmor frequency; in that case, it can satisfy the resonance condition.

For that reason, the relation 𝜏𝑐 ~1/𝜔 must be satisfied. On the other hand, fluctuations

in the z-direction on any time scale can contribute to the spreading of magnetization

on the xy-plane and thereby to transverse relaxation. In summary, only fast fluctuations

have an effect on T1 while both fast and slow fluctuations can affect T2.95 Very fast

𝜏𝑐 ≪ 1/𝜔 fluctuations are ineffective to cause either longitudinal or transverse

relaxation.

To illustrate the consequences, we consider a typical experimental situation when one

performs temperature dependent spin relaxation experiment. Temperature changes the

molecular motions (often through an Arrhenius-type dependence). This leads to the

variation illustrated in Figure 2.2. At high temperatures with short correlation times,

both relaxation times are long and of approximately the same value because the

24

molecular motions are not effective; this regime is called extreme narrowing. Upon

decreasing temperature and increasing correlation times, both relaxation times

decrease. Yet, upon further change, the longitudinal relaxation time experiences a

minimum that signifies that the molecular motions reached the range where 𝜏𝑐 ~1/𝜔.

Upon further decrease of temperature, T1 starts to increase as the molecules motions

become less effective to cause transitions among the spin energy levels. Yet, T2

continues to decrease since it can be affected also by slow fluctuations. In the slow

motion regime of 𝜏𝑐 ≫ 1/𝜔, often experienced in large molecules or in the solid state,

a typical experimental situation is T1 ≫ T2.95, 96

Figure 2.2 The schematic dependence of the relaxation times T1 and T2 on the motional

correlation time 𝜏𝑐.

The transverse relaxation time T2 determines the decay of the precessing nuclear

magnetization and thereby the decay of the FID signal, too. Through the Fourier

transform between the FID and the spectrum, it defines the width of the spectral peaks;

the shorter the T2, the broader the peaks. However, the line width in the spectra is also

25

contributed by the static inhomogeneity of the magnetic field. The spin-echo pulse

sequence,97 shown in Figure 2.3, can be used to separate those two effects.

Figure 2.3 The spin echo pulse sequence.

In this pulse sequence, a 90° rf pulse turns the magnetization to the xy-plane where the

magnetization starts to precess. If the magnetic field is inhomogeneous, the nuclear

magnetization from the different regions precess with different speed, with some

magnetization components going ahead and some lagging behind. Individually, each

component is also affected by transverse relaxation. As a consequence of the

inhomogeneity, the magnetization fans out in the xy-plane. After a duration of τ, a 180°

rf pulse is applied that inverts the orientation of the magnetization components in the

xy-plane so that the relative order of those components changes: the slowly-precessing

components are placed ahead and fast ones will be put at the end. Yet, they retain their

relative precessing speed which means that the fast magnetization components

gradually catch up with the slow ones. In other words, the magnetization that was

originally fanning out during the first half of the pulse sequence is now gathered and

refocused. This apparent re-appearance of decaying transverse magnetization is called

the spin echo. If one detects the FID signal starting at the time τ, the resulting spectral

peaks are still broadened by the magnetic field inhomogeneity but if one records their

decay upon increasing τ, one obtains their true T2 relaxation time. Similarly, one has

to use a separate pulse sequence to measure T1. It should note that the separation

between fluctuating and static random fields is not clear-cut. Roughly speaking, the

spin echo experiment refocused that effect of those fields that fluctuate on the time

scale is slow as compared to τ.

26

2.2.4. NMR diffusion experiments

Self-diffusion, the Brownian motion without any applied potential gradient, is the

random translational motion of molecules. It is the most fundamental form of

molecular transport, taking part in all chemical phenomena. In an isotropic

homogeneous system, the effect of self-diffusion is typically formulated in terms of

the mean-square displacement by the Einstein relation:

⟨(𝒓𝑡 − 𝒓0)2⟩ = 6𝐷𝑡 (2.13)

where r0 is the initial position for a molecule and rt is its position after time t. In this

relation, D is the self-diffusion coefficient,92, 98, 99 expressed in units of m2s-1. In a

small-molecular liquid at room temperature, D is typically in the range of 10-9 – 10-12

m2s-1,92, 99 and is related to the molecular size as given by Stokes-Einstein equation:

𝐷 =𝑘𝐵𝑇

6𝜋𝜂𝑅ℎ (2.14)

where 𝑘𝐵 is the Boltzmann constant, T is the absolute temperature, 𝜂 is the viscosity

of the solution at the temperature T, and 𝑅ℎ is the hydrodynamic radius of the molecule

or particle. In this equation, the viscosity is another molecular property influenced

heavily by intermolecular interactions.

For a given molecule, adsorption (for example, to other molecules of larger size or to

surfaces) or self-association into an aggregate may change the detected diffusion

coefficient. Namely, under those conditions and assuming that the association is

persistent on the time scale of the experiment performed, the measured diffusion

coefficient is going to be that of the total associated system that, under the conditions

specified, is lower than that for molecules without association. Hence, one can for

example detect both micelle formation,92, 99 and the attachment of molecules to a CNT

surface.100, 101 The diffusion coefficient is also influenced if the molecules exchange

among the environments of different self-diffusion coefficients. In that case, one may

obtain a population average of that property.

NMR spectroscopy provides a non-invasive method to determine the self-diffusion

coefficient of native molecules. In other words, NMR-based diffusion measurements

27

do not require the samples to be chemically or physically disturbed, or labeled probe

molecules, or macroscopic potential gradients to initiate net transport. The reason for

that is that in NMR diffusion experiments the diffusing atoms/molecules are marked

with regard to their position by applying, during the course of suitable NMR

experiments, spatially dependent magnetic fields, also called magnetic field gradients.

In this manner, the Larmor frequency becomes spatially dependent. Since the nuclear

spin transitions involve very small energy differences, this has no influence on any

other relevant degrees of freedom.

Specifically, if the applied magnetic field gradient G is spatially independent (called

“linear gradient”), the resonant frequency at position r becomes:

𝜔𝑟 = 𝛾𝐵0 + 𝛾𝑮 ∙ 𝒓 (2.15)

The pulsed-field-gradient spin-echo (PGSE)102 experiment, based on the modification

of the spin-echo experiment, takes advantage of such labeling of the spins to determine

the self-diffusion coefficient.

In the PGSE pulse sequence shown Figure 2.4a, two identical gradient pulses of

magnitude g and duration 𝛿 are added to the spin echo experiment, one before and one

after the refocusing pulse. The two gradient pulses are separated by a period that is

called the diffusion time and is denoted by Δ. If the spin-bearing entities were

immobile and kept their position over Δ, the experiment would work just like the spin

echo experiment: the transverse magnetization spread out by the field inhomogeneity

caused by the first gradient pulse would be completely refocused. The condition for

such complete refocusing is that the resonance frequencies of the spins are identical at

the time of the two gradient pulses. If the spin-bearing entities randomly move, the

resonance frequencies before and after the refocusing pulses are not the same but differ

by a random factor. Hence, the magnetization components from the different nuclei do

not align with each other in the xy-plane but will be spread out in that plane depending

on the random factor to mark the change in their position by diffusion. Since the NMR

signal is proportional to the total transverse magnetization, if the magnetization

components are spread out, their vector sum decreases and so does the NMR signal.

28

Figure 2.4 The pulsed-field-gradient spin-echo (a) and stimulated-echo (b) experiments.

Detailed analysis shows that the decrease of the NMR signal acquired in the PGSE

experiment is related to the diffusion coefficient (because that governs the random

displacement of spin-bearing entities) as given by the so-called Stejskal-Tanner

equation102

𝑆

𝑆0= 𝑒𝑥𝑝 {−𝛾2𝛿2g2𝐷 (𝛥 −

𝛿

3)} (2.16)

where 𝑆 and 𝑆0 denote the signal intensities obtained with and without the gradient

pulses, respectively. The signal attenuation can be monitored upon increasing the

gradient, and the signal decay yields D. The huge advantage with PGSE NMR

diffusion experiments is that the detected spectrum is of high resolution with resolved

peaks for different chemical entities. Hence, by following the diffusional decay of a

particular peak in the spectrum, the diffusion coefficient of the corresponding

29

particular molecule can be selectively obtained, even in complex mixtures. This has

been exploited throughout the work in this thesis.

