Nanoparticle-Polymer-Composites the solution and spray drying … · 2018-04-17 · the solution and spray drying process with an emphasis on colloidal interactions ... 6.2.2 Yield

Nanoparticle-Polymer-Composites

the solution and spray drying process with an emphasis on

colloidal interactions

Der Fakultät für Maschinenbau, Verfahrens- und Energietechnik

der Technischen Universität Bergakademie Freiberg

eingereichte

DISSERTATION

zur Erlangung des akademischen Grades

Doktor-Ingenieur

(Dr.-Ing.)

vorgelegt

von Dipl.-Ing. Martin Rudolph

geboren am 10.03.1983 in Frankenberg/Sa.

Freiberg, den 11.09.2012

I

In Liebe und tiefer Dankbarkeit

Mei e F au Julia u d ei e „Apfel äu he Malte

II

Abstract

When it comes to preparing nanoparticle-polymer-composites for the synthesis of functional

materials, the homogeneous distribution of the individual filler particles can be a tough

challenge due to strong attractive interactions. In this work a process chain for the

preparation of highly filled nanoparticle-polymer-composites is presented and the colloidal

interactions are investigated. The central element of this process chain is an organic solvent

based mixture of stabilized nanoparticles and the dissolved polymer to be spray dried. The

unit processes comprise: nanoparticles synthesis in water, phase transfer of the particles to

the organic solvent by a liquid-liquid phase transfer with an amphiphilic substance, mixing

the dissolved polymer with the transferred particles and spray drying the prepared solution

to withdraw the solvent rapidly. The bottleneck to obtain well dispersed composites are the

type of adsorbing stabilizing amphiphile as well as the complex particle interactions within

the organic solvent including the dissolved polymer. The nanoparticles investigated in the

experiments are super-paramagnetic magnetite nanoparticles with a primary particle size of

approximately 15 nm. They are transferred to dichloromethane with the amphiphiles being

different fatty acids (C8 to C18, saturated and unsaturated). A physical model describing the

deagglomeration phenomenon upon chemical adsorption of certain fatty acids onto

agglomerated particles is presented and numerically assessed. Ricinoleic acid is found to

produce the most stable nanoparticle suspension. Four different thermoplastic unbranched

chain polymers, namely poly(methyl methacrylate) PMMA, poly(bisphenol A carbonate) PC,

poly(vinyl butyral) PVB and poly(styrene) PS are chosen as the continuous phase of the

composites due to their mechanical properties, common use and solubility. Adding PMMA,

PC and PS lead to reduction of the primary particle concentration in the stabilized

nanoparticle dispersion independent from the fatty acid used. The spray dried and injection

molded composites exhibit large agglomerates of the nanoparticles fillers with sizes in the

lower micron range. In the dispersions PVB leads to adsorption on the nanoparticles and

further stabilization of the suspension. Composites with the polymer PVB show well

dispersed particles which are almost entirely deagglomerated and homogeneously

distributed. It is most probably due to the hydroxyl groups in the PVB structure that

adsorption and stabilization can occur. The other polymers are non-adsorbing and cause

flocculation possibly by depletion mechanisms. Prove of concept shows that the process

chain results in well dispersed composites when compared to the classic melt compounding

method.

III

Kurzfassung (German)

Die Herstellung von Nanonpartikel-Polymer-Kompositen für die Synthese von funktionalen

Materialien stellt auf Grund von starken attraktiven Wechselwirkungen der Nanopartikel

eine große technologische Herausforderung. In der vorliegenden Arbeit wird eine

Prozesskette für die Darstellung hochgefüllter Nanopartikel-Polymer-Komposite vorgestellt,

wobei näher auf die kolloidalen Wechselwirkungen eingegangen wird. Das zentrale Element

der Prozesskette ist ein, in einem Sprühtrocknungsprozess zu verarbeitendes organisches

Lösungsmittel, welches stabilisierte Nanopartikel und gelöste Polymere beinhaltet. Die

Einzelprozesse der Kette sind: Nanopartikelsynthese in der wässrigen Phase, Transfer der

Nanopartikel in das organische Lösungsmittel in einem Flüssig-Flüssig-Phasentransferprozess

durch amphiphile Substanzen, Herstellung einer Mischung von gelösten Polymeren mit den

transferierten und stabilisierten Partikeln und Sprühtrocknung des Lösungssystems für den

schnellen Entzug des Lösungsmittels. Die Schwerpunkte, um gut dispergierte Komposite zu

erhalten, sind sowohl die Art des adsorbierenden und stabilisierenden Amphiphils, als auch

die komplexen Partikelwechselwirkungen im organischen Lösungsmittel mit gelösten

Polymeren. Die Nanopartikel des Experimentalteils sind super-paramagnetische Magnetit

Nanopartikel mit einer Primärpartikelgröße von etwa 15 nm. Diese werden mit

unterschiedlichen Fettsäuren als Amphiphile (C8 bis C18, gesättigt und ungesättigt) in das

Lösungsmittel Dichlormethan transferiert. Ein physikalisches Modell wird vorgestellt und

numerisch gewertet, welches das Phänomen der Deagglomeration beschreibt, wenn

bestimmte Fettsäuren beim Phasentransfer auf den agglomerierten Nanopartikeln

chemisorbieren. Die Rizinolsäure führt hier zu den stabilsten Nanopartikelsuspensionen. Vier

unterschiedliche, geradlinige, unverzweigte, thermoplastische Polymere, namentlich

Poly(Methyl Methacrylat) PMMA, Poly(Bisphenol A Karbonat) PC, Poly(Vinyl Butyral) PVB

und Poly(Styrol) PS werden, auf Grund ihrer mechanischen Eigenschaften, ihrer weiten

Verbreitung und guten Löslichkeit, als kontinuierliche Phase der Komposite verwendet. Die

Zugabe von PMMA, PC und PS führt zur Abnahme der Primärpartikelkonzentration in den

stabilisierten Nanopartikeldispersionen, unabhängig von der Art der Fettsäure. Die

sprühgetrockneten und spritzvergossenen Komposite weisen große Agglomerate im Bereich

von wenigen Mikrometern auf. Das Polymer PVB führt in den Dispersionen zur Adsorption

auf der Nanopartikeloberfläche und weiterer Stabilisierung der Suspension. Komposite mit

dem Polymer PVB zeigen gut dispergierte Füllstoffpartikel, welche fast komplett

agglomeriert und homogen verteilt vorliegen. Wahrscheinlich sorgen die Hydroxylgruppen in

der Struktur des PVBs dazu, dass Adsorption und Stabilisierung auftritt. Die anderen

Polymere absorbieren nicht und führen möglicherweise auf Grund von Verarmungsflockung

zur Destabilisierung. Es wird gezeigt, dass die entwickelte Prozesskette zu besser

dispergierten Kompositen, im Vergleich zur klassischen Schmelze Einarbeitung führt.

IV

Acknowledgements / Danksagung

Diese Arbeit entstand während meiner Tätigkeit als wissenschaftlicher Mitarbeiter am

Institut für Mechanische Verfahrenstechnik und Aufbereitungstechnik der TU Bergakademie

Freiberg von Oktober 2008 bis März 2012. Das von mir bearbeitete, DFG finanzierte Projekt

„E t i klu g ei e P ozesskette fü die “ these u d Ve a eitu g ho hgefüllte Pol e -

Na opa tikelko posite in Kooperation mit dem Institut für Polymerwerkstoffe und

Ku ststoffte h ik de TU Clausthal liefe te de „“toff und den wissenschaftlichen Spielplatz

für meine Untersuchungen und wissenschaftlichen Erkenntnissen.

Ich bedanke mich bei meinem Betreuer und Doktorvater Herrn Prof. Dr.-Ing. Urs A. Peuker

für das mir übertragene, hoch interessante Projekt. Er hatte in meiner Promotionszeit stets

ein offenes Ohr für Ideen, gab mir weitreichende Freiheiten und übte die in der

Wissenschaft so nötige Kritik. Dankbar bin ich ihm auch sehr für die vielen Möglichkeiten

meine Arbeit in Journalen und auf internationalen Tagungen vorgestellt haben zu dürfen.

Hochachtungsvoll bedanke ich mich bei Herrn Prof. Dr.-Ing. Wolfgang Peukert der

Universität Erlangen-Nürnberg für die Bereitschaft zum Korreferat.

Da die Arbeit schon umfangreich genug ist und auch um niemandem durch Vergessen

ungerecht zu werden, möchte ich es vermeiden nun folgend namentlich Dank zu verstreuen.

Mein besonderer Dank geht an sämtliche Mitarbeiter des Instituts für Mechanische

Verfahrenstechnik und Aufbereitungstechnik auch denen die es mal waren. Vielen Dank

auch an viele weitere Wissenschaftler und unzählige Studenten (insbesondere den Hiwis,

Studienarbeitlern und Diplomanden) der ehrwürdigen TU Bergakademie Freiberg. Für die

spannende Kollaboration danke ich den Mitarbeitern vom PUK der TU Clausthal. Des

Weiteren bedanke ich mich bei vielen Wissenschaftlern und Studenten aus aller Welt für

tolle Arbeiten sowie Diskussionen und Anregungen bei Treffen und Tagungen.

„Eine wissenschaftliche Arbeit kann nur so gut werden wie die Kritik die man an ihr ausübt.“

Der letzte Absatz der Danksagung gehört wie gewohnt den Mitmenschen, die das Leben so

lebenswert machen. Zuallererst bedanke ich mich daher bei meiner Frau Julia für eine

wunderschöne Beziehung und dem faszinierendstem „P ojekt meines Lebens, unseren

wunderbaren Sohn Malte. Meinen Eltern danke ich für viel Freiheit und währende

Unterstützung. Vielen Dank für die Wegbegleitung auch meinem Bruder Thomas und seiner

Frau Franziska, sowie meinen lieben Schwiegereltern. Ebenso wertschätze ich all die vielen

lieben und aufrichtigen Menschen um mich herum.

V

Contents

1 INTRODUCTION 1

2 NANOPARTICLE-POLYMER-COMPOSITES – A STATE OF THE ART 4

2.1 Terminology 5

2.1.1 Nanomaterials 5

2.1.2 Polymers 7

2.1.3 State of Dispersion 8

2.2 Composite Synthesis/Preparation 9

2.2.1 Separate Synthesis of Particles and Polymers 9

2.2.2 Synthesis of Nanoparticles within the Polymer 12

2.2.3 Synthesis of the Polymer with the Presence of Nanoparticles 12

2.3 Industrial Products 13

3 THESIS MOTIVATION – DEVELOPMENT OF A MODULAR PROCESS CHAIN FOR

COMPOSITE PREPARATION 15

3.1 Process Steps 15

3.1.1 Nanoparticle Synthesis 16

3.1.2 Liquid-Liquid Phase-Transfer 16

3.1.3 Polymer Addition 17

3.1.4 Spray Drying 18

3.1.5 Powder Agglomeration 19

3.2 Composition 20

3.2.1 Composite Composition 20

3.2.2 Colloid Composition 23

4 FOCUS ON THE PHASE TRANSFER OF NANOPARTICLES 25

4.1 Theory 25

4.1.1 Phase Transfer 25

4.1.2 Adsorption of Surfactants 26

4.1.3 Steric Stabilization 29

4.2 Physical Model of Deagglomeration at the Interface 32

VI

4.2.1 Gedankenexperiment 32

4.2.2 Geometrical Model 33

4.2.3 Numerical Results 35

4.3 Experimental Results 37

4.3.1 Visualization of the Phase Transfer 37

4.3.2 Particle (Agglomerate) Size Distribution 38

4.3.3 Primary Particle Concentration 39

4.3.4 Inert Decomposition of Chemisorbed Ricinoleic Acid on Magnetite Nanoparticles 42

5 NANOPARTICLES AND POLYMERS IN AN ORGANIC SOLVENT 49

5.1 Theory 50

5.1.1 Polymers in Solution 50

5.1.2 Solubility 55

5.1.3 Phenomena in Nanoparticle Polymer Mixtures 57

5.2 Pair Interactions – DLVO-like Consideration 58

5.3 Experimental Results 62

5.3.1 Influence of the Polymer 63

5.3.2 Influence of the Surfactant 71

5.3.3 Stabilization by Adsorption of PVB 75

5.3.4 Influence of Mechanical Dispersing Methods on the Stability 81

5.3.5 Kinetics of Flocculation at Low Nanoparticle Concentration and high PMMA concentrations 83

5.3.6 Influence of the Solvent 86

6 HIGHLY FILLED COMPOSITES 91

6.1 Theory 91

6.1.1 Spray Drying 91

6.1.2 (Micro) Injection Molding 97

6.1.3 Image Processing and the Mathematical Description of the State of Dispersion 98

6.1.4 Filler Concentration – Agglomerate Concentration – Stability Relation 101

6.2 Experimental Results – Spray-Dried Microparticle Composites 104

6.2.1 Compositional Separation 104

6.2.2 Yield of Product 108

6.2.3 Influence of the Polymer 109

6.2.4 Influence of the Surfactant 110

6.2.5 Increasing Filler Concentration F 113

VII

6.3 Experimental Results – Injection Molded Composites 116

6.3.1 PMMA vs. PC vs. PVB 117

6.3.2 Increasing Filler Concentration – PMMA composite 123

6.3.3 Identity of Agglomerates 125

6.3.4 Solution and Spray Drying Process vs. Melt Mixing 126

7 GENERAL CONCLUSIONS AND OUTLOOK 129

APPENDIX A1

REFERENCES R1

VIII

List of Symbols

A parameter, area

B constant

c concentration, speed of light

C scattering cross section

CH HAMAKER constant

d differential, thickness, diameter

D surfactant ratio, distance, solubility distance

E extinction, energy

f frequency

F filler concentration, force, group contributor

G GIBBS free energy

h specific enthalpy, PLANCK constant

H enthalpy

i imaginary unit

ID interparticle distance

k BOLTZMANN constant, constant,

imaginary part of the complex refractive index K constant

l length

L length

m mass

M molar mass

n number, refractive index

N number, degree of polymerization

NA LOCHSCHMIDT constant (AVOGADRO constant)

p pressure

PI polydispersity index

q size ratio

Q cumulative distribution, heat

R radius, gas constant

R2 coefficient of determination

s distance

S surface area, specific surface area, entropy

IX

SP structural parameter

t time

T absolute temperature

u error

UE electrophoretic mobility

v volume, velocity

V volume

w mass concentration

W energy of interaction

x variable, particle size

X ratio

z thickness

Z atomic number (number of protons)

α angle, constant, heat transfer coefficient

β constant, mass transfer coefficient

surface coverage

length, thickness, solubility distance

difference, concentration of surfactants, group contributor

dielectric constant

zeta potential

dynamic viscosity

θ angle, state

θ temperature

wavelength

mass fraction, ratio

circumference to diameter ratio for a circle, FLORY-HUGGINS parameter

density, specific weight

σ constant, specific surface energy

τ correlation time

φ relative volume concentration

Φ grafting density

φ volume concentration

χ FLORY interaction parameter

X

List of Abbreviations

Abs absorption

AdG ALEXANDER-DE-GENNES theory

AFM atomic force microscopy

ALR air to liquid ratio

AO ASAKURA and OOSAWA theory

ATR attenuated total reflection

BET BRUNAUER-EMMETT-TELLER theory

BSE back scattering electron detection

CA caprylic acid

cov coefficient of variance

DCM dichloromethane, methylene chloride

DFG Deutsche Forschungsgemeinschaft

DLS dynamic light scattering (cf. PCS, QELS)

DLVO DERJAGUIN-LANDAU-VERWEY-OVERBEEK theory

DMT DERJAGUIN-MULLER-TOPOROV model

DTG differential thermal gravimetry

EA ethyl acetate

FA fatty acid

FTIR FOURIER transform infrared spectroscopy

FWHM full width half maximum

HSP HANSEN solubility parameters

IR infrared (spectroscopy)

JKR JOHNSON-KENDALL-ROBERTS model

LA linoleic acid

MA myristic acid

MMA methyl methacrylate

NNLS non-negative least squares

Nu NUSSELT number

OA oleic acid

Oh OHNESORG number

PC poly(bisphenol A carbonate)

pc-AFM phase contrast atomic force microscopy

XI

PCS photon cross correlation (cf. DLS, QELS)

Pe PECLÉT number

pH negative decimal logarithm of the hydrogen ion activity

PM planetary ball mill

PMMA poly(methyl methacrylate)

Pr PRANDTL number

PS poly(styrene)

PSD particle size distribution

PT phase transfer experiment

PTFE poly(tetrafluoro ethylene)

PVA poly(vinyl alcohol)

PVAc poly(vinyl acetate)

PVB poly(vinyl butyral)

QELS quasi elastic light scattering (cf. DLS, PCS)

RA ricinoleic acid

Re REYNOLDS number

Sc SCHMIDT number

SE secondary electron detection

SEM scanning electron microscopy

Sh SHERWOOD number

ST spray drying experiment

TEM transmission electron microscopy

TEOS tetraethyl orthosilicate

TGA thermogravimetric analysis

US ultrasound, ultrasonication

UT Ultra-Turrax®

UV/VIS ultraviolet and visual light spectrometry

vdW VAN DER WAALS

We WEBER number

XRD x-ray diffraction

YSZ yttrium stabilized zirconium

XII

List of Indices

0 original, pristine, contact, unit, number weighted

3 volume weighted

50 median value of a distribution

A adsorption, area

abs absolute

c contour

C=C carbon-carbon double bond

C-C carbon-carbon single bond

d dispersive

ext extinction

g glass transition

G gyration

h hydrogen bonding

H enthalpic

i parameter, fraction

int intensity weighted

m molar, average

M mixing, MARK-HOUWINK-SAKURADA

max maximum

min minimum

n by number, number specific

p polar

rel relative

s segment

S entropic

scat scattering

sp specific

t total, HILDEBRANDT

w by weight, weight specific

wb wet bulb

Θ theta state

XIII

List of figures

figure 1: Schemes of the three main types of nanomaterials (from left to right) nanotubes with diameters smaller 100 nm, nanoplatelets with thickness smaller 100 nm and nanoparticles with diameters smaller 100 nm ....................................................... 5

figure 2: (left) Model geometry for the calculation of the relative number of atoms/clusters at the surface Nsurface/Ntotal, cf. eq. (72), (right) evaluation of eq. (72) for gold Au and magnetite Fe3O4 with given atom/cluster sizes Rcluster ..................................... 6

figure 3: Definition of the state of dispersion (poorly dispersed in the upper left, well dispersed in the lower right) by both a good distribution (more homogeneous from left to right) as well as a thorough deagglomeration (more successful from top to bottom) ......................................................................................................... 8

figure 4: Scheme of composite formation by mixing a dry nanoparticle powder with a polymer melt, the mixer represents any type of high shear mixing device such as an extruder ............................................................................................................ 10

figure 5: Scheme of composite formation by mixing a nanoparticle solvent based colloid with a polymer solution, the nanoparticles can be well dispersed e.g. by adsorption of end-grafted molecules .......................................................................................... 10

figure 6: Scheme of composite formation by compounding a nanoparticle and a polymer powder with high forces in a mill, typically a high energy ball mill ...................... 11

figure 7: Scheme of composite formation by attraction of oppositely charged particles and polymers in water and withdrawing the water for composite synthesis ............. 11

figure 8: Scheme of composite formation by mixing a pre-courser (symbolized as small dots in the left frame) with a polymer in solution and initiation of nanoparticle synthesis ................................................................................................................ 12

figure 9: Scheme of composite formation by dispersing nanoparticles in a monomer solution (on the left represented as an emulsion droplet) and subsequent polymerization and withdrawing of all solvents ............................................................................ 13

figure 10: Synthesis of the nanoparticles by a precipitation reaction in an aqueous environment .......................................................................................................... 16

figure 11: Transfer of the nanoparticles from the aqueous to an organic phase (immiscible solvent) by adsorption of amphiphilic molecules ................................................. 16

figure 12: Mixing the organic solvent based hydrophobic nanoparticles with a polymer solution in the same solvent ................................................................................. 17

figure 13: Spray Dryer with a two-fluid nozzle in the co-current regime. A small fraction of large particles falls through the drying cylinder by gravity separation, the largest fraction of particles by mass is the product collected at the coarse output of a cyclone by centrifugal force separation. ............................................................... 18

figure 14: Press agglomeration of the spray dried fine composite micropowder and granulation in a rotor-mill for an improved powder handling .............................. 19

figure 15: Surfactant ratio D, cf. eq. (4) with Msurfactant = 282.46 g/mol, Ssurfactant = 2.4·10-19 m2

[95] and nanoparticle = 5.2·106 g/m3 for magnetite Fe3O4 ....................................... 21

figure 16: Interconnection of the composition values: filler concentration F, surfactant ratio D a d olu e o e t atio of the pol e φpolymer for the given specific weights

XIV

of the three components using eq. (7), volume concentrations have to be larger than 0.26 which is the limit for closest packing of spheres .................................. 22

figure 17: Interparticle distance as a function of the filler concentration for monodisperse and homogeneously distributed fillers of different size (D = 0.2 and specific weights as mentioned in figure 16). The inset shows the geometry for the body-centered cubic space. ............................................................................................ 22

figure 18: Interconnection of the polymer concentration in the organic solvent based mixture cpolymer, the solids concentration csolid and the filler concentration F in the solvent free composite material, cf. eq. (11). ....................................................... 24

figure 19: Principle structure of a fatty acid molecule with a hydrophilic carboxyl head group and a hydrophobic alkyl chain with certain length and functionalities (double bonds, functional groups). .................................................................................... 26

figure 20: Principle of Adsorption of Fatty Acids on Magnetite: (a) monolayer covered particle, (b) before adsorption in aqueous solution (mediating water layer in blue) with high pH resulting in negatively charged magnetite surface and disassociated carboxyl group [99], (c) chelating bidentate bond between fatty acid and magnetite [95] and (d) monodentate mononuclear configuration [116] ............................................................................................................................... 29

figure 21: Geometries of two types of end-grafted molecule covered interacting particle surfaces at distance D (left) brush type as expressed in eq. (17), (right) mushroom type in eq. (18) .................................................................................... 30

figure 22: Schematics representing the gedankenexperiment of an agglomerated nanoparticle doublet passing the liquid-liquid interface where fatty acid molecules adsorb and push the particles apart by a disjoining force when tails of opposing end-grafted molecules overlap [23] ...................................................... 32

figure 23: Representation of the geometrical model of particles with radii R in contact, covered by a layer of molecules with the thickness . The defined region of interest, which attributes to the repulsive force is highlighted. Calculations are based on the distance of overlap of the layers D which are a function of the angle α. At αmin D = and at αmax D = 0. [23] .................................................................. 33

figure 24: Angle dependent pressure between the spheres with radii of 7.5 nm and the pa a ete la e thi k ess i steps of . nm, from 0.4 nm to 2.4 nm, for a rather high degree of adsorption of s = 0.5 nm according to eq. (23), [23] ......... 35

figure 25: Repulsive forces according to eq. (24) (left) as a function of the adsorption dista e s a d ith the pa a ete of la e thi k ess , ight as a function of the la e thi k ess ith the pa a ete adso ptio dista e s i o pa iso to the constant absolute VAN DER WAALS force (horizontal line, cf. eq. (26)), [23] ........... 36

figure 26: relative repulsive force (cf. eq. (27)) as a function of adsorption distance s and la e thi k ess , [ ] ............................................................................................ 36

figure 27: Vials of phase transfers of magnetite nanoparticles from an aqueous phase (upper half) to a DCM phase (below liquid interface) with the grafting molecules/surfactants as defined in the image above the vials all at pH 9 in the aqueous phase except for myristic acid with a second phase transfer at pH 8 [23] ............................................................................................................................... 38

figure 28: Intensity weighted particle size distributions of the samples in figure 27 applying analytical centrifugation with a cut-off size of 30 nm, additionally the primary particles distribution is determined from the stable colloid using DLS [23] ......... 39

XV

figure 29: Mass concentration of primary particles after phase transfer derived from colloidal interactions studied in 6.3.2, values presented in table 8, [23]............................ 40

figure 30: Solubility plot of the fatty acids pristine (FA - filled blue symbols) and grafted to the magnetite surface (FA-Fe3O4 - open blue symbols) calculated using the group contribution method in A.11, compared to the solvent DCM (red circle), values from [140] .............................................................................................................. 41

figure 31: Correlation of primary particle concentration given in figure 29 and the solubility distance between the fatty acid capped magnetite and DCM calculated using eq. (90) and values found in A.11.......................................................................... 42

figure 32: ATR-FTIR results of pristine ricinoleic acid, pristine magnetite as well as ricinoleic acid adsorbed on magnetite [21] .......................................................................... 43

figure 33: (left) first derivative of the TGA results (DTG) of pure ricinoleic acid (RA), ricinoleic acid coated magnetite (RA-Fe3O4) and Aerosil® 200 (RA-SiO2) in inert atmosphere [21], (right) TGA of RA-Fe3O4 with the mass losses at the three distinct steps, the error bars show the 95 % quantile of three measurements ................................. 44

figure 34: FTIR of the evolving gases for the major steps of decomposition of the three samples in figure 33, (left) RA including the ATR-FTIR spectrum at room temperature in the top graph, (middle) RA-Fe3O4 and (right) RA-SiO2 [21] ......... 45

figure 35: Powder diffractograms of pristine precipitated and fatty acid grafted magnetite RA-Fe3O4 and of the mixed iron oxide residue FeOx after inert gas TGA with ide tified ajo diff a tio a gles Θ of ag etite Fe3O4, wüstite FeO, Ferrite α-Fe and Hematite Fe2O3. [21] .............................................................................. 47

figure 36: Polymer coils in solution for different polymer solvent interactions (left) real chain in a good solvent, (middle) ideal chain with equal interaction of polymer segments and solvent molecules, (right) real chain in a bad solvent. The symbols accou t fo : χ – the FLORY interaction parameter, RG – the radius of gyration, ls – the segment length, N – the number of monomer units and TΘ – the theta temperature .......................................................................................................... 52

figure 37: Graphical visualization of the radius of gyration RG normalized with the segment length ls, depending on the FLORY i te a tio pa a ete χ a d the u e of segments NS, cf. eq. (40) ....................................................................................... 53

figure 38: Concentration regimes of polymers in solution (left) dilute solution, (middle) solution at the overlap concentration c*polymer and (right) semi dilute regime in accordance with [156] ........................................................................................... 54

figure 39: Principle types of colloidal regimes between neutral nanoparticles and neutral polymers in an organic solvent (a) depletion flocculation, (b) depletion stabilization, (c) bridging flocculation, (d) steric stabilization, cf. [118, 126, 127, 134, 155, 160] ........................................................................................................ 57

figure 40: Scheme of two interacting particles of diameter 15 nm surrounded by grafted molecules with a length of 2.0 nm in a solution of polymer coils with a diameter of 7.5 nm which do not adsorb at the particle surface resulting in a depletion layer surrounding the particles (dashed circles) [23] ............................................ 59

figure 41: DLVO-like addition of colloidal interactions with regard to VAN DER WAALS and depletion attraction (by ASAKURA and OOSAWA AO) as well as BORN and steric-osmotic repulsion with R = 7.5 nm, RG = 4.0 nm, s = 0.6 nm, = 1.5 and φ = 1.0, [23] ........................................................................................................................ 61

figure 42: Distance dependent total interaction, following eq. , left fo a i g [ ] (right) for varying φ and constant parameters mentioned in figure 41. .............. 61

XVI

figure 43: Light extinction at 600 nm of diluted samples E600 nm and gravimetrically determined primary particle concentration wPrimary as a function of the PMMA concentration at constant nanoparticle concentrations; the inset is a photograph displaying the samples (b) only holding the primary particles after centrifugation with increasing polymer concentration from left to right .................................... 63

figure 44: Light extinction at 600 nm of diluted samples E600 nm and gravimetrically determined primary particle concentration wPrimary as a function of the PC concentration at constant nanoparticle concentration; the inset is a photograph displaying the samples (b) only holding the primary particles after centrifugation with increasing polymer concentration from left to right .................................... 64

figure 45: Representation of the data in figure 43 and figure 44 as a function of: (left) the elati e pol e o e t atio φ usi g the o e lap o e t atio s dete i ed

with the intrinsic viscosity in A.9.1 and (right) the number concentration of polymer coils cpolymer/Mn. ...................................................................................... 65

figure 46: Light extinction at 600 nm of diluted samples E600 nm and gravimetrically determined primary particle concentration wPrimary as a function of the PVB concentration at constant nanoparticle concentration; the inset is a photograph displaying the samples (b) only holding the primary particles after centrifugation with increasing polymer concentration from left to right .................................... 66

figure 47: Correlation of the photometric extinction E600 nm and the primary particle concentration wPrimary as well as the total particle volume concentration for the extinctio easu e e t φ, t o diffe e t slopes fo the desta ilizatio ith PMMA and PC and the stabilization with PVB, w0

Primary = 91.7% located at the intersection of the linear models (dashed line) [20] ............................................. 67

figure 48: Photometric primary particle investigation as a function of the polymer concentration for both destabilizing polymers PMMA and PC as well as for the stabilizing PVB and three different phase transfer batches: PT101104, PT100902II and PT100305 ........................................................................................................ 69

figure 49: Normalized extinction curves for the three phase transfer batches as a function of (left) PMMA as well as (right) PC, the fitted lines follows the mathematical model in eq. (61) ............................................................................................................... 69

figure 50: Primary particle concentrations of the five investigated samples in nanoparticle polymer mixtures with different concentrations of the polymer PMMA in DCM. The solid lines present mathematical fits of eq. (61), published in [23] ............... 72

figure 51: Primary particle maps by combining results of the investigations in figure 50 and table 8 with eq. (62) for ricinoleic (RA), linoleic (LA) and oleic acid (OA) with the same scale in the three dimensions ...................................................................... 73

figure 52: Primary particle concentrations of the five investigated samples in nanoparticle polymer mixtures with different concentrations of the polymer PVB in DCM; presented in context (scaling) with figure 50 above, published in [23]................ 74

figure 53: The median diameter of the number weighted particle size distribution x50,0 of diluted samples (b) measured with DLS as a function of the pristine concentrations cpolymer of PMMA, PC and PVB in the samples (a) ........................ 76

figure 54: Numerically obtained volume weighted particle size distribution from the DLS measurements of the (b) samples with the polymers (left) PMMA and (right) PVB; the polymer concentrations cPMMA and cPVB refer to the undiluted samples (a) ........................................................................................................................... 76