In practice, the different experimental parameters, such as the diffusion time Δ, the

gradient strength g and length 𝛿 have to be set so that the signal is detectable and the

decay is well-sampled. For a small molecule with a large diffusion coefficient, a small

gradient might be suitable to preserve the NMR signal; for a large molecule with a

small diffusion coefficient, larger and longer gradients and a long diffusion time must

be used to make sure that there is significant signal attenuation. However, for large

molecules or aggregates the transverse relaxation time can, as discussed above, be very

short and thereby the one cannot set suitably long diffusion time with gradient pulses

limited in strength because the signal would then be lost by transverse relaxation. To

alleviate this problem, the stimulated-echo pulse sequence (Figure 2.4b) has been

introduced. Here, the magnetization is stored not in the xy-plane but along the z axis

during the diffusion time. Since z-magnetization is influenced by longitudinal

relaxation and because, for large molecules and aggregates, the longitudinal relaxation

time is often much longer than the transverse relaxation time, much more signal is left

for a given diffusion time. Most of the diffusion experiments in this thesis work were

performed with the stimulated-echo pulse sequence or with one of its varieties.

2.3. Additional characterization techniques

2.3.1. Small-angle X-ray scattering (SAXS)

SAXS is a powerful method in colloidal science to investigate the size, shape and

internal structure of particles in the size range from nanometers up to several hundred

nanometers.103 Normally, it can be applied to systems consisting of randomly oriented

or/and statistically distributed structures of colloidal dimensions, such as proteins,104

polymer particles105 and molecular aggregates in solutions106 (e.g., micelles). During

a SAXS experiment, an X-ray beam irradiates the sample and electrons in the atoms

coherently scatter the incident radiation in directions away from the directions of the

incident beam. The intensity of scattered X-rays can be measured as a function of the

angle away from the direction of the incident radiation. In principle, the solvent in the

system provides a background scattering which is almost constant at small angles

30

where the scattering from larger structures is significant; since the particles of interest

in the system have different composition and thereby different electron density (that

is, as compared to the surrounding solvent) the scattering relative to the solvent

background is containing the sought structural information.

SAXS is based on the coherent scattering of X-rays. Such scattering events result in

scattered X-rays that are phase coherent with the incident radiation. Hence,

interference phenomena can arise; in crystalline lattices, this result in diffraction. For

randomly arranged objects, the result is that the scattered intensity becomes the

orientationally averaged Fourier transform of the autocorrelation of the electron

density. The latter quantity depends on the geometry (and, in sufficiently dense

systems, interparticle arrangement yielding the so-called structure factor) of the

scattering particles (often summarized as the so-called form factor).

To extract the structure from the angular dependence of the scattered radiation

intensity, the data can be analyzed by either model-independent approaches or by

direct modeling. The former procedure necessarily involves inverse Fourier

transformation.107 In direct modeling, a geometrical model of the objects is assumed

and the scattering intensity is calculated first, then the parameters of the model can be

optimized by, for example, least-squares fitting of the calculated function to the

experimental data.108 Nowadays, various models are available to provide information

on aggregate shape and aggregation number. In this thesis, the SAXS analysis relied

on a model based on a spherical core-shell structures.109, 110

31

3. Summary

In this thesis, the behavior of amphiphilic molecules associated in aqueous bulk

solution and to nanotube surfaces was detected and analyzed by NMR (chemical shift,

spin relaxation and self-diffusion coefficient) experiments, in some cases in

combination with SAXS experiments. In bulk solution, we focus on the temperature-

dependent properties of some micellar systems (formed by Brij surfactants) and the

solubilizate release from such micelles upon decreasing temperature (Papers I and II).

Whether or not a small amphiphilic molecule, propofol, self-associates in water

solution was also investigated (Paper III). A large part of the work concerned aqueous

dispersions of CNTs, where the adsorption behavior of the amphiphilic polymer

(Pluronic F127) and of different kinds of small surfactant molecules were studied with

emphasis on the question how does the adsorption affect the dispersibility (Papers IV,

V and VI). Furthermore, the CNT dispersions in an epoxy resin were prepared and the

underlying mechanism for the rheological behavior was investigated (Paper VII).

3.1. Amphiphilic molecules associated in solution

Core-shell (with alkyl cores and polyethylene oxide shell) micelles were formed by

dissolving either pure Brij S20 (C18E20), pure Brij S100 (C18E100) or their 1:1

mixture in water. Furthermore, the micelles formed by Brij S20 were used as a carrier

to be loaded by HMDSO.

3.1.1. Micelle core freezing

Brij micelles exhibit the interesting feature of having the hydrophobic chains in their

micellar core frozen below a given temperature. Core freezing was detected by us from

the temperature-dependent line-width data shown in Figure 3.1. Clearly, the core peaks

broadened a lot below 6-8 °C that is a consequence of the shortening of the transverse

relaxation time T2 because of the slowing down of molecular mobility and the loss of

segmental flexibility. That this is a local feature in the core and does not caused by,

32

for example, a large change in the overall micellar mobility is shown by the absence

of any change in the data for the shell moieties.

Figure 3.1 The temperature-dependent 1H NMR line width for moieties within the micellar

core and shell parts in pure Brij S20 (A), pure Brij S100 (B) and 1:1 mixture of those two

(C).

Micelles prepared with different Brij compositions behaved slightly differently from

each other (Figure 3.1), although the line width increased by decreasing temperatures

in all three systems. In Brij S20 micelles, the core freezing commenced sharply at

around 7 °C while in Brij S100 micelles the slowing down of dynamics was seemingly

more gradual but ending with a sharp transition below 3 °C. Since the hydrophobic

chains of Brij S20 and Brij S100 are the same, the observed differences indicate that

the longer PEO blocks in Brij S100 hinder the ordering of the core alkyl chains.

Considering the high mobility and strong hydration of PEO peaks, this is quite

plausible. In the case of the mixture, the presence of a second component has not made

33

core freezing more spread out but, if anything, contributed to sharpening that

transition.

The freezing of the micellar core was also detected in the longitudinal relaxation data,

see Figure 3.2. The relative change in those data upon freezing is smaller than that for

the line widths that can be explained by having smaller effect of the alkyl chain

ordering on fast and small-amplitude local segmental motions. The continuous change

of the data for the PEO chains signifies a conventional continuous change in the

thermally activated processes in the shell part.111

Figure 3.2 The temperature dependent longitudinal relaxation times for moieties within the

micellar core and shell parts in pure Brij S20 (A), pure Brij S100 (B) and 1:1 mixture of

those two (C).

There are some additional details to notice when comparing the shell data in Figures

3.1 and 3.2. Firstly, while the longitudinal relaxation time is decreasing by decreasing

temperature, the line width keeps constant. The reason for the insensitivity of the latter

34

parameter to a slow gradual change in dynamics is that for the very flexible PEO chain

the transverse relaxation is quite slow and therefore the line width is dominated other

effects like field inhomogeneity and spin-spin coupling. Secondly, the line width of

PEO moieties from Brij S20 micelles is larger than that in Brij100. This could be the

result of having shorter PEO blocks in Brij S20 and the PEO segments closest to

micellar core probably exhibit slower dynamics than the segments further out.

NMR spectroscopy was more sensitive to the core freezing features than SAXS.

However, SAXS experiment could reveal that, in the mixture, one obtains not only

core freezing but also a phase separation of the molecular constituents. Namely, the

SAXS data obtained for the mixture below the core freezing point were well

reproducible by assuming two different micellar aggregates with shell sizes dictated

by the different PEO block lengths.110 As shown by Figure 3.3, above the core freezing

point, the same structural model was clearly invalid. Yet, the data still revealed a core-

shell geometry. These observations are consistent with the two surfactants forming a

single mixed aggregate above core freezing.

Figure 3.3 The discrepancy (𝜒2) between the scattering data obtained in the 1:1 mixture of

Brij S20 and Brij S100 and the linear combination of the data for the two pure constituents.

The discrepancy is shown for repeated heating/cooling cycles.