XVII

figure 55: (left) PVB adsorption on the RA-Fe3O4 ag etite a opa ti les su fa e as a function of the PVB concentration in the mixture cPVB, (right) layer thickness of the adsorbed PVB layer, calculated using the data in figure 53 ........................... 77

figure 56: Density of the adsorbed PVB layer as a function of the PVB solution concentration cPVB with the fitted data from figure 55 using eq. (65) and a nanoparticle diameter xnanoparticle of 15 nm and a nanoparticle specific weight of 5200 g/l ...... 78

figure 57: Relative mass loss in the third degradation step for inert gas TGA between 600°C and 900 °C for PMMA and PVB and Ra-Fe3O4 as a function of the polymer concentration ........................................................................................................ 79

figure 58: Colloidal stability of PVB mixed with RA-Fe3O4 nanoparticles at four different mass ratios in solutions of PMMA with the concentration cPMMA .................................. 80

figure 59: Scheme of the hypothetical model of adsorption of PVB on the surface of magnetite nanoparticles carrying chemically grafted fatty acid surfactant molecules as well; (left) train adsorption of PVB at vacant magnetite surface sites, (right) surface hydroxide groups shall interact with the hydroxyl groups of the PVB backbone. ................................................................................................ 80

figure 60: Primary Particle Concentration as determined gravimetrically (TGA) and by light extinction at 600 nm (UV/VIS) with eq. (59) for a DCM based mixture of RA-Fe3O4 and PC with D = 0.2 and cpolymer = 51.2 g/l (F = 0.3) for four different mixing procedures (US – sonotrode ultrasonication, UT – Ultra Turrax) ......................... 81

figure 61: Destabilization curve of RA-Fe3O4 (PT091227II) in DCM with PMMA, without and with 1 min sonotrode ultrasonication (US) ........................................................... 82

figure 62: Dispersing a phase transfer batch without polymer in DCM using a planetary ball mill, (left) intensity weighted particle ................................................................... 83

figure 63: Time development of the intensity weighted frequency particle size distribution (in logarithmic scaling) determined with DLS with respect to the viscosity of the polymer solution for a magnetite dispersion without pristine agglomerates at cnanoparticles = 1.2 g/l and cpolymer = 58 g/l; (left) destabilizing PMMA (right) stabilizing PVB ....................................................................................................... 84

figure 64: Time dependent light extinction to monitor agglomeration of a RA-Fe3O4 dispersion in DCM with dissolved PMMA, the polymer is mixed with the stable nanoparticle dispersion at t = 0 min ..................................................................... 85

figure 65: Correlation of extinction rates (set-in and maximum slopes of the lines in figure 64) with the PMMA concentration ....................................................................... 85

figure 66: Primary particle concentration wPrimary (left) and normalized photometric extinction E600nm/E0,600nm (right) as function of the PMMA concentration for the solvents: DCM and EA ........................................................................................... 87

figure 67: 100 % completed phase transfers of magnetite nanoparticles originating from the same precipitation batch (a) to dichloromethane DCM (PTDCM120227) by gravity driven transport and stirred emulsification in a beaker, (b) to MMA (PTMMA120227) and (c) to styrene ST (PTST120227) by mixer-settler extraction in separation funnels, (left) strong emulsion formation for MMA and ST, (right) after breaking the emulsion by reducing the pH to 6.0 with 10 ml 1N HCl .......... 88

figure 68: Extinction based determination of the primary particle concentration of RA-Fe3O4 in DCM (PTDCM120227) and MMA (PTMMA120227) under the presence of PMMA with the polymer concentration on the abscissa; (left) absolute (right) normalized values; the extinction ratio for the polymer free point of MMA to DCM is 8.6 % .......................................................................................................... 89

XVIII

figure 69: Extinction based determination of the primary particle concentration of RA-Fe3O4 in DCM (PTDCM120227) and ST (PTST120227) under the presence of PS with the polymer concentration on the abscissa; (left) absolute (right) normalized values; the extinction ratio for the polymer free point of ST to DCM is 80.5 % ............... 90

figure 70: Principle scheme of an external mixing two fluid nozzle with turbulent atomization, adopted from [214] .......................................................................... 92

figure 71: Time progression of mass and temperature of a drying droplet with constant surface area due to film formation, taken from [225] with the steps A through D described in the text .............................................................................................. 95

figure 72: Drying progression of a droplet with suspended small particles and shell formation, taken from [230] ................................................................................. 96

figure 73: (left) principal scheme of an injection molding device (right) top view of an optical micrograph of a PMMA-RA-Fe3O4 composite with F = 0.3 and D = 0.2 of a test structure ................................................................................................................ 98

figure 74: Image processing steps (from left to right) of a phase contrast AFM image on the left showing dark magnetite and light polymer phases, binarized after thresholding and with a watershed, automatic detection and measurement of >800 individual particles (including aggregates and agglomerates) and finally the VORONOI diagram of the binarized image .............................................................. 99

figure 75: Coefficients of variance normalized with the first value covmax for the VORNOI and the line method and a set of simulated images with (from left to right) improving state of dispersion taken from fig. 7 in [55], covmax is 1.08 for the VORONOI method and 1.42 for the line method ................................................................. 101

figure 76: Visualization of the volume (area) fractions of the primary and agglomerated particles as well as the total volume fraction of nanoparticles applying eqs. (77) - (80), parameters w0

Primary and A are chosen in connection with colloidal stability investigations of PMMA and PC in 6.3.1 ............................................................. 103

figure 77: Relative agglomerate concentration as a function of the total filler concentration F and the polymer concentration cpolymer with the parameters of colloidal stability of eq. (80) applying eqs. (77) - (80) ..................................................................... 103

figure 78: Schematic set-up of the co-current spray dryer used in this thesis with three particle fractions (cylinder, cyclone and filter) where the cyclone fraction is the product of the process and should be the largest fraction in mass, the solvent is recovered in a condenser and the dry air recycled passing a heating device .... 104

figure 79: (left) representative particle size distributions of the microparticles of the three fractions of the spray dryer measured with laser diffraction and (right) TGA curves of the corresponding samples with an initial composition of PMMA RA-Fe3O4 with F = 0.3 and DRA = 0.2; the yields are 11 %, 74 % and 15 % for cylinder, cyclone and filter, respectively ............................................................................ 105

figure 80: (left) TGA measured mass residue at 600 °C of various spray drying experiments with RA-Fe3O4 and DRA = 0.2 for the three fractions of the spray dryer at different initial filler concentrations of magnetite F; (right) relative magnetite concentration of the cylinder or the filter fractions compared to the corresponding cyclone fraction with the cyclone fraction residual mass on the abscissa ................................................................................................................ 106

figure 81: High temperature mass loss (residual mass at 600 °C compared to residual mass after magnetite reduction at 900 °C, compare with investigations in 5.3.4) of the TGA analyzed particles of various spray drying experiments with RA-Fe3O4 based

XIX

on 38, 51 and 23 samples for cylinder, cyclone and filter materials, respectively ............................................................................................................................. 107

figure 82: BSE-SEM images of the particles of the cyclone fraction of a PMMA-based composite with RA-Fe3O4 DRA = 0.2 and F = 0.3, the image on the right is a close-up of a region in the approximate center of the image on the left, magnification is 3,000x ............................................................................................................... 108

figure 83: Yield of product at the coarse exit of the cyclone as a function of the filler concentration F for PMMA-based composites with RA-Fe3O4 and DRA = 0.2 .... 109

figure 84: Inverted BSE-SEM images of composites with RA-Fe3O4 at F = 0.3 and DRA = 0.2 for the destabilizing polymers PMMA and PC as well as the stabilizing polymer PVB [27] ...................................................................................................................... 110

figure 85: BSE-SEM images of spray-dried composites of (from left to right) RA, LA, OA, MA and CA coated magnetite (appearing light for high back scatter electron densities of iron atoms) in PMMA, magnification of 2000x and 10000x in the upper and lower row, respectively [23] ................................................................................ 111

figure 86: BSE-SEM images of spray-dried composites of (from left to right) RA, LA, OA, MA and CA coated magnetite (appearing light for high back scatter electron densities of iron) in PVB, magnification of 2000x and 10000x in the upper and lower row, respectively; to be compared to the images in figure 85 [23] ............................ 112

figure 87: Inverted BSE-SEM images of individual spray-dried microparticles with similar size from the cyclone fraction with RA-Fe3O4 at DRA = 0.2 at (from left to right) F = (0.3, 0.5, 0.8) with comparable contrast to visualize the impact of the increasing magnetite concentration, magnification 20,000x ............................. 114

figure 88: Volume weighted PSD of the three spray dryer fractions at the cylinder, cyclone and filter for the PMMA-based composites of RA-Fe3O4 with DRA = 0.2 and the filler concentrations F = (0.3, 0.5, 0.8) ................................................................ 114

figure 89: Bright field optical microscopy images, (from left to right) PMMA, PC and PVB for (top row) F = 0.3 and (bottom row) F = 0.5, lens magnification: 20x ................. 117

figure 90: Inverted BSE-SEM images (from left to right) PMMA, PC and PVB for (top row) F = 0.3 and (bottom row) F = 0.5, magnification: 5,000x .................................... 120

figure 91: Inverted BSE-SEM images (from left to right) PMMA, PC and PVB for (top row) F = 0.3 and (bottom row) F = 0.5, magnification: 50,000x .................................. 122

figure 92: (top row) inverted BSE-SEM images of PMMA-based RA-Fe3O4 composites with DRA = 0.2 and (from left to right) F = (0.3, 0.4, 0.5, 0.6), (bottom row) binary image of the SEM images above for particle detection, magnification: 2,000x . 123

figure 93: Inverted BSE-SEM images of PMMA-RA-Fe3O4 composites with DRA = 0.2 and (from left to right) F = (0.3, 0.4, 0.5, 0.6), magnification: 24,000x................................ 124

figure 94: Visualization of agglomerates of flocculated magnetite nanoparticles (a) under optical microscope of a flocculated suspension, (b) as dark spots in composite microparticles and (c) in a cross-section of an injection molded composite material with inverted BSE-SEM; al images with the same scaling; the diagram depicts the size distribution of the agglomerates in the suspension compared to the cross-section showing very similar sizes [27] ............................................... 126

figure 95: Comparison of Inverted BSE-SEM of cross-sections of (left) a melt compounded sample of RA-Fe3O4 compounded with PMMA and (right) one that was spray-dried and injection molded with PMMA and RA-Fe3O4; for both samples F = 0.3 and DRA = 0.2 ........................................................................................................ 127

XX

figure 96: Chemical structures of the fatty acids: (a) caprylic acid CA, (b) myristic acid MA, (c) oleic acid OA, (d) linoleic acid LA, (e) ricinoleic acid RA ........................................ A4

figure 97: (left) Thermal decomposition in N2 atmosphere with a heating rate of 20 K/min (from left to right, red to blue) PMMA, PVB, PS, PC; (right) contact angle with water on a thin film using sessile drop analysis .................................................... A6

figure 98: Chemical structure of the repeating unit of poly(methyl methacrylate) ................ A6

figure 99: Chemical structure of the repeating unit of poly(bisphenol A carbonate) [247] .... A7

figure 100: Chemical structure of the three repeating units of poly(vinyl butyral) with n butyral, m alcohol and p acetate subunits [247] .................................................. A8

figure 101: Chemical structure of the repeating unit of poly(styrene) .................................... A9

figure 102: Chemical structure of dichloromethane .............................................................. A11

figure 103: Chemical structure of ethyl acetate .................................................................... A11

figure 104: Chemical structure of methyl methacrylate ........................................................ A11

figure 105: Chemical structure of styrene ............................................................................. A12

figure 106: Comparison of a micrograph of a fractured composite surface of RA-Fe3O4 in PMMA with F = 0.3 and DRA = 0.2 (left) with detection of the secondary electrons and (right) when detecting the back scattered electrons with much better phase contrast (the lighter sections are due to strong back scattering of the heavy iron atoms in the magnetite nanoparticles) ............................................................... A15

Figure 107: pc-AFM image of a PMMA-RA-Fe3O4 composite with F = 0.3 ............................ A17

figure 108: (left) powder diffractogram (XRD) of the washed and dried co-precipitated magnetite, all peaks correspond to the magnetite crystal system, the four major peaks are used for calculation of the crystallite size (right) using the Williamson-hall plot [260] ...................................................................................................... A20

figure 109: (left) transmission electron micrograph of the precipitated magnetite nanoparticles in a PMMA matrix with F = 0.3, (right) particle size distribution (number frequency) as obtained from image analysis of the TEM image .......... A20

figure 110: (top) Intensity weighted cumulative particle size distribution of phase transferred magnetite particles as determined with the cuvette centrifuge and (bottom) volume weighted particle size distribution of the supernatant after centrifugation without polymer as measured with DLS compared to the TEM investigation of encapsulated magnetite nanoparticles (inset) [20] ............................................ A21

figure 111: Particle size distributions of precipitated magnetite in water with full ion strength after the reaction with given zeta-potential (blue line measured with analytic centrifugation) and after washing reducing the ion concentration increasing the absolute zeta-potential (green line measured with DLS) .................................... A22

figure 112: Steps of a gravity driven phase transfer using a organic solvent which is heavier than water, e.g. DCM .......................................................................................... A23

figure 113: Three representative sets of TGA analyses of the composition of the solids in the pristine dispersion - samples (a) and in the supernatant - samples (b) to determine the primary particle concentration wPrimary (numbers explained in table 25) [20] ....................................................................................................... A27

figure 114: Configuration of the lab-scale spray dryer with inert gas flow (left) photograph after spray drying a PMMA-based composite with RA-Fe3O4 and F = 0.3, (right) schematic drawing with: a) external mixing two fluid nozzle, b) ventilator for drying gas circulation, c) heater, d) condenser; c)-d) cannot be seen in the photograph on the left ........................................................................................ A33

XXI

figure 115: (left) dielectric properties of magnetite, data extracted from [271] as a function of the e e g E, hi h is ight al ulated fo a ele gths ........................... A35

figure 116: real and imaginary part of the refractive index of magnetite as calculated from the data in figure 115 using eq. (90) ................................................................... A36

figure 117: Extinction, scattering and absorption cross-sections Cext, Cscat and Cabs of magnetite nanoparticles with a diameter of 15 nm and optical properties defined in figure 116 in dichloromethane with a refractive index of 1.4242 as a function of the wavelength, notice that scattering and absorption values are 100- and 0.5-fold, i.e. extinction is mainly due to absorption.................................................. A37

figure 118: dynamic viscosities of the polymers of this thesis in DCM as well as PMMA in EA as a function of the polymer concentration, determined with UBBELOHDE viscosimetry ......................................................................................................... A38

figure 119: Reduced viscosity over polymer concentration for various polymers in DCM and PMMA in EA. The lines show the linear fit to evaluate the intrinsic viscosity following eq. (46) ................................................................................................. A38

figure 120: (left) correlation coefficient and (right) frequency distribution intensity weighted of one single DLS experiment for ricinoleic acid transferred particles containing agglomerates at time steps t1 through t4 which are 2 minutes apart each; additionally the result for the sample after centrifugation containing only primary particles.................................................................................................. A41

XXII

List of tables

table 1: A selection of industrial polymer-nanoparticle composites which are already on the market ................................................................................................................... 14

table 3: median particle size x50,int, fraction of primary particles from the intensity weighted distribution Qint, primary particle concentration w0

Primary, calculated solubility distance between the fatty acid coated magnetite (FA-Fe3O4) and the solvent (DCM) DFA-Fe3O4 – DCM using eq. (28) and A.11 as well as the calculated FLORY-HUGGINS parameter χ using eq. (52) ....................................................................... 41

table 4: Factors given in equation (29) with errors representing 1.96 times standard deviation, the resulting oxygen to iron ratio in the iron oxide after reduction is calculated to be 0.76 ± 0.09, compared to the value in pristine magnetite which is 1.33. ....................................................................................................................... 46

table 5: Compositional minerals analysis of the FeOx residue, which is made up of four identified mineral structures. The mass concentrations are calculated from the diffractogram using the XRD device software and the molar concentrations follo f o φ= i/Mi ⁄∑ i/Mi), the resulting iron to oxygen ratio is 0.99 ......... 48

table 6: Impact of increasing parameters (the other parameters shall be constant) of the total DLVO-like interaction on the colloidal stability as a consequence of a changing absolute attraction energy |Wtotal,min| and distance Dtotal,min ↑↑↑/↓↓↓ p og essi el , ↑↑/↓↓ li ea l , ↑/↓ deg essi el

increasing/decreasing), starting values for the parameters as good approximates [23] ........................................................................................................................ 62

table 7: Results for the mathematical fit parameters using eq. (61) of each batch for both polymers as well as for the three batches combined; estimation of the initial primary particle concentrations w0

Primary of batches PT101104 and PT100921 using the initial extinction values E0

600 nm and the given w0Primary of batch

PT100305 (see figure 47) ....................................................................................... 70

table 8: Results of the first order exponential decay fit parameters A and w0Primary for all five

samples investigated ............................................................................................. 72

table 9: Granulometric data: median microparticle size, specific surface area calculated from the particle size distribution, BET surface and structural parameter SP of the spray-dried samples with PMMA as the matrix polymer [23] ............................ 111

table 10: Granulometric data: median microparticle size, specific surface area calculated from the particle size distribution, BET surface and structural parameter SP of the spray-dried samples with PVB as the matrix polymer [23] ................................. 113

table 11: Specific Weight and dynamic viscosity (cf. A.9) of the spray-dried dispersion and granulometric data of the cyclone particle fraction for different filler concentrations of PMMA-based RA-Fe3O4 composites with DRA = 0.2, calculated spe ifi eight , edia pa ti le size , , spe ifi su fa e a eas al ulated with the PSD SPSD (cf. eq. (76)) and measured with BET SBET, SAUTER diameter x3,2 calculated with PSD and structural parameter SP (cf. eq. (76)) .......................... 115

table 12: Summary of data obtained from the binarized images of figure 89, coefficient of variance of the VORONOI polygons covVORONOI (cf. 6.1.3), median size of the number weighted distribution of FERET diameters xFERET 50,0, area fraction of the detected

XXIII

agglo e ates φAgglomerates (related to the overall volume / area fraction of the filler nanoparticles in parentheses), agglomerated particle concentration from the primary particle concentration reported in 5.3.1 with wAgglomerates = 1 - wPrimary and number of agglomerates detected .............................................................. 118

table 13: Summary of data obtained from the binarized images of figure 90, coefficient of variance of the VORONOI polygons covVORONOI (cf. 6.1.3), median size of the number weighted distribution of FERET diameters xFERET 50,0, area fraction of the detected agglo e ates φAgglomerates (related to the overall volume / area fraction of the filler nanoparticles in parentheses), agglomerated particle concentration from the primary particle concentration reported in 5.3.1 with wAgglomerates = 1 - wPrimary and number of agglomerates detected .............................................................. 121

table 14: Summary of data obtained from the binarized images in figure 92, coefficient of variance of the VORONOI polygons covVORONOI (cf. 6.1.3), median size of the number weighted distribution of FERET diameters xFERET 50,0, area fraction of the detected agglo e ates φAgglomerates (related to the overall volume / area fraction of the filler nanoparticles in parentheses), agglomerated particle concentration from the primary particle concentration reported in 5.3.1 with wAgglomerates = 1 - wPrimary and number of agglomerates detected .............................................................. 124

table 15: Summary of data obtained from the binarized images in figure 95, coefficient of variance of the VORONOI polygons covVORONOI (cf. 6.1.3), median size of the number weighted distribution of FERET diameters xFERET 50,0, area fraction of the detected agglo e ates φAgglomerates (related to the overall volume / area fraction of the filler nanoparticles), and number of agglomerates detected ............................. 127

table 16: Chemicals used for magnetite nanoparticle synthesis ............................................. A3

table 17: List of fatty acids used for the experiments in this thesis ........................................ A3

table 18: General physical properties of the polymers used in this thesis[154, 247] ............. A5

table 19: HSP values of the polymers used in this thesis ......................................................... A6

table 20: Polymer chain properties of the PMMA batch Diakon CLG 902............................... A7

table 21: Polymer chain properties of the PC batch Makrolon 2407 ...................................... A8

table 22: Polymer chain properties of the PVB batch used, including the composition of the three functional units displayed in figure 100 ...................................................... A9

table 23: List of solvents used in the experiments of this thesis ........................................... A10

table 24: General physical properties of the solvents ........................................................... A10

table 25: Concentrations cpolymer and cnanoparticles of the investigated dispersions of polymers, nanoparticles in DCM as an organic solvent with a surfactant (fatty acid) to nanoparticle mass ratio D of 0.2, as well as the resulting filler concentration F of the particles in a composite synthesized with the dispersion withdrawing the solvent assuming specific weights of 5.2 g/cm3 and 1.2 g/cm3 for the magnetite nanoparticles and the polymer as well as the surfactant layer, respectively. .... A25

table 26: Important setting parameters of the lab scale spray dryer .................................... A33

ta le : E aluatio of the i t i si is osities [η], the o e lap o e t atio *polymer using

eq. (42), the radius of gyration RG using eq. (46), the hydrodynamic radius determined with DLS and the ratio of hydrodynamic radius to radius of gyration X using eq. (41) ....................................................................................................... A39

table 28: Polymer solubility in DCM for PMMA, PC and PVB with HSP values from table 19 and table 24, solubility distance D1,2 using eq. (50), FLORY interaction parameter χ using eq. (52) ....................................................................................................... A39

XXIV

table 29: Solubility of PMMA in the solvents dichloromethane, ethyl acetate and methyl methacrylate with HSP values from table 19 and table 24, solubility distance D1,2 using eq. (50), FLORY interaction parameter χ using eq. (52) .............................. A40

table 30: Solubility of PS in dichloromethane and styrene with HSP values from table 19 and table 24, solubility distance D1,2 using eq. (50), FLORY interaction parameter χ using eq. (52) ....................................................................................................... A40

table 31: Molar group parameters of the group contribution method of the groups relevant for the structure of fatty acids ............................................................................ A44

table 32: Number of specific groups in the fatty acids used ................................................. A44

table 33: Number of specific groups in the grafted fatty acids, neglecting influence from the chemically bound complex .................................................................................. A45

table 34: HSP of FA calculated with the group contribution method and the number of groups from table 32 and solubility distance in DCM DFA-Fe3O4-DCM with eq. (50) and HSP for DCM in A.1.4 .................................................................................... A45

table 36: solubility distances of fatty acid caped Fe3O4 in different solvents using eq. (50) as well as FLORY interaction parameter using eq. (52) and the HSP of the solvents reported in A.1.4 and of the FA-Fe3O4 listed above ............................................ A46

1

1 Introduction

For the stability and growth of modern economies it is a crucial fact that the research and

development in new materials plays a very important role [1]. Especially the scientific and

technological advances in the field of nanoparticles, which manifest manifold physical

properties of various types offer new perspectives for material applications in merely any

field of industry. Classic polymer products can be turned into high performance functional

nanoparticle-polymer-composites by incorporation of nanoparticles [2-13]. When preparing

these composites one has to take care of the special interfacial phenomena and resulting

particle interactions to prevent undesired nanoparticle clustering to agglomerates and

aggregates [14, 15]. To achieve a high state of dispersion is the main task when wanting to

prepare nanoparticle-polymer-composites. This means a homogeneous distribution of

primary nanoparticles within the continuous polymer matrix is desired. There are numerous

techniques and strong research efforts to meet this demand.

In this doctoral thesis one technique of nanoparticle-polymer-composite preparation is

focused on in detail. It is the process chain which is based on solution blending of stabilized

hydrophobic nanoparticles with dissolved polymers and spray drying of this complex mixture

[4, 7, 16-18] to obtain highly filled composites. These composites are based on thermoplastic

polymers where the nanoparticle filler concentration by volume exceeds 10 %. With the

motivation to develop a process chain for composite preparation, this thesis critically

investigates and analyzes the process units with a focus on colloidal interactions. It is meant

to explain the physical-chemical limitations of preparing well disperse composites using the

solution and spray drying method.

The work is structured in seven chapters (including the introduction and general conclusions

and an outlook in chapters 1 and 7, respectively) and an appendix including the materials

and methods.

The various pathways of preparing nanoparticle-polymer-composites are introduced in

chapter 2 where the main terms: nanomaterials, polymers and state of dispersion are

introduced and a list of commercial composites is presented as well.

In chapter 3 the unit processes of the developed process chain are specified and the limiting

compositional parameters of the solution based mixture and the final composite are

explained.

Chapter 4 theo eti all a d e pe i e tall fo uses o the u it p o ess li uid-liquid phase

t a sfe . It ill p ese t a e ph si al odel des i i g the effe t of ph si -chemical

2

particle deagglomeration upon surfactant adsorption at a liquid interface in 4.2.

Experimental results on fatty acid adsorption on and deagglomeration of magnetite

nanoparticles are given in 4.3.

The following chapter 5 is dedicated to the phenomena in non-aqueous organic solvent

based nanoparticle-polymer-mixtures. In paragraph 5.2 a short discussion on a DLVO-like

treatment of the governing particle interactions is presented. Experimental results on the

colloidal stability investigations based on two newly developed methods of determining

quantitatively the primary particle concentration are presented in paragraph 5.3.

Finally solvent-free highly filled nanoparticle-polymer-composites are investigated and

discussed in chapter 6. Here it is differentiated between composite microparticles from the

spray dryer in paragraph 6.2 and cross-sections of injection molded composites in paragraph

6.3.

Most results have been published in the following peer reviewed papers (1-5) and

conference proceedings (6-9):

1) Phase-contrast atomic force microscopy for the characterization of the

distribution of nanoparticles in composite materials, In: Chemie Ingenieur Technik

82 (2010), p. 2189-2195 [19]

2) Coagulation and stabilization of sterically functionalized magnetite nanoparticles

in an organic solvent with different technical polymers, In: Journal of Colloid and

Interface Science 357 (2011), p. 292-299 [20]

3) A TGA/FTIR perspective of fatty acid adsorbed on magnetite nanoparticles -

decomposition steps and magnetite reduction, In: Colloids and Surfaces A 397

(2012), p. 16-23 [21]

4) Nanocomposites based on technical polymers and sterically functionalized soft

magnetic magnetite nanoparticles: synthesis, processing, and characterization, In:

Journal of Nanomaterials 2012 (2012), Article ID 670531 [22]

5) Phase transfer of agglomerated nanoparticles - deagglomeration by adsorbing

grafted molecules and colloidal stability in polymer solutions, In: Journal of

Nanoparticle Research 14 (2012), p. 990 [23]

6) Synthesis of highly filled nanomagnetite polymeric composites via sterically

stabilized organosols and the spray drying process, WCPT6, Nuremberg, april 26th

– 29th 2010 [24]

3

7) Processing and characterization of highly filled polymer nanoparticle composites

for micro injection molding applications, ECCM14, Budapest, june 7th – 10th 2010

[25]

8) On the significance of nanoparticle interactions for the synthesis of highly filled

polymer-nanoparticle-composites with the solution and spray-drying process, In:

V.S. Litvinenko (Ed.), International Forum-Competition of Young Researchers -

Topical Issues of Subsoil Usage, Ministry of Education and Science of the Russian

Federation, St. Petersburg, april 20th – 22nd 2011, ISBN: 978-5-94211-506-7, pp.

278-281 [26]

9) Nanoparticles in organic solvents with polymers - stability and consequences upon

material synthesis through spray drying and melt moulding, In: G. Tiddy, R.B.H.

Tan (Eds.), NanoFormulation, The Royal Society of Chemistry, 2012, ISBN: 978-1-

84973-378-6, p. 177-187 [27]

4

2 Nanoparticle-Polymer-Composites – A State of the

Art

The 20th century is often referred to as the century of plastics (macromolecular/polymeric

materials)1, because synthetic polymers have been developed and widely applied improving

materials technologies in nearly any sector of industry ever since. In the 19th century already

first signs of polymerization reactions have been reported. However, in 1907 the first

synthetic polymeric material, a thermoset, has been presented by Joe Baekeland and new

polymeric materials have been developed and synthesized continuously. In order to improve

and influence their properties other classes of materials are incorporated, such as soot in

rubber for tires and wood chips in Bakelite. One such class of materials being incorporated

are the ever emerging nanomaterials.

The 21st century is sometimes called the century of nanotechnology [28] which has been

proposed by RICHARD FEYNMAN in 1959 already with his famous talk The e’s Plenty of Room

at the Botto .

It is a logical consequence to combine nanomaterials and polymers in so called polymer-

nanoparticle-composites. And while more and more nanomaterials are being discovered and

developed, substantial research is focused on synthesizing nanocomposites making use of

the special properties of the nanomaterials. Therefore one can find quite a lot of review

articles on nanomaterial-polymer-composites. However, they are somewhat limited to only

focusing on special classes of nanomaterials, their properties, applications and/or syntheses

routes. As an example the Polymer Nanocomposites Handbook [29] is mainly focused on

clays being nanoplatelets. Other review articles on platelet type nanomaterials in a polymer

matrix are found in [5, 30, 31]. A thorough review on nanoparticles, their preparation and

composite formation with a focus on properties is given by HANEMANN and SZABÓ in [32].

CAMENZIND et al. deliver a thorough review on nanocomposites with flame-synthesized

nanoparticles as the disperse phase [9]. Polymer nanoparticle composites for optical and

magnetic applications with a focus on properties and ways of syntheses are found in [8].

Special focus on transparent nanocomposites regarding the nanoparticle surface design for a

successful incorporation is found in a review by ALTHUEs et al. [33]. Finally an interesting and

recommendable review on modeling the properties of polymer nanoparticle composites is

found in [34].

The purpose for incorporating nanomaterials is to change and advance the properties of the

polymer, as mentioned above. These changes can be manifold, e.g.:

1 http://www.plasticsnews.com/century/bakelite.html, march 26th 2012

5

- I p o i g Me ha i al st e gth a d stiff ess, heali g cracks and increase scratch

resistance,

- Improving the thermal stability, flame retardancy,

- Influencing the chemical stability,

- Improving the barrier properties e.g. for gases,

- Influencing the electric conductivity,

- Introducing catalytic properties,

- Improving and Influencing the optical properties, e.g. refractive index, color,

luminescence or

- Incorporating special magnetic properties, such as super-paramagnetism.

In this thesis the nanomaterial focused on are super-paramagnetic magnetite nanoparticles,

which are to be dispersed in a thermoplastic polymer. Super-paramagnetism means that the

particles are magnetic in a magnetic field but show no remanence and therefore no

magnetization loss on hysteresis. A selected set of papers focused on nanocomposites with

magnetic properties can be found in the references [11, 35-44].

This chapter gives an overview on the state of the art in the science of synthesizing

composite materials. They are consisting of nanomaterials dispersed in a polymeric matrix,

which are in this study limited to thermoplastics. Before summarizing the methods for

incorporation nanomaterials in 2.2, the basic terminology, which has already been used

above is introduced next.

2.1 Terminology

2.1.1 Nanomaterials

Nanomaterials are clearly defined in ISO/TS 27687:2008 as objects with at least one

dimension smaller than 100 nm. Furthermore one has to distinguish between objects with

different aspect ratios where one, two or three dimensions can meet the definition of a

nanomaterial. The images in figure 1 schematically display the three main types of

nanomaterials with distinguishable aspect ratios.

figure 1: Schemes of the three main types of nanomaterials (from left to right) nanotubes with diameters

smaller 100 nm, nanoplatelets with thickness smaller 100 nm and nanoparticles with diameters

smaller 100 nm

6

Nanotubes exhibit one very long dimension (length) which can go up to several micrometers,

e.g. carbon nanotubes [45]. Nanoplatelets are planar nanomaterials which are thinner than

100 nm, e.g. clays [46] or graphene [47]. Very often nanomaterials in general are referred to

as nanoparticles. However, the term nanoparticle clearly defines an object with a low aspect

ratio, i.e. nearly spherical and a size smaller than 100 nm.

One reason why nanoparticles and nanomaterials, in general, show special physical and

chemical properties, is their large surface area. A simple estimate to calculate the relative

number of atoms/clusters at the surface of a nanoparticle is:

3

Abulk

clustercluster

3

cluster

total

surface

6

2

1

11

N

MR

R

R

N

N

.

(1)

Here the number of atoms at the surface Nsurface related to all atoms in the particle Ntotal is a

function of the particle radius R and the characteristic length of a unit, which the particle is

made of, i.e. atoms or clusters Rcluster. The cluster radius Rcluster can be estimated with the

molar mass of a cluster Mcluster, the bulk density of the particle bulk and AVOGADRO’s number

NA. In figure 2 a graphical scheme of the estimation as well as a graph with the results of the

calculation for magnetite and gold are presented.

figure 2: (left) Model geometry for the calculation of the relative number of atoms/clusters at the surface

Nsurface/Ntotal, cf. eq. (72), (right) evaluation of eq. (72) for gold Au and magnetite Fe3O4 with given

atom/cluster sizes Rcluster

Certainly the surface atoms/clusters will not be identical in chemical and physical properties

to the bulk atoms/clusters for they are not entirely surrounded by atoms/clusters of the

same type and must be saturated (no dangling bonds or radicals) towards the surrounding

medium. As a consequence the particles exhibit a high surface energy which leads to strong

interactions such as aggregation (due to high reactivity and sintering) or agglomeration

which has to be dealt with when processing these materials. Magnetite is chosen in figure 2

7

since it is the model nanoparticle system applied in the experiments of this thesis and the

particle radius is R ≈ 7.5 nm, which leads to 10 % of all Fe3O4 clusters located at the surface.