35

3.1.2. Release of solubilizate from micelle upon core freezing

Micelles are widely used to solubilize hydrophobic molecules. As a particular aspect,

they can also act as carriers of hydrophobic drugs or other agents that can be released

under the stimuli of pH, light, or temperature, or the combination thereof.112-119 The

condition for release is that those stimuli change the micellar state so that it becomes

less favorable for holding the solubilizate. Hence, the core freezing of Brij micelles

discussed above seemed to be worth testing as a possible triggered release mechanism.

HMDSO was chosen as a hydrophobic model molecule to be loaded in Brij S20 (1 w%

in heavy water) micelles. An experimental protocol was developed to investigate the

saturation point of the micelles with HMDSO; since the uptake kinetics was slow, the

protocol involved an over-saturation of the solution with HMDSO and after which the

superfluous non-solubilized HMDSO was repeatedly removed via phase

separation/centrifugation.

Figure 3.4 The amount of solubilized HMDSO as a function of the initial HMDSO loading

in the micellar solution. The insert shows the initial part of the loading curve.

As shown by Fig. 3.4, the saturation level was approx. 0.25 mg/mL that corresponds

(see below) to approx. 10% HMDSO in the total core volume. Figure 3.5 displays that

the core freezing, as witnessed by the line width of the core methylene peaks, was

suppressed, even well below core saturation, from 7-8 °C to 2-3 °C.

36

Figure 3.5 The temperature-dependent line width of the core methylene peak in Brij S20

micelles with and without added HMDSO. The different symbols belong to samples

prepared by different initial loadings.

Core freezing does not only lead to slowing down the molecular motions of the alkyl

chains in the micellar core. In addition, the temperature dependence of the 1H chemical

shifts (see Fig. 5 in Paper II) shows that the core becomes denser and the alkyl chains

presumably attain all-trans conformations. Again, in line with what was discussed in

the previous section, the 1H chemical shift of the shell moieties shows no discernible

effect upon core freezing, thereby indicating that the shell remains in a well hydrated

and highly mobile state even if the core attains limited mobility.

Another 1H chemical shift effect permits us to assess the amount of HMDSO within

the micelles and being out in the aqueous phase. Namely, the 1H chemical shift is

different in those two environments. Hence, upon decreasing the temperature toward

the core freezing point, the HMDSO peak splits into two as shown by Figure 3.6, where

the peak around 0 ppm arises from HMDSO loaded in the micelles and the peak at

approx. -0.05 ppm belongs to HMDSO that has been released into aqueous

environment.

37

Figure 3.6 The 1H NMR spectra of HMDSO at saturation level in Brij S20 micelles (A) with

the spectral segment with HMDSO peak enlarged (see dashed delimiters in A) at 20 °C (B),

7 °C (C), 4 °C (D) and 1 °C (E).

Upon lowering the temperature, the intensity of the HMDSO peak at -0.05 ppm

becomes higher, indicating cumulative release all the way down to the core freezing

point, to which point we return later. The HMDSO for which this peak arises from

cannot be in a molecularly dispersed state, because the equilibrium HMDSO solubility

in water is approx. 0.0005 mg/mL. Two more observations, the slight turbidity of the

liquid at the temperatures where the split spectra are observed and the slow (hours)

38

decrease of the peak intensity by time, indicate that the corresponding HMDSO state

is small droplets dispersed in water. The decrease of signal intensity is a consequence

of those droplets floating to the top of the sample columns and out of the sensitive

probe volume.

Figure 3.7 Surfactant (from shell moieties) and HMDSO self-diffusion coefficients at

different initial loadings as obtained by 1H diffusion NMR experiments.

The fact that we see two HMDSO peaks in Figure 3.6 indicates that the HMDSO

exchange between micelles and droplets is slow on the spectral time scale (∼40 ms).

On the other hand, the self-diffusion coefficients of HMDSO and the micellized

surfactant (see Figure 3.7) indicated that, above the onset of the release, the exchange

between molecularly dissolved HMDSO and the micelles was fast on the diffusion

time scale (∼200 ms). This apparent controversy can be resolved (see the relevant

discussion in more detail in Paper II) by considering that the exchange between the

environments (A) HMDSO solubilized in the micelles and (B) HMDSO in droplets, is

proceeding through (C) HMDSO molecularly dissolved in water. The small size (recall

the solubility of 0.0005 mg/mL) of environment C makes it act as a bottleneck.

39

From the spectra as in Figure 3.6 that relative proportions of HMDSO in the two

dominant environments, solubilized and released, can be extracted by fitting two

Lorentzian peaks to the experimental points. The results are presented in Figure 3.8.

Figure 3.8 The HMDSO content (%) within () and out of () micelles at different

temperatures, at different initial loadings (A) 2.5 mg/mL, (B) 1.5 mg/mL and (C) 0.5

mg/mL.

In all cases, one detects a fairly stable situation at high temperatures but upon

approaching the core freezing point of unloaded Brij S20 micelles, released HMDSO

appears in the system. This release accelerates in the temperature region down to the

core freezing point below which the micelles retain HMDSO at approx. 15% of the

saturation level. In samples that were originally unsaturated (Figure 3.8 C) the initial

levels are more stable and the transition is sharper than those in saturated samples.

Polydispersity, both in size and in HMDSO content, might be the underlying reason

for this observation.

40

Additional insight into the release process can be obtained by expressing the HMDSO

content in the micelle core through its volume fraction 𝜙:

𝜙 =𝑉𝐻𝑀𝐷𝑆𝑂 ∙ 𝑐𝐻𝑀𝐷𝑆𝑂

𝑉𝐶18 ∙ 𝑐𝑠 + 𝑉𝐻𝑀𝐷𝑆𝑂 ∙ 𝑐𝐻𝑀𝐷𝑆𝑂 (3.1)

where 𝑉𝐻𝑀𝐷𝑆𝑂 and 𝑉𝐶18 are the molar volumes of HMDSO (0.212 L/mol) and C18

chains (0.314 L/mol), respectively and 𝑐𝐻𝑀𝐷𝑆𝑂 and 𝑐𝑠 are the molar concentrations of

HMDSO and surfactant, respectively. As shown in Figure 3.9, the volume fraction in

saturated micelles is approx. 9.5%. This value and the decrease of it upon decreasing

temperature has been confirmed by SAXS experiments, see details in Paper II.

Figure 3.9 The temperature dependence of the volume fraction of HMDSO within the

micellar core.

The most interesting feature in Figure 3.9 is that the HMDSO content during the

release process follows the same master curve irrespective of the initial HMDSO

content. This result suggests that the release of HMDSO from micelle is not only

related to core freezing but is also associated with a loading capacity, which is

temperature dependent. Once the loading capacity is lower than the solubilizate

content in the micelle, solubilizate is released.

41

3.1.3. The state of propofol in aqueous solution

As described in section 2.2.2, the chemical shift depends on both the intra- and

intermolecular environments for a particular atom. Hence, changes in the

intermolecular environment in form of self-association should be reflected in the

chemical shift. Propofol, an anaesthetic (see Fig. 3.10), is an amphiphilic molecule and

in gas phase it was indicated to form trimers and tetramers in association to water

molecules.120 Since the association of propofol at surfaces and in pores turned out to

be an open issue, it turned out that a question arose concerning its self-association in

bulk water. We note in passing that propofol was also one molecule we tested (and

ruled out) as solubilizate to be released upon core freezing as discussed in the previous

section.

Figure 3.10 The molecular structure and 1H NMR peak assignment of propofol. The

chemical shift was calibrated with the water peak at 4.78 ppm.

To investigate this point, the chemical shift of propofol was recorded in aqueous

solutions with different propofol concentrations. Since propofol contains an aromatic

ring, one would expect large changes in the chemical shift upon any conceivable self-

association. As shown in Figure 3.11, all chemical shifts within the propofol are

constant irrespective the concentration range which indicates that propofol remains a

single-molecular solute. Furthermore, the intensity of NMR peaks of propofol are

proportional to its concentration, excluding the presence of large aggregates (for which

the NMR signal could be lost because of large broadening).

42

Figure 3.11 The 1H NMR chemical shift recorded for propofol at different concentrations in

water, the symbols correspond to atom of the same color in Figure 3.10.

3.2. Block copolymers adsorbed at carbon nanotube surfaces

As described above, CNTs can be dispersed in aqueous media with the help of Pluronic

F127 as a dispersant. Several factors can markedly affect the dispersions created, such

as the initial loading of CNT and the sonication time. Dispersions can also be prepared

by other surfactants, and competitive adsorption between different surfactants and

F127 can reveal the relative strength of binding of those molecules to CNTs is another

factor that influences the dispersion process. These phenomena and some insights into

them via the work presented in this thesis are shortly discussed below. Furthermore,

the effect of dispersed CNTs on the rheology of epoxy resins and the molecular

features affecting this are deliberated as the final topic of this thesis.

3.2.1. The dispersion process of carbon nanotubes

In some previous studies, one has exploited the rather plausible and consistent

postulate that Pluronic F127 molecules that are attached to nanotube surfaces still

exhibit mobile sections and corresponding narrow NMR spectra.62, 100 Based on that

postulate, one analyzed the bound fraction of F127 molecules in SWNT dispersions

43

and investigated, for example, how does that bound fraction exchange with the F127

molecules free in solution.

In Paper V we investigate in detail whether or not the state of F127 molecules attached

to SWNT surfaces matches those earlier expectations. Indeed, we verify that the slow

tail in the diffusional decay of F127 molecules in SWNT dispersions arises from F127

molecules that are attached to the SWNT surfaces by their PPO block. Figure 3.12

summarizes the unambiguous evidence. To collect this evidence we, again, relied on

the chemical selectivity of NMR spectroscopy – namely, that the signal arises from the

PPO blocks presents a single-component diffusional decay while the same decay for

the signal from the PEO block is two-exponential. This can only happen if a strongly

attached and thereby dynamically strongly hindered PPO block fails to exhibit

sufficient motional averaging of spin couplings and thereby its 1H NMR signal suffers

from large spectral broadening so that it becomes undetectable.

Figure 3.12 The 1H NMR diffusional decays of the peaks arising from the different polymer

blocks of Pluronic F127, in both neat solution and in SWNT-F127 dispersion.

In addition to this confirmation, a careful analysis of the relative spectral intensities of

the different peaks in the F127 spectrum yielded that the block polydispersity of F127

played a role in the preparation of dispersions. Namely, we found that F127 polymers

44

with relatively large hydrophobic blocks dominantly adsorb on such bundles of

nanotubes that do not become exfoliated and subsequently dispersed. Assumedly, this

can be a consequence of competitive adsorption of different F127 varieties on the

hydrophobic bundle surfaces.

Similarly to a large group of other ionic surfactants,121 the amount of SWNTs in

dispersions exhibits a characteristic sigmoidal dispersion curve (see Figure 1 in Paper

IV) as a function of the logarithmic surfactant concentration. The question is what are

the physical/molecular mechanisms that provide this curve with its characteristic and

apparently generic form. Specifically, (i) what provides the apparent threshold (cdc,

critical dispersibility concentration) at which the SWNT concentration in the

dispersions starts to grow and (ii) what is the mechanism that limits the plateau value

(the maximum dispersed SWNT concentration, cSWNT,max) to far below the initial

amount of added SWNT.

Figure 3.13 (a) Initial F127 concentration vs the F127 concentration in the dispersion as

estimated from NMR intensity of the ethylene oxide signal in the 1H NMR spectrum. The

lines are exponential fits to guide the eye. (b) The dispersion curves plotted as a function of

the polymer content in the supernatant. Data in both (a) and (b) are provided at different

initial SWNT loadings cSWNT-init.

Regarding question (i), by following the dispersant concentration in the dispersions

via 1H NMR spectral intensities we could show that initially, the dispersants are

45

adsorbed to the pristine SWNT particles/bundles mixed into the solution. Apparently,

the concentration in the dispersant must then reach a threshold value. If the dispersion

curves are re-scaled with the surfactant concentration in the dispersant as the control

parameter, they seem to coincide (see Figure 3.13b) for different initial loadings of

SWNT. Moreover, 1H diffusion studies along the dispersion curves (see Figure 5 in

Paper IV) indicate that, in the dispersion, most of the dispersants remain dissolved in

the bulk, with only a few % providing surface coverage for the defoliated small bundles

and individual nanotubes. Hence, the threshold behavior in the dispersion curves

seems to originate from kinetic demands: that sufficient amount of surfactant must be

around to be able to provide coverage to freshly exposed surfaces in a sufficiently rapid

manner. If this not happens, the openings created by ultrasonication close back.

Regarding question (ii), we performed experiments with increasing sonication time

and found that the one obtains similar dispersion curves with the plateau value

increasing by increasing sonication time. The conclusion drawn was that cSWNT,max is

set on one hand by the particle/bundles size distribution that is moving to smaller sizes

upon increasing sonication time and, on the other hand, by the centrifugation step that

sets up a size limit above which the particles/bundles are precipitated. Yet, simply

increasing the ultrasonication time is not a necessarily recipe to follow to prepare

SWNT dispersions – besides making smaller bundles, the nanotubes are also become

broken by sonication which limits their applicability.

3.2.2. Competitive adsorption of amphiphilic molecules at CNT

surfaces

As mentioned earlier, the F127 PEO signal in SWNT dispersion exhibits a bi-

exponential diffusional decay from which the amount of F127 adsorbed at the SWNT

surfaces in dispersion can be estimated.62, 100 From those studies, one could also

conclude that the PPO groups of the adsorbed F127 molecules provide a complete

coverage of the available SWNT surfaces. In Paper VI, this feature is exploited to

assess the competitive adsorption of surfactants to SWNT. Namely, after having

prepared dispersions by F127 (as detailed in the previous sub-section), one can add

another sort of surfactant to the dispersion and record the diffusional decay of F127

46

PEO signal (Figure 3.14). As earlier, the slowly decaying component provides the

F127 adsorbed on SWNT surface. If this component decreases with the increasing of

concentration of the added second surfactant, it indicates that the added surfactant

replaces F127 on the SWNT surface.

Figure 3.14 The diffusional decay of the normalized 1H NMR signal from the PEO peak in

F127 as a function of concentration of SDS (a) and DTAB (b) added to SWNT-F127

dispersion. The lines are fits of bi-component decays described with the customary Stejskal-

Tanner equation with b = γ2δ2g2(Δ-δ/3), see eq 2.16.

As shown by Figure 3.14, different surfactants show markedly different abilities to

replace F127. In other words, they show markedly different strength of binding to the

SWNT surfaces. To provide a more quantitative analysis, the fraction of adsorbed

F127 on the SWNT surfaces, fF127 was extracted by fitting the experimental F127

diffusional decay with a bi-exponential expression.62, 100 From the obtained values

shown in Figure 3.15, the apparent surface coverage 𝜎F127 of SWNTs in the dispersion

can also be assessed. Both the fF127 and 𝜎F127 strongly decrease with adding SDS in the

dispersion, while the addition of DTAB hardly influences those two parameters.

Hence, SDS, an anionic surfactant, binds to SWNT surfaces significantly stronger than

F127. The binding of DTAB, a cationic surfactant, is on the other hand significantly

weaker than that of F127. A more quantitative analysis over a wider range of

surfactants is presented in detail in Paper VI.

47

Figure 3.15 Fraction fF127 of F127 molecules adsorbed on the SWNT surface and the

corresponding apparent surface coverage 𝜎F127 by F127 molecules as a function of the

concentration of SDS (a) and DTAB (b) added to system originally dispersed by F127. The

dashed lines are linear fits.

3.2.3. Polymer-CNT composite: structure and mobility

CNT-polymer composites represent one current and important applications of CNTs.

In one preparation pathway, CNTs are dispersed into a monomeric/oligomeric resin

and the resin is subsequently polymerized. One limiting factor for this process is the

very high viscosity the dispersion may exhibit. This feature and its structural/molecular

details constituted on important motivation for the experiments in Paper VII.

Figure 3.16 Light microscopy images of epoxy resin loaded with CNTs at (a) 0.05, (b) 0.10

and (c) 0.15 wt%.