2.1.2 Polymers

The term polymer stands for a wide range of materials consisting of large molecules. These

are made up of small repeating units, which are called monomers. A polymer is

characterized by a great variety of chemical composition, confirmations, properties and

applications which are only very briefly summarized here. If the repeating units are of the

same type, one is referring to homopolymers. The macromolecule can be chainlike or

branched. Individual macromolecules can fruthermore be cross linked with each other.

Thermoplastics are polymers which are not crosslinked, they can undergo a shaping process

under high temperatures. They are meltable and soluble in certain solvents. Crosslinked

polymers are categorized as gels, elastomers and duroplastics depending on the flexibility of

the network, in solvents they do not dissolve but swell.

If the polymer is made up of two types of monomers, they are referred to as co-polymers.

Three and more monomers in a macro molecule are possible as well. Polymers with a net

charge in aqueous solution are called polyelectrolytes. The individual macromolecules in a

polymer can be ordered and semi-crystalline in solid state. Typical transparent polymers are

amorphous. The impact of plastics and their wide range of materials applications are for

example described in [48].

In the present work the focus is on amorphous thermoplastics, which are soluble in organic

solvents and widely used in industrial applications. Examples for such amorphous

thermoplastics are:

- Poly(styrene) (PS),

- Poly(methyl methacrylate) (PMMA),

- Poly(bisphenol A carbonate) (PC),

- Poly(vinyl acetate) (PVAc),

- Poly(vinyl butyral) (PVB) and

- Poly(vinyl alcohol) (PVA).

Important characteristic dimensional properties of a polymer are [49]:

- molecular mass of the monomer M0,

- molecular mass of the polymer M,

- degree of polymerization N = M/M0,

- weight and number average molar mass Mw and Mn with

8

- index of polydispersity PI = Mw/Mn ≥ 1.

2.1.3 State of Dispersion

An important term related to the science of nanoparticle-polymer-composites is the state of

dispersion. In the literature there is however sometimes a confusing usage of this term. Here

it shall be clarified that the state of dispersion is a combination of the deagglomeration of

the disperse phase, i.e. the particles and their distribution within the continuous phase, i.e.

the polymer. The homogeneous distribution of the particles is achieved by mixing processes,

e.g. enforced turbulences and diffusion. The task of a deagglomeration process is to

overcome the attractive forces between the particles which are mainly adhesive forces (VAN

DER WAALS interactions)2. The images of figure 3 depict the interplay of distribution and

deagglomeration to describe the state of dispersion. Certainly most mechanical processes to

disperse particles achieve to both deagglomerate and distribute at once, for they are based

on high shear turbulent systems [50-53]. It will be shown in this work however, that

deagglomeration mechanisms, which will not influence the distribution, can exist as well.

figure 3: Definition of the state of dispersion (poorly dispersed in the upper left, well dispersed in the lower

right) by both a good distribution (more homogeneous from left to right) as well as a thorough

deagglomeration (more successful from top to bottom)

Another term related to the state of dispersion is the stability, which includes the factor

time. One can consider a dispersion to be sta le if a good state of dispe sio does ot decrease with time which can happen in both directions in figure 3 starting at the bottom

right. The distribution is disproved by spatial concentration which can be induced by a force

field, such as gravity. Agglomeration is a consequence of particle collisions which exhibit a

higher energy than the repulsive barrier between the particles.

The mathematical description of the state of dispersion of particles in a composite material

can be difficult, because both distribution and deagglomeration have to be considered.

2 If there are strong cohesive forces to be overcome then one should speak of deaggregation instead, which is a comminution process and consumes more energy.

9

There is an ongoing discussion of adequate measures to quantify the state of dispersion, e.g.

using tessellation (VORONOI polygons) [54, 55] or other methods [56-58], cf. 6.1.3.

2.2 Composite Synthesis/Preparation

This chapter offers an overview on the various possible process principles to

synthesize/prepare nanoparticle-polymer-composites. The focus is on composites for

advanced materials excluding specialized composites like functional core-shell particle

systems. The reviewed articles are characterized in three main groups:

- composites where the disperse (nanoparticles) and continuous phase (polymer) are

synthesized prior to the composite synthesis process,

- composites where the disperse phase is synthesized in situ of an as-synthesized

polymer and

- composites where the continuous phase is synthesized around the as-synthesized

disperse one.

It shall be kept in mind that all experiments in this doctoral thesis are related to the method

Mi i g i a pol e solutio , which is introduced in chapter 2.2.1.

2.2.1 Separate Synthesis of Particles and Polymers

When there is a desire to combine both existing nanoparticles as well as readily synthesized

polymeric materials there are quite different techniques to achieve a well-dispersed

composite. Four routes of synthesis are presented here in this paragraph.

Mixing in a polymer melt

Thermoplastic polymers which are emphasized in this study melt at elevated temperatures,

typically around 100 °C to 300 °C. In extrusion based processes, e.g. compounding or

injection molding, high shear is introduced in a highly viscous polymer melt to distribute and

deagglomerate solid particles, which are typically introduced in a side-stream. This

compounding is already well established in polymer technologies when it comes to disperse

microparticle polymer composites [59, 60]. Nanoparticle-polymer composites have been

processed using this simple technique as well [5, 30]. In the following figure 4 this method is

schematically depicted.

10

figure 4: Scheme of composite formation by mixing a dry nanoparticle powder with a polymer melt, the mixer

represents any type of high shear mixing device such as an extruder

Obviously the state of dispersion depends on the shear rate [61] and other parameters such

as residence time, where an optimum of the parameters is investigated in [62]. The problem

in succeeding with a good state of dispersion when distributing nanoparticle agglomerates is

that the shear forces are too low to act at scales below 100 nm where the deagglomeration

should occur. Hence, usually the homogneization is good but the deagglomeration is poor.

Mixing in a polymer solution

To overcome the problem of a poor deagglomeration another technique is simply to mix a

stable nanoparticle dispersion with a solution of a polymer. As will be shown in 5.1.1, the

soluble polymers will occur in the shape of coils dispersed in the solvent. The nanoparticles,

moreover the nanomaterials are well deagglomerated within the same solvent. A schematic

description of this method is presented in figure 5.

figure 5: Scheme of composite formation by mixing a nanoparticle solvent based colloid with a polymer solution,

the nanoparticles can be well dispersed e.g. by adsorption of end-grafted molecules

When it comes to dispersing nanoplatelets, namely clays, this is a common process in order

to exfoliate (deagglomerate) the clay layers [30, 63-66]. This method has been introduced by

BANERT and PEUKER for spherical particles such as magnetite dispersed in an organic solvent,

namely dichloromethane and mixed with polymer solutions of PMMA or PVB [4, 7, 67].

Furthermore TiO2 and ZrO2 have been well dispersed by solvent based mixtures with pristine

and modified PC [68]. Besides synthesizing bulk composites the technique of mixing

nanoparticles in a polymer solution is important in film casting methods, e.g. when

dispersing fullerenes in PMMA [69]. In order to achieve a good result one has to prevent

agglomeration induced by the polymers and consequently concentrate on the solubility of

the compounds [70].

11

This method seems straight forward, however the interactions between well dispersed

nanoparticles and dissolved polymers can be very critical as will be emphasized in this

present thesis, which focuses on this solution based process.

Milling

A common method for synthesizing metal alloys with nanocrystalline phases is high energy

ball milling [71]. Briefly, dry powders are mixed and well dispersed in the chamber of a dry

media mill, such as a planetary ball mill. Certainly it seems worthwhile adopting this process

to synthesize well dispersed nanoparticle-polymer-composites out of a dry nanoparticle

powder with a polymer powder by high energy mixing. This has been reported in [72, 73],

where fumed silica is being dispersed in a PMMA matrix. However, the main problem here is

the undesired change of the bulk polymer properties, mainly reduction of the chain length of

the polymer [72-75] [76]. In figure 6 a scheme of this method is depicted.

figure 6: Scheme of composite formation by compounding a nanoparticle and a polymer powder with high

forces in a mill, typically a high energy ball mill

Nanotubes have also been reported to distribute and deagglomerate well in PE using the

principle of ball mill mixing [77]. Besides the aforementioned problem of influencing the bulk

properties of the polymer, certainly this method is difficult to scale-up for a higher material

throughput and is therefore not of interest for technical processes.

Hetero-coagulation

A fourth type of process for this kind of composite synthesis is hetero-coagulation [78]. A

principle scheme is presented in figure 7.

figure 7: Scheme of composite formation by attraction of oppositely charged particles and polymers in water

and withdrawing the water for composite synthesis

Certainly this is a very exclusive method for composite synthesis which is limited for water

based systems and opposite surface charges for the nanoparticles and the particulate

12

polymers within the same aqueous environment (pH, salt types and ionic strengths).

However, it was reported to be a successful way to synthesize fire retarded composites [78].

2.2.2 Synthesis of Nanoparticles within the Polymer

Aside from the four techniques for composite synthesis described in chapter 2.2.1 it is also

considerable to synthesize the nanoparticle fillers individually within a solvent swollen

polymer network. Such a network can be crosslinked polymers or porous polymeric

structures. For this, the building blocks to form nanoparticles need to be well dispersed

within the polymer, e.g. salts for a co-precipitation or pre-coursers for a sol-gel-synthesis

based formation, respectively. A scheme for this principle of composite synthesis is

presented in figure 8.

figure 8: Scheme of composite formation by mixing a pre-courser (symbolized as small dots in the left frame)

with a polymer in solution and initiation of nanoparticle synthesis

Good results regarding the state of dispersion are reported for sol-gel based systems such as

the formation of silica from TEOS pre-coursers [79] in a nylon polymer.

The nanoparticles used in this thesis are magnetite nanoparticles synthesized in a co-

precipitation reaction, as explained in detail later. ALI-ZADE has reported on the composite

synthesis of co-precipitated magnetite nanoparticles in porous polystyrene or collagen

structures with excellent results concerning nanoparticle distribution [80, 81].

In the reviewed literature the obstacles of this method have not been reported. However, it

is necessary for a successful synthesis to be given a polymer three dimensional structure

with pores that are accessible for the reagents. Therefore, the method obviously is not

suitable for thermoplastic polymers which are not cross-linked but of interest in this work,

cf. 2.1.2. Furthermore the nanoparticle synthesis in the presence of the polymer must be

undesirably different from a controlled synthesis without the polymer.

2.2.3 Synthesis of the Polymer with the Presence of Nanoparticles

Contrary to 2.2.2 one can also synthesize the matrix polymer around the already existing

nanoparticles. A straight forward way to achieve this is by dispersing and stabilizing

nanoparticles in a monomer or monomer solution and synthesizing the polymer by a

polymerization reaction. In figure 9 the scheme for this method is presented.

13

figure 9: Scheme of composite formation by dispersing nanoparticles in a monomer solution (on the left

represented as an emulsion droplet) and subsequent polymerization and withdrawing of all solvents

In [82] fatty acid stabilized magnetite nanoparticles are dispersed in methyl methacrylate

droplets of an emulsion. Magnetite PMMA composites are achieved by radical modified

suspension polymerization with a filler content of up to 11 % by weight. GYERGYEK et al.

report on the synthesis of magnetite-PMMA composites as well, applying in-situ dispersion

polymerization of MMA in n-decane [35]. The composites prepared there reach a filler

concentration of up to 48 % by weight with a high saturation magnetization due to the

super-paramagnetic properties of the oleic acid coated nanoparticles. Quantum dot filled

composites of zinc sulfide ZnS in acrylate polymers have been prepared in [83] using

dispersion polymerization. The transparency of the resulting composite reveals the high

state of dispersion. It is stressed that, in order to obtain a good incorporation and a well

dispersion, the nanoparticle surface needs to be designed to have a good interaction with

the polymer and the polymer precursors preventing particle agglomeration. Another way to

synthesize composites by polymerization can be based on two component systems, e.g. for

thermosetting polymers [84]. For this the particles are dispersed in the epoxy resin and

subsequently the hardener phase is added.

In [85] and as well in [36] highly filled magnetite thermoplastic polymer composites are

synthesized by miniemulsion polymerization. Both studies reveal agglomeration of the

sterically stabilized nanoparticles induced by the polymerization. This may conclude that

attractive interactions are introduced by the polymers that have not been present in the

nanoparticle monomer dispersion. When the goal is to create a highly disperse composite,

cf. 2.1.3, these attractions need to be dealt with and prevented. Furthermore the presence

of the nanoparticles might influence the polymer synthesis, so that one may be losing

control over desired polymer bulk properties.

2.3 Industrial Products

This final paragraph presents selected nanoparticle-polymer composites which are already

products on the market. They have been researched in [31] and on the world wide web 1,3,4.

14

table 1: A selection of industrial polymer-nanoparticle composites which are already on the market345

company product application properties

Südchemie Nanofil® polymer additive flame retardant

Bayer Durethan® coatings strength, barrier

properties, gloss

Honeywell Aegis® packaging barrier properties against

water and air

Noble Polymers Forte®,

Nubrid® automotive

strength, temperature

resistancy

Polykemi Scancomp® automotive, home care strength, low density,

gloss, scratch resistance

Mitsubishi Gas

Chemical Company M9 beverage packaging high barrier properties

Grado Zero Espace Absolut Black polymer additive homogeneous surfaces,

intense black coloring

RAS Materials AgPure polymer additive antimicrobial

Silanotex Nano-Silber polymer additive antimicrobial

ApNano Materials NanoG® military, packaging low density, strength,

energy adsorption

Evonik Hanse

Chemie Nanocone®

electronics, dental care,

medical technology

strength, transparent,

flexibility

Evonik nanoresins

AG Nanocryl® coatings

scratch resistance,

transparency, strength

NanoSky Nanoterra

soil® road construction

non-toxic,

environmentally friendly,

longer lasting roads

Polyone Nanoblend® polymer additive improved mechanical

and thermal properties

Chemtura Polybond® packaging barrier properties

TurboBeads® (several) biochemical, medical,

chemistry magnetic

3 www.nanoproucts.de, January 27th 2011 4 www.nanocompositech.com/commercial-nanocomposites-nanoclay.htm, January 27th 2011 5 http://www.turbobeads.com, June 10th 2012

15

3 Thesis Motivation – Development of a Modular

Process Chain for Composite Preparation

This chapter presents and discusses the background and motivation of this thesis. It explains

the topics of the following chapters and defines certain limits, e.g. material, composition and

processes.

From october 2008 to march 2012 the German research foundation (DFG – Deutsche

Forschungsgemeinschaft) supported the project PE1160/7-1 devoted to The Development

of a Process Chain for the Synthesis and Processing of Highly Filled Polymer-Nanoparticle-

Co posites [86]. The contribution of this thesis is to answer the urging questions related to

this project. Concluding, the project report shall present and discuss the How to and the

thesis addresses the Why so. So before the Whys can even be addressed the reader of this

thesis must be given the Hows in this chapter. Therefore the paragraph 3.1 presents each

process step pointing out the relation to the subsequent chapters. In paragraph 3.2

important compositional parameters are introduced since in some cases of this research

there are up to four components to deal with.

The starting point for the project are papers by ZHOU, PEUKER, BANERT, HICKSTEIN and

MACHUNSKY [7, 18, 67, 87-90]. These introduce the solvent based process for preparing

nanoparticle-polymer-composites (cf. 2.2.1) as well as the liquid-liquid phase transfer of

nanoparticles are introduced. It shall be emphasized in this context, that the scientific

contributions of this thesis, based on the various publications, cf. chapter 1, are mainly

concerned with special colloidal interactions.

3.1 Process Steps

The solvent based process for the synthesis of nanoparticle-polymer composites developed

in the research project comprises several individual process steps, namely:

- particle synthesis,

- particle functionalization and particle transport to an organic solvent,

- addition of the dissolved polymer,

- quick evaporation of the solvent in a spray dryer and finally

- composite powder agglomeration.

These steps are briefly introduced in the following paragraphs.

16

3.1.1 Nanoparticle Synthesis

figure 10: Synthesis of the nanoparticles by a precipitation reaction in an aqueous environment

The first step is to synthesize the nanoparticles in an aqueous milieu by a precipitation

process. The model system of nanoparticles is the super-paramagnetic iron oxide magnetite

Fe3O4. Super-paramagnetism is a special property of iron oxides in the nanoscale. These iron

oxides are well magnetizable in a magnetic field (high saturation magnetization), yet the

magnetic character is lost when the magnetic field is zero (no remanence). In the appendix

A.3 the procedure of the co-precipitation as well as the nanoparticle characterization by

TEM, XRD, Zeta Potential and DLS are presented there. Due to the high ionic strength in the

aqueous phase the particles agglomerate and settle rapidly in a force field, e.g. gravity. The

surface characteristic of the particles is hydrophilic and they carry a net charge which is

negative. However, the particles both need to be hydrophobic in order to incorporate in the

hydrophobic polymer matrix of a composite and should be deagglomerated to obtain a high

state of dispersion. These demands are fulfilled using the process step of liquid-liquid phase-

transfer.

3.1.2 Liquid-Liquid Phase-Transfer

figure 11: Transfer of the nanoparticles from the aqueous to an organic phase (immiscible solvent) by

adsorption of amphiphilic molecules

The liquid-liquid phase-transfer process of particles is in principle similar to the widely

applied liquid extraction of ionic or molecular species [91]. One could therefore also refer to

it as particle extraction. Details of this method and new scientific findings offered by this

thesis are to be found in chapter 4.

17

Just briefly, in this second step of the process chain the surface of the nanoparticles is

hydrophobized and the agglomerates are partially deagglomerated by entering an organic

solvent, which is immiscible with water. This is achieved by adsorption of amphiphilic fatty

acid molecules at the liquid interface. The organic solvent mainly used in this research is

dichloromethane with a higher specific weight compared to water. Thus it is located below

the water phase underlying gravity. The transport of the particles to the interface is achieved

by settling. Without the amphiphilic substance at the interface of the liquids, the particles

rest on top of the organic phase due to their hydrophilicity.

The profit of this liquid-liquid phase-transfer is that there is no need for drying the

nanoparticles and subsequently resuspending them in the organic phase. Therefore

undesired agglomeration or even aggregation of the particles is prevented.

The result of phase transfer is, depending on the fatty acid used, a deep black long term

stable colloid. The stabilization mechanism of the nanoparticles is the steric repulsion

between the chains of the adsorbed fatty acid molecule layers. It is introduced in chapter 4,

as well. When using dichloromethane as a carrier solvent there are a wide range of soluble

thermoplastic polymers, such as PMMA, PC and PVB, to mix with the colloid, which is

accounted for in the next unit process.

3.1.3 Polymer Addition

Adding a polymer solution to a stable colloidal dispersion of phase transferred particles

seems to be a rather simple procedure. However, it turns out to be the bottleneck of the

entire process chain, because the state of dispersion is mostly affected in this step, given

that the previous phase transfer process delivers a stable dispersion. Chapter 5 covers this

issue in greater detail.

figure 12: Mixing the organic solvent based hydrophobic nanoparticles with a polymer solution in the same

solvent

Just briefly noted here, a soluble polymer is dissolved in the same type of solvent which the

particles are transferred to. When dissolved, the individual polymer chains will coil-up with a

coil dimension that is in the same order of magnitude compared to the size of the stabilized

18

particles. Depending on whether the polymer can adsorb on the surface of the particles the

colloid can be destabilized and undesired flocculation may occur. However, if this is the case

then not all of the particles will flocculate and settle. There is still a fraction of long term

stable primary particles left. Visually the destabilized nanoparticle polymer dispersion may

be as deep black as before and seem stable. Hence, one would not necessarily be aware of

the problem of flocculation and carry on with the next step. This next step is the quick

evaporation of the solvent in a spray dryer in order to preserve the high state of dispersion

in a nanoparticle colloid polymer mixture.

3.1.4 Spray Drying

An important step in the solvent based processing of nanoparticle-polymer-composites is

the rapid removal of the solvent, because the high state of dispersion of the nanoparticles in

the polymer solution has to be preserved. This can be achieved by quick evaporation in a

spray dryer. In addition with this process one can design composite microspheres which are

already a desired final product, e.g. functional particles such as magnetic beads for

bioseparation purposes [4, 92].

figure 13: Spray Dryer with a two-fluid nozzle in the co-current regime. A small fraction of large particles falls

through the drying cylinder by gravity separation, the largest fraction of particles by mass is the

product collected at the coarse output of a cyclone by centrifugal force separation.

The spray drying unit process itself is based on the following individual steps:

At first, a feed dispersion, namely the nanoparticles in the polymer solution, is

atomized into small droplets with several micrometers in size using a nozzle or more

generally an atomizer. This atomization characterized by a droplet size distribution is

influenced by the atomizer design (e.g. pressurized gas two fluid nozzle or rotary

atomizer and the dimensions of these devices), the dispersion properties (e.g.

viscosity, surface tension and specific weight) and the process parameters (e.g. mass

flow of the dispersion, gas pressure and temperature).

Secondly, the droplets are contacted with a hot gas at temperatures above or close

to the evaporation temperature of the solvent. This is drying step, is based on heat

19

and mass transport between the gas and the drying droplets and also depends on

various parameters. The flow regime of the droplet and gas can be co-current or

counter-current and is defining the temperature gradient and thus the kinetics of

drying. Several properties of the droplet components in terms of thermal

conductivity, heat of evaporation and heat of crystallization as well as interactions

between the components influence the kinetics of drying and the final morphology of

the dry particles and with it the bulk powder properties of the product.

The third individual process of a spray dryer is the separation of the dry particles

from the solvent carrying hot gas flow. Very large partiocles are gathered at the spray

drying vessel. Usually the majority of the material is to be found at the coarse outlet

of an aerocyclone. Small particles will be trapped in a filter. The solvent carrying gas

flow is either released or recycled by a condenser. The condenser captures the

solvent and then heats the gas up again to act as the drying gas.

Pressurized gas two fluid nozzles produce a very fine powder which is hard to process

further, if desired. In order to improve the bulk powder properties, such as flowability and

specific weight of the bulk powder, an agglomeration process is needed. Such a process has

also been developed within the research project and is briefly explained in the next

paragraph.

3.1.5 Powder Agglomeration

The task of powder agglomeration is to improve flowability and increase the powder bulk

density by increasing the grain size. This can be achieved by press agglomeration and

subsequent granulation in a rotor sieve mill [93].

figure 14: Press agglomeration of the spray dried fine composite micropowder and granulation in a rotor-mill

for an improved powder handling

The process has been developed as part of a diploma thesis by ANNE HORSCHIG [94]. Grain

sizes within the powder are increased from 10 µm to 1 mm. This leads to an increase of the

bulk density by a factor of 10. Pressures up to 1.6 MPa are applied in a tabletting press with

20

a diameter of 50 mm and a final tablet thickness of 10 mm. Using a JENIKE shear tester it is

shown that the flowability and thus process ability improves significantly.

3.2 Composition

This paragraph introduces the compositional parameters and limitations of the solvent

based process described above. There are two considerations regarding the composition

within the solvent free composite in paragraph 3.2.1 as well as the composition of the

solvent based dispersion in paragraph 3.2.2.

3.2.1 Composite Composition

An important concentration characterizing the final composite is the filler concentration F,

defined in eq. (2). It is taking into account the masses of the nanoparticles, the surfactant

and the polymer mnanoparticles, msurfactant and mpolymer, respectively.

polymerlesnanopartic

lesnanopartic

polymersurfactantlesnanopartic

lesnanopartic

composite

lesnanopartic

1 mmD

m

mmm

m

m

m F

(2)

The surfactant ratio D describes a fixed ratio of the surfactants (fatty acids in this study) to

the nanoparticles in eq. (3), with the mass of the surfactants and nanoparticles msurfactant and

mnanoparticles, respectively.

lesnanopartic

surfactant

m

mD (3)

This parameter D is limited by the amount of fatty acids adsorbed onto the nanoparticles

and can be transformed into a function of the size of the nanoparticles xnanoparticle in eq. (4)

and figure 15, with Snanoparticle and Vnanoparticle being the surface and volume of the particle,

respectively.

lenanoparticlenanoparticAsurfactant

surfactant

lenanoparticlenanoparticAsurfactant

surfactantlenanopartic

lenanopartic

surfactant

6

xNS

M

VNS

MS

m

mD

(4)

Eq. (4) is illustrated in figure 15. The surfactant parameters necessary are the surface of the

adsorption site Ssurfactant, hi h is i the o de of … ·10-19 m2 as well as the molecular mass

of the surfactant Msurfactant, which is 282.46 g/mol for oleic acid.

21

figure 15: Surfactant ratio D, cf. eq. (4) with Msurfactant = 282.46 g/mol, Ssurfactant = 2.4·10

-19 m

2 [95] and

nanoparticle = 5.2·106 g/m

3 for magnetite Fe3O4

The specific weight of the composite composite can be calculated with the parameters F and D

as well as the components individual specific gravities as shown in eq. (5).

polymersurfactantlesnanoparticcomposite

1

DFFDFF (5)

Another important parameter regarding the composite is the volume fraction of the

nanoparticles φnanoparticles in eq. (6).

1

polymersurfactantlesnanoparticlesnanopartic 1

111

D

F

D

(6)

Accordingly the volume fraction of the polymer is defined in eq. (7).

11

surfactant

polymer1

nano

polymerpolymer 1

111

11

DDFD

F

(7)

Assuming the particles are spherical and monodisperse, the volume fraction of the polymer

can in no case be smaller than a threshold value of 0.26, which is the volume fraction of

voids in the closest packing of spheres. The relation of eq. (7) is depicted in figure 16 for

given specific gravities of the three components. With less surfactants per nanoparticles D

and higher filler concentrations F, the polymer volume fraction decreases. In other words,

higher theoretical filler concentrations can be achieved for lower amounts of surfactants. If

the polymer volume fraction approaches the threshold value of 0.26 percolation would

occur, which is the unavoidable contact of well dispersed individual nanoparticles.

22

figure 16: Interconnection of the composition values: filler concentration F, surfactant ratio D and volume

o e tratio of the pol er φpolymer for the given specific weights of the three components using

eq. (7), volume concentrations have to be larger than 0.26 which is the limit for closest packing of

spheres

The distance between the particles would approximate zero. This interparticle distance ID

can easily be calculated with a given packing structure of the spherical particles. For a body-

centered cubic packing ID (red arrow in the inset drawing in the graph of figure 17) is a

function of the particle diameter x and the volume fraction of the nanoparticles φnanoparticles

as expressed in eq. (8).

1433

3/1

lesnanopartic

xID (8)

Combining eqs. (8) and (6) the interparticle distance ID is plotted as a function of F for three particle diameters x in figure 17 on a log-lin-scale.

figure 17: Interparticle distance as a function of the filler concentration for monodisperse and homogeneously

distributed fillers of different size (D = 0.2 and specific weights as mentioned in figure 16). The inset

shows the geometry for the body-centered cubic space.

The interparticle distance approaches zero for filler concentrations F slightly larger than 0.8,

because the packing of spheres is not the closest in eq. (8).

23

3.2.2 Colloid Composition

For the solvent based process of composite synthesis there is an additional solvent

component to be considered when looking at the composition of the complex dispersion of

surfactant stabilized nanoparticles in a polymer solution. Crucial parameters regarding the

processing of the complex colloid are presented in this paragraph.

An economically as well as ecologically important parameter is the concentration of solids

within the solvent csolid. It is defined in eq. (9) using the mass of the solvent msolvent and the

mass of the composite mcomposite as a sum of the masses of nanoparticles, surfactants and

polymer.

solventcomposite

compositesolid mm

mc (9)

The lower this value is, the more solvent has to be processed which increases the economic

cost for processing and the ecologic impact due to higher energy consumptions and

potential solvent release to the environment.

As will be shown in the experimental part of this thesis the concentration of the polymer in

the colloid cpolymer, given in eq. (10), influences the process ability as well as the stability of

the colloid. Generally, higher polymer concentrations will tremendously increase the

viscosity of the colloid, which is an important factor for the spray drying process, cf. 3.1.4

and chapter 6. The impact of cpolymer on the colloidal stability is the focus of chapter 5. For

destabilizing polymers the stability increases with lower cpolymer.

solid

solidsolidsolventpolymer 1

1

c

DFccc

(10)

The main parameter for synthesizing a certain composite is the filler concentration, defined

in eq. (2), which can be expressed as a function of csolid and cpolymer, presented in eq. (11).

D

cc F

1

11

solvent

solidpolymersolvent

(11)

In figure 18 the relation in eq. (11) for D = 0.2 and the specific weight of the solvent DCM

solvent = 1330 g/l is visualized. Due to suppressed droplet formation, a polymer

concentration higher than 60 g/l for PMMA dissolved in DCM cannot be processed in a spray

drying step, cf. chapter 6.

24

figure 18: Interconnection of the polymer concentration in the organic solvent based mixture cpolymer, the solids

concentration csolid and the filler concentration F in the solvent free composite material, cf. eq. (11).

After having introduced the process chain developed as a practical background of this thesis

as well as the most important compositional parameters, in the following chapter theoretical

and experimental investigations and advances concerning the liquid-liquid phase transfer of

magnetite nanoparticles are presented.

25

4 Focus on the Phase Transfer of Nanoparticles

This chapter is dedicated to the liquid-liquid phase transfer of hydrophilic nanoparticles, as

part of the process chain in 3.1.2. After introducing the theoretical background and the

present literature in paragraph 4.1, a new physical model for a deagglomeration

phenomenon of this process is presented in paragraph 4.2. Subsequently in 4.3 experimental

results are discussed. These cover the study on the state of dispersion after phase transfer

for different types of fatty acids adsorbed on magnetite nanoparticles as well as the

spectroscopic coupled thermo gravimetric study on the chemisorption of ricinoleic acid.

4.1 Theory

After having synthesized nanoparticles in an aqueous co-precipitation reaction, as is the case

in this work (cf. appendix A.3), it is necessary to modify the surface to become hydrophobic

and to transfer the particles to an organic solvent phase (cf. appendix A.4).

Hydrophobization is achieved by grafting amphiphilic substances, such as carboxylic acids /

fatty acids, onto the particle surface. Transfer of the particles to the organic solvent phase

can e.g. be realized by washing and subsequent removal of the water and redispersing the

dried nanoparticles in the solvent. However, besides high energy consumption due to the

additional drying and mechanical redispersion steps and loss of material within the washing

step, the particles are endangered by oxidation and strong aggregation. A convenient one-

step process is the direct transfer of the nanoparticles to the water immiscible organic

solvent across the liquid-liquid interface, cf. 4.1.1. The amphiphilic molecules are dissolved

within the solvent phase and will be located at the interface for they have a surface active

character and can hence be referred to as surfactant molecules. The hydrophilic head group

points towards the aqueous phase where the adsorption onto the nanoparticle surface

occurs, cf. 4.1.2. After adsorption and hydrophobization has occurred the transferred

nanoparticles are furthermore stabilized by the adsorbed species through steric stabilization,

cf. 4.1.3.

4.1.1 Phase Transfer

The liquid-liquid phase transfer of nanoparticles is in principle an extraction of particles as

compared to the well-established extraction of ion-complexes or molecules [91]. The

mechanism of nanoparticle transfer has been described by MACHUNSKY and PEUKER [88].