Dominantly monomeric epoxy resin was loaded with different CNTs (the CNTs were

first pre-dispersed and then dried). Their large-scale structure consists of aggregates

48

as revealed by the light microscopy images in Figure 3.16. The aggregates, whose size

is around tens of microns, increase in volume fraction as the CNT concentration

increases. Yet, considering the composition (∼0.1 wt%), they must be rather loose with

most of the aggregate volume filled by epoxy.

To get more insight, 1H diffusion NMR experiments were performed on the epoxy

molecules. As shown by Figure 3.17, the diffusional decays were clearly not single-

exponential but seemed to exhibit a slowly-decaying tail that indicated the presence of

epoxy molecules that diffused much slower than those in the bulk. To extract the

fraction p of epoxy molecules that resided in the environment with slow diffusion, we

fitted the data by a simple bi-exponential decay.62, 100

Figure 3.17 The diffusional decay of the 1H NMR signal of the epoxy-specific aromatic

protons in CNT-loaded epoxy resin with various CNT concentrations. The lines are bi-

exponential fits, with the exception of CNT-free epoxy (single exponential fit).

In bulk epoxy without any CNT, the self-diffusion coefficient of the epoxy was

obtained to Dneat=5.05×10-12m2/s, a value that is more than an order of magnitude

lower than that in alkanes of comparable size. This points to strong intermolecular

interactions among the epoxy molecules. Regarding the slow component, in contrast

to the F127 molecules in aqueous dispersions,62, 100 this could not be identified as

surface-adsorbed epoxy. Namely, the small epoxy molecules could not be divided up

49

into CNT-attached and mobile blocks. Hence, the slow component was identified

instead as epoxy within the aggregates, where a strong attractive interaction with the

CNT surfaces leads to slowing down of molecular transport. The strong interaction

with the CNT surface was confirmed by SAXS data that indicated the intermolecular

order of epoxy changed in the presence of the CNTs. As a further support for this

hypothesis, the diffusional decays showed no discernible dependence on the diffusion

time that indicated no significant exchange of epoxy between the bulk and within-

aggregates states. This was consistent with the aggregate sizes as revealed by

microscopy.

Figure 3.18 (a) The variation of p, the fraction of epoxy within the CNT aggregates and (b)

Dslow/Dneat, the epoxy diffusion coefficient within the aggregates normalized with Dneat of

neat epoxy.

Upon increasing CNT content, the proportion of the aggregate-inhibited epoxy p

increased, first quickly and then toward an apparent saturation, see Figure 3.18. At the

same time, the within-aggregate diffusion coefficient slowed down which indicated

that at higher CNT content the assumed aggregates were becoming denser. As shown

in detail in Paper VII, the variation of p with increasing CNT content had some

similarities with the variation of various rheological parameters. This was rationalized

by positing that aggregate-aggregate interactions and jamming are the features that

control rheology. Such behavior is probably facilitated by having the aggregates

stabilized by the strong attractive interactions between the CNT scaffold and the epoxy

filler.

50

4. Concluding remarks

In this work, two types of colloidal systems, micelles and CNT dispersions, were

studied by various NMR spectroscopic methods in combination with other techniques

such as SAXS (micellar system) and TGA/UV-vis (CNT dispersions). The obtained

results reveal molecular-level information about phase transitions of micellar cores

and about dispersant-CNT interactions. In both cases, the chemical selectivity of NMR

spectroscopy was a crucial advantage: by exploiting it we could investigate particular

molecular components in complex mixtures.

In the Brij-type micellar systems studied here, we could reveal the structural details of

two separate phase transition phenomena, both of which appears in connection to the

freezing of the alkyl core of the micelles upon decreasing temperature. Regarding the

first of those phase transitions, the spontaneous separation of different sorts (that is,

Brij S100 and Brij S20) of surfactants into distinct aggregates, NMR spectroscopy

could accurately pinpoint freezing in the mixed surfactant system while SAXS

indicated the presence of one aggregate type above core freezing and two aggregate

types below core freezing. In pure micelles, we could show that the presence of longer

PEO block leads to a significant reduction in the core freezing temperature but also to

a more complex (and, assumedly, more heterogeneous) freezing behavior. In micelles

loaded with a hydrophobic solubilizate, the phase transition was again from a mixed

phase, namely the mixed micellar core, into two pure phases, the micelles practically

devoid of solubilizate and solubilizate droplets in water. Universal release curves

obtained at different initial solubilizate loadings were rationalized in terms of a

temperature-dependent loading capacity of micelles.

In most studies of CNT dispersions, Pluronic F127 was used as a model dispersant. As

we have argued above, this molecule adsorbs to CNT surfaces via its hydrophobic PPO

block but its terminal PEO blocks remain extended into water and highly mobile. For

that reason, they provide easily detectable NMR spectra even if the F127 molecule,

via its PPO block, is attached to a CNT. This feature could be exploited in NMR

diffusion experiments to obtain a quantitative measure of the amount of the adsorbed

F127 molecules within a dispersion. Hence, the distribution of the dispersant over

various molecular states could be followed in detail along the dispersibility curves that

51

recorded the amount of CNT that could be put into dispersion with at a given

concentration of the dispersant. The role and effect of the ultrasonication in the

dispersion process has been clarified and the results strengthened the notion that the

dispersant concentration required to disperse is a kinetic requirement: enough

surfactant must be around to be quickly able to cover newly explored CNT surfaces.

Yet, the strength of binding of different possible dispersants is still an important factor

that could be assessed by measuring the competitive adsorption of different dispersants

and F127 at the surface of CNTs.

NMR diffusion experiments were also helpful in another system, constituted by CNTs

dispersed in an epoxy resin. Surprisingly, we detected that a fraction of the epoxy

diffused much slower in such dispersions which was explained by the presence of

randomly associated CNT aggregates. Within those, the slow epoxy diffusion was

attributed to strong interactions between epoxy and the CNT surface. The interaction

of the aggregates with each other was found to be responsible for the observed

rheological behavior.

52

5. List of abbreviations

cdc Critical dispersibility concentration

cmc Critical micelle concentration

cpp Critical packing parameter

CNT Carbon nanotube

DMF N,N-dimethylformamide

DTAB Dodecyltrimethylammonium bromide

FID Free induction decay

HMDSO Hexamethyldisiloxane

MWNT Multi-walled carbon nanotube

NMP N-methyl-2-pyrrolidone

ODCB O-dichlorobenzene

PEO Poly(ethylene oxide)

PGSE Pulsed field gradient spin Echo

PPO Poly(propylene oxide)

rf Radiofrequency

SAXS Small-angle X-ray scattering

SDS Sodium dodecyl sulfate

SWNT Single-walled carbon nanotube

TGA Thermogravimetric analysis

UV-vis Ultraviolet-visible spectroscopy

53

6. Acknowledgements

This work would be impossible without lots of help and support from many people.

Therefore I would like to address my deeply thanks to:

Prof. István Furó for accepting me as a PhD student in your department. Thanks for

your excellent supervision, your patience and sharing your knowledge and great ideas.

Things were not always going well during last five years, but you always kept passion

and encouraged me to overcome all the difficulties. Thanks for your trust and help,

otherwise I could not reach the finish line. Thank you also for lots of help during the

writing process.

Prof. Sergey Dvinskikh for giving support whenever I got problems with the software

or spectrometers. Thank you for sharing your knowledge, answering all my questions

and helping me to improve my experimental skills.

Dr. Pavel Yushmanov for helping to deal with all the instrumental problems. Thank

you for building various useful things in the lab. Thanks for your intelligent

suggestions and improvements on many devices used in this work. I also thank you for

the nice time we shared in our office during the last but important period of my doctoral

study.

Prof. Mark Rutland for reviewing this thesis and making valuable comments.

Lena Skowron for your kind help not only in administrative stuffs but also in my daily

life. Vera Jovanovic and Dr. Ulla Jacobsson for your kind assistance in the

administrative matters. Johanna Hagerman, Alexandra Rudyk Kinnander and

Daniel Tavast for answering the questions related to thesis preparation.

Dr. Ricardo M. F. Fernandes for your guidance of all the carbon nanotube work.

Thank you for sharing your knowledge and ideas, answering my questions with the

greatest patience. Thanks for discussing and giving valuable suggestions in this work.