Magnetite nanoparticles with a diameter of about 15 nm are transferred from water to

dichloromethane with the surfactant material oleic acid, which is known to adsorb

26

chemically on the magnetite surface [21, 95-97]. The transport of the particles to the

interface is achieved by settling of agglomerated nanoparticles. Without surfactants at the

interface the magnetite particles will not enter the hydrophobic solvent phase for the

surface properties of magnetite in water are of hydrophilic nature [98]. Fatty acids other

than oleic acid have proven to result in a successful transfer of the particles as well, however

qualitatively different states of dispersions are observed [99]. MÉRIGUET et al. report on the

direct diffusion driven phase transfer of water stable maghemite ( -Fe2O3) nanoparticles into

cyclohexane and nonane using didodecylammonium bromide as an ionic surfactant [100].

Both in [99, 100] the problem of emulsion formation is mentioned which is due to the fact

that the surfactant molecules also act as good emulsifiers. Diffusion driven liquid-liquid

phase transfer has also been reported for gold nanoparticles [101-104]. PRAKASH et al. show

that it is also possible to reverse the phase transfer from an organic solvent phase into water

for magnetite nanoparticles and quantum dots [105] using a surfactant in the aqueous phase

which will cause bilayer formation with the already existing first layer on the nanoparticle

formation and thus stabilization.

4.1.2 Adsorption of Surfactants

Surfactants in general are surface active substances, i.e. they change the surface (or

interfacial) energy by concentrating at a surface or an interface. In the sense of this study

surfactants are amphiphilic molecules which comprise both of a hydrophilic and a

hydrophobic (or lipophilic) section. One class of such amphiphilic molecules are the fatty

acids, which consist of a hydrophilic polar carboxylic head group and a hydrophobic aliphatic

tail.

Fatty Acids

Fatty acid is a more widely used term for the carboxylic acids extracted from fats

(triglycerides), after chemical removal of the glycerin. The structure can be well explained

with the head-tail model, as depicted in figure 19, where the dotted line stands for a linear

hydrocarbon rest R of different length.

figure 19: Principle structure of a fatty acid molecule with a hydrophilic carboxyl head group and a hydrophobic

alkyl chain with certain length and functionalities (double bonds, functional groups).

27

The phase behavior of fatty acids in aqueous solutions are described in [106]. At a pH above

6, depending on the length and saturation of the hydrocarbon rest, the carboxylic group

disassociates leaving behind a negatively charged carboxylate molecule.

This will also be the case if the molecules are located at an interface of water and an

immiscible organic solvent [107]. Disassociation is given by the following equilibrium formula

in eq. (12), where the negative charge of the carboxylate is delocalized between the two

oxygen atoms.

OHCOOROHCOOHR 32 (12)

The carboxyl group is the functionality, which will adsorb at the inorganic nanoparticle

surface [108]. This adsorption of fatty acid molecules will be introduced and discussed in the

next paragraph.

An important feature of the fatty acids is their maximum length when fully stretched, which

will be important for the phase transfer and the stability of functionalized nanoparticles. The

maximum length can be calculated with equation (13).

CCCCCC

CCCC lnln

2

sinmax

(13)

In this equation nC-C and lC-C are the number and length of C-C-bonds, αC-C is the angle

between three neighboring sp3-hyridized C-atoms and nC=C and lC=C are the number and

length of C=C-bonds. Values for the bond lengths and angles are lC-C = 0.154 nm,

lC=C = 0.150 nm and αC-C = 109.47° [109]. The ratio of C-C and C=C bonds, characteristic for a

fatty acids, is given by the lipid number. The fatty acids used in this study and their molar

mass M, maximum length max, lipid number and special functional groups beside the

carboxyl group are presented in table 2.

table 2: List of the fatty acids used in this study with maximum chain lengths as defined in eq. (13)

Name Abbreviation M in g/mol Length

max in nm

Lipid Number Special

Groups

Ricinoleic Acid RA 298.5 2.16 C18:1 -OH (C12)

Linoleic Acid LA 280.5 2.19 C18:2

Oleic Acid OA 282.5 2.16 C18:1

Myristic Acid MA 228.4 1.63 C14:0

Caprylic Acid CA 144.2 0.88 C8:0

28

Adsorption

Adsorption, in the present context, is an interaction of a substance (molecule, ion, complex,

etc.) with a solid surface [110-112]. Depending on the strength and reversibility of the

interaction physical and chemical adsorption are distinguished. For physical adsorption the

heat of adsorption is Hphysisorption < 40 kJ/mol. Heats of adsorption for chemisorptions are in

the order of chemical reactions Hchemisorption > 80 kJ/mol [112].

Typically adsorption is characterized by the amount of adsorbed molecules , which is the

amount (mass or number) of adsorbts related to the mass or surface area of the adsorbent,

so that [ ] = (g/g, mol/g, g/m2, mol/m2). It is usually investigated as a function of the

concentration c of the adsorbtive in the solution. Plotting over c results in adsorption

isotherms which can reveal the type of adsorption, whether it is physical or chemical and

furthermore giving clues on the arrangement of the substance in mono- or multilayers.

Chemical adsorption is present if by reducing the concentration of the adsorbent the

adsorbed amount does not follow the adsorption isotherm for increasing concentrations.

Analytical descriptions of surfactant adsorption isotherms are given in [112]. A very

prominent analytical formula is the Langmuir isotherm in eq. (14).

cK

cK

A

Amax 1

(14)

Here max is the maximum amount of adsorbt, when the entire surface of the adsorbent is

covered with a monolayer of adsorbt molecules. The constant KA is a function of the free

enthalpy of adsorption.

Fatty Acids Adsorption on Magnetite Surfaces

It is a well observed fact that fatty acids chemically adsorb on the surface of magnetite [4, 7,

20, 21, 24, 88, 95-97, 99, 108, 113-117]. Once the fatty acids are adsorbed, they can well be

described by a brush-like structure, where the hydrophobic tails point toward the solvent

[118, 119], as presented in figure 20a. Before the fatty acid adsorbs, typically at high pH the

negatively charged magnetite surface [98] approaches the negatively charged carboxylate,

as depicted in figure 20b. The charges are screened by the intermediate water layer, which

enables the approach of adsorbent and adsorpt [116]. However, the mechanism of

adsorption is presently discussed and no one well accepted theory exists. Three different

possible ways are presented in general. The most cited is the occurrence of chelating

bidentate binding by ZHANG et al. [95], which is presented in figure 20c. ROONASI et al. assume

a bidentate mononuclear bond [96]. Monodentate mononuclear configurations are

presented by MACHUNSKY et al. and most recently by CHERNYCHOVA et al. [99, 116]. However,

in MACHUNSKY et al. the carbonyl group is preserved, which is not expected for disassociated

29

specimens, which are represented by carboxylate groups, as approved by several different

analytical investigations [116].

figure 20: Principle of Adsorption of Fatty Acids on Magnetite: (a) monolayer covered particle, (b) before

adsorption in aqueous solution (mediating water layer in blue) with high pH resulting in negatively

charged magnetite surface and disassociated carboxyl group [99], (c) chelating bidentate bond

between fatty acid and magnetite [95] and (d) monodentate mononuclear configuration [116]

Adsorption isotherms with the adsorbent magnetite are published by KOROLEV et al. in [120]

with the fatty acids oleic (C18:1), linoleic (C18:2) and linolenic acid (C18:3) in cyclohexane

and heptane. It reports on Langmuir type adsorption with similar constants KA, cf. eq. (14).

The parameter max depends on the solvent. It increases or decreases with more double

bonds for heptane or cyclohexane, respectively.

The maximum amount of adsorpt max is an important value, when it comes to determining

the area of adsorption of a single molecule Smolecule, which furthermore is necessary to know

for assessing the grafting density of monolayer adsorption φ. For Smolecule, values ranging

from 20 Å2 to 38 Å2 are reported in the literature [21, 95, 96, 108]. The following relation in

eq. (15) can be applied for determining the surface of an adsorbed molecule head group

Smolecule, grafting density φ or maximum amount of adsorpt in a dense monolayer max, with

the specific surface of the particle Sparticle [21]. In this relation the following experimentally

accessible units are used: [ max] = mol/g and [Sparticle] = m2/g.

Amax

particlemolecule N

SS (15)

Once adsorption of the amphiphilic molecule occurred, the particles are hydrophobic and

the end-grafted molecule tails point toward the solvent causing a steric repulsion upon

contact of two approaching particles. This steric stabilization is introduced in the next

paragraph.

4.1.3 Steric Stabilization

One aim of the phase transfer of nanoparticles is the stabilization in the organic solvent

phase. Due to a typically low dielectric constant , double layer repulsion can be neglected

and is therefore not the source of stabilization against VAN DER WAALS attraction [121].

However, due to overlap of surfactant or polymer layers on interacting nanoparticles

30

surfaces another, so called steric repulsion defines stability [99, 122-125]. In COSGROVE [126],

BREZESINSKI and MÖGEL [127] and DÖRFLER [112] there is an enthalpic and an entropic reason

presented for describing stabilization. An expression based on the Gibbs free energy of

overlapping polymer layers GM is given in equation (16), using the entropic and enthalpic

FLORY-HUGGINS pa a ete s S a d H, respectively [112, 128]. Steric stabilization occurs for

GM > 0.

Vsolvent

S dNV

VG A2

2

SHM 2

kT (16)

The molar segment volumes of polymer and the solvent are expressed by 2SV and 2

solventV ,

respectively. Furthermore, the density of the polymer surface layers is given by the integral

in eq. (16). The following three cases are distinguished for stabilization:

- pure entropic stabilization for ( S < 0)^( H < 0 ^ H/ S < 1),

- pure enthalpic stabilization for S > ^ H > ^ H/ S > 1) and

- combined entropic / e thalpi sta ilizatio fo S < 0)^ H > 0).

The first theoretical explanation for steric stabilization with small end-grafted molecules,

taki g i to a ou t the le gth of elati el sho t od-like ole ules is gi e MACKOR in

the 1950s [129, 130] and discussed by OVERBEEK in an introductory paper of 1966 [131].

MACKOR showed theoretically how carbon black nanoparticles are stabilized in hydrocarbons

by 2 nm long molecules, yet not by 1 nm long molecules [129].

For the analytical expression of interaction energies two structural regimes of the end-

grafted adsorbed molecules are generally distinguished, the brush and the mushroom type,

as depicted in figure 21 [118].

figure 21: Geometries of two types of end-grafted molecule covered interacting particle surfaces at distance D

(left) brush type as expressed in eq. (17), (right) mushroom type in eq. (18)

31

Most recently published by BUTT in [118] is a good summary of analytic expressions for the

interaction energies due to steric repulsion of polymer covered surfaces and the brush and

mushroom regime, referred to as the ALEXANDER-DE-GENNES [119, 132, 133] and DOLAN-

EDWARDS models [123], respectively.

The following eqs. (17) and (18) define the repulsive interaction energy W between two

spherical particles as a function of the geometrical parameters in figure 21, particle distance

D, brush length and radius of gyration of the mushroom RG. Please note that the DERJAGUIN

approximation [134] has been applied, yet the particle radius does not appear in the

relations.

122

52

71

35

8)(

4

7

4

5

3brush D

DsTkDW (17)

GR

D

eTks

DW

2mushroom

136 (18)

The distance between two adsorbed molecules s is related to the surface coverage in a

monolayer as given in eq. (15), with [ ] = molecules/m2.

2

1

s (19)

The well-established ALEXANDER-DE-GENNES theory (AdG) [119, 132, 133], which leads to

eq. (17), offers an analytical expression for the pressure between two planar surfaces in

distance D covered by end-grafted polymers with chain length and a separation of s

between the molecules on the surface in eq. (20) [133].

4

3

4

9

3 2

2)(

D

Ds

TkDP (20)

The length of the grafting layer is a function of the maximum length of the molecules (cf.

table 2) as well as the solubility of the molecules which influences their conformation as

stated in the experimental part in paragraph 4.3. Furthermore is a function of s as well

[107, 133].

It is interesting to note that in the parenthesis there is a strong positive term which causes

the planes to push apart and a smaller negative pressure. The first term is due to

unfavorable structural ordering upon compaction of the layers (note that the theory

assumes that the brushes of the two particles do not penetrate). The negative term accounts

for the favorable elastic deformation of the prior stretched molecules.

32

In the next paragraph this model for the pressure is applied for a physical model which

describes a phenomenon occurring when agglomerated nanoparticles are phase transferred,

namely the deagglomeration upon surfactant adsorption.

4.2 Physical Model of Deagglomeration at the Interface

For the phase transfer of magnetite nanoparticles from an unstable aqueous phase to an

organic solvent with the amphiphiles oleic or ricinoleic acid it has been observed that the

agglomerates break-up and are deagglomerated, when crossing the interface [7, 20, 88, 99,

114]. This phenomenon is accounted for in this chapter. At first a gedankenexperiment is

presented of what is thought to occur at the liquid interface. Based on the hypothetical ideas

of this experiment a physical model, applying a simplified sphere geometry, is developed and

numerical results are presented.

Experimental investigations with different fatty acids and the assessment of the primary

particle concentration are presented in paragraph 4.3.

4.2.1 Gedankenexperiment

In figure 22 a set of schematics depict the basic ideas and hypothetical events occurring at

the liquid-liquid interface.

figure 22: Schematics representing the gedankenexperiment of an agglomerated nanoparticle doublet passing

the liquid-liquid interface where fatty acid molecules adsorb and push the particles apart by a

disjoining force when tails of opposing end-grafted molecules overlap [23]

At first a strong attractive force (LONDON VAN DER WAALS type) is holding the particles together,

which means they are agglomerated. Electric double layer repulsion is suppressed due to a

high electrolyte concentration. At the liquid interface the amphiphiles adsorb on the surface

of the particles as a brush type. Close to the contact of the particles the molecule tails will

start overlapping with a resulting repulsive pressure described best by the ALEXANDER-DE-

GENNES theory (AdG). Under certain circumstances, which are discussed below, the repulsive

33

forces will overcome the attractive force and deagglomeration occurs. In the next section a

geometrical model based on this gedankenexperiment is introduced leading to the

calculation of a disjoining force based on overlapping spheres and the AdG.

4.2.2 Geometrical Model

To make use of the AdG in eq. (20), for calculation of a repulsive force, the model geometry

of particles with equal radii R in contact with a layer of molecules is simplified and shown

in figure 23.

figure 23: Representation of the geometrical model of particles with radii R in contact, covered by a layer of

molecules with the thickness . The defined region of interest, which attributes to the repulsive force

is highlighted. Calculations are based on the distance of overlap of the layers D which are a function

of the angle α. At αmin D = and at αmax D = 0. [23]

As a logic approach for calculating the integral force in the region of interest in figure 23,

angles α are introduced. The integration limits αmin and αmax are defined in eq. (21).

R

a

RRRa

Ra

minmax,1minmax,

222

min

2max

sin2

42422

2

(21)

The distances amin and amax are horizontal lines reaching from the point of contact of the

hard spheres with radius R to the point where the vertical distance between the hard

spheres equals and where the circumferences of the (R + ) spheres intersect, respectively.

The distance D between the true particle surfaces (without the adsorpt layer) can be

formulated as a function of the introduced angle α in eq. (22).

2

cos12RD (22)

34

Combining the AdG in eq. (20) with eq. (22) leads to eq. (23).

4

3

4

9

3 2

2cos12

2cos12

2)(

R

Rs

TkP (23)

This is an expression for the local disjoining pressures at a given angle α. For α ≥ αmax the

pressure is zero and at the angle αmin the pressure should be the strongest.

Consequently, the repulsive disjoining force Frepulsion is calculated, integrating the local

pressure P(α) times the differential angle dependent area dA(α), which the local pressure is

acting on from αmin to αmax, in eq. (24).

max

min

)(d)(repulsion

APF (24)

Again by applying sphere geometrics, the differential angle dependent area dA(α) is assessed

with eq. (25). This area describes an infinitesimal small ring segment on the hard sphere

surface with radius R.

2cos

2cos

2cos)(22

sin2

22)(d

2

2

2

2

dαRR

dxxD

RA

d

(25)

Due to the fact that the differential dα is part of the cos-function, an analytical result of the

integral in eq. (24) is not presented.

To account for the attractive force between the particles, a simple analytical expression of

VAN DER WAALS interactions is used, which disregards contact deformations (HERTZ, JKR, DMT)

as well as repulsion from electrostatic interactions or hydration forces [110], given in

eq. (26).

20

Hder Waalsvan 12 a

RCF

(26)

For the HAMAKER constant CH a value of 2.2 E-20 J is assumed which has recently been

reported for magnetite in organic solvents [135]. The radius of the particle R is taken to be

7.5 nm in accordance to the actual size of the particles in this work (cf. A.3.3). The contact

distance a0 shall be 0.2 nm in agreement with [110]. This yields a force of

Fvan der Waals = 3.4 10-10 N.

35

The repulsive force in eq. (24) will only be successful in deagglomerating the particles, if it is

larger than the absolute attractive force in eq. (26). It is therefore convenient to introduce a

relative repulsive force x(Frepulsion) in eq. (27).

der Waalsvan

der Waalsvan repulsion

repulsion F

FFFx

(27)

If this relative force of repulsion is higher than zero, the particles will deagglomerate. On the

other hand the particles will remain agglomerated in the organic phase after

hydrophobization occurred for values below zero.

4.2.3 Numerical Results

The numerical results of the expressions in 4.2.2, presented in this paragraph, are obtained

using Matlab® version R2011b. At first, the local pressure as a function of the angle α and

the parameter layer thickness is visualized in figure 24 for s = 0.5 nm and R = 7.5 nm. This

adsorption distance of s = 0.5 nm corresponds to a surface coverage of = 4 molecules/nm2

and with this a molecule surface head area of 25 Å2, a typical value found in the literature for

fatty acids [21, 95, 96, 110]. The range of the layer thickness corresponds to the dimensions

of stretched fatty acids, as presented in table 2.

figure 24: Angle dependent pressure between the spheres with radii of 7.5 nm and the parameter layer

thi k ess i steps of 0. nm, from 0.4 nm to 2.4 nm, for a rather high degree of adsorption of

s = 0.5 nm according to eq. (23), [23]

The left and right end of each curve in figure 24 is located at αmin and αmax, respectively. The

highest pressure is occurring at αmin with an increase for larger .

In figure 25 the results of the integration of eq. (24) using eqs. (23)and (25) are presented as

a function of the adsorption distance s and the layer thickness . For comparison the

absolute value of the VAN DER WAALS attraction force is included in the graphs as a horizontal

line. The repulsive force increases with increasing and thus longer molecules and with

higher surface coverage of the molecules, i.e. smaller s. It is important to note that both

36

values are not only a function of the end-grafted molecule geometry but also on the

solubility dependent conformation, whether they are fully stretched.

figure 25: Repulsive forces according to eq. (24) (left) as a function of the adsorption distance s and with the

para eter of la er thi k ess , right as a fu tio of the la er thi k ess ith the parameter

adsorption distance s in comparison to the constant absolute VAN DER WAALS force (horizontal line, cf.

eq. (26)), [23]

The following figure 26 shows a three dimensional graph of the relative repulsion in eq. (27)

with the variables s and . In the black triangular on the bottom left, the particles are not

disjoined, for the repulsive force is lower than the attractive force.

figure 26: relative repulsive force (cf. eq. (27) as a fu tio of adsorptio dista e s a d la er thi k ess , [23]

Certainly eq. (26) is a very simple expression for the attractive force as a level, where it is

decided, whether the particles are disjoined or not. The actual attractive force could deviate

in both directions. Yet it is fascinating that the repulsive disjoining force of the introduced

model has a similar order of magnitude and could be reliable. In future research studies it

could hence be tried to evaluate such a disjoining force by a smart experimental set-up, e.g.

using colloidal probe microscopy [136].

37

In the next paragraph the phase transfer of magnetite nanoparticles from water to the

solvent dichloromethane (DCM) is investigated for the fatty acids (FA) presented in table 2.

4.3 Experimental Results

Following the experimental procedures for nanoparticle synthesis and phase transfer in the

appendices A.3 and A.4, dichloromethane based magnetite colloids have been prepared with

the fatty acids in table 2. In 4.3.1 the phase transfer results are visualized to draw first

conclusions on the colloidal stability and the success of the deagglomeration, described

above. The agglomerate size distribution of the samples is investigated in 4.3.2. Derived

from complex colloidal investigations in 5.3.2, the primary particle concentration is

presented in 4.3.3 with a discussion on the solubility of the molecules. Finally, in paragraph

4.3.4 the adsorption of ricinoleic acid is investigated with a thermal gravimetric and

spectroscopic method.

4.3.1 Visualization of the Phase Transfer

In figure 27 six test-tubes with completed phase transfers of magnetite from the lighter

water phase to the heavier immiscible DCM phase using the fatty acids in table 2, are

presented. Due to reproducibility difficulties with the precipitation process (cf. 5.3.1), only

one magnetite precipitation batch is being used for all samples of this study. The first test-

tube on the far left displays the case when no fatty acid is dissolved in DCM and the particles

will settle to and remain on the interface because of their hydrophilic surface characteristics.

Consequently the liquid-liquid interface is found in the middle part of each test tube,

pointed out on the right test tube. All fatty acids adsorb at the particle interface and lead to

hydrophobization and thus ability to enter the non-aqueous dichloromethane phase.

One can notice sediments in the solvent phase for the four tubes to the right with OA, MA

and CA. Furthermore there are stable organic solvent soles with no visible sediment

formation for the second and third tube with RA and LA as well as a stable aqueous sole for

the third tube to the right with MA and an aqueous pH of 9.0. For RA and LA this concludes

that the agglomerates of the aqueous phase are completely disjoined when the fatty acid

molecules adsorb at the interface. For the other systems it seems like the majority of the

agglomerates of the aqueous phase are preserved and settle within the organic phase. The

special feature for MA is that a stable nanoparticle dispersion is formed in the aqueous

phase at high pH of 9.0.

38

figure 27: Vials of phase transfers of magnetite nanoparticles from an aqueous phase (upper half) to a DCM

phase (below liquid interface) with the grafting molecules/surfactants as defined in the image above

the vials all at pH 9 in the aqueous phase except for myristic acid with a second phase transfer at

pH 8 [23]

Gravimetric analyses show that only a small amount of roughly 10 % of particles is not found

in the solvent phase in this case. When the pH is reduced to 8.0 using 1M HCl all particles

enter the DCM phase and no stable aqueous sole is formed. An explanation can be, that

adsorption takes place at the interface and dissociated not adsorbed molecules enter the

upper aqueous phase forming a bilayer around already hydrophobic nanoparticles and

sterically stabilizing those in the high ionic aqueous phase. The bilayer formation should also

be possible for CA since it is even smaller and should thus have a similar chance to enter the

aqueous phase entirely as it is mentioned in [99]. Yet this might be the case but due to the

short length of the molecule stabilization in water may not be possible by bilayer formation

and settling and reentering the organic solvent phase will dominate.

Due to the intense light absorbing properties of magnetite nanoparticles, cf. appendix A.8, it

is not clear whether there are agglomerated nanoparticles also for the RA and LA samples

and if those two samples have a different content of deagglomerated supernatant

nanoparticles.

4.3.2 Particle (Agglomerate) Size Distribution

In addition to the photographs and observations above the intensity weighted particle size

distribution of the agglomerates in the water phase (solid line without symbols on the right)

as well as the transferred nanoparticle assemblies (lines with symbols) using the analytical

centrifugation method are presented in figure 28. The line on the left represents a dynamic

light scattering result of the stable supernatant after centrifugation with only primary

particles present. Dynamic light scattering must not be applied for the agglomerate

containing samples for it does not lead to reliable and reproducible information, cf. A.10.

39

figure 28: Intensity weighted particle size distributions of the samples in figure 27 applying analytical

centrifugation with a cut-off size of 30 nm, additionally the primary particles distribution is

determined from the stable colloid using DLS [23]

The median size of aqueous agglomerates is 1.3 µm and in any case larger than the particles

found after phase transfer with median size values of 0.3 µm, 0.5 µm, 0.6 µm, 0.7 µm and

1.1 µm for RA, LA, OA, MA and CA, respectively. There is an unknown fraction of particles

smaller than the resolution limit (at about 30 nm) of the analytical centrifuge which stands

for primary particles. This fraction is especially noticeable for RA and LA. The size of these

particles is obtained by analyzing the particles in the supernatant after centrifugation with

DLS, resulting in a median intensity weighted size of 24 nm.

The quantitative assessment of the weight fraction of these primary particles wPrimary is

presented in the next paragraph, when assessing the colloidal stability in mixtures with

PMMA, derived from results in 5.3.2.

4.3.3 Primary Particle Concentration

The primary particle mass concentration wPrimary is a quantitative measure, which can be

experimentally assessed in nanoparticle polymer mixtures, following the procedures in

appendix A.5. The colloidal stability investigations presented in 5.3.2 can, by means of

extrapolation using eq. (61), deliver the primary particle mass concentration under the

absence of a dissolved polymer.

For the phase transfer samples introduced above, the initial primary particle concentration

w0

Primary, without dissolved polymer, is presented in figure 29.

40

figure 29: Mass concentration of primary particles after phase transfer derived from colloidal interactions

studied in 5.3.2, values presented in table 8, [23]

The values correlate well with the agglomerate sizes x and primary particle fractions

Qint(xPrimary) in figure 28. What the images in figure 27 cannot visualize though, is the

significant difference of stable primary particles of the RA and LA samples. Ricinoleic acid

achieves to deagglomerate almost all of the aqueous agglomerates, whereas the other C18

acids linoleic and oleic acid with similar maximum chain length max (cf. table 2) lead to less

primary particles, thus are less efficient in deagglomerating. Following the physical model of

deagglomeration in 4.2, either the grafting density is lower for LA and OA (and with this the

distance between adsorpts s would be higher), or the actual layer thickness is lower, due

to reduced solubility and thus less stretching, cf. 5.1.2. Obviously both could be the case as

well. MA and CA are smaller molecules and thus lower disjoining forces, due to smaller are

expected in any case.

In order to judge on the solubility of the end-grafted fatty acids on the magnetite

nanoparticles (FA-Fe3O4) in the solvent dichloromethane (DCM), the HANSEN solubility

distance DFA-Fe3O4-DCM given by eq. (28) is evaluated [70, 137-139].

2DCMh,OFe-FAh,

2DCMp,OFe-FAp,

2DCMd,OFe-FAd,

DCMOFe-FA

43

43

43

43

4

D (28)

For this the HANSEN solubility parameters (HSP) relating to disperse, polar and hydrogen

bonding interactions d, p and h, respectively, are determined by calculations presented in

A.11. Since the carboxyl group is chemically attached to the magnetite surface it is more

convenient to calculate the HSPs without this group (FA-Fe3O4). Disperse parameters d are

similar for the components presented and therefore not depicted in the p - h solubility plot

in figure 30.

41

figure 30: Solubility plot of the fatty acids pristine (FA - filled blue symbols) and grafted to the magnetite surface

(FA-Fe3O4 - open blue symbols) calculated using the group contribution method in A.11, compared to

the solvent DCM (red circle), values from [140]

On the solubility plot, a better solubility is represented by a shorter distance between the

fatty acid species and the solvent DCM. A better solubility results in the fatty acids to be

more stretched towards the solvent, because fatty acid – solvent interactions are more

profitable from a thermodynamic perspective. This is the case especially for ricinoleic acid,

which is due to the OH-group and the double bond. Calculated solubility distances are listed

in table 3, together with the median size x50,int and primary particle fraction Qint(40 nm) from

the size distribution in figure 28 and the initial primary particle mass fraction w0

Primary in

figure 29. Furthermore, using eq. (52), the FLORY interaction parameter is calculated and

listed in table 3, as well. More details on this parameter are given in 5.1.2. Generally, it is to

note here, that the smaller this value, the better the solubility, thus the more favorable the

fatty acid – solvent interaction. A value of χ = 0.5 is found for equal interactions of fatty acid

– fatty acid and fatty acid – solvent.

table 3: median particle size x50,int, fraction of primary particles from the intensity weighted distribution Qint,

primary particle concentration w0

Primary, calculated solubility distance between the fatty acid coated magnetite

(FA-Fe3O4) and the solvent (DCM) DFA-Fe3O4 – DCM using eq. (28) and A.11 as well as the calculated FLORY-HUGGINS

parameter χ using eq. (52)

fatty acid x50,int

in µm

Qint(40 nm)

in %

w0

Primary

in %

DFA-Fe3O4 – DCM

in MPa-1/2

χ

in -

RA 0.3 30.8 96.3 ± 2.3 2.98 0.03

LA 0.5 7.1 35.1 ± 4.4 7.74 0.23

OA 0.6 1.9 8.0 ± 1.8 8.16 0.29

MA 0.7 1.6 2.1 ± 0.7 10.08 0.39

CA 1.1 1.3 2.8 ± 0.7 10.16 0.40

42

All end-grafted fatty acids have a FLORY interaction parameter smaller than 0.5. However, the

best solubility is found for magnetite with grafted ricinoleic acid. The correlation of initial

primary particle concentration w0

Primary and HANSEN solubility distances DFA-Fe3O4 – DCM is

presented in figure 31.

figure 31: Correlation of primary particle concentration given in figure 29 and the solubility distance between

the fatty acid capped magnetite and DCM calculated using eq. (90) and values found in A.11

As mentioned above, the solubility determines the actual layer thickness , used for the

model in 4.2. For same max, as is the case for RA, LA and OA a lower solubility would reduce

stretching behavior and thus the actual layer thickness . The impact of grafting density and

therefore adsorpt distance s would in the future need to be evaluated by adsorption

isotherms of the different fatty acids, especially for the C18 specimens RA, LA and OA, which

has not been done for the studies presented here.

4.3.4 Inert Decomposition of Chemisorbed Ricinoleic Acid on

Magnetite Nanoparticles

In this section, the previously published investigation on the thermal degradation of

ricinoleic acid capped magnetite nanoparticles using ATR-FTIR, FTIR coupled TGA and XRD is

reviewed [21]. The study is part of this thesis’ research program, evaluating the composition

of composites of fatty acid capped magnetite nanoparticles with polymers using thermo

gravimetry. It is found that chemically grafted ricinoleic acid causes the reduction of

magnetite at temperatures between 600 °C and 900 °C in inert atmosphere and the best

measure for the magnetite content is the residual mass at 600 °C, as opposed to higher

temperatures, which is often found in the literature.

Starting Point Hypotheses from the Literature

In several references, fatty acid capped magnetite nanoparticles are analyzed with TGA and

different conclusions are drawn on residual masses at temperatures of 600 °C or 900 °C after

heating in inert atmosphere [39, 85, 95-97, 105, 115, 141-148]. In most cases the different

43

mass losses, usually three, are attributed to physic- and chemisorbed bilayers of the fatty

acid [95, 105, 142-148]. The residual mass is in the majority of the investigations attributed

to magnetite. However there are investigations that prove that there is a very dominant

reduction of magnetite in inert atmosphere and vacuum for temperatures higher than about

600 °C depending on the size of the magnetite crystals and with this on the surface area of

the nanoparticles [97, 115]. Along with the reduction of magnetite a severe evolution of CO2

is observed. The reduction must therefore be achieved by residual carbonaceous

compounds at this temperature which were reported by ROONASI and HOLMGREN [96] when

looking at the IR spectrum of residues derived after heating to 550 °C. This concludes that

neither the residual mass at 600 °C nor at 900 °C will represent the magnetite concentration.

Furthermore it is evidence that the decomposition mechanism proposed by ZHANG et al. [95]

which is very often cited is not consistent.