Thanks for your encouragement all the way.

54

Dr. Yuan Fang for your kind help from the beginning of my PhD days, both in studies

and daily life. Thank you for encouraging me to go through the very hard time before.

Thanks for supporting me whenever I need it.

My dear friends and colleagues in our corridor for creating a very nice working

atmosphere. Thank you Marianne, Erik, Fredrik, Evgeny, Bulat, Boris, Zoltan,

Tore, Joakim, Björn, Camilla G., Max, Ge, Himanshu, Christofer, Rena, Malisa,

Vasantha, Maria, Kristina, Camilla T., Michal, Mika for all the good time we

shared in fika and after-works. Thank you Dr. Robert Corkery for inspiring

discussions when I got stuck on some difficulties in my work. Thank you for sharing

your knowledge, experiences and giving helpful suggestions. Thank you also Zahra

Alaei for the support and companionship. Thanks for the time we spent on discussing

and working in the lab. I really enjoyed the time we worked together.

My friends in Sweden: Zhizhi, Hongduo, Miaomiao, Hao, Qiong, Hui, Bin, Yujie,

Wei, Duo, Dongming,Weifeng, Xiaoyan, Fan, Geng for all the good time we shared

and all the helps you gave me.

My parents for your love and support all the time. I love you!

55

7. References

1. Holmberg, K., et al., Surfactants and Polymers in Aqueous Solution, Second Edition.

2002: John Wiley & Sons.

2. Rubinstein, M. and R.H. Colby, Polymer physics 2003: Oxford University Press.

3. Fruijtier-Pölloth, C., Safety assessment on polyethylene glycols (PEGs) and their

derivatives as used in cosmetic products. Toxicology, 2005. 214(1): p. 1-38.

4. Harris, J.M., Poly(Ethylene Glycol) Chemistry: Biotechnical and Biomedical

Applications. 2013: Springer Science & Business Media.

5. Lee, J.H., H.B. Lee, and J.D. Andrade, Blood compatibility of polyethylene oxide

surfaces. Progress in Polymer Science, 1995. 20(6): p. 1043-1079.

6. Smyth, H.F., C.P. Carpenter, and C.S. Weil, The toxicology of the polyethylene

glycols. Journal of the American Pharmaceutical Association, 1950. 39(6): p. 349-354.

7. Zalipsky, S., Chemistry of polyethylene glycol conjugates with biologically active

molecules. Advanced Drug Delivery Reviews, 1995. 16(2): p. 157-182.

8. Zhu, J., Bioactive modification of poly(ethylene glycol) hydrogels for tissue

engineering. Biomaterials, 2010. 31(17): p. 4639-4656.

9. Behrens, M.A., J. Bergenholtz, and J.S. Pedersen, Temperature-Induced Attractive

Interactions of PEO-Containing Block Copolymer Micelles. Langmuir, 2014. 30(21): p. 6021-

6029.

10. Jensen, G.V., et al., Structures of PEP–PEO Block Copolymer Micelles: Effects of

Changing Solvent and PEO Length and Comparison to a Thermodynamic Model.

Macromolecules, 2012. 45(1): p. 430-440.

11. Ribeiro, M.E.N.P., et al., Solubilisation capacity of Brij surfactants. International

Journal of Pharmaceutics, 2012. 436(1): p. 631-635.

12. Sommer, C., J.S. Pedersen, and V.M. Garamus, Structure and Interactions of Block

Copolymer Micelles of Brij 700 Studied by Combining Small-Angle X-ray and Neutron

Scattering. Langmuir, 2005. 21(6): p. 2137-2149.

13. Tang, J., et al., Key Structure of Brij for Overcoming Multidrug Resistance in Cancer.

Biomacromolecules, 2013. 14(2): p. 424-430.

14. Alexandridis, P., Amphiphilic copolymers and their applications. Current Opinion in

Colloid & Interface Science, 1996. 1(4): p. 490-501.

15. Batrakova, E.V. and A.V. Kabanov, Pluronic block copolymers: Evolution of drug

delivery concept from inert nanocarriers to biological response modifiers. Journal of

Controlled Release, 2008. 130(2): p. 98-106.

56

16. Bromberg, L.E. and E.S. Ron, Temperature-responsive gels and thermogelling

polymer matrices for protein and peptide delivery. Advanced Drug Delivery Reviews, 1998.

31(3): p. 197-221.

17. Kabanov, A.V., E.V. Batrakova, and V.Y. Alakhov, Pluronic® block copolymers as

novel polymer therapeutics for drug and gene delivery. Journal of Controlled Release, 2002.

82(2): p. 189-212.

18. Kabanov, A.V., et al., Pluronic® block copolymers: novel functional molecules for

gene therapy. Advanced Drug Delivery Reviews, 2002. 54(2): p. 223-233.

19. Alexandridis, P., Poly(ethylene oxide) poly(propylene oxide) block copolymer

surfactants. Current Opinion in Colloid & Interface Science, 1997. 2(5): p. 478-489.

20. Nagarajan, R., Micellization, mixed micellization and solubilization: The role of

interfacial interactions. Advances in Colloid and Interface Science, 1986. 26(Supplement C):

p. 205-264.

21. Lombardo, D., et al., Amphiphiles Self-Assembly: Basic Concepts and Future

Perspectives of Supramolecular Approaches. Advances in Condensed Matter Physics, 2015.

22. Ramanathan, M., et al., Amphiphile nanoarchitectonics: from basic physical

chemistry to advanced applications. Physical Chemistry Chemical Physics, 2013. 15(26): p.

10580-10611.

23. Chattopadhyay, A. and E. London, Fluorimetric determination of critical micelle

concentration avoiding interference from detergent charge. Analytical Biochemistry, 1984.

139(2): p. 408-412.

24. Ebel, C., Sedimentation velocity to characterize surfactants and solubilized membrane

proteins. Methods, 2011. 54(1): p. 56-66.

25. Lipfert, J., et al., Size and shape of detergent micelles determined by small-angle x-

ray scattering. Journal of Physical Chemistry B, 2007. 111(43): p. 12427-12438.

26. Rosen, M.J. and S. Aronson, Standard Free-Energies of Adsorption of Surfactants at

the Aqueous Solution-Air Interface from Surface-Tension Data in the Vicinity of the Critical

Micelle Concentration. Colloids and Surfaces, 1981. 3(3): p. 201-208.

27. Söderman, O., P. Stilbs, and W.S. Price, NMR studies of surfactants. Concepts in

Magnetic Resonance Part A, 2004. 23A(2): p. 121-135.

28. Xu, R.L., et al., Light-Scattering Study of the Association Behavior of Styrene

Ethylene-Oxide Block Copolymers in Aqueous-Solution. Macromolecules, 1991. 24(1): p. 87-

93.

29. Jones, M.C. and J.C. Leroux, Polymeric micelles - a new generation of colloidal drug

carriers. European Journal of Pharmaceutics and Biopharmaceutics, 1999. 48(2): p. 101-111.

30. Alexandridis, P. and T.A. Hatton, Poly(Ethylene Oxide)-Poly(Propylene Oxide)-

Poly(Ethylene Oxide) Block-Copolymer Surfactants in Aqueous-Solutions and at Interfaces -

Thermodynamics, Structure, Dynamics, and Modeling. Colloids and Surfaces a-

Physicochemical and Engineering Aspects, 1995. 96(1-2): p. 1-46.

57

31. Bohorquez, M., et al., A Study of the Temperature-Dependent Micellization of

Pluronic F127. Journal of Colloid and Interface Science, 1999. 216(1): p. 34-40.

32. Kulthe, S.S., et al., Mixed micelle formation with hydrophobic and hydrophilic

Pluronic block copolymers: Implications for controlled and targeted drug delivery. Colloids

and Surfaces B: Biointerfaces, 2011. 88(2): p. 691-696.

33. Perry, C.C., et al., Fluorescence of commercial Pluronic F127 samples: Temperature-

dependent micellization. Journal of Colloid and Interface Science, 2011. 354(2): p. 662-669.

34. Gaucher, G., et al., Block copolymer micelles: preparation, characterization and

application in drug delivery. Journal of Controlled Release, 2005. 109(1-3): p. 169-188.

35. Kataoka, K., A. Harada, and Y. Nagasaki, Block copolymer micelles for drug delivery:

design, characterization and biological significance. Advanced Drug Delivery Reviews, 2001.

47(1): p. 113-131.

36. Kwon, G., et al., Block copolymer micelles for drug delivery: loading and release of

doxorubicin. Journal of Controlled Release, 1997. 48(2-3): p. 195-201.

37. Christian, S.D. and J.F. Scamehorn, Solubilization in Surfactant Aggregates. 1995:

CRC Press.

38. Letchford, K. and H. Burt, A review of the formation and classification of amphiphilic

block copolymer nanoparticulate structures: micelles, nanospheres, nanocapsules and

polymersomes. European Journal of Pharmaceutics and Biopharmaceutics, 2007. 65(3): p.

259-269.

39. Iijima, S., Helical Microtubules of Graphitic Carbon. Nature, 1991. 354(6348): p. 56-

58.

40. Britz, D.A. and A.N. Khlobystov, Noncovalent interactions of molecules with single

walled carbon nanotubes. Chemical Society Reviews, 2006. 35(7): p. 637-659.

41. Vaisman, L., H.D. Wagner, and G. Marom, The role of surfactants in dispersion of

carbon nanotubes. Advances in Colloid and Interface Science, 2006. 128: p. 37-46.

42. Grady, B.P., Carbon Nanotube-Polymer Composites: Manufacture, Properties, and

Applications. 2011: Wiley.

43. Baughman, R.H., A.A. Zakhidov, and W.A. de Heer, Carbon nanotubes - the route

toward applications. Science, 2002. 297(5582): p. 787-792.

44. De Volder, M.F.L., et al., Carbon Nanotubes: Present and Future Commercial

Applications. Science, 2013. 339(6119): p. 535-539.

45. Popov, V.N., Carbon nanotubes: properties and application. Materials Science and

Engineering: R: Reports, 2004. 43(3): p. 61-102.

46. Wu, Z., et al., Transparent, Conductive Carbon Nanotube Films. Science, 2004.

305(5688): p. 1273-1276.

58

47. Spitalsky, Z., et al., Carbon nanotube–polymer composites: Chemistry, processing,

mechanical and electrical properties. Progress in Polymer Science, 2010. 35(3): p. 357-401.

48. Terrones, M., Science and technology of the twenty-first century: Synthesis, properties

and applications of carbon nanotubes. Annual Review of Materials Research, 2003. 33: p.

419-501.

49. Bergin, S.D., et al., Towards solutions of single-walled carbon nanotubes in common

solvents. Advanced Materials, 2008. 20(10): p. 1876-1881.

50. Kim, K.K., et al., Design of dispersants for the dispersion of carbon nanotubes in an

organic solvent. Advanced Functional Materials, 2007. 17(11): p. 1775-1783.

51. Balasubramanian, K. and M. Burghard, Chemically Functionalized Carbon

Nanotubes. Small, 2005. 1(2): p. 180-192.

52. Kharissova, O.V., B.I. Kharisov, and E.G.D. Ortiz, Dispersion of carbon nanotubes

in water and non-aqueous solvents. Rsc Advances, 2013. 3(47): p. 24812-24852.

53. Backes, C., Noncovalent Functionalization of Carbon Nanotubes: Fundamental

Aspects of Dispersion and Separation in Water. 2012: Springer-Verlag.

54. Wang, H., Dispersing carbon nanotubes using surfactants. Current Opinion in Colloid

& Interface Science, 2009. 14(5): p. 364-371.

55. Islam, M.F., et al., High weight fraction surfactant solubilization of single-wall carbon

nanotubes in water. Nano Letters, 2003. 3(2): p. 269-273.

56. Poulin, P., B. Vigolo, and P. Launois, Films and fibers of oriented single wall

nanotubes. Carbon, 2002. 40(10): p. 1741-1749.

57. Wallace, E.J. and M.S.P. Sansom, Carbon nanotube self-assembly with lipids and

detergent: a molecular dynamics study. Nanotechnology, 2009. 20(4).

58. Strano, M.S., et al., The role of surfactant adsorption during ultrasonication in the

dispersion of single-walled carbon nanotubes. Journal of Nanoscience and Nanotechnology,

2003. 3(1-2): p. 81-86.

59. Nativ-Roth, E., et al., Physical adsorption of block copolymers to SWNT and MWNT:

A nonwrapping mechanism. Macromolecules, 2007. 40(10): p. 3676-3685.

60. Backes, C., et al., High Population of Individualized SWCNTs through the Adsorption

of Water-Soluble Perylenes. Journal of the American Chemical Society, 2009. 131(6): p. 2172-

2184.

61. Hunter, C.A. and J.K.M. Sanders, The Nature of Pi-Pi Interactions. Journal of the

American Chemical Society, 1990. 112(14): p. 5525-5534.

62. Fernandes, R.M.F., et al., Surface Coverage and Competitive Adsorption on Carbon

Nanotubes. Journal of Physical Chemistry C, 2015. 119(38): p. 22190-22197.

63. O'Connell, M.J., et al., Reversible water-solubilization of single-walled carbon

nanotubes by polymer wrapping. Chemical Physics Letters, 2001. 342(3-4): p. 265-271.

59

64. Star, A., et al., Preparation and properties of polymer-wrapped single-walled carbon

nanotubes. Angewandte Chemie-International Edition, 2001. 40(9): p. 1721-1725.

65. Gao, J., et al., Selective Wrapping and Supramolecular Structures of Polyfluorene-

Carbon Nanotube Hybrids. Acs Nano, 2011. 5(5): p. 3993-3999.

66. Moore, V.C., et al., Individually suspended single-walled carbon nanotubes in various

surfactants. Nano Letters, 2003. 3(10): p. 1379-1382.

67. Domun, N., et al., Improving the fracture toughness and the strength of epoxy using

nanomaterials - a review of the current status. Nanoscale, 2015. 7(23): p. 10294-10329.

68. Kim, J.A., et al., Effects of surface modification on rheological and mechanical

properties of CNT/epoxy composites. Carbon, 2006. 44(10): p. 1898-1905.

69. Nadiv, R., et al., Optimal nanomaterial concentration: harnessing percolation theory

to enhance polymer nanocomposite performance. Nanotechnology, 2017. 28(30).

70. Nadiv, R., et al., The multiple roles of a dispersant in nanocomposite systems.

Composites Science and Technology, 2016. 133: p. 192-199.

71. Reales, O.A.M. and R.D. Toledo, A review on the chemical, mechanical and

microstructural characterization of carbon nanotubes-cement based composites. Construction

and Building Materials, 2017. 154: p. 697-710.

72. Shtein, M., et al., Fracture behavior of nanotube-polymer composites: Insights on

surface roughness and failure mechanism. Composites Science and Technology, 2013. 87: p.

157-163.

73. Yang, K., et al., Effects of carbon nanotube functionalization on the mechanical and

thermal properties of epoxy composites. Carbon, 2009. 47(7): p. 1723-1737.

74. Allaoui, A. and N. El Bounia, Rheological and Electrical Transitions in Carbon

Nanotube/Epoxy Suspensions. Current Nanoscience, 2010. 6(2): p. 158-162.

75. Bauhofer, W. and J.Z. Kovacs, A review and analysis of electrical percolation in

carbon nanotube polymer composites. Composites Science and Technology, 2009. 69(10): p.

1486-1498.

76. Du, F.M., et al., Nanotube networks in polymer nanocomposites: Rheology and

electrical conductivity. Macromolecules, 2004. 37(24): p. 9048-9055.

77. Hu, G.J., et al., Low percolation thresholds of electrical conductivity and rheology in

poly(ethylene terephthalate) through the networks of multi-walled carbon nanotubes.

Polymer, 2006. 47(1): p. 480-488.

78. Sandler, J.K.W., et al., Ultra-low electrical percolation threshold in carbon-

nanotube-epoxy composites. Polymer, 2003. 44(19): p. 5893-5899.

79. Shtein, M., I. Pri-bar, and O. Regev, A simple solution for the determination of pristine

carbon nanotube concentration. Analyst, 2013. 138(5): p. 1490-1496.