Molecular Vibration Spectroscopy of RA, Fe3O4 and RA-Fe3O4

Before looking at the thermal degradation of ricinoleic acid capped magnetite (RA-Fe3O4),

ATR-FTIR is applied to investigate the type of binding of RA to magnetite, cf. 4.1.2. For this,

the IR spectra of pristine ricinoleic acid, pristine magnetite and ricinoleic acid capped

magnetite samples are presented in figure 32 for the wavenumber range of 1900 cm-1 to

1000 cm-1.

figure 32: ATR-FTIR results of pristine ricinoleic acid, pristine magnetite as well as ricinoleic acid adsorbed on

magnetite [21]

For pristine magnetite (low absorption of IR) there are weak bands at 1630 cm-1 and roughly

1120 cm-1, which can be attributed to surface bound H2O, because the sample is measured

in ambient environment and can be compared to the findings of CAI et al. [149]. The distinct

band at 1700 cm-1 for ricinoleic acid is due to the vibration of the carbonyl C=O as part of the

carboxyl-group. This band is significantly reduced for the RA-Fe3O4 sample. The new bands

for ricinoleic acid bound magnetite at about 1530 cm-1 and 1425 cm-1 reveal the chelating or

monodentate mononuclear chemical binding of the RA from its carboxylate on the

44

magnetite surface [116]. Consequently, it is affirmed that ricinoleic acid chemically adsorbs

on the magnetite surface.

TGA of Ricinoleic Acid, Pristine and Physically or Chemically Bound

Pristine ricinoleic acid (RA), ricinoleic acid coated magnetite by the phase transfer procedure

(RA-Fe3O4) with mRA/mFe3O4 = 0.2 and ricinoleic acid mixed with fumed silica Aerosil® 200

with 12 nm primary particle size (RA-SiO2) with mRA/mSiO2 = 0.2 are heated in a pure nitrogen

atmosphere from room temperature to 900 °C at 20 K/min. The differential mass losses are

presented in the left graph in figure 33. In the right hand graph, the thermogram of RA-Fe3O4

is depicted. Fumed silica is used as a material with a similar particle size like the magnetite of

this study, yet chemical adsorption of fatty acids has not been reported and it is assumed to

not occur.

figure 33: (left) first derivative of the TGA results (DTG) of pure ricinoleic acid (RA), ricinoleic acid coated

magnetite (RA-Fe3O4) and Aerosil® 200 (RA-SiO2) in inert atmosphere [21], (right) TGA of RA-Fe3O4

with the mass losses at the three distinct steps, the error bars show the 95 % quantile of three

measurements

The decomposition (evaporation) of the pure fatty acid takes place in a single step at a peak

temperature of about 350 °C. The decomposition however is slightly broader towards

smaller temperatures. Gases leaving the gas FTIR analysis at this temperature have a distinct

scent of ricinoleic acid.

For both particulate samples there are three distinct steps of decomposition, and the first

two match in temperature range. The first decomposition is located at lower temperatures

than the decomposition of pristine RA. The second step coincides with the pure fatty acid.

RA is entirely decomposed up to 400 °C, whereas a third decomposition is occurring for the

RA at SiO2 at around 460 °C and the majority of the RA-Fe3O4 degradation is happening at a

temperature of about 800 °C. The residual concentration at 600 °C for RA-SiO2 is 79% and

thus only slightly lower than the specified value of 83.3% which is most probably due to

surface bound water, since fumed silica very hygroscopic. The residual content of RA-Fe3O4

45

at 900 °C is 71.19 % and much lower than the magnetite content, concluding that some are

even all of the magnetite must be reduced.

The investigation of the FTIR of the evolving gases in the individual steps in the next

paragraph allows for a better judgment on the mechanism of decomposition.

FTIR of Evolving Decomposition Gases

The gas phase FTIR spectra of the major decomposition steps for the samples of figure 33

are presented in figure 34 in the wavenumber range of 4000 cm-1 to 1500 cm-1, including the

ATR-FTIR spectrum of pristine RA.

figure 34: FTIR of the evolving gases for the major steps of decomposition of the three samples in figure 33,

(left) RA including the ATR-FTIR spectrum at room temperature in the top graph, (middle) RA-Fe3O4

and (right) RA-SiO2 [21]

Due to the before mentioned broader decomposition towards temperatures below 350 °C

two IR spectra obtained from the decomposition gases at 300 °C and 350 °C are evaluated

for pristine RA. The common features of both gas phase spectra are the dominant symmetric

and asymmetric C-H stretching modes for sp3-hybridized carbon at 2934 cm−1 and 2865 cm−1

and for sp2 hybridized carbon (double bond) at 3016 cm-1, the C=O carbonyl stretching

bands at 1778 cm−1 and 1741 cm−1 as well as the OH gas phase stretching (from the hydroxyl

group) at 3577 cm−1. These modes clearly point towards the ricinoleic acid molecules or

fractions of these molecules in the gas phase. However at the lower temperature of 300 °C

there also appears a mode at 2705 cm−1 which indicates C-H stretching at aldehyde

functionalities. An explanation for this is the autoxidation of the fatty acid resulting in

aldehyde and alkane volatile compounds with hydroperoxides as intermediates [150, 151].

For the hydroxyl group containing unsaturated ricinoleic acid there might also be alcohols

that are formed as volatiles of the autoxidation. The typical alkane as well as alcohol bands

are already resolved in the before mentioned gas phase ricinoleic acid bands. The effect of

46

autoxidation is well known from food chemistry and leads to degradation of fatty acids, they

go ad . Most of the gas phase at °C however is already made up of ricinoleic acid.

For RA-Fe3O4 at the lower temperature step with a peak value of about 255 °C in figure 34

(middle), it is to recognize that the spectrum shows a very strong evolution of aldehydes

identified with the C-H stretching bands at 2710 cm−1 and 2813 cm−1. In discrepancy to the

spectra obtained for the ricinoleic acid presented above the carbonyl stretching shows a

singular band at 1741 cm−1 assuming that the aldehyde content is much higher. This would

conclude that autoxidation is either catalyzed by the iron compound [150] or more effective

because of the high surface area exposed to atmosphere before TGA. A possible hydroxyl

band of alcohols and the ricinoleic acid at 3577 cm−1 is superimposed by stretching bands of

gaseous water. Furthermore there is a high signal of bands related to CO2 at roughly

2300 cm−1 originating from C=O stretching vibrations (which means a relatively low content

for CO2 is a highly absorbing molecule). In the second decomposition step with a peak

temperature of about 380 °C no more bands pointing to aldehyde are found. However the

carbonyl band at 1732 cm−1 is lower than the ones shown for RA. Furthermore one cannot

clearly identify strong C-H stretching for sp2 hybridized carbon from the unsaturated fatty

acid. It has been reported in [96, 97, 152] that dehydrogenation also starts to take place in

this temperature range which obviously cannot be detected in the IR spectra. It is assumed,

that in the second step physically bound fatty acid is detached and evaporates but the

unsaturated bonds disappear in the gas phase because of reactions with the hydrogen which

results from dehydrogenation of chemically grafted acids. As a consequence of

dehydrogenation, carbonaceous residues must be left on the nanoparticle surface, which are

able to reduce magnetite at a peak temperature of about 760 °C. The gases evolved at this

reaction are COx. This is also what is found looking at the evolved gas spectrum. In the range

of 2360 cm−1 and 2309 cm−1 CO2 stretching bands and for 2183 cm−1 as well as 2108 cm−1 CO

stretching bands dominate. For the last step of degradation the following reaction pathway

is proposed in eq. (29).

FeOFeCOCOCOFe edc2bcb43a nnnnnnn (29)

In [21] the molar numbers na through ne have been determined and presented in table 4.

table 4: Factors given in equation (29) with errors representing 1.96 times standard deviation, the resulting

oxygen to iron ratio in the iron oxide after reduction is calculated to be 0.76 ± 0.09, compared to the value in

pristine magnetite which is 1.33.

na 1.89 ± 0.29 nd 1.36 ± 0.27

nb 1.05 ± 0.17 ne 4.32 ± 1.01

nc 1.15 ± 0.32

47

For RA-SiO2, in figure 34 (right), one special feature is the strong absorption bands of

desorbing water which has been bound to SiO2. For the first two decomposition steps the C-

H stretching bands between 3000 cm−1 and 2700 cm−1 are similar to the ones presented for

RA-Fe3O4. This concludes that again autoxidation of RA plays a role and must be due to the

high surface area of the fumed silica which is accessible for oxidation of the fatty acid. For

RA-Fe3O4 however, the aldehyde related stretching bands are more pronounced which must

be due to the possible catalyzing effect of iron. Contrary to the last step of decomposition

for RA-Fe3O4 a band at 3087 cm−1 appears, which is due to stretching vibrations of C-H at sp2

hybridized carbon. This points towards the missing dehydrogenation of chemically grafted

RA molecules and rather the strong physically adsorbed RA is detached from the SiO2

surface. The occurrence of CO2 and CO vibrations is only weak when compared to the last

step of RA-Fe3O4. In summary the decomposition of RA-SiO2 is following the order: first

release of volatiles of RA autoxidation at 290 °C, then decomposition and detachment of

loosely bound RA at 370 °C and finally at 460 °C desorption of tightly, yet physically bound

ricinoleic acid.

Powder X-Ray Diffraction of Pristine Fe3O4 and the TGA-residue FeOx

In this paragraph the results of the powder diffraction analysis of pristine magnetite after

the precipitation and of the iron oxide residue after TGA in inert atmosphere to 900°C and

cooling to room temperature, are presented. This is important to support the findings above

and especially for verification of the reaction mechanism in eq. (29). Both diffractograms are

depicted in figure 35.

figure 35: Powder diffractograms of pristine precipitated and fatty acid grafted magnetite RA-Fe3O4 and of the

mixed iron oxide residue FeOx after inert gas TGA with identified major diffraction angles (2Θ) of

magnetite Fe3O4, wüstite FeO, Ferrite α-Fe and Hematite Fe2O3. [21]

Before the thermal analysis, the only crystal structure identified is magnetite Fe3O4 with

broad diffraction peaks due to the small crystallite size. This size is calculated to be 14.8 nm

using the Williamson-Hall method, cf. A.3.1. As expected from the argumentations above in

the TGA residue wüstite and ferrite are identified amongst magnetite and hematite. The

48

occurrence of magnetite and hematite can have different reasons. Magnetite found could

either be unreduced specimens, a product of maghemite reduction or a product of oxidized

residue which might happen between TGA and XRD analysis. Hematite can be obtained by

the following solid state reaction in eq. (30) [97].

32OFeFeFeO3 (30)

In order to check the before discussed mass balance to quantify the reduction reaction in

eq. (29) the mineral composition in terms of mass and molar concentrations of the four

identified minerals are presented in table 5. The two major components (molar

concentration) are wüstite and ferrite, as expected. The occurrence of magnetite and

hematite has been addressed above, however it is found that the oxygen to iron ratio is 0.99

and thus higher than in the mass balances for the TGA results in table 4. The deviations

might be caused by oxygenation reactions before XRD and after TGA. If one would neglect

the occurrence of magnetite, the oxygen to iron ratio would be 0.78 and within the margin

of error of the result in table 4.

table 5: Compositional minerals analysis of the FeOx residue, which is made up of four identified mineral

structures. The mass concentrations are calculated from the diffractogram using the XRD device software and

the olar o e tratio s follo fro φ= i/Mi ⁄∑ wi/Mi), the resulting iron to oxygen ratio is 0.99

Mineral i Formula Molar weight

Mi in g/mol

Mass concentration

wi in -

Molar concentration

φi in -

Magnetite Fe3O4 231.53 0.41 0.18

Wüstite FeO 71.84 0.32 0.46

Ferrite Fe 55.85 0.15 0.28

Hematite Fe2O3 159.66 0.12 0.08

It has to be mentioned, that there is also the possibility to obtain maghemite -Fe2O3 even in

the RA coated sample before TGA, however it cannot clearly be distinguished from

magnetite using XRD analyses only. However oxidation of the precipitated magnetite is

expected not to play a significant role, for fresh samples have been investigated and

oxidation has been proven to be slow [153]. From [153] it can also be deducted, that the

mixed iron / iron oxide, formed by reduction, might be performed, resulting in a core-shell

structure with the magnetite in the center of the residual compound.

49

5 Nanoparticles and Polymers in an Organic

Solvent

In chapter 4 it was shown, that by a phase transfer process it is possible to graft the

amphiphilic fatty acids chemically on the surface of magnetite nanoparticles. This leads to

hydrophobization and steric stabilization of the particles in a water immiscible organic

solvent. The right choice of molecule (high grafting density with a small adsorpt distance s

and thick adsorption layer with the length ) furthermore achieves to physic-chemically

deagglomerate the particles into primary particles with a few nanometers in diameter.

Ricinoleic acid has proven to be the best choice when preparing a stable dichloromethane

based organosol with a primary particle concentration of more than 95 % by weight. Such a

colloid can be chosen as the source to prepare nanoparticle-polymer-composites by mixing

with a dissolved polymer and subsequent drying.

In this chapter it will be shown, that it is by far not a simple task to preserve the state of

dispersion when mixing a stable colloid with a polymer solution, as introduced as an

important step of the process chain to prepare nanoparticle-polymer composites in 3.1.3.

The resulting complex colloid underlies special types of interaction, which are being

introduced theoretically and assessed experimentally.

Classic colloid stabilization by electric double layer interactions is not of importance in such

systems, however, in 5.2 a DLVO-like treatment of the governing interactions is presented.

Beforehand it will be introduced how polymers behave in solution in 5.1.1. The solubility of

polymers and particles in solvents are defined in 5.1.2 and the phenomena occurring in

solvent based nanoparticle-polymer mixtures are pointed out in 5.1.3.

The state of dispersion is experimentally assessed in the experimental section in 5.3. At first,

the influence of the polymer is investigated for the well dispersed ricinoleic acid coated

magnetite nanoparticles with the polymers poly(methyl methacrylate) PMMA,

poly(bisphenol A carbonate) PC and poly(vinyl butyral) PVB in 5.3.1. The influence of

different fatty acids and different solvents in presented in 5.3.2 and 5.3.6, respectively.

Section 5.3.3 is addressing the mechanisms of stabilization when using PVB. Finally, a

preliminary study on the kinetics of flocculation is presented in 5.3.5.

50

5.1 Theory

5.1.1 Polymers in Solution

This section introduces the behavior of long chain molecules in solution, namely

thermoplastic homopolymers, which are not cross-linked. These macromolecules will

typically form coils, which can be described as soft colloids, with dimensions depending on

the total chain length, as presented by the molar mass of a polymer, the segment length,

and the interaction between polymer segments and the solvent molecules as well as the

polymer concentration [109, 154-157].

Thermodynamics

The dissolution of polymers in solvents is a classic thermodynamic problem of mixing, with

the Gibbs free energy of mixing Gmix given in eq. (31) [109, 111, 156, 158, 159].

mixmixmix STHG (31)

From statistical thermodynamics, as part of the FLORY-HUGGINS theory, the entropy and

enthalpy of mixing are defined in eqs. (32) and (33), respectively [109].

TkNH polymersolventmix (32)

polymerpolymersolventsolventmix lnln NNkS (33)

In these equations Nsolvent and Npolymer are the number of solvent and polymer molecules,

φsolvent and φpolymer are the volume concentrations of the solvent and the polymer and χ is

the FLORY HUGGINS interactions parameter. This is the only parameter which is component

specific, depending on the material combination of polymer and solvent.

Combining eqs. (32) and (33) with eq. (31) leads to eq. (34).

polymersolventpolymerpolymersolventsolventmix lnln NNNTkG (34)

The last term is enthalpic and the first two contribute to the entropic effect of mixing.

Multiplying eq. (34) with AVOGADRO’s u e NA and regarding φsolvent + φpolymer = 1 leads to

the partial molar free energy of mixing in eq. (35).

2polymerpolymer

solventpolymermmix

111ln

NTRG (35)

51

The osmotic pressure in a polymer solution can be defined by eq. (36) [109].

solvent

mix G (36)

After some expansion and with a series expansion eqs. (35) and (36) can be combined to

eq. (37) [158].

...2

11 2polymer2

spolymerpolymerm c

M

Nvc

MTR (37)

The second term of this virial equation of the osmotic pressure stands for the intermolecular

interactions. This term equals zero for χ = 0.5, is positive when χ = (0, 0.5) and negative for

χ > 0.5. Concluding, the following can be said about the FLORY interaction parameter.

- For χ = 0.5 the polymer-polymer interactions are equal to the polymer-solvent

interactions, which is often referred to as the Θ-state. The solvent temperature

where this relation is found is called Θ-temperature TΘ. Dissolved polymers will in

this case be deforming into random coils behaving as ideal chains, cf. figure 36.

- Typically when the temperature of a solvent is larger than the Θ-temperature, the

solubility of a polymer increases, which means that the polymer-solvent interactions

are in favor. This is the case for χ = (0, 0.5) and the solvent is then called a good

solvent. The polymer coils will eventually expand.

- The opposite is true for a bad solvent, where χ > 0.5 and T < TΘ. In this case polymer-

polymer interactions are in favor and the coil would contract, or not even dissolve in

the first place.

Below in 5.1.2 it will become obvious that the FLORY-HUGGINS parameter is related to the

solubility of a polymer in a solvent. But first it will be shown how the coil parameters can be

described using the FLORY theory.

Coil Dimensions

In figure 36 a polymer coil in solution is schematically depicted for different polymer-solvent

to polymer-polymer interactions.

52

figure 36: Polymer coils in solution for different polymer solvent interactions (left) real chain in a good solvent,

(middle) ideal chain with equal interaction of polymer segments and solvent molecules, (right) real

chain in a bad solvent. The s ols a ou t for: χ – the FLORY interaction parameter, RG – the radius

of gyration, ls – the segment length, N – the number of monomer units and TΘ – the theta

temperature

The i age i the iddle ep ese ts the so alled ideal hai , hi h is a a do oil, he e there are equal polymer-polymer and polymer-solvent interactions, the so called Θ-state.

Fu the o e the seg e ts do ot feel ea h othe along the polymer backbone [160].The

FLORY interaction parameter χ, describing the interaction, equals 0.5 in this case, cf. eq. (37).

A measure for the size of the coil is the radius of gyration RG, which for an ideal chain is given

by eq. (38) [161].

21sG 6

1sNlR (38)

Here ls is the length of a segment or sometimes also referred to as KUHN length or

persistence length. Consequently Ns is the number of segments in the polymer chain. The

entire length of a polymer, walking along its backbone, is given by the contour length Lc in

eq. (39).

00

0ssc lM

MlNlNL (39)

Usually the molar masses M and M0 of the polymer and its monomer units, respectively, are

given. Derived from this is the actual number of monomer units N, the degree of

polymerization. The length of one unit is l0 and equals 0.25 nm for vinylic monomers, such as

present in PMMA. Depending on the flexibility of the polymer, due to the monomer-

monomer interactions, the segment length is a multiplicity of the unit length. For PMMA,

which is one of the polymers used in this study ls/l0 = 8.7 [118].

Real Chains can be described with the excluded volume v, the volume around a chain

segment which is not accessible for another segment of the same chain. It is positive for

good solvents when the segments repel each other and can often be approximated with

v = ls3. For bad solvents the segments attract and the excluded volume is defined as negative.

53

As a consequence of the real segment interactions, in a good solvent the coil will be swollen.

In a bad solvent the coils shrink, in a sense the polymer segments approach each other

e ause the sol e t is too old .

The radius of gyration obviously also depends on the degree of polymerization and thus N,

as well as M [118]. For ideal chains at the Θ-state the radius of gyration increases with the

square root of the molar mass and RG ~ N1/2, cf. eq. (38). The swollen coils in a good solvent

increase less degressively with the molar mass and a good approximation is RG ~ N3/5 [109,

157, 161]. Finally the coils in bad solvents increase in size more degressively, with RG ~ N1/3

[162]. A general expression for the radius of gyration in theta and better solvents using the

FLORY-HUGGINS interaction parameter χ is given in eq. (40) [161].

5221

s21

ssG 215.611309.0 NNlR (40)

The relation normalized with the segment length ls is depicted in figure 37. It shows that the

radius of gyration RG is a multiple of the segment length ls strongly increasing with the

number of segments Ns.

figure 37: Graphical visualization of the radius of gyration RG normalized with the segment length ls, depending

on the FLORY i tera tio para eter χ a d the u er of seg e ts NS, cf. eq. (40)

The radius of gyration is directly experimentally accessible with sophisticated neutron or

electron scattering techniques only [157]. However the size of polymer coils can also be

determined by means of dynamic light scattering, cf. A.2.1. With this method however the

hydrodynamic radius RH is measured instead. A relation of hydrodynamic radius and radius

of gyration is discussed in [163], represented in eq. (41). Notably, the hydrodynamic radius is

in any case smaller than the radius of gyration.

1,G

H XR

RX (41)

54

The ratio X is smaller in good solvents compared to theta solvents, with an average value of

X = 0.77.

Overlap Concentration

The behavior of polymers in solution very much depends on the concentration of the

polymer coils. Three regimes are typically distinguished and depicted in figure 38. In the

dilute regime the individual polymer coils, when homogeneously distributed are not in

contact. This occurs right at the overlap concentration c*polymer. Polymer solutions with

concentrations higher than the overlap concentration are referred to as semi-dilute and the

individual coiled chains will interpenetrate.

figure 38: Concentration regimes of polymers in solution (left) dilute solution, (middle) solution at the overlap

concentration c*polymer and (right) semi dilute regime in accordance with [156]

The overlap concentration, with [c*polymer] = g/l, can be estimated in terms of the so-called

intrinsic viscosity [ ] in eq. (42) [157].

1*polymer c (42)

General remarks on the viscosity of polymer solutions as well as the intrinsic viscosity are

introduced in the following section.

Viscosity

The increase of the dynamic viscosity of a dilute colloidal system with increasing volume

concentration of the dispersant φ is a famous EINSTEIN relation and presented in eq. (43). The

dispersant free dynamic viscosity of the pure solvent is given by 0.

5.210 (43)

This holds also true for dilute polymer solutions, for the polymer coils are nothing but soft

colloid particles [109, 158]. For polymer solutions, typically a specific viscosity sp is defined

with the relation in eq. (44).

55

polymer

3A

0

0sp

345.25.2 c

M

RN G (44)

To characterize a polymer solution, the specific viscosity is determined as a function of the

polymer concentration cpolymer. With this it is possible to assess the intrinsic viscosity [ ],

which is a function of the molar mass, as given by the MARK-HOUWINK-SAKURADA relation in

eq. (45) [156].

MK

cc M

polymer

sp

0polymer

l im (45)

Here the factor KM is a polymer-solvent pair specific parameter, which is e.g. listed in [154].

The MARK-HOUWINK-SAKURADA exponent α is a function of χ and equals 0.5 for theta solvents

or 0.7 – 0.8 for good solvents. For rigid polymers it is larger than unity [156]. Practically the

intrinsic viscosity can be determined as the y-intercept plotting sp/cpolymer against cpolymer.

Combining eqs. (44) and (45) leads to an expression to estimate the radius of gyration with

the intrinsic viscosity in eq. (46).

3

AG 10

3

N

MR

(46)

5.1.2 Solubility

In the previous paragraph the FLORY-HUGGINS theory has been introduced as a way to

describe the solubility of polymers in solvents. In general, the solubility describes how a

certain species (a number of atoms, molecules or even particles) tend to mix with another

species. This is typically achieved if the two species are similar, or in othe o ds alike , as i the fa ous sa i g like dissol es like . The ua tifi atio of the si ila it is ealized at first by the HILDEBRANDT approach [137, 154], introducing the HILDEBRANDT solubility

parameters t, a material constant which is the square root of the volume specific energy of

cohesion, as shown in eq. (47).

21

m

cohesivet

V

E (47)

This parameters, with [ t] = MPa1/2, is material specific and a function of the temperature.

The HILDEBRANDT approach now claims, that the likelihood of mixing of two substances is

higher, if the difference of their solubility parameters is small. The enthalpy of mixing in

terms of solubility parameters is then given by eq. (48) [137].

56

2polymert,t,solventpolymersolventmmix VH (48)

Even though widely used for polymer solubility in solvents, the HILDEBRANDT parameters

prediction of solubility are limited, for the main problem is that the cohesive energies and

thus the parameters are a sum of different types of interactions, namely dispersive Ed, polar

Ep and hydrogen bonding Eh. By regarding these contributions HANSEN introduces the so-

called HANSEN solubility parameters (HSP) d, p and h [137, 138, 164].

2222

hpdcohesive

hpdt

V

E

V

E

V

E

V

E

(49)

The improvement of the HANSEN approach is, that not only solubility of solubles but also

dispersibility of insolubles can be predicted [137]. Two substances are alike (e.g. a solvent

dissolves a polymer, a particle mixes well in a polymer matrix) if the distance of their

solubility D12 is small, with eq. (50).

2h,2h,12

p,2p,12

d,2d,112 4 D (50)

In A.11 the HSP of surfactants on the surface of particles are calculated using the group

contribution method. HSPs of solvents are well known and listed e.g. in [140]

Combining eqs. (32) and (48) it becomes obvious, that the HILDEBRANDT and the FLORY-HUGGINS

parameters are related which can be seen in eq. (51) [109].

2polymert,t,solventm

m,solvent

T

R

v (51)

With β typically taken as 0.35 [109]. In [139] a similar relation is given for the HSP, as shown

in eq. (52).

4T

212

m

solventm, D

R

v (52)

The e pi i al o e tio fa to α is determined to be approximately 0.6 [139]. The molar

volume of the solvent vm,solvent can be calculated dividing the molar mass by the specific

weight Msolvent/ solvent, which results in 63.86 cm3/mol for dichloromethane, used in this

study.

57

5.1.3 Phenomena in Nanoparticle Polymer Mixtures

The colloidal stability of dispersions of nanoparticles will dramatically be affected by addition

of dissolved polymers [124, 125, 161, 165-175]. Four general effects have to be distinguished

as shown in figure 39.

figure 39: Principle types of colloidal regimes between neutral nanoparticles and neutral polymers in an organic

solvent (a) depletion flocculation, (b) depletion stabilization, (c) bridging flocculation, (d) steric

stabilization, cf. [118, 126, 127, 134, 155, 160]

The polymer-nanoparticle interactions can lead to polymer adsorption or the polymer can be

non-adsorbing [155]. For both cases the state of dispersion can be improved by stabilization

or will decrease due to destabilizing effects. Non-adsorbing polymers typically destabilize the

nanoparticle dispersion by depletion [134, 160, 161, 165, 166, 176-178], which will be

emphasized in the following paragraph 5.2, schematized in figure 39a. Typically at low

concentrations of non-adsorbing polymers, depletion stabilization can occur as well [155,

170], shown in figure 39b. The polymers may also be able to adsorb on the nanoparticle

surface. In this case again stabilization as well as destabilization is possible. Stabilization can

be explained by introduction of steric repulsive interactions due to the adsorbed polymer, cf.

4.1.3 and figure 39d. This is the case if a polymer molecule only adsorbs on the surface of a

single particle. Bridging flocculation is a result of multi-particle adsorption of the polymer

molecules, due to many anchoring points in the polymer backbone [111, 118], depicted in

figure 39c.

The destabilizing depletion effects can furthermore take place when the polymer

concentration of adsorbing polymers (situation in figure 39d) is much higher than the

amount necessary for adsorption [125, 179]. In case the solvency changes, the effect of

58

stabilization or destabilization can also change within the four options presented [118, 180,

181]

Phase Separation due to Depletion

If a colloidal system is destabilized through depletion, case (a) in figure 39, not all particles

need to be destabilized but a phase of stable primary particles and a phase of flocculated

particles can co-exist [134, 161, 168, 169, 180, 182, 183]. This can occur certainly as well,

while polymerization takes place [85, 184]. In general, polymer blends exhibit phase

separation as well [157, 185]. The more polydisperse the polymer is, the higher the amount

of flocculated particles becomes, as discussed in [186] showing that the depletion effect is

increased with polydispersity. The structure of agglomerates as a result of depletion

flocculation is presented in [15, 187-190]. A quantitative description of the depletion

interaction is presented in the next paragraph in eq. (56) and figure 40.

5.2 Pair Interactions – DLVO-like Consideration

The well established DLVO theory, namely the addition of VAN DER WAALS and electrostatic

double-layer interactions, enjoys great popularity when it comes to explaining and predicting

colloidal stability in aqueous dispersions [110]. However, in non-aqueous dispersions with

low dielectric constant , where double-layer interactions are negligible, classic DLVO does

not represent the governing forces. Nevertheless, a DLVO-like addition of governing

interaction mechanisms will reveal minima and maxima in the interaction potentials, which

can be used to explain colloidal stability. This is based on the assumption, that the

interactions are independent from each other and additive.

For the experimental system in 5.3 with non-adsorbing polymers such as PMMA in a

dispersion of sterically stabilized magnetite nanoparticles, one has to consider the attractive

pair potentials of LONDON type VAN DER WAALS [191] and due to osmotic depletion of polymer

coils in the vicinity of the functionalized nanoparticle surface described by ASAKURA and

OOSAWA [134, 165]. The opposing repulsive interactions are due to steric-osmotic pressures

in either brush-type [110, 133] or mushroom-type [123] adsorbed polymer layers, cf. 4.1.3,

as well as the BORN mechanism of overlapping electron shells of surface atoms [192].

The following analytical definitions for the separate interactions are all multiples of the

energy kT. In figure 40 a schematic drawing of the mixed system of sterically stabilized

nanoparticles in a solution of polymer coils is presented.

59

figure 40: Scheme of two interacting particles of diameter 15 nm surrounded by grafted molecules with a length

of 2.0 nm in a solution of polymer coils with a diameter of 7.5 nm which do not adsorb at the particle

surface resulting in a depletion layer surrounding the particles (dashed circles) [23]

The dashed circle around the particles represents the depletion layer and is approximately

the size of the functionalized particle plus twice the polymer coil size. When these layers

overlap, the particles will be pushed closer to each other. The driving force for this attractive

interaction is the osmotic pressure in the zone of overlap, where the polymer coil

concentration is lower than in the particle surrounding volume, which is due to the depletion

of the coils.

The LONDON type VAN DER WAALS interaction WvdW between two spheres with radii R at a

contact distance D is presented in eq. (53), depending on the HAMAKER constant CH of the

particles in the solvent [191].

2

2

22H

vdW

2

4ln

2

2

4

2

6)(

R

D

R

D

R

D

R

D

R

D

R

DTk

CDW (53)

The radius of the nanoparticles investigated in this work is about 7.5 nm and CH is taken to

e . ∙ -20 J, as recently published by FAURE et al. [135].

An approximation for BORN repulsion is given in [192] for a sphere-plate interaction and

presented in eq. (54) with the parameter σ = 0.5.

77

6

Born

6

2

8

7560)(

D

DR

DR

DR

Tk

CDW H

(54)

The interaction of steric repulsion between brush polymer covered surfaces with the layer

thickness and adsorption distance s is defined in eq. (55) [110].

60

D

es

DW

2brush-steric 100)( (55)

This relation is valid for D/(2 ) = [0.2, 0.9] [110] and a simplified version of eq. (17). A good

approximation for the layer thickness is the length of the molecules, which are presented in

chapter 4 for the fatty acids used in this study. A commonly found literature value for s is

about 0.6 nm, which is equal to an adsorption area per molecule of 36 Å2.

The pair attraction potential between two particles in a polymer solution is presented with

eq. (56) [134]. This simple expression assumes the depletion layer thickness to be equal to

the radius of gyration of the polymer coil.

R

Dq

R

DqqDW

42

3

2)(

2

3depletion (56)

In this case the nanoparticle radius or hard sphere size is the nanoparticle crystal plus its

surface molecule layer. The variable q stands for the ratio of the radius of gyration of the

polymer coils RG and the nanoparticle radius (R + ). The dimensionless concentration φ,

given in eq. (57), is the relative volume concentration of the polymer coils in a nanoparticle

free system and can be defined with the polymer concentration cpolymer and the critical

concentration c*, where the coils start overlapping, defined in eq. (42).

*Polymer

c

c (57)

The critical concentration of overlap in return is roughly the inverse of the intrinsic viscosity

[ ], cf. eq. (42), which can be obtained with low concentration rheological experiments

[193]. As can be seen from the determination of the viscosity in A.9, the concentration for

overlap for the PMMA used in this study, dissolved in DCM, is 32.2 g/l.

In figure 41 all four interactions are presented together with the additive total interaction

Wtotal, combining eqs. (53) - (56) to eq. (58). The variables are defined in the figure caption

and base on magnitudes to be faced with in the present study.

)()()()()( depletionbrush-stericBornder Waalsvan total DWDWDWDWDW (58)

61

figure 41: DLVO-like addition of colloidal interactions with regard to VAN DER WAALS and depletion attraction (by

ASAKURA and OOSAWA AO) as well as BORN and steric-osmotic repulsion with R = 7.5 nm, RG = 4.0 nm,

s = 0.6 nm, = 1.5 and φ = 1.0, [23]

Furthermore in figure 42 the impact of grafting layer thickness (left diagram) as well as

elati e pol e o e t atio φ ight diag a o the total i te a tio are visualized.

figure 42: Distance dependent total interaction, following eq. (58), (left) for varying [23] (right) for varying φ

and constant parameters mentioned in figure 41.

Contrary to classic DLVO interactions, in the left graph in figure 42 there is no repulsive

interaction maximum which manifests stability. However, the steric osmotic repulsion leads

to a steep increase of interaction with the point of zero interaction increasing with the layer

thickness. An important feature of the colloidal behavior in this DLVO-like total interaction, is

the magnitude of the always present interaction minimum located at roughly twice the layer

thickness . With an increasing polymer concentration the absloute interaction minimum

increases almost linearly, remaining at roughly the same interactions distance.

Generally, the tendency to agglomerate, upon particle impact, will increase with a higher

absolute attraction minimum |Wtotal,min|. Furthermore, the resulting agglomerate structure

will be more compact, the closer this attraction is to the surface of the particles Dtotal,min. The

following table 6 summarizes the effects of the influencing parameters in eqs. (53)-(56) on

|Wtotal,min| and Dtotal,min and with this on the expected colloidal stability.

62

table 6: Impact of increasing parameters (the other parameters shall be constant) of the total DLVO-like

interaction on the colloidal stability as a consequence of a changing absolute attraction energy |Wtotal,min| and

distance Dtotal,min ↑↑↑/↓↓↓ progressi el , ↑↑/↓↓ li earl , ↑/↓ degressi el i reasi g/de reasi g , starting values for the parameters as good approximates [23]

parameter

increasing

(constant value)

unit |Wtotal,min| Dtotal,min colloidal

stability

R (7.5) nm ↑↑↑ ↓↓ ↓

CH ( . ∙ -20) J ↑ ↓↓ ↓

(1.0) nm ↓↓↓ ↑↑ ↑

s (0.5) nm ↑ ↓↓ ↓

RG (4.0) nm ↓ ↑ ↑

φ (0.5) - ↑↑ ↓ ↓

Consequently the most effective way to increase stability is realized using longer end-grafted

brush type molecules and thus thicker layers . This can also be achieved by improving the

solubility (solvent molecule interactions) so that the end-grafted layers stretch further,

which has also been discussed in 4.3.3. Besides, solvent polymer interactions are an

important factor influencing the polymer coil size, which increases with increasing solubility,

cf. [181] and eq. (40). However, this will also effect the polymer volume concentration,

which for a fixed cPolymer will increase due to an increase of the intrinsic viscosity [ ].

Furthermore, the solubility of adsorbed surfactants is interconnected with the maximum

length of the surfactant max and thus the layer thickness . Another important factor related

to the solubility of the polymers as well as of the tails of the end-grafted fatty acid molecules

is the temperature. Typically the higher the temperature, the better the solubility. Yet, on

the other hand, higher temperatures increase particle impact probability and strength of

impact, which could reduce the colloidal stability.

5.3 Experimental Results

This section is presenting experimental results assessing the colloidal stability of mixtures of

stable magnetite nanoparticles from the phase transfer in chapter 4 interacting with

dissolved polymers. The main parameter, which is focused on below, is the primary particle

concentration wPrimary, which is a quantitative measure and determined following the

procedures described in the appendix A.5. At first the influence on colloidal stability using

different polymers with technical quality is discussed in 5.3.1. Using different fatty acid

surfactants, corresponding to the study in 4.3, the impact on colloidal stability is investigated

in 5.3.2. Paragraph 5.3.3 presents an in-depth investigation on the mechanism of

63

stabilization using the polymer PVB. The following paragraph 5.3.4 addresses the question,

how the mechanical mixing procedure influences the destabilization using PMMA or PC.

Finally, 5.3.5 and 5.3.6 discuss preliminary studies on the coagulation kinetics for

destabilizing polymers, and the influence of different solvents on colloidal stability,

respectively.

5.3.1 Influence of the Polymer

In this section and the included subsections, the fatty acid surfactant used for magnetite

nanoparticle transfer to DCM and functionalization, is ricinoleic acid (RA). The reason for this

choice is, that in 4.3.3 it was shown, that RA-Fe3O4 has the highest initial primary particle

concentration and thus most stable system.

Poly(methyl methacrylate) and Poly(bisphenol A carbonate)

In figure 43 the impact of different PMMA concentrations on the spectroscopically and

gravimetrically determined primary particle concentration E600 nm and wPrimary are displayed,

respectively.

figure 43: Light extinction at 600 nm of diluted samples E600 nm and gravimetrically determined primary particle

concentration wPrimary as a function of the PMMA concentration at constant nanoparticle

concentrations; the inset is a photograph displaying the samples (b) only holding the primary

particles after centrifugation with increasing polymer concentration from left to right

Both lines decrease with increasing polymer concentration. This means that the PMMA coils

are destabilizing the nanoparticle dispersion and lead to coagulation of a fraction of the

nanoparticles, presumably by depletion. Yet the residual primary particle phase remains long

term stable. All (b) samples of this thesis (see explanation in A.5) have been stored for over

two years without any sediment or turbidity recognizable. As presented in A.1.3 and A.9 the

PMMA coil size is about half the size of the ricinoleic acid grafted nanoparticles, so the

parameter q in this system is smaller than one. At the highest polymer concentration of

52 g/l there are only about 35 % of primary particles in the mixture (lighter curve in figure

64

43), whereas without polymer addition the initial primary particle concentration is

approximately 90 %, approving the results in 4.3.3. Since both curves show a similar

decreasing trend, the expected linearity of E600 nm and wPrimary, as stated in appendix A.5.3 is

satisfied, which will be discussed below in connection with figure 47.

The stability results for PC are plotted in figure 44 with the same scaling like figure 43.

Similar to PMMA the primary particle concentration is decreasing with increasing polymer

concentration, yet this destabilizing effect is stronger. At a concentration of 52 g/l only about

10 % of the particles are stable primary particles and the (b) sample is much more diluted

compared to PMMA. The PC coil size, however, is similar to PMMA, cf. A.1.3 and A.9.

Therefore, if depletion is claimed as the mechanism of destabilization, which is assumed and

introduced in 5.2, the parameter q would not explain this difference. This concludes that the

elati e pol e o e t atio φ ust e the dete i i g pa a ete , for the nanoparticle

concentration is constant.

figure 44: Light extinction at 600 nm of diluted samples E600 nm and gravimetrically determined primary particle

concentration wPrimary as a function of the PC concentration at constant nanoparticle concentration;

the inset is a photograph displaying the samples (b) only holding the primary particles after

centrifugation with increasing polymer concentration from left to right

There are two reasonable possibilities to normalize the abscissae in figure 43 and figure 44

for a better comparison of differences in the destabilization for PMMA and PC. On the one

hand the rough estimations of overlap concentrations for both PMMA and PC in table 27

(see appendix A.9.1), with c*PMMA = 32.2 g/l and c*

PC = 52.6 g/l, can be used to calculate the

relative polymer concentration, yet the uncertainty of the overlap concentrations obtained

is very high. Nevertheless, this graph is presented in figure 45 (left). The second option is, to

compare PMMA and PC, by plotting the coil concentration defined as cpolymer/Mn, with

[cpolymer/Mn] = mmol/l. The corresponding graph is show in figure 45 (right). This strategy is

similar to the comparison based on the relative polymer concentration, since the coil

concentration defines the relative polymer concentration for polymers with the same coil

size.

65

figure 45: Representation of the data in figure 43 and figure 44 as a function of: (left) the relative polymer

concentration φ using the overlap concentrations determined with the intrinsic viscosity in A.9.1 and

(right) the number concentration of polymer coils cpolymer/Mn.

The obtained results for normalization are contradictory at first. By taking into account the

confidence range of the data points it is possible that at least the normalization towards the

coil concentration on the right leads to comparable results. Still remaining differences can be

due to different polydispersities of the polymers, which are reported to be PIPMMA = 2.14 and

PIPC = 2.56, cf. A.1.3. Another differing factor can be, that indeed the coil size for PMMA is

slightly larger compared to PC, which would increase the depletion effect at the same coil

number concentration, explaining the slight shift of the curves to the left for figure 45 (right).

The relative polymer concentration plot on the left in figure 45 is non-conclusive for the

mean values. At a similar coil size and a lower molecular weight for PC the overlap

concentration should be smaller for PC compared to PMMA. However, the experimentally

obtained overlap concentrations are not significantly different by statistics.

Regarding composite preparation, the six measurement points for the wPrimary curves above

(figure 43 thorugh figure 45)correspond to the filler concentrations F = (0.8, 0.7, 0.6, 0.5, 0.4,

0.3) from left to right for csolid = 0.05, cf. eq. (11). Consequently, when aiming at synthesizing

composites with a higher filler concentration F, the primary particle concentration wPrimary is

increasing, this is a positive effect for highly filled composites. However, when withdrawing

the solvent in the drying step of composite synthesis, the polymer concentration increases,

due to an increasing csolid. This would lead to depletion flocculation again, depending on the

time scale of drying and flocculation. In case drying is quicker than agglomeration, then the

state of dispersion and thus wPrimary should be preserved.

Poly(vinyl butyral)

The third investigated polymer is PVB. Its corresponding colloidal stability results are

presented in figure 45.

66

figure 46: Light extinction at 600 nm of diluted samples E600 nm and gravimetrically determined primary particle

concentration wPrimary as a function of the PVB concentration at constant nanoparticle concentration;

the inset is a photograph displaying the samples (b) only holding the primary particles after

centrifugation with increasing polymer concentration from left to right

Contrary to PMMA and PC, destabilization is not encountered but stabilization seems to

occur since there is an increase in the fraction of particles in the supernatant. Furthermore

the increasing trend between the two curves differs, so the E600 nm to wPrimary relation for the

PVB stabilization must be different compared to PMMA and PC. The extinction of the diluted

(b) samples increases strongly, which is supported by the visual impression of the

photograph in figure 46.

Only judging on this plot, it is not quite sure whether agglomerates of the pristine dispersion

are deagglomerated physic-chemically by adsorption of the PVB on the particle surface,

similar to the mechanism introduced in 4.2 or whether these agglomerates are captured in

larger PVB structures prevented from settling by a similar specific weight as compared to the

carrying solvent. The backbone of the PVB used holds about 25% of vinyl-alcohol monomer

units and naturally a few vinyl-acetate groups, cf. A.1.3. These groups may lead to strong

attractive hydrogen bond interactions with the nanoparticles either directly with the plain

magnetite surface [194, 195] or with the hydroxyl groups of adsorbed ricinoleic acid both

resulting in additional stabilization by steric hindrance, cf. 4.1.3. The following sections 5.3.2

and 5.3.3 will shed more light on the possible mechanism of stabilization by PVB resulting in

a hypothetical binding model in figure 59.

Next it is to be discussed which information can be drawn from the relation between the

two ways to obtain the primary particle concentration, the UV/VIS extinction determination

and the TGA study.

Extinction vs. Gravimetric Analysis

In figure 47 the 600 nm extinction values of all three polymer samples are plotted against

the mass primary particle concentration wPrimary, which relates to the volume concentration

67

of nanoparticles in the diluted dispersion of the (b) samples φnanoparticles, as expressed by

eq. (93) in A.5.3.

figure 47: Correlation of the photometric extinction E600 nm and the primary particle concentration wPrimary as well

as the total particle volume concentration for the extinction measurement φ, two different slopes for

the destabilization with PMMA and PC and the stabilization with PVB, w0

Primary = 91.7% located at the

intersection of the linear models (dashed line) [20]

A good linear relation of light extinction over particle concentration is shown for both

destabilizing polymers (PMMA and PC) with deviations caused by erroneous measurement

results from the TGA investigations. This finding suggests a constant extinction cross section

Cext (cf. A.8) and thus equal particle scattering sizes with no influence of scattering by the

polymer. It is argued, that this furthermore shows, that there is no formation of larger

structures of particles interconnected by strong adsorbing polymers. As a conclusion

depletion flocculation must be the dominant factor of destabilization.

The expected linear relations obtained from the destabilizing polymers are given in eqs. (59)

and (60).

Primary600nm 0084.06830.0 wE (59)

lesnanopartic600nm 7.893.7291 E (60)

The factor 7291.3 in eq. (60) results in a scattering cross-section Cext of 2.97·10-18 m-2, using

eq. (92) with x = 15 nm and d = 1 cm. This is about 10-times smaller than the value calculated

in A.8, possibly due to the unknown influence of the grafted fatty acid layer and limited

reliable optical data for magnetite nanoparticles.

If the stabilization by PVB was only due to deaggregation of agglomerates without

adsorption of the PVB, the same linearity compared to PMMA and PC would be assumed, i.e.

the lighter line in figure 47 should simply extend the darker one. This is obviously not the

case but instead a very steep increase of extinction over particle concentration is observed

68

and both lines intersect at the primary particle concentration of the pristine sample without

polymers in solution which is calculated to be 91.7%. As a deduction from the steep slope of

the PVB s ste ’s e ti tio values, the scattering cross-section deviates from the constant

value of the destabilizing situation which can be attributed to structure formation

influencing light scattering behavior. Dynamic light scattering experiments would be

expected to support this assumption because of changing hydrodynamic radii, which is

discussed in 5.3.3.

Varying Phase Transfer Batches

All the investigations above have been realized with one magnetite precipitation batch,

following the procedure in A.3. This batch again was used for one phase transfer batch, with

the sample name PT100305 (phase transfer on march 5th 2010), following the procedure in

A.4. The reason for sticking to one batch for investigations of the colloidal stability and its

influences above is that the subjective quality of magnetite precipitation batches is not

reproducible. Some of the subjective markers are:

- the speed of sedimentation of magnetite (should be rapid but can be very slowly),

- the height of sediment (should be compact but can be very loose),

- the color of the supernatant (should be clear but can be slightly bluish) and

- the suspension pH (should be in the range 8.9 … 9.1).

The reasons for this problem of reproducibility are not clarified and have not been assessed

as part of the research of this thesis. However, the sources of fluctuation can probably be

limited to: a) the unstable (not quantified) quality of deionized water from the purification

device, b) the alterations in and aging of the manually fabricated high speed stirrer of the

reaction vessel as well as c) possible aging of the technical quality ricinoleic acid. Therefore,

by now the stirrer has been replaced and the purification vessel reactivated every time the

p e ipitatio got too fa off . If the precipitation behaves differently it is assumed, that the

iron salt conversion to magnetite is not complete, which the bluish appearance of the

supernatant and the settling behavior due to different particle interactions can be hints for.

Not reacted iron ions will furthermore influence the adsorption of the fatty acids [108, 196,

197] at the liquid-liquid interface and thus influence the state of dispersion in the organic

phase. When using precipitations which are too far off of the specifications mentioned

above it can occur that the phase transfer is incomplete even after a long standing time.

Even though the precipitations used in this work met the visual inspections and proper

suspension pH values and the phase transfer was only used when completed within less than

6 hours standing time, it is not impossible for the conditions of adsorption to differ, with yet

not described consequences on the state of dispersion of the particles in the organic solvent.

69

In figure 48 the colloidal stability of three different phase transfer batches (PT101104,

PT100902II and PT100305) by addition of PMMA, PC or PVB by means of extinction analysis

of diluted supernatants are compared. The different phase transfer batches originate from

different precipitation batches as well, but are prepared with the same type of RA and DCM

at DRA = 0.2 and RA = 0.02.

figure 48: Photometric primary particle investigation as a function of the polymer concentration for both

destabilizing polymers PMMA and PC as well as for the stabilizing PVB and three different phase

transfer batches: PT101104, PT100902II and PT100305

Looking at the differences between the polymers for each phase transfer batch, the results

are consistent with the findings above. Both PMMA and PC are destabilizing the colloid, with

a stronger impact by PC, when comparing on the basis of [cpolymer] = g/l. PVB stabilizes all of

the dispersions with a similar endpoint. The difference between the batches is the initial

extinction E0

600 nm without any polymer. This value is proportional to the initial primary

particle concentration and decreases in the order PT100305, PT101104 and PT100902II.

Comparing the batches for the same polymer type (same color of lines in figure 48) it can be

noticed that the curves show the same progression. Therefore the data is normalized with

the initial extinction for PMMA and PC in figure 49.

figure 49: Normalized extinction curves for the three phase transfer batches as a function of (left) PMMA as well

as (right) PC, the fitted lines follows the mathematical model in eq. (61)

70

Normalization for PVB is not satisfactory, for it is stabilizing and the state of dispersion

improves independently from the initial particle concentration, yet with a similar endpoint at

which no sediment or turbidity is noticeable after centrifugation. The normalization of the

ordinate axis enables to visualize, that all three batches have a comparable progression for

both destabilizing polymers. A mathematical model to fit the results is inferred from the PC-

curves, with a first order exponential decay function, represented by eq. (61).

Polymer

0Primary

Primary0600nm

600nm cAe

w

w

E

E (61)

The parameter A is the decaying rate. In table 7 the fitted results for the three batches

separately and combined are presented for both polymers. Additionally w0Primary is estimated

for PT101104 and PT100902II from the given initial extinction and the known ratio of E0600 nm

and w0Primary for PT100305. The high margin of error comes from the uncertainty of the

normalized data points of the extinction curve and is not a measure for the quality of the fit.

For judgment of the fit quality, the coefficient of determination is given as well, recognizing

the uncertainty of the data points.

table 7: Results for the mathematical fit parameters using eq. (61) of each batch for both polymers as well as

for the three batches combined; estimation of the initial primary particle concentrations w0

Primary of batches

PT101104 and PT100921 using the initial extinction values E0

600 nm and the given w0

Primary of batch PT100305

(see figure 47)

sample PMMA PC E

0600 nm

in -

w0

Primary

in % A in l/g R2 A in l/g R

2

PT101104 0.029

± 0.027 0.993

0.061

± 0.009 0.993

0.366

± 0.093

51.1

± 16.3

PT100902II 0.023

± 0.042 0.970

0.073

± 0.026 0.993

0.216

± 0.008

30.1

± 3.1

PT100305 0.028

± 0.021 0.956

0.047

± 0.045 0.996

0.658

± 0.043 91.7

combined 0.028

± 0.015 0.969

0.062

± 0.008 0.989 - -

For all samples R2 is sufficiently high to argue on a reasonable mathematical model. In any

case the decaying rate is higher for PC as compared to PMMA. As suspected in figure 48 the

initial primary particle concentration is the highest for PT100305. Sample PT100902II only

contains 30 % of initial primary particle content as a result of a less successful phase transfer.

The destabilization mechanism proves to be not a function of the primary particle

concentration but only on the type of polymer.

71

Even though the initial primary particle concentration is not reproducible, due to the

implications at the beginning of this subsection, the colloidal stability investigation of this

paragraph is very well reproducible for all three polymers.

Future studies will need to investigate this reproducibility problem to guarantee composite

synthesis with a reliable prediction of the primary particle content.

5.3.2 Influence of the Surfactant

In 4.3.3 the initial primary particle concentrations w0Primary for phase transfers with different

fatty acids but the same precipitation batch have been presented. Those values originate

from the study presented in this section, dedicated to the influence of the type of fatty acid

surfactants on the colloidal stability of sterically stabilized magnetite nanoparticles in

polymer solutions both destabilizing (PMMA) and stabilizing (PVB).

Above and in [20, 26, 27] it is shown, that for ricinoleic acid (RA) coated magnetite

nanoparticles in DCM, PMMA leads to flocculation and phase separation because of non-

adsorbing interactions contrary to PVB, which adsorbs and stabilizes the colloid by additional

steric repulsive interactions. Ricinoleic acid is a C18 unsaturated hydroxilized fatty acid,

double bond at C(9) and with a hydroxyl group at C(12). The task of this section is to assess,

whether PMMA and PVB behave similarly for the other fatty acids used for the phase

transfer process in 4.3, namely the C18 acids linoleic acid (LA) with two double bonds at C(9)

and C(12) and oleic acid (OA) with one double bond at C(9) as well as the saturated C14

myristic acid (MA) and C8 caprylic acid (CA).

All the colloidal stability investigations in this section are only based on the gravimetric

method of determination of the primary particle concentrations and are published in [23].

Poly(methyl methacrylate)

In figure 50 the results of colloidal stability of the DCM based magnetite dispersions with

different fatty acids and varying polymer concentration of PMMA are presented. The

ordinate axis depicting the primary particle concentration wPrimary is in logarithmic scale and

an approximately linear decrease over the PMMA concentration cPMMA can be noticed for all

samples. Deviations from this finding for small primary particle concentrations of samples

MA and CA may result from a higher margin of error for the TGA method applied.

Nevertheless, a decrease means that independent from the type of fatty acid PMMA leads to

depletion phase separation where the magnetite concentration of the phase with primary

particles decreases, as it is the case in 5.3.1 as well.

72

figure 50: Primary particle concentrations of the five investigated samples in nanoparticle polymer mixtures

with different concentrations of the polymer PMMA in DCM. The solid lines present mathematical fits

of eq. (61), published in [23]

The data is fitted with the mathematical relation given in eq. (61). Both A and w0Primary are

fitted, for the initial primary particle concentration without polymer cannot be directly

measured with the gravimetric method, where the degradation of the polymer is crucial but

results from linear fitting log(wPrimary) versus cpolymer. The fitted results are summarized in

table 8.

table 8: Results of the first order exponential decay fit parameters A and w0

Primary for all five samples

investigated

sample fatty acid A in l/g R2 w

0Primary in %

PT110802_III Ricinoleic Acid (RA) 0.037 ± 0.001 0.999 98.5 ± 2.5

PT110802_II Linoleic Acid (LA) 0.035 ± 0.005 0.891 32.2 ± 5.0

PT110802_I Oleic Acid (OA) 0.034 ± 0.009 0.718 8.4 ± 1.6

PT110802_IV Myristic Acid (MA) 0.065 ± 0.015 0.784 2.0 ± 0.7

PT110802_V Caprylic Acid (CA) 0.002 ± 0.009 -0.238 2.0 ± 0.3

The colloidal stability decreases in the order RA, LA, OA, CA and MA. The slope of decrease is

independent from the fatty acids and nearly constant for RA, LA and OA with A ≈ 0.036 l/g

and statistically comparable to the results in table 7 (deviations for MA and CA are assumed

to be due to the uncertainty of the measurement). However, the initial primary particle

content varies significantly, even though the same precipitation batch has been used and is

not the influencing factor as it was the case in table 7. An explanation for the different

primary particle concentrations is given in chapter 4 (discussed in paragraph 4.3.3) using the

model of deagglomeration presented in paragraph 4.2.

In summary, it can be stated that the mathematical model (eq. (61)) is a well description of

the colloidal destabilization and the course of stability does neither depend on the initial

primary particle concentration nor the type of fatty acid.

73

With a given mathematical relation of wPrimary versus cpolymer, the map in figure 18 – cpolymer as

a function of the composite solids concentration in the solvent based mixture csolid and the

filler concentration F in the solvent free composite – can be formulated as a primary particle

map for destabilizing polymers with the parameters A and w0

Primary. The combination of

eqs. (11) and (61) leads to eq. (62)

solid

solventsolid

-1

11

0PrimaryPrimary

solvent

solid0Primary

Primarysolvent

1

11

ln

c

DFAc

eww

DA

cw

wA

F

(62)

The primary particle maps for the C18 fatty acids RA, LA and OA in a DCM based PMMA

solution with eq. (62) and the parameters A and w0Primary from table 8 are presented in figure

51.

figure 51: Primary particle maps by combining results of the investigations in figure 50 and table 8 with eq. (62)

for ricinoleic (RA), linoleic (LA) and oleic acid (OA) with the same scale in the three dimensions

With these maps the primary particle concentration can be predicted, showing that the best

results regarding the state of dispersion in the composite are to be expected for high filler

concentrations in the composite and low solid concentrations in the solvent based mixture

to be processed. Filler concentrations are defined by the specifications of the desired

composite product. The second is a question of the economic and ecological process design,

since low csolid means an increased usage and processing of solvents which can be costly and

environmentally unfriendly.

In the next subsection the influence of fatty acid surfactants on the interactions in PVB

solutions is investigated.

Poly(vinyl butyral)

An urging question to be answered is, whether the type of the surfactant fatty acid will

influence the colloidal interactions of stabilized magnetite nanoparticles with PVB, which has

been described as a stabilizing polymer for RA covered magnetite nanoparticles above in

74

5.3.1. In figure 52 the primary particle concentration as a function of three different PVB

concentrations in a DCM based system is presented for all five surfactants. For all samples an

increase in the concentration is noticeable, which for the C18-acids concludes, that

independent from the degree of saturation as well as an additional hydroxyl group, the

stability is increased by the addition of PVB. It is supposes that its adsorption is the

mechanism of stabilization. This leads to the absence of attractive depletion interactions and

further stabilization by the polymer. At the highest PVB concentration the apparent primary

particle concentration in the supernatant is almost 100 % for all samples. Consequently the

stabilization is not limited to the hydroxyl group bearing ricinoleic acid for which adsorption

will be confirmed in the following paragraph. When only looking at ricinoleic acid it could

have been concluded, that the hydroxyl groups of the fatty acid interact with the once

present in the PVB structure. In general, the polarity of the hydroxyl groups in the PVB

structure is seen as the cause of adsorption. Independent from the structure of the fatty

acid, it could be unoccupied magnetite nanoparticle surface which offers adsorption sites for

the OH-bearing PVB, as it has been reported by HSIANG et al. and JEAN et al. [194, 195].

figure 52: Primary particle concentrations of the five investigated samples in nanoparticle polymer mixtures

with different concentrations of the polymer PVB in DCM; presented in context (scaling) with figure

50 above, published in [23]

Myristic acid shows the most dramatic increase in apparent primary particle concentration.

Only small amounts of PVB already stabilize the system, whereas for the other fatty acids a

quasi linear increase in stability is noticed for the logarithmic plot. For a DLVO like discussion

of this type of polymer colloid interaction, the effect of depletion flocculation is absent, so

that: Wtotal = WvdW + WBorn + Wsteric and Wdepletion = 0 (cf. 5.2). The PVB molecules may not

adsorb as brushes, hence the additional steric component could only be described by DOLAN

and EDWARDS theory instead of the AdG, which is valid for the brush regime only, cf. 4.1.3.

75

5.3.3 Stabilization by Adsorption of PVB

In the previous two paragraphs it is shown that in DCM based solutions with the polymer

PVB the colloidal stability of fatty acid stabilized magnetite nanoparticles improves

independently from both the initial concentration of primary particles and the type of fatty

acid. It is argued that the stabilization mechanism is due to adsorption which results in an

additional steric stabilization, case (d) in figure 39. In literature, the ability of PVB to stabilize

non-aqueous ferritic particle dispersions has been described already and adsorption

occurred [194, 195]. Therefore, in this paragraph the mechanism of stabilization of the fatty

acid coated magnetite nanoparticles is assessed. At first an assumption linked to the findings

in figure 47 is checked, which is that the primary particle size would increase when PVB

adsorbs on the particles.

Primary Particle Size

Primarily, in order to check whether the nanoparticles in the supernatant meet the

definition of primary particles (cf. A.5) the further diluted samples (b) from the extinction

measurement are characterized using dynamic light scattering. In the graph in figure 53 the

results of this set of measurements are presented plotting the number weighted median

particle sizes x50,0 over the polymer concentration cpolymer in the corresponding samples (a).

At zero polymer concentration a particle size of about 20 nm is measured for all systems as

expected. There is a slight drop in median particle size for the destabilizing polymer

dispersions especially for PC. This effect might be caused in the following possible ways. As

mentioned in 5.2, depletion interactions are stronger for larger particles and thus leaving

behind the smaller ones in the supernatant. Furthermore, the polymer coils might influence

the structure of the fatty acid layer around the nanoparticles compressing it and with this

lowering its hydrodynamic radius. Additionally, the measurement could simply be influenced

by the less scattering but faster fluctuating smaller coils with respect to the magnetite

nanoparticles. Nevertheless, for the destabilizing polymers the assumption of primary

particles in the supernatant is met. A very interesting development of particle size over

polymer concentration is noticed for the PVB and functionalized magnetite nanoparticle

dispersion. The particle size increases with higher initial polymer concentration, which can

well be described with a Langmuir type adsorption model and a fitted maximum particle size

xmax of 52.5 nm as shown in figure 53.

76

figure 53: The median diameter of the number weighted particle size distribution x50,0 of diluted samples (b)

measured with DLS as a function of the pristine concentrations cpolymer of PMMA, PC and PVB in the

samples (a)

The increase in primary particle size for PVB is attributed to the occurrence of the coverage

of the nanoparticles with PVB via adsorption. However this effect may not account for the

large increase of extinction over primary particle concentration in figure 47, which is

assumed to be caused by larger structures of nanoparticles and PVB that are not covered by

the number weighted particle size distribution investigated in figure 53, but by plotting the

numerically obtained (NNLS method applied in the Zetasize software [198, 199]) volume

weighted particle size distribution (frequency type µ3(x)) data of the DLS measurements in

figure 54. For PVB there are fractions larger 100 nm noticed with an increased amount for

increasing PVB concentrations. The increasing viscosity effect by the polymer must be

neglected, though, because the samples investigated are the diluted (b) samples with a

polymer concentration too low to induce such a strong viscosity effect.

figure 54: Numerically obtained volume weighted particle size distribution from the DLS measurements of the

(b) samples with the polymers (left) PMMA and (right) PVB; the polymer concentrations cPMMA and

cPVB refer to the undiluted samples (a)

Since the nanoparticle in PVB solutions are long term stable, the structures must be loosely

bound with a resulting low overall specific weight because they do not settle under the

experimental centrifugation conditions of this study (A.5) and no supernatant is noticed after

77

more than one year sample storage. Yet by definition they should not be accounted for as

primary particles and might be as problematic for material synthesis as the agglomerates in

the sediment. Nevertheless, this explains the extreme increase in extinction values in figure

47. In order to support the assumption of PVB adsorption and thus its stabilizing behavior, in

the next subsection, the polymer adsorption is quantified.

Adsorption Study

The results of the adsorption of PVB on the nanoparticle surface, as described in A.5.4, are

plotted in the left diagram in figure 55. The straight line is the case for 100 % adsorption and

the fitted line follows the assumption of a Langmuir Type adsorption, as discussed below.