60

80. Matarredona, O., et al., Dispersion of single-walled carbon nanotubes in aqueous

solutions of the anionic surfactant NaDDBS. Journal of Physical Chemistry B, 2003. 107(48):

p. 13357-13367.

81. Xu, H.X., B.W. Zeiger, and K.S. Suslick, Sonochemical synthesis of nanomaterials.

Chemical Society Reviews, 2013. 42(7): p. 2555-2567.

82. Grossiord, N., et al., Time-dependent study of the exfoliation process of carbon

nanotubes in aqueous dispersions by using UV-visible spectroscopy. Analytical Chemistry,

2005. 77(16): p. 5135-5139.

83. Blanch, A.J., C.E. Lenehan, and J.S. Quinton, Optimizing Surfactant Concentrations

for Dispersion of Single-Walled Carbon Nanotubes in Aqueous Solution. Journal of Physical

Chemistry B, 2010. 114(30): p. 9805-9811.

84. Hennrich, F., et al., The mechanism of cavitation-induced scission of single-walled

carbon nanotubes. Journal of Physical Chemistry B, 2007. 111(8): p. 1932-1937.

85. Lucas, A., et al., Kinetics of Nanotube and Microfiber Scission under Sonication.

Journal of Physical Chemistry C, 2009. 113(48): p. 20599-20605.

86. Pagani, G., et al., Competing mechanisms and scaling laws for carbon nanotube

scission by ultrasonication. Proceedings of the National Academy of Sciences of the United

States of America, 2012. 109(29): p. 11599-11604.

87. Blanch, A.J., C.E. Lenehan, and J.S. Quinton, Parametric analysis of sonication and

centrifugation variables for dispersion of single walled carbon nanotubes in aqueous solutions

of sodium dodecylbenzene sulfonate. Carbon, 2011. 49(15): p. 5213-5228.

88. Fagan, J.A., et al., Length fractionation of carbon nanotubes using centrifugation.

Advanced Materials, 2008. 20(9): p. 1609-1613.

89. Zhang, R. and P. Somasundaran, Advances in adsorption of surfactants and their

mixtures at solid/solution interfaces. Advances in Colloid and Interface Science, 2006. 123: p.

213-229.

90. Keeler, J., Understanding NMR spectroscopy. 2end ed. 2010: John Wiley and Sons,

Ltd.

91. Hore, P.J., Nuclear Magnetic Resonance. 1995: Oxford University Press.

92. Furo, I., NMR spectroscopy of micelles and related systems. Journal of Molecular

Liquids, 2005. 117(1-3): p. 117-137.

93. Nordstierna, L., I. Furó, and P. Stilbs, Mixed Micelles of Fluorinated and

Hydrogenated Surfactants. Journal of the American Chemical Society, 2006. 128(20): p.

6704-6712.

94. Szutkowski, K., P. Stilbs, and S. Jurga, Proton Chemical Exchange in Aqueous

Solutions of Dodecylammonium Chloride:  Effects of Micellar Aggregation. The Journal of

Physical Chemistry C, 2007. 111(43): p. 15613-15619.

61

95. Farrar, T.C. and E.D. Becker, Pulse and Fourier transform NMR: introduction to

theory and methods. 1971: Academic Press.

96. Kowalewski, J. and L. Maler, Nuclear Spin Relaxation in Liquids: Theory,

Experiments, and Applications. 2006: CRC Press.

97. Hahn, E.L., Spin Echoes. Physical Review, 1950. 80(4): p. 580-594.

98. Price, W.S., NMR Studies of Translational Motion: Principles and Applications. 2009:

Cambridge University Press.

99. Stilbs, P., Fourier transform pulsed-gradient spin-echo studies of molecular diffusion.

Progress in Nuclear Magnetic Resonance Spectroscopy, 1987. 19: p. 1-45.

100. Fernandes, R.M.F., et al., Lateral Diffusion of Dispersing Molecules On Nanotubes

As Probed by NMR. Journal of Physical Chemistry C, 2014. 118(1): p. 582-589.

101. Frise, A.E., et al., Polymer Binding to Carbon Nanotubes in Aqueous Dispersions:

Residence Time on the Nanotube Surface As Obtained by NMR Diffusometry. Journal of

Physical Chemistry B, 2012. 116(9): p. 2635-2642.

102. Stejskal, E.O. and J.E. Tanner, Spin diffusion measurements: spin echoes in the

presence of a time-dependent field gradient. Journal of Chemical Physics, 1965. 42: p. 288-

292.

103. Brumberger, H., Modern Aspects of Small-Angle Scattering. 2010: Springer

Netherlands.

104. Lipfert, J. and S. Doniach, Small-angle X-ray scattering from RNA, proteins, and

protein complexes. Annual Review of Biophysics and Biomolecular Structure, 2007. 36: p.

307-327.

105. Ballauff, M., et al., Small-angle X-ray scattering on latexes. Macromolecular

Chemistry and Physics, 1996. 197(10): p. 3043-3066.

106. Ballauff, M., SAXS and SANS studies of polymer colloids. Current Opinion in Colloid

& Interface Science, 2001. 6(2): p. 132-139.

107. Pedersen, J.S., Small-Angle Scattering from Surfactants and Block Copolymer

Micelles, in Soft Matter Characterization, R. Borsali and R. Pecora, Editors. 2008, Springer

Netherlands. p. 191-233.

108. Pedersen, J.S., Analysis of small-angle scattering data from colloids and polymer

solutions: modeling and least-squares fitting. Advances in Colloid and Interface Science,

1997. 70: p. 171-210.

109. Pedersen, J.S. and C. Svaneborg, Scattering from block copolymer micelles. Current

Opinion in Colloid & Interface Science, 2002. 7(3-4): p. 158-166.

110. Plazzotta, B., et al., Core Freezing and Size Segregation in Surfactant Core-Shell

Micelles. Journal of Physical Chemistry B, 2015. 119(33): p. 10798-10806.

62

111. Nelson, T.R. and S.M. Tung, Temperature dependence of proton relaxation times in

vitro. Magnetic Resonance Imaging, 1987. 5(3): p. 189-199.

112. Nakayama, M., J. Akimoto, and T. Okano, Polymeric micelles with stimuli-triggering

systems for advanced cancer drug targeting. Journal of Drug Targeting, 2014. 22(7): p. 584-

599.

113. Neradovic, D., C.F. van Nostrum, and W.E. Hennink, Thermoresponsive polymeric

micelles with controlled instability based on hydrolytically sensitive N-isopropylacrylamide

copolymers. Macromolecules, 2001. 34(22): p. 7589-7591.

114. Oerlemans, C., et al., Polymeric Micelles in Anticancer Therapy: Targeting, Imaging

and Triggered Release. Pharmaceutical Research, 2010. 27(12): p. 2569-2589.

115. Phillips, D.J. and M.I. Gibson, Towards being genuinely smart: 'isothermally-

responsive' polymers as versatile, programmable scaffolds for biologically-adaptable

materials. Polymer Chemistry, 2015. 6(7): p. 1033-1043.

116. Qin, S.H., et al., Temperature-controlled assembly and release from polymer vesicles

of poly(ethylene oxide)-block-poly(N-isopropylacrylamide). Advanced Materials, 2006.

18(21): p. 2905-2909.

117. Rapoport, N., Physical stimuli-responsive polymeric micelles for anti-cancer drug

delivery. Progress in Polymer Science, 2007. 32(8-9): p. 962-990.

118. Rijcken, C.J.F., et al., Triggered destabilisation of polymeric micelles and vesicles by

changing polymers polarity: An attractive tool for drug delivery. Journal of Controlled

Release, 2007. 120(3): p. 131-148.

119. Wohl, B.M. and J.F.J. Engbersen, Responsive layer-by-layer materials for drug

delivery. Journal of Controlled Release, 2012. 158(1): p. 2-14.

120. Leon, I., et al., Water Encapsulation by Nanomicelles. Angewandte Chemie-

International Edition, 2014. 53(46): p. 12480-12483.

121. Fernandes, R.M.F., et al., Dispersing Carbon Nanotubes with Ionic Surfactants under

Controlled Conditions: Comparisons and Insight. Langmuir, 2015. 31(40): p. 10955-10965.