Furthermore, only considering the number weighted median particle sizes of figure 53, the

PVB layer thickness z(cPVB) = (x50,0(cPVB) – x050,0) is calculated and plotted in the right diagram

in figure 55, also fitted with a Langmuir type curve, which is discussed below, as well.

figure 55: (left) PVB adsorption on the RA-Fe3O4 magnetite a oparti les surfa e as a fu tio of the PVB

concentration in the mixture cPVB, (right) layer thickness of the adsorbed PVB layer, calculated using

the data in figure 53

For the destabilizing polymers PMMA and PC, adsorption is been detected (as checked for

samples with cpolymer = 25 g/l, not presented in figure 55), supporting the assumption of

destabilization through depletion interactions (case (a) in figure 39), rather than bridging

flocculation (case (c) in figure 39). The lack of adsorption for PMMA and PC furthermore

verifies the experimental procedure applied showing that the washing with DCM is efficient

in removing non-adsorbed polymers. The adsorption of PVB, which has been assumed in the

light scattering and particle size analyses above as well as the layer thickness z can be

described with a Langmuir type isotherm with the relations and fitted parameters (constants

k and kz as well as maximum adsorption max,PVB and maximum layer thickness zmax,PVB) in

eqs. (63) and (64).

04.0 and g/g497.0,1 ΓPVBmax,

PolyΓ

PolyΓPVBmax,

kΓck

ckΓΓ (63)

78

12.0 and nm72.16,1 zPVBmax,

Polyz

PolyzPVBmax,

kzck

ckzz (64)

Knowing the adsorbed amount of PVB and the layer thickness z one can calculate the PVB

layer density PVB_adsorbed as a function of the polymer concentration cpolymer, with eq. (65).

This corresponds to the PVB concentration at the nanoparticle surface [160].

3lenanopartic

3lenanopartic

lenanopartic3

lenanopartic

shell

edPVB_adsorbedPVB_adsorb

2 xzx

x

V

m

(65)

In figure 56 the adsorbed PVB layer density is plotted as a function of the polymer

concentration in the solution cPVB, considering eqs. (63) and (64).

figure 56: Density of the adsorbed PVB layer as a function of the PVB solution concentration cPVB with the fitted

data from figure 55 using eq. (65) and a nanoparticle diameter xnanoparticle of 15 nm and a nanoparticle

specific weight of 5200 g/l

The progression of the curve in figure 56 shows that the density of the adsorbed layer

decreases with increasing polymer concentration and thus increasing layer thickness. This is

a conclusive result, when compared to the review of FLEER for polymers at an interface [160].

The steep decrease of the curve let’s assume that the first PVB polymers adsorb with several

anchoring points along its backbone (the train configuration). The density of free, non-

adsorbed PVB coils in solution is defined by the coil dimensions, due to solubility (cf. 5.1.1)

and can be calculated with eq. (66).

A3

coil

PVBPVB_coil

26

NR

M

(66)

With a molar weight of the PVB of MPVB = 32,000 g/mol (see A.1.3) and a coil dimension

represented by the measured hydrodynamic radius of RPVB-coil = 8.5 nm (see A.9) the coil

density yields 20.7 g/l, which is lower than the density of the adsorbed PVB, as expected.

79

Thermo gravimetric Study

In 4.3.4 and [21] it was shown, that the mass loss between 600 °C and 900 °C for inert gas

TGA analyses can be attributed to residues of adsorbed ricinoleic acid reducing the

magnetite nanoparticles to FeOx. However the influence of a polymer on that high

temperature degradation step has not been reported in that context. In this subsection the

relative mass loss at the high temperature step of degradation is presented as a function of

the polymer concentration in the dispersion cpolymer for PMMA and PVB, depicted in figure

57.

figure 57: Relative mass loss in the third degradation step for inert gas TGA between 600°C and 900 °C for

PMMA and PVB and Ra-Fe3O4 as a function of the polymer concentration

There is no statistically significant difference for the data points of the PMMA curve.

However, a clear increasing trend for the adsorbing PVB can be noticed with a similar

degressive behavior as mentioned above in figure 55. The results for PC are not shown, due

to an unclear influence of the high temperature residues and complex thermal

decomposition of pristine poly(bisphenol A carbonate) [200].

The results in figure 57 allow for the assumption that the high temperature mass loss could

at least be used qualitatively to judge on the adsorption behavior of organic matter on

magnetite. This needs to be evaluated in a future research.

Destabilization in PVB-PMMA mixtures

So far it is shown in various experiments and on different samples, that the polymer PVB

stabilizes a dispersion of fatty acid grafted magnetite nanoparticles with xnanoparticles ≈ 15 nm

in the solvent dichloromethane by adsorption. PMMA tends to destabilize such particle

dispersions. It is now interesting to know, whether the presence of PVB in a dispersion with

dissolved PMMA (assuming it is the desired matrix material) could reduce or even prevent

destabilization. In figure 58 there are four stability curves for RA-Fe3O4 (PT100902II) in a

DCM solution of PMMA with different amounts of PVB added, before adding the PMMA. A

80

PVB content of mPVB/mFe3O4 = 0.3 equals a PVB concentration of cPVB = 7.3 g/l for which the

stabilizing effect already is apparent in figure 48.

figure 58: Colloidal stability of PVB mixed with RA-Fe3O4 nanoparticles at four different mass ratios in solutions

of PMMA with the concentration cPMMA

The destabilizing effect of PMMA is not reduced by the addition of PVB, rather it is

enhanced. The decrease of the relative extinction is even more pronounced with increasing

PVB content. This is e e logi , he a k o ledgi g the PMMA’s e ha is of destabilization as a depletion induced effect. The effective particle diameter increases with

increasing PVB content (cf. figure 53) and with this the attractive energy between the

particles is increased, supporting the results in 5.2.

Hypothetical Model of Adsorption of PVB

Based the various knowledge of adsorption of PVB gathered and argued above, a

hypothetical model of adsorption is proposed and visualized in figure 59.

figure 59: Scheme of the hypothetical model of adsorption of PVB on the surface of magnetite nanoparticles

carrying chemically grafted fatty acid surfactant molecules as well; (left) train adsorption of PVB at

vacant magnetite surface sites, (right) surface hydroxide groups shall interact with the hydroxyl

groups of the PVB backbone.

PVB adsorbs directly on vacant surface sites which are not occupied by fatty acid surfactant

molecules. The interaction may occur between surface hydroxide groups of the magnetite

81

nanoparticles and the hydroxyl groups of the PVB structure. In this case physical adsorption

by hydrogen bonding would be probable. Since PVB bears several OH-groups the adsorption

of the polymer coil takes place as a train adsorption. Certainly for OH-bearing ricinoleic acid

the PVB adsorption may also take place between bound RA and PVB.

5.3.4 Influence of Mechanical Dispersing Methods on the Stability

When faced with agglomeration of nanoparticles it is often convenient to apply dispersive

stress by mechanical agitation, in order to overcome the attractive forces between the

particles [50-53, 201-203]. Devices which are typically used produce a flow field with high

shear to disperse particles, such as the rotor-stator-mixer Ultra-Turrax (UT) or a cavitation

inducing sonotrode (US) [50, 204-206]. Another common method especially for dispersing

nanoparticles smaller than 100 nm is agitated ball milling [52, 202, 207-210]. However, shear

or ball milling does not only lead to deagglomeration but can induce agglomeration as well

[211]. Furthermore it is shown in [52] that the agglomerate limit down to which dispersing

acts is rather large with approximately 100 nm compared to the 20 nm primary particle size

of the present study. In this paragraph it will be discussed, whether the state of dispersion

can be improved by applying high-shear devices (UT and US) or planetary ball milling (PM).

In the first experiment, a mixture of RA-Fe3O4 (PT110211) with a solution of PC in DCM at

cPC = 52 g/l (F = 0.3) is treated with simple stirring, sonotrode ultrasound (200 W input

power), Ultra-Turrax T25 (level 1) each for 15 min and compared to a quickly mixed sample

with a retention time close to 0 min. After the mixing procedure, the primary particle

concentration is assessed as described in A.5. For the extinction measurement, the relation

in eq. (59) is used to calculate wPrimary. The results are presented in figure 60. Due to the

pessimistically assumed uncertainty for the gravimetric determination (cf. A.5.2), the

differences between the four types of dispersing are not statistically significant.

figure 60: Primary Particle Concentration as determined gravimetrically (TGA) and by light extinction at 600 nm

(UV/VIS) with eq. (59) for a DCM based mixture of RA-Fe3O4 and PC with D = 0.2 and cpolymer = 51.2 g/l

(F = 0.3) for four different mixing procedures (US – sonotrode ultrasonication, UT – Ultra Turrax)

82

For the determination of the primary particle concentration using repetitive extinction

measurements at 600 nm (green columns in figure 60), the primary particle concentration

for 0 min mixing is significantly higher compared to the others. This means that none of the

shear inducing mixing procedures applied can improve the state of dispersion of this

depletion flocculated sample. The erroneous of the composition of the mixture is not

accounted for i the data’s u e tai ties, but would have concluded, that all four samples do

not vary statistically.

Another experiment on the influence of cavitation by ultrasonication is related to the

colloidal stability of RA-Fe3O4 in DCM (PT091227II) with the destabilizing polymer PMMA.

The same experiment has conducted twice, one without mechanical agitation (no US) and

one appl i g so ot ode ult asou d fo ’ ’ U“ prior to the procedure in A.5.2. Both

experimental results are plotted in figure 61.

figure 61: Destabilization curve of RA-Fe3O4 (PT091227II) in DCM with PMMA, without and with 1 min

sonotrode ultrasonication (US)

The trend is the same for both curves due to the destabilizing effect of PMMA, presumably

by depletion flocculation. However, the primary particle concentrations for the cavitation

agitated mixtures are all lower than the simple mixtures. This again shows that the

dispersing device is in this case not dispersing but rather decreasing the colloidal stability

slightly.

The final experimental set-up includes a planetary ball mill, cf. A.6. Without addition of a

polymer it is tested, whether it is possible to reduce the fraction of agglomerates after the

phase transfer. The dispersion is conducted up to 25 min and every 5 min a sample is

analyzed with dynamic light scattering. The results of intensity weighted particle size

distributions (frequency distribution of µint(x)) as well as the development of the second

cumulant median particle size x2nd cumulant over dispersing time tPM are depicted in the two

graphs in figure 62.

83

figure 62: Dispersing a phase transfer batch without polymer in DCM using a planetary ball mill, (left) intensity

weighted particle

Even the planetary ball milling dispersing method, which has been successfully applied to

disperse diamond nanoparticles with primary sizes much lower than 20 nm [209, 210], does

not improve the state of dispersion of the transferred magnetite nanoparticles. Based on the

increase of the second cumulant particle size, agglomeration is induced by the planetary ball

mill dispersion. Such favored agglomeration has been accounted for by excessive dispersing

devices, such as planetary ball milling, as mentioned in [203].

In summary, all three experimental setups in this paragraph showed, that the dispersing

devices applied, cannot improve the state of dispersion. For all samples a disapprovement is

noticeable instead. This concludes that the destabilizing mechanisms are not overcome by

mechanical dispersing. Due to the compositional limitations of the solution and spray drying

process it is not feasible to introduce another stabilizing agent to stabilize the mechanically

dispersed particles, as it is often done [212].

5.3.5 Kinetics of Flocculation at Low Nanoparticle Concentration and

high PMMA concentrations

In the previous paragraphs, information is gathered on the colloidal stability of variably fatty

acid stabilized magnetite nanoparticles in dichloromethane based dispersions and different

kinds of dissolved polymers. It was found that PMMA as well as PC destabilize the stable

fatty acid grafted magnetite nanoparticle dispersions, presumably by depletion flocculation.

PVB is most probably stabilizing the dispersions further by adsorption and additional steric

interactions.

In this section the kinetics of (depletion) flocculation are assessed by time dependent

particle size measurements as well as light extinction over time. Both methods rely on a low

particle concentration, due to the strong absorption of magnetite in the visible

electromagnetic spectrum. The magnetite nanoparticle sample is RA-Fe3O4 obtained from

84

PT100305 (as used in 5.3.1) after withdrawing the agglomerates with a magnet and

centrifugation and determining the particle and fatty acid concentration gravimetrically,

resulting in wnanoparticles = 0.0113 g/g and DRA = 0.276. In each experiment the nanoparticle

concentration is reduced to cNanoparticles = 1.2 g/l, compared to the 24 g/l in the processed

systems, which have been considered in the discussions above. So the particle mass

concentration is reduced by a factor of 20, which means that the particle number is reduced

by a factor of 8000, which is 203. The polymer concentration is chosen so that flocculation

can be monitored with the methods applied, with cPMMA > 50 g/l.

It is to show, that indeed the primary particles flocculate to large agglomerates and that the

speed of flocculation is a function of the polymer concentration, which defines the attractive

strength of interaction. It furthermore gives rise to the hope that flocculation could be

suppressed by quickly mixing and evaporation which has not been achieved in experiment

yet.

In figure 63 the intensity weighted frequency particle size distributions over time of the

stable RA-Fe3O4 dispersion immediately after mixing with both the destabilizing polymer

PMMA and the stabilizing polymer PVB to cpolymer = 58 g/l are displayed. The measured

polymer viscosities from A.9 are used to obtain DLS results.

figure 63: Time development of the intensity weighted frequency particle size distribution (in logarithmic

scaling) determined with DLS with respect to the viscosity of the polymer solution for a magnetite

dispersion without pristine agglomerates at cnanoparticles = 1.2 g/l and cpolymer = 58 g/l; (left)

destabilizing PMMA (right) stabilizing PVB

For PMMA after about 10 min larger agglomerates with sizes >100 nm appear and the

fraction of smaller particles is reduced. Due to the limitations of DLS mentioned in A.10, the

information obtained is not to be taken quantitatively. Yet it can for sure be stated, that

agglomeration sets in. Contrary, as expected, PVB stabilizes the dispersion. The particles are

slightly larger than the primary particles at t = 0 for PMMA, which is due to the adsorption of

the PVB onto the nanoparticles, cf. figure 53.

85

A more straightforward way to assess the kinetics of flocculation is by extinction

measurements, for the initial slope dE/dt is proportional to the agglomeration rate [183,

213]. Here the kinetics is investigated using time dependent extinction at 600 nm with a

Perkin Elmer UV/VIS photo spectrometer. In figure 64 four agglomeration curves with E600 nm

on the ordinate and the time up to 50 min on the abscissa are displayed. The PMMA

concentration is varied between 59 g/l and 72 g/l. Lower concentrations did not show a

significant increase of extinction over more than one hour.

figure 64: Time dependent light extinction to monitor agglomeration of a RA-Fe3O4 dispersion in DCM with

dissolved PMMA, the polymer is mixed with the stable nanoparticle dispersion at t = 0 min

For the initial slopes are zero for all samples, the set-in and maximum slopes of the

extinction curves of figure 64 are presented in figure 65. The set-in slope is defined here as

dE/dt, where d2E/dt

2 has a first order maximum.

figure 65: Correlation of extinction rates (set-in and maximum slopes of the lines in figure 64) with the PMMA

concentration

Summarizing the results, the higher the PMMA concentration is, the faster the coagulation

becomes. Experiments on the coagulation by adding PC failed, since it was too rapid and in

addition the extinction values went into saturation (at around E = 4).

86

It is very difficult to draw conclusions from the kinetics of flocculation for the higher

concentrated systems, which are applied to synthesize nanoparticle-polymer-composites,

since there is no information on the appearance of the entire phase diagram. Experiments of

directly mixing the nanoparticle dispersions and the polymer solutions prior to atomizing

and solvent evaporation in the spray drying step with residence times lower than 2 s let

suggest that in systems with high nanoparticles concentrations, flocculation is much more

rapid. This would be expected when acknowledging that the coagulation rate is a function of

the particle concentration. Kinetics investigation in the way presented, however, are not

applicable for these systems because of the high absorption of light of magnetite

nanoparticles, cf. A.8. A feasible way to investigate this in the future would be using X-Ray

investigations instead of visible light extinctions.

5.3.6 Influence of the Solvent

This paragraph discusses the influence of the solvent on the colloidal stability with and

without polymers dissolved. All samples so far were based on the organic solvent

dichloromethane. For a technical scale process it would be profitable to substitute this

health and environmentally hazardous solvent. Furthermore it is intriguing to find out how a

monomer based particle dispersion is influenced by addition of the polymerized monomer,

for instance methyl methacrylate (MMA) and PMMA or styrene (ST) and PS. This would show

that the destabilization occurring with the polymerization method of composite synthesis

(cf. 2.2.3) as e.g. reported in [85] is caused by the same destabilizing effect studied above for

PMMA and PC in DCM.

The follo i g t o su se tio s i t odu e to the esults o tai ed i t o sepa ate stude t’s theses which are both based on techniques developed in this doctoral thesis. They are both

related to using different solvents for the process of composite synthesis. The first one

shows how the usually applied solvent dichloromethane can be substituted by the less

hazardous yet good solvent ethyl acetate (EA). The second work is on the phase transfer of

magnetite nanoparticles to DCM, MMA and ST as well as the colloidal stability under the

presence of PMMA for DCM and MMA samples and PS for DCM and ST samples.

Dichloromethane Substitution with Ethyl Acetate

In a diploma thesis of TINA BREMERSTEIN [214] it is evaluated how EA, a good solvent for

PMMA, could substitute DCM in the process chain of nanoparticle-polymer-composite

preparation by the solution and spray drying method. The results on the colloidal stability

and the consequences on the spray drying results are discussed in a manuscript to be

submitted [215]. The substitution occurs after the phase transfer step by solvent mixing of

RA-Fe3O4 in DCM with EA and low pressure rotary evaporation of the DCM.

87

The results on the colloidal stability for both solvents by determining the primary particle

concentration with the TGA as well as the UV/VIS method are presented in figure 66. Due to

a difference in the magnetite concentration in both solvents, the extinction is normalized

with the initial extinction value. By comparison on the basis of the gravimetrically

determined primary particle concentration (figure 66 (left)) one can notice smaller yet not

significantly smaller values for EA and a good agreement on the progression of both curves.

The same analogy can be noticed for the normalized extinction curves in the graph in figure

66 (right). In A.9.1 it is reported that the solubility of PMMA in both solvents is similar with

no significant difference. Nevertheless, taking the calculated solubility by literature reported

HANSEN solubility parameters (in table 29) for granted, PMMA is better soluble in DCM with a

Flory interaction parameter χ of 0.10 as compared to EA with χ = 0.31. As a consequence the

polymer coils in DCM are expected to be more extended and at the same cPMMA and thus

same coil number concentration, depletion will be stronger in the DCM solution. However,

there is also a rather large uncertainty in the HANSEN solubility parameters of polymers.

figure 66: Primary particle concentration wPrimary (left) and normalized photometric extinction E600nm/E0,600nm

(right) as function of the PMMA concentration for the solvents: DCM and EA

In summary it is shown that for similar solubilities the destabilization of a given nanoparticle

species is similar for the same dissolved polymer. Furthermore it is shown, that DCM can be

replaced by EA, yet DCM is still necessary for the phase transfer step since the water

solubility of EA is too high.

In the next subsection DCM is replaced in the phase transfer step by MMA and ST for the

PMMA and PS composite preparation purpose, respectively.

Phase Transfer of Magnetite Nanoparticles to Methyl Methacrylate and Styrene

and the Colloidal Stability with PMMA and PS

The basis of this subsection is the experimental student research of ROBERT HARTMANN [216],

which has been presented on a conference in St. Petersburg in 2012 [217]. The working

hypothesis of this work is, that the polymer PMMA should also destabilize a MMA based RA-

Fe3O4 dispersion, following that the problems of destabilization discovered in this doctoral

88

thesis are as well of importance for the composite synthesis based on dispersing particles in

a monomer with subsequent polymerization, cf. 2.2.3. To extent the investigations, another

common monomer system is tested as well, which is styrene (ST) with poly(styrene) (PS).

Contrary to the previous subsection, the solvents MMA and ST, both lighter than water, are

to replace DCM, heavier than water, entirely, meaning beginning with the particle phase

transfer. To achieve comparable results, all experiments are based on a single homogeneous

mixture of three precipitation batches, which is used for phase transfers to DCM, MMA and

ST, with DRA = 0.2. The batch phase transfer experiments for the lighter solvents MMA and

ST are conducted in typical separation funnels, which are used for batch mixer-settler

extractions as well. All phase transfers are carried out by emulsification, which is stirring for

DCM and shaking for MMA and ST. In any case stable emulsions are formed, most

dominantly for styrene. An obvious mechanism of stabilization of the droplets is

electrostatic, due to the disassociated RA at the solvent-water interface. This means, that

reducing the pH would reduce disassociation behavior and thus stability. Therefore, to break

the stable emulsion, 1N HCl is added to obtain pH 6.0. Photographic images of the three

completed phase transfers before and after breaking the emulsion by adding HCl are

displayed in figure 67.

figure 67: 100 % completed phase transfers of magnetite nanoparticles originating from the same precipitation

batch (a) to dichloromethane DCM (PTDCM120227) by gravity driven transport and stirred

emulsification in a beaker, (b) to MMA (PTMMA120227) and (c) to styrene ST (PTST120227) by mixer-

settler extraction in separation funnels, (left) strong emulsion formation for MMA and ST, (right)

after breaking the emulsion by reducing the pH to 6.0 with 10 ml 1N HCl

After completed phase transfer, the particle free water phase is located at the top for DCM

and at the bottom for both MMA and ST. Settling droplets of RA-Fe3O4 in MMA are noticed

in the image b on the right of figure 67. It is assumed, that transferred agglomerates (due to

ineffective physicochemical deagglomeration, cf. 4.2) lead to local specific gravities, which

are higher than water. These zones of high specific weight will eventually drip off the liquid-

liquid interface. For the purpose of testing the colloidal stability when mixing in a polymer

89

solution all transferred particles are extracted together with DCM, MMA or ST and only the

water is removed after three washing steps.

For the MMA based system, the effect of dissolved PMMA is investigated with the UV/VIS

extinction method (see A.5.3) using the newer photo spectrometer Cary 60 from Agilent

Technologies. The comparative results for DCM and PMMA as well as MMA and PMMA are

depicted in the graphs in figure 68 both for absolute extinction at 600 nm (left) and for the

normalized extinction (right).

figure 68: Extinction based determination of the primary particle concentration of RA-Fe3O4 in DCM

(PTDCM120227) and MMA (PTMMA120227) under the presence of PMMA with the polymer

concentration on the abscissa; (left) absolute (right) normalized values; the extinction ratio for the

polymer free point of MMA to DCM is 8.6 %

At first it is to notice, that both lines are decreasing, meaning the polymer leads to

destabilization, which is expected for DCM, based on all findings so far. This means that

PMMA remains a non-adsorbing polymer in MMA as well. For MMA the initial extinction is

only 8.6 % of the initial value for DCM, which corresponds to the relation of the initial

primary particle concentrations after phase transfer of MMA and DCM, cf. eq. (61).

Consequently, the deagglomeration is less effective for RA in MMA, compared to RA in DCM.

Following the discussion of the new model of deagglomeration in 4.2 this could be due to a

lower coverage of RA on the nanoparticle surface, leading to a lower adsorption distance s,

or due to a smaller adsorption thickness because of the poorer solubility of the RA tails of

RA-Fe3O4 in MMA with χ = 0.06, compared to DCM with χ = 0.03, in table 36. Aside this

different behavior of DCM and MMA for phase transfer of magnetite nanoparticles using the

fatty acid ricinoleic acid, the polymer PMMA leads to a similar deagglomeration, when

looking at the normalized extinction in figure 68 (right). Similar to the observations in 5.3.1,

regarding eq. (61), this means that the mechanism of destabilization is similar in both cases.

Supposing depletion is the destabilizing mechanism (see 5.2), the relative coil volume

o e t atio φ a d the oil size RG should be similar in MMA and DCM. The coil number

concentration certainly is the same, for the same polymer, with this the same number of

segments, is used in both experiments. The only clue on the coil dimension is based on the

90

calculated FLORY interaction parameters in table 29 together with eq. (40). Since the

solubility of PMMA is better in DCM than in MMA with the calculated FLORY interaction

parameters χ of 0.10 to 0.24, respectively, the polymer coils in DCM should be bigger and

with this the relative volume concentration would be bigger, as well. With this, the depletion

destabilization should be stronger for DCM, yet in figure 68 the insignificant difference

points in the other direction.

In figure 69 the results of 600 nm extinction and normalized extinction versus concentration

of PS in a DCM and a ST dispersion of RA-Fe3O4 are displayed. In table 36 it is shown that

the solubility of RA-Fe3O4 is much better in DCM compared to ST with χ of 0.03 and 0.29,

respectively. In table 30 it is stated, that the PS solubility is similar in DCM and ST.

figure 69: Extinction based determination of the primary particle concentration of RA-Fe3O4 in DCM

(PTDCM120227) and ST (PTST120227) under the presence of PS with the polymer concentration on

the abscissa; (left) absolute (right) normalized values; the extinction ratio for the polymer free point

of ST to DCM is 80.5 %

Again both curves have a decreasing trend, which means PS is a destabilizing polymer, just

like PMMA. The destabilizing trend, found in figure 69 (right) is similar for both solvents and

even similar to PMMA in DCM and MMA in figure 68 (right). This is plausible, looking at the

almost equal FLORY interaction parameters of PS in DCM and ST, following the similar coil

dimensions. As it was the case comparing MMA and DCM, the difference between DCM and

ST is the phase transfer and the deagglomeration efficiency. The initial concentration of

primary particles in ST after phase transfer is only 80.5 % of the concentration in DCM, so

the deagglomeration is less effective with the solvent styrene and the fatty acid ricinoleic

acid. Even though the RA-Fe3O4 solubility is poorer in ST compared to MMA, the primary

particle concentration after phase transfer is much higher. In this case the correlation of

w0

Primary versus the solubility distance D1,2 for different fatty acids, reported in 4.3.3 and

figure 31 does not seem to be reliable in this case. It would have predicted a higher initial

primary particle concentration for MMA compared to ST. Therefore the discussion based on

HANSEN solubility parameters is not helpful in this case.

91

6 Highly Filled Composites

In chapter 2 various methods of composite syntheses and preparations are introduced of

which only one is applied in this thesis, namely the method of solution blending of

separately synthesized nanoparticles and polymers of paragraph 2.2.1 as depicted in figure

5.

This chapter summarizes important findings on the solvent free nanoparticle-polymer-

composites prepared from the complex colloidal dispersion discussed in chapter 5 above,

using the methods of spray drying and subsequent (micro) injection molding. The volume

concentration of the nanoparticles is in any case higher than 10 %, therefore they are

referred to as highly filled composites. Parameters describing the composition are

introduced above in 3.2. Paragraph 6.1 offers a brief introduction to the important

theoretical backgrounds necessary to understand the discussions on the experimental

results in 6.2 and 6.3. It will cover both methods of composite preparation by spray drying

(cf. 6.1.1) as well as injection molding (cf. 6.1.2) and introduce ways to evaluate the

dispersion of particles in a cross-section of a composite by image processing and geometrical

mathematical methods (cf. 6.1.3). The results of spray drying of organic solvent based

nanoparticle polymer mixtures are presented in paragraph 6.2. In the last paragraph of this

chapter, cross-sections of injection molded composites are investigated and evaluated

varying the matrix polymer and filler concentration in paragraph 6.3.

6.1 Theory

6.1.1 Spray Drying

Spray Drying is a well-established industrial process to produce particulate solids from

solutions, suspensions and emulsions [218]. The industrial applications are diverse, so that

spray dryers are found e.g. in the pharmaceutical sector or the food industry (especially for

dairy products) and ceramics industry as well [219-222].

The principle mechanism of spray drying is best divided into the following three individual

process steps:

a) atomization of the liquid (solution, suspension or emulsion),

b) particle formation by drying of the individual droplets and

c) solids separation from the drying gas, using e.g. aero-cyclones and/or filters.

92

Below, the first two steps are discussed theoretically, for they are defining the properties of

the final product of spray drying. Those properties are determined by the powder bulk

product which is characterized by a particle size distribution and a particle morphology

distribution. In case there are several solid components in the feed liquid, their distribution

within a dry particle is of interest as well. Hence, particle morphology and components’ distribution are discussed below as well.

Atomization of a Liquid

Atomization of a bulk liquid into small droplets (spherical liquid particles) is a

thermodynamically unfavorable process which acquires input of energy. This energy is

necessary to dynamically deform the liquid opposing the viscous friction characterized by

the viscosity and to create new surface area (interface of liquid and surrounding gas) which

is characterized by a specific surface energy σ. There are different techniques available to

atomize a liquid categorized by the energy input [223, 224]. Most often nozzles and rotary

atomizers are applied.

In this section only the atomization using an external mixing two fluid nozzle, applied in the

experiments of this thesis, is considered. A scheme of such a nozzle is depicted in figure 70.

figure 70: Principle scheme of an external mixing two fluid nozzle with turbulent atomization, adopted from

[214]

The liquid is flowing through the inner cylinder with the diameter dnozzle under a relatively

low pressure. At the exit a pressurized gas flow expands resulting in a high gas velocity and a

turbulent break-up of the liquid film. A few micrometers below the exit point the liquid film

is already atomized into very small droplets with SAUTER diameters x3,2 on the micrometer

scale. Since it is a turbulent process, the modeling of the droplet size, characterized with x3,2,

is typically achieved empirically applying dimensional analysis. This results in the

dimensionless REYNOLDS number Re, WEBER number We, gas WEBER number Wegas, OHNESORG

number Oh and liquid mass loading of the gas µ [220]. The REYNOLDS number Re is the ratio of

inertia to viscous forces acting on the liquid and defined in eq. (67).

93

liquid

liquidnozzlerel dv

Re (67)

In the numerator the characteristic velocity and length are the relative velocity between gas

and fluid vrel and the inner diameter of the liquid part of the nozzle dnozzle, respectively.

Furthermore, the specific weight of the liquid liquid is characteristic for inertia and the

dynamic viscosity liquid for the viscous forces.

The dimensionless WEBER number We is the ratio of inertia forces of the liquid (of the gas for

the gas WEBER number Wegas) to the surface forces at the interface gas fluid and presented in

eq. (68).

gasnozzle2rel

gas

liquidnozzle2rel

dv

We

dvWe

(68)

The denominator is made up of the specific surface energy σ with [σ] = J/m2 = N/m, which is

depending on the interactions between the gaseous and the liquid material specious. It can

thus be described as the energy necessary to increase the surface area by 1 m2.

Another result of the dimensional analysis is the OHNESORG number Oh which is a

combination of Re and We and defined in eq. (69).

liquidnozzle

liquid

ρdσRe

WeOh

(69)

Low OHNESORG numbers are typical for high surface and inertia forces compared to lower

viscous forces. The final dimensionless parameter is the ratio of liquid mass flow to gas mass

flow in eq. (70), which is also often found as the inverse value, termed ALR (air to liquid

ratio).

ALRm

m 1

gas

liquid

(70)

Based on many experimental results, MULHEM et al. present an empirical relation for the

SAUTER diameter of droplets generated by an external mixing two fluid nozzle in eq. (71)

[220].

4.0

gas

0622.0nozzle3,2 21.0

We

Ohdx

(71)

94

Combining eq. (71) with eqs. (68), (69) and (70) one can better judge on the strength of the

individual influencing parameters in eq. (72). Please note that the units are defined to be

following SI conventions with a negative exponent of the empirical exponent presented.

0311.0liquid

0.0622liquid

3689.04.0gas

4.0

gas

liquid5689.08.0rel2,3 21.0

m

mdvx nozzle

(72)

The influencing parameters are ordered with increasing absolute exponent values.

Conclusion can be drawn, that a change in the first parameters has the stronger effect on

the droplet size. The relative velocity vrel is typically increased by raising the gas pressure and

thus the gas velocity, which will reduce the droplet size. Smaller nozzle diameters will result

in smaller droplets as well. The last four parameters are material specific and can

technologically be influenced by the temperature (pressure for the specific weight of the

gas) only. When working with liquids that are more viscous with higher liquid, the droplet

size is expected to be higher, the other values being constant.

Drying of a Droplet

In a spray dryer the atomized droplets come in contact with a hot gas in different possible

regimes. The spray dryer used in this study is based on a co-current flow of spray and hot

drying gas. Other possibilities are counter-current and cross-flow. For the co-current regime

the characteristic is that drying is very rapid in the beginning and at the end of the drying

step the peak temperature is relatively low, which is important for heat sensitive materials,

such as the organics used in this study. In the co-current regime, the final particle

temperature is as high as the inlet temperature of the hot gas.

In general the drying of a droplet is a phenomenon of combined heat and mass transport.

Additionally solids formation has to be accounted for. The mass transported is the solvent

which is transported across the droplet surface Sdroplet out of the particle to the drying gas. A

simple model for the convective mass transport at the droplet surface including a driving

force and a transport coefficient is presented in eq. (73).

gas solvent,surface solvent,droplet

solvent ppTRS

m (73)

The pressures in the parenthesis are the partial pressures of the solvent at the surface and in

the drying gas. The higher their difference is, the more solvent is transported in time. R and T

are the gas constant and total temperature, respectively. Finally, the mass transfer

coefficient β needs to be evaluated with the help of dimensional analysis using empirical

relations of SHERWOOD number Sh, SCHMIDT number Sc and REYNOLDS number Re. The

SHERWOOD number Sh represents the ratio of mass transport by convection to mass transport

95

by diffusion. Whereas, the SCHMIDT number Sc relates the kinematic viscosity to the diffusion

coefficient and REYNOLDS number is described above.

Similar to mass transport, the transport of heat flowing from the drying gas into the droplet

is characterized by a transport coefficient and a driving force, which is the temperature

difference. The droplet area specific heat flow due to convection is given in eq. (74).

gasparticledroplet

TTS

Q (74)

Again the transport coefficient, which in this case is the heat transfer coefficient α, is

assessed with the help of the empirical relation of three different dimensionless numbers.

The REYNOLDS number is the same like above and instead of SHERWOOD and SCHMIDT number

for mass transport, the NUSSELT number Nu and PRANDTL number Pr are introduced for

transport of heat. Similar to the SHERWOOD number, the NUSSELT number relates heat

transport by convection to heat transport by conduction (diffusion). A heat transport

analogy to the SCHMIDT number is the PRANDTL number, defined as the ratio of kinematic

viscosity and heat conductivity.

A typical time progression of mass and temperature of a drying droplet, with hypothetical

constant surface area, is depicted in figure 71.

figure 71: Time progression of mass and temperature of a drying droplet with constant surface area due to film

formation, taken from [225] with the steps A through D described in the text

The first step A is characterized by the heating of the droplet up to a constant wet-bulb

temperature θwb with an acceleration of the mass flow of the solvent. In step B the mass

flow is constant and high at the constant wet-bulb temperature, which is defined in eq. (75),

assuming a sufficiently high conductivity of heat inside of the drying droplet. For the case

that the surface area of the drying droplet is changing the mass flow is expected to decrease

instead of being constant.

96

gasdroplet

solventsolventwb

S

hm (75)

So far only one parameter is not described, which is the specific heat of evaporation of the

solvent hsolvent. Once the solvent mass transport is hindered by a solids layer (shell

formation) on the dried droplet surface, the temperature increases in step C and the mass

flow decelerates. At step D the temperature of gas and particle are similar and the solvent

content in the particle is approaching equilibrium, no more drying occurring. Certainly the

progression of mass and temperature very much depend on the solids formation while the

droplet is drying influencing heat and mass transport. Altogether the particle morphology is

defined by these progressions as well.

Particle Morphology

The particle design by spray drying is extremely variable especially due to the manifold

possible particle morphologies, depending on various process and material properties.

Consequently, there are many publications dealing with particle morphology in connection

with spray drying [226-229] yet only a few offer models of particle formation. HANDSCOMB

et al. introduce a model of a drying droplet with suspended small particles and shell

formation discussing the resulting possible morphologies [230]. Taken from this publication

is the graphic in figure 72.

figure 72: Drying progression of a droplet with suspended small particles and shell formation, taken from [230]

At very low solids concentration of less than 1 % by weight no particles are formed from the

drying d oplet. Ve high d i g te pe atu es ill lead to puffed pa ti les ith dia ete s considerably larger than the initial droplet diameter. At moderate drying temperatures the

flexibility of the forming shell distinguishes solid particle formation for rigid dry shells and

hollow particles. Depending on the elastic behavior of the flexible shell the hollow particles

can be spherical or shriveled. In case the shell is very dense the hollow spheres can burst and

leave blistered particles.

97

An important dimensionless number determining the particle morphology is the PECLET

number Pe. It is the ratio of evaporation rate of the solvent solvent and diffusion rate of the

solutes Dsolutes [226]. At high PECLET numbers the evaporation rate is much higher than the

diffusion of the solutes resulting in surface enrichment of the solutes and shell formation.

In need for a descriptive morphology parameter of the composites prepared in 6.2, the

structural parameter SP is introduced in [23] and presented in eq. (76).

i

3,i1

m,icomposite

PSDBET

PSD 6with,100 xS

S

SSP (76)

The structural parameter SP is defined with the BET determined mass specific surface SBET

and the calculated mass specific surface from the particle size distribution SPSD. It ranges

from zero to 100. Smaller SP values account for higher structuring, which can be due to

nanoparticle agglomerates, increased micro porosities or general shape deviations from a

sphere. Solid spheres with smooth surfaces will lead to SP = 100. The specific weight of the

composite microparticle composite with F = 0.3 and D = 0.2 amounts to 1.51 g/cm3, with

magnetite = 5.20 g/cm3, fatty acid = 0.90 g/cm3 and polymer = 1.20 g/cm3. Taken from the particle

size distributions determined with laser diffraction the average particle size xm,i and the

weight fraction i of fractions i are used for calculating the specific surface, in this case

assuming spherical particles. It can be stated that, the smaller the structural parameter SP,

the more irregular the shape/morphology of the particles. In addition to the microparticle

morphology, liberated, not encapsulated nanoparticles will result in a higher BET surface and

thus reduce the SP value as well.

Compositional Separation

In case there are several solid components in the spray-dried liquid, a separation in

composition can occur [220, 225, 226]. If one component is bigger than the droplets formed

in the spray drying process, than there obviously will be a depletion of this component in the

smaller particles formed and thus a compositional segregation depending on the particle size

[220]. Separation within one drying droplet is due to different mobilities (diffusivity) of the

individual components and thus different PECLET numbers, as defined above [226].

Components can specifically concentrate on the particle surface because of this. If there are

slightly asymmetric drying conditions for a single droplet than indentations in the particles

can occur with higher concentrations of components with lower diffusivity [225].

6.1.2 (Micro) Injection Molding

One application of spray-dried composite microparticles, composed of nanoparticles and

thermoplastic polymers, is the preparation of microstructured components for

98

microsystems. This is achieved by micro injection molding, a polymer processing technique

[231-233]. Simple injection molding is a widely used shaping method in the plastics industry.

In general, thermoplastic polymer granules are fed in an extruder, which is heating up the

granules to obtain a polymer melt. This melt is intensively mixed by the screw extruder

device and at the end injected through a needle into a structured molding tool, which is a

negative form of the final product [231]. The injection occurs at very high pressures of

several tens of bar. In figure 73 (left) there is a schematic drawing of an extruder fed on the

right with the composite particles prepared by spray drying, transported and molten from

right to left and finally injected into a microstructured tool on the left. In figure 73 (right)

there is an optical micrograph in top-view of a micro injection molded test structure with a

nanoparticle-polymer-composite PMMA-RA-Fe3O4 with F = 0.3 and D = 0.2. There are well

reproduced edge structures proving a good result of micro injection molding using a rather

high filled composite.

figure 73: (left) principal scheme of an injection molding device (right) top view of an optical micrograph of a

PMMA-RA-Fe3O4 composite with F = 0.3 and D = 0.2 of a test structure

In micro injection molding the molding tool is microstructured with elements that have

thicknesses or lengths of a few micrometer only [233]. Therefore, when processing

composites, the size of the filler component of the thermoplastic polymer needs to be

sufficiently smaller than the smallest structures to guarantee the same composition within

the microstructured elements. Well dispersed nanoparticles satisfy this condition.

6.1.3 Image Processing and the Mathematical Description of the State

of Dispersion

In paragraph 2.1.3 the te state of dispe sio is introduced and descriptively discussed. It

is shown that the dispersion is characterized by homogeneity and a deagglomeration

component. When judging the actual dispersion of a particulate component in a matrix,

typically a cross-section is investigated but in many studies only comparatively and

descriptively evaluated. Often only the size of the disperse component (particles, aggregates

and agglomerates) is the sole quantity [72, 79, 81, 82, 234, 235], even though it is only

judging the deagglomeration success and not the homogenization. Researching the

literature for mathematical measures describing objectively and quantitatively the state of

99

dispersion leads to fe fi di gs ithout a o e a d o l universal measure [54-56, 58, 236,

237]

The premise to evaluate the dispersion, based on cross-section imaging, is an objective

reproducible image processing procedure for extracting and detecting the disperse

component. Certainly this very much also depends on the quality of the primary image of

the cross-section and the contrast of the phases, which however is not part of this

paragraph, thus assuming image quality is sufficient. In this paragraph the image processing

procedure of binarization and particle detection is introduced and different objective

quantifications of the state of dispersion are proposed. Image acquisition for a good contrast

and subsequent image processing is discussed in A.2.2 and [19], respectively.

There are three steps of processing (binarization, particle recognition, VORONOI tessellation)

presented in figure 74 which will be discussed below.

figure 74: Image processing steps (from left to right) of a phase contrast AFM image on the left showing dark

magnetite and light polymer phases, binarized after thresholding and with a watershed, automatic

detection and measurement of >800 individual particles (including aggregates and agglomerates)

and finally the VORONOI diagram of the binarized image

Binarization

The very first image processing step is generating a grey value image with either 256 or

65536 grey values, corresponding to 8-bit and 16-bit images, respectively. The highest grey

value is white and the zero value typically corresponds to black. It is assumed that the darker

pixels to correspond to the particulate phase, as is the case in figure 74 on the left showing

an 8-bit grey value image. The next step is the binarization, which sets all pixels below a

certain value to zero (particles) and above to one (matrix). This is achieved by thresholding.

The second image in figure 74 is the result of setting the threshold to a grey value of 200.

What is also done in the second image is a binary image tool, called watershed, which

segments very structured elements [238]. This should not have been done, when wanting to

preserve the ability to recognize connected particles as agglomerates.

With the as-prepared image an automated particle recognition and size determination can

be achieved using the standard freeware program in scientific image processing ImageJ

100

[238]. Otherwise one can also define the area filling ratio φA by dividing the number of pixels

with value zero (particles) by the total number of image pixels.

Particle Recognition and Size Determination

The third image in figure 74 depicts the result of particle recognition with 836 individual

particles and a median number weighted FERET size of 100 nm. The FERET size determined by

the program ImageJ is the longest distance between any two points of the individual

particle. Another possible diameter is the area equivalent sphere diameter.

VORONOI Polygons

A next processing step is the generation of so-called VORONOI polygons around the

segmented particles. These polygons have the property that every point within a polygon is

closest to the finite element surrounded by the polygon [54, 239]. Such tessellations can be

helpful in determining quantitative measures for the state of dispersion and have been

promising tools for this purpose [19, 55, 240].

State of Dispersion Measures

Ce tai l it is diffi ult t i g to put the easu e of state of dispe sio i o e u e , maybe as impossible as trying to put the particle size distribution in a single value. Yet it is

necessary to start defining objective ways to quantify the state of dispersion.

One proposal is based on the horizontal distance between particles, which will be referred to

as the li e ethod [56, 241, 242]. Depending on the state of dispersion the coefficient of

variance cov of all separation distances li will be low for a good dispersion and high for a

poor dispersion. The VORONOI polygons introduced above can be used similarly [55], defined

he e as the VORONOI ethod . The cov of the polygon areas Ai will be low for a good

dispersion and high for a poor one.

Both methods are compared in figure 75 evaluating a set of simulated dispersion with

improving homogeneity from left to right. One can see that the VORONOI method is more

sensible when compared to the line method.

101

figure 75: Coefficients of variance normalized with the first value covmax for the VORNOI and the line method and

a set of simulated images with (from left to right) improving state of dispersion taken from fig. 7 in

[55], covmax is 1.08 for the VORONOI method and 1.42 for the line method

A better quantitative resolution is noticeable for the cov of the areas of the VORONOI

polygons (darker bars in figure 75) as compared to the cov of the distances of the line

method. Thus the VORONOI method is expected to offer a better measure for the state of

dispersion as it is based on a 2D analysis instead of the 1D and directional based line

method. Applying the VORONOI method to the four images in figure 3, where the state of

dispersion was introduced, results in covVORONOI = (0.581, 0.254, 0.538, 0.129) from top left

(agglomerated and demixed) to bottom right (deagglomerated and mixed).

6.1.4 Filler Concentration – Agglomerate Concentration – Stability

Relation

In paragraph 5.3.1 it is shown that for destabilizing polymers PMMA and PC, the primary

particle concentration wPrimary decreases with increasing polymer concentration in the

dispersion cpolymer. For a constant solids concentration in the dispersion csolid the filler

concentration of a dry composite material F prepared from such a dispersion increases with

decreasing polymer concentration cPolymer, cf. eq. (11) on page 23. This concludes that with

increasing filler concentration the relative agglomerate concentration φAgglomerates / φtotal,

agglomerates related to all nanoparticles, is reduced. It shall be assumed that the area

concentration of particles in a planar cross-section is equal to the volume concentration of

particles within the composite part. Since a higher filler concentration also leads to more

nanoparticles in total, the absolute agglomerates concentration in a composite cross-section

φAgglomerates does not necessarily need to decrease.

This paragraph explains theoretically how to predict the concentration of agglomerated

nanoparticles and primary nanoparticles within a solid composite cross-section from stability

experiments in dispersion. It is presumed in these calculations, that the state of dispersion is

not affected by the composite preparation through spray drying and injection molding,

102

which certainly is questionable and can be reconsidered when discussing experimental

results.

First of all, the agglomerate volume or area fraction (concentration) is defined in eq. (77)

with the total volume fraction and the fraction of primary particles.

PrimarytotalesAgglomerat (77)

Both the primary and the total particle fractions can be defined with the filler concentration

F, the surfactant ratio D and the specific weights of the three solid components,

nanoparticles, surfactants and the polymer, in eq. (78).

1

arytotal/Primpolymersurfactantlesnanoparticarytotal/Prim 1

111

D

F

D

(78)

As introduced in 3.2.2, the total filler concentration is a function of the polymer cpolymer and

the solids concentration csolid in the dispersion as well as D and solvent. Furthermore, by

definition of the primary particle concentration wPrimary in A.5, the filler concentration of the

primary particles is the product of total filler concentration F = Ftotal and wPrimary. Both

ingredients of eq. (78) are presented in eq. (79).

PrimarytotalPrimary

solvent

solidpolymertotal 1

11

1

1

wFF

D

cc

DF

(79)

Finally, as demonstrated in paragraph 5.3, the primary particle concentration wPrimary in

dispersions with destabilizing polymers can be formulated as a function of the polymer

concentration in the dispersion cpolymer with two parameters: the initial primary particle

concentration w0Primary and an exponential decay rate A.

polymer0PrimaryPrimary

cAeww

(80)

In figure 76 there is a graphic evaluation of eqs. (77) through (80) plotting the volume

fractions φtotal, φPrimary and φAgglomerates as functions of F and cpolymer for parameters of the

destabilizing polymers PMMA and PC reported in 5.3.1 with csolid = 0.05 and w0Primary = 0.9 as

well as APMMA = 0.025 and APC = 0.050.

103

figure 76: Visualization of the volume (area) fractions of the primary and agglomerated particles as well as the

total volume fraction of nanoparticles applying eqs. (77) - (80), parameters w0

Primary and A are chosen

in connection with colloidal stability investigations of PMMA and PC in 5.3.1

It shall be recognized that the agglomerate fraction φAgglomerates increases up to a filler

concentration of 0.65 for APMMA and 0.70 for APC and dropping for higher filled composites.

Nevertheless, the relative agglomerate concentration is decreasing with increasing filler

concentration and thus decreasing polymer concentration in the dispersion for all

destabilizing colloidal parameter conditions, as plotted in figure 77 for selected values of

initial primary particle concentration and decay rate.

figure 77: Relative agglomerate concentration as a function of the total filler concentration F and the polymer

concentration cpolymer with the parameters of colloidal stability of eq. (80) applying eqs. (77) - (80)

It is shown that there is a reproducibility problem for magnetite nanoparticle precipitation

batches in 5.3.1 which will affect the initial primary particle concentration. This effect as well

as the aforementioned impact of different polymers on the decay rate can be acknowledged

when wanting to predict the fraction of agglomerates in a composite supposing one has

knowledge on the parameters and, furthermore, the state of dispersion is not changed by

spray drying and/or injection molding.

104

6.2 Experimental Results – Spray-Dried Microparticle

Composites

This paragraph is dedicated to discuss the results of the spray drying of dichloromethane

based dispersions of fatty acid grafted magnetite nanoparticles with different polymers as

well as fatty acids and for variable filler concentrations. Spray drying of non-aqueous

formulations can be found in the literature for example in [66, 226, 243-246].

The following investigations are based on BSE-SEM imaging (cf. A.2.2), granulometry with

laser diffraction and BET analysis (cf. A.2.1) as well as compositional analysis with TGA (cf.

A.2.4).

6.2.1 Compositional Separation

At first an interesting phenomenon concerning the spray drying of the destabilized PMMA

sytems is discussed, which is the compositional segregation. The spray-dried particles can be

collected in three fractions in the spray dryer. In figure 78 one can see a schematic drawing

of the spray dryer set-up used for this study with the location of the three particle fractions:

cylinder, cyclone and filter.

figure 78: Schematic set-up of the co-current spray dryer used in this thesis with three particle fractions

(cylinder, cyclone and filter) where the cyclone fraction is the product of the process and should be

the largest fraction in mass, the solvent is recovered in a condenser and the dry air recycled passing a

heating device

The majority of the particles are found in the cyclone with a yield of about 80 % by weight.

The other particles are located in the spray cylinder in front of or in the bag filter behind the

cyclone. When collecting the particles, which appear brown to black depending on the filler

concentration, one can see the darkest appearance for the cylinder and the lightest for the

filter fraction. Since the color black is induced by the strong light absorption of the magnetite

105

particles (cf. A.8), this subjective finding suggests that the magnetite concentration varies

within those fractions.

PMMA-based Composites

In order to investigate the previous assumption the particle size distributions of these

fractions are determined with laser diffraction and the composition is analyzed with TGA,

assuming that the magnetite concentration is approximately equal to the residual mass at

600 °C, corresponding to the findings in 4.3.4 and [21]. A representative set of these two

analyses for the three fractions of a PMMA composite with RA-Fe3O4 at F = 0.3, DRA = 0.2 and

csolid = 0.07 is given in the two graphs of figure 79.

figure 79: (left) representative particle size distributions of the microparticles of the three fractions of the spray

dryer measured with laser diffraction and (right) TGA curves of the corresponding samples with an

initial composition of PMMA RA-Fe3O4 with F = 0.3 and DRA = 0.2; the yields are 11 %, 74 % and 15 %

for cylinder, cyclone and filter, respectively

The particles of the filter fraction are the finest, which is expected. Based on the median

values, the particles in the cylinder of this volume weighted sum distribution Q3(x) are the

largest with about 8 µm compared to 3 µm and 5 µm for the filter and cyclone material,

respectively, which can be confirmed by SEM imaging. Looking at the residual masses at a

temperature of 600 C, before the last step of decomposition (due to magnetite reduction, cf.

4.3.4) in the right graph of figure 79, the magnetite concentration is the highest for the

cylinder and the lowest for the filter sample, as it had been assumed. The values are 43 %,

32 % and 23 % for cylinder, cyclone and filter, respectively. Combining the information of

particle size and magnetite concentration in figure 79 one finds out that larger particles have

a higher magnetite concentration. Furthermore the composition of the cyclone fraction

correlates well with the set value of F = 0.3 and thus meets the specifications. Therefore the

filler concentration in the small filter material is too low and in the larger cylinder fraction

too high. For the deviations in magnetite concentration are similar and the yield of both

fractions is similar, the total material composition meets the specification again. In summary

there is a compositional separation depending on the spray-dried microparticle size. This

106

must be occurring already during the atomization step. It is supposed, that large depletion

induced floccules are situated within the larger droplets. In other words, the atomized

smaller droplets which will eventually be the filter material after drying are depleted of

magnetite for they are smaller than the large filler agglomerates, following the results of

MULHEM et al. in [220].

This separation phenomenon is studied for a number of spray drying experiments with

PMMA, the fatty acid ricinoleic acid (RA-Fe3O4) at DRA = 0.2 and varying filler concentrations

F. The individual results are presented in two graphs in figure 80.

figure 80: (left) TGA measured mass residue at 600 °C of various spray drying experiments with RA-Fe3O4 and

DRA = 0.2 for the three fractions of the spray dryer at different initial filler concentrations of

magnetite F; (right) relative magnetite concentration of the cylinder or the filter fractions compared

to the corresponding cyclone fraction with the cyclone fraction residual mass on the abscissa

The data is scattering rather strongly, which must be due to different precipitation and

phase transfer batches used and the varying resulting state of dispersions in the polymer

free dichloromethane suspensions, cf. 5.3.1. Looking at the graph on the left in figure 80, it is

to notice, that the cyclone magnetite concentration (green upward triangle), measured with

the 600 °C TGA residue, correspond well to the calculated expected filler concentration F.

They approximately lie on the line through the origin. All samples from the cylinder and the

filter are above and below this line, respectively. This confirms the argumentation from

above, showing that the cylinder material is enriched in magnetite due to large agglomerates

and consequently there is a lower content of magnetite in the filter material. The graph in

figure 80 (right) show the measured magnetite concentrations of the cylinder and filter

particles related to the concentration of the corresponding cyclone material as a function of

the magnetite concentration of the cyclone material. The deviations of the cylinder material

are increasing with decreasing magnetite concentration and thus increasing polymer

concentration in the spray-dried dispersion. This trend goes along well with the findings in

chapter 5, for a higher polymer concentration in the solvent based dispersion and thus lower

magnetite concentration in the dried composite leads to increased magnetite agglomerates

which will gather in the larger particles of the spray dryer. Again the uncertainty in the data

107

presented is most probably due to the deviations in the quality of the different precipitation

and phase transfer batches, cf. paragraph 5.3.1.

PC- and PVB-based Composites

The amount of data for the destabilizing polymer PC and the stabilizing polymer PVB is too

little to be presented conclusively. Yet, for PC a strong compositional separation is

noticeable. For PVB neither the visual inspection (difference in light absorption) nor

compositional analyses reveal a statistically significant compositional separation within the

differently sized particle fractions.

High Temperature Degradation

Looking again at the PMMA based composites studied in this paragraph, there is another not

yet described difference between the three fractions related to the inert atmosphere

decomposition between 600 °C and 900 °C, which has been attributed to the reduction of

magnetite to lower oxygen containing iron compounds by residual carbon from adsorbed

species, cf. paragraph 4.3.4 and [21]. For all samples presented in figure 80 the relative mass

loss between 600 °C and 900 °C is determined as (w600 °C-w900 °C)/w600 °C. The average values

and the 95 % confidence level are presented in figure 81.

figure 81: High temperature mass loss (residual mass at 600 °C compared to residual mass after magnetite

reduction at 900 °C, compare with investigations in 4.3.4) of the TGA analyzed particles of various

spray drying experiments with RA-Fe3O4 based on 38, 51 and 23 samples for cylinder, cyclone and

filter materials, respectively

The lowest values is found for the cylinder material, which is yet not statistically significantly

lower than the average relative mass loss of the cyclone particles. A significant difference

occurs for the filter material, with the highest relative mass loss. In [21] it is argued, that the

residual carbon responsible for the magnetite reduction in this temperature window is from

the chemically bound fatty acid molecules. This would mean that the magnetite

nanoparticles captured in the fine filter fraction bear more chemically bound fatty acids. This

is logical assuming the magnetite of the filter fraction are mostly made up of primary

108

nanoparticles and not the agglomerates from the phase transfer or the polymer addition

with a high specific surface area covered by chemically adsorbed fatty acid molecules. Thus it

is assumed that the grafting density of organic matter on agglomerated magnetite

nanoparticles is lower and hence the high temperature mass loss is lower as well.

Separation of Components in a Single Microparticle

With the help of BSE-SEM imaging another separation phenomenon can often be observed

as shown in figure 82 for a PMMA-based composite. This micrograph reveals asymmetric

truncated particles with high magnetite content on one side of the composite microparticle.

figure 82: BSE-SEM images of the particles of the cyclone fraction of a PMMA-based composite with RA-Fe3O4

DRA = 0.2 and F = 0.3, the image on the right is a close-up of a region in the approximate center of the

image on the left, magnification is 3,000x

This type of separation could occur at the moment when the droplet is formed and/or after

the atomization step while drying of the particle, cf. last subsection of paragraph 6.1.1. It is

supposed that the dimension of the magnetite agglomerate visible corresponds to the

agglomerate dimension in the destabilized dispersion. This will be considered again below in

paragraph 6.3.3.

6.2.2 Yield of Product

It is shown in the previous paragraph that only the cyclone material should be considered as

the product of the composite preparation process using the method of spray drying an

organic solvent based mixture of stabilized nanoparticles and a dissolved polymer, especially

for destabilizing polymers. This is because the compositional specifications of the material in

the cylinder and in the filter are too far off of the specified values for filler concentrations F.

Therefore in this paragraph the yield of cyclone material is investigated depending on the

filler concentration F for RA-Fe3O4 and DRA = 0.2. The results are presented in figure 83

plotting the yield against the filler concentration, where the yield is the mass of material in

the cyclone coarse exit related to the entire solid content spray-dried.

109

figure 83: Yield of product at the coarse exit of the cyclone as a function of the filler concentration F for PMMA-

based composites with RA-Fe3O4 and DRA = 0.2

It can be seen that the yield of the product at the cyclone increases with increasing filler

content from 70 % up to 90 %. This is due to a decreasing yield in the cylinder. There are two

explanations for this. On the one hand, the stickiness of the microparticles is reduced with

increasing filler concentration so the particles will not get stuck in the cylinder. However, this

would mean that the coarse particles would be found in the cyclone, which is not the case

when looking at the fraction size distributions as presented below in 6.2.5. On the other

hand, chapter 5 teaches that with increasing filler concentration, thus decreasing polymer

concentration in the spray-dried dispersion, the stability of the nanoparticles improves and

the content of agglomerates is reduced. With this, the reduction of amount of coarse

particles in the cylinder could be due to a reduced amount of magnetite nanoparticle

agglomerates, which could influence formation of large particles as reported in [220].

Besides the better yield with increasing filler concentration the powder bulk properties are

affected in a way that the bulk powder is denser and shows a better flowability. Furthermore

the electrostatic charging of the particles whilst collecting in the cyclone is reduced, which is

due to the reduction of the dielectric polymer content. This improves the discharge of

material out of the collection vessel.

6.2.3 Influence of the Polymer

The investigations on the colloidal stability of dichloromethane dispersions of RA-Fe3O4 with

the polymers PMMA, PC and PVB are discussed in paragraph 5.3.1. In this chapter it is to

show that the colloidal stability impact can be correspondingly visualized in the dispersion of

magnetite nanoparticles within the spray-dried composite microparticles. Therefore, the

distribution of magnetite nanoparticles in the polymer matrix after quick solvent

evaporation in the spray dryer with a polymer concentration of cpolymer = 52 g/l is presented.

Composite microparticles with the composition by weight of: 30 % Fe3O4, 6 % ricinoleic acid,

64 % polymer (PMMA, PC or PVB), are shown in figure 84. In the center area of each image a

110

single spherical spray-dried microparticle is depicted with a diameter of about 2 . The

images are inverted back-scattering scanning electron micrographs (BSE-SEM). The heavy

iron atoms in the magnetite nanoparticles scatter electron back more efficiently than the

other atoms in the composite (carbon, oxygen and hydrogen) resulting in higher electron

densities which appear black when inverting the intensities of the BSE-SEM image. The

microparticles rest on a sticky carbon patch which appears lightly grey.

figure 84: Inverted BSE-SEM images of composites with RA-Fe3O4 at F = 0.3 and DRA = 0.2 for the destabilizing

polymers PMMA and PC as well as the stabilizing polymer PVB [27]

Phase separation leads to large agglomerates of magnetite apparent in the PMMA and even

more so in the PC composite microparticles. For PC, much less smaller magnetite spots can

be detected, taking into account that the resolution with PC is poorer compared to PMMA

because of the enhanced electrostatic charging of PC. For PVB, regarding the low resolution

of the electron microscope, no agglomerates are to be found and the microparticle appears

much more homogeneous. Using a higher resolution system, the distribution of magnetite in

the cross-section of injection molded samples is presented in paragraph 6.3.1, below.

6.2.4 Influence of the Surfactant

It is shown in paragraph 4.3 how different fatty acids, namely ricinoleic acid (RA), linoleic

acid (LA), oleic acid (OA), myristic acid (MA) and caprylic acid (CA) influence the state of

dispersion after the phase transfer process step. Continuing in paragraph 5.3.2, the impact

of these different fatty acids on the colloidal stability in solutions of PMMA and PVB is

discussed. In this paragraph the influence of the fatty acid surfactant type on the composite

preparation with the spray drying method is presented. The results are published as well in

[23].

PMMA-based Composites

In figure 85 BSE-SEM images of microparticle composites are presented with F = 0.3 in the

destabilizing polymer PMMA at two magnifications. When processing composites with this

composition, the polymer concentration in the solvent based mixture being atomized is

52 g/l. In figure 50 of paragraph 5.3.2 investigating the colloidal stability, this composition is

found in the far right points and the primary particle concentrations are 14.9 %, 6.0 %, 2.4 %,

111

~0 % and 1.7 % for RA, LA, OA, MA and CA, respectively. The other nanoparticles are

agglomerated in yet unspecified sizes.

figure 85: BSE-SEM images of spray-dried composites of (from left to right) RA, LA, OA, MA and CA coated

magnetite (appearing light for high back scatter electron densities of iron atoms) in PMMA,

magnification of 2000x and 10000x in the upper and lower row, respectively [23]

For all samples one can notice a phase separation as well as heterogeneous distribution of

the nanoparticles, which form rather large agglomerates (white spots). The largest of these

agglomerates are found for MA and CA samples. A dramatic phase separation is observed for

the MA sample, which also shows the lowest primary particle concentration for the solvent

based polymer concentration. In any case the findings coincide with the destabilization of

PMMA with a high polymer concentration in figure 50.

In table 9 the granulometric parameters median size x50,3 obtained with laser diffraction and

the calculated specific surface area SPSD as well as the measured specific surface area

obtained with the BET method SBET are presented for the different fatty acid coated

magnetite PMMA composite microparticles together with the structural parameter SP,

accounting for morphology as introduced in paragraph 6.1.1.

table 9: Granulometric data: median microparticle size, specific surface area calculated from the particle size

distribution, BET surface and structural parameter SP of the spray-dried samples with PMMA as the matrix

polymer [23]

fatty acid x50,3 in µm SPSD in m2/g SBET in m2/g SP

RA 3.37 1.51 2.77 54.5

LA 3.08 1.74 3.90 44.6

OA 3.17 1.72 4.00 43.0

MA 3.95 1.35 12.44 10.6

CA 3.08 1.69 8.00 21.1

The particle sizes do not vary significantly, only MA stands out with larger particles, which

also leads to a smaller calculated SPSD. It is supposed that the larger agglomerates for MA as

compared to CA are due to the reduction of the pH from 9 to 8 in order to guarantee a

112

complete phase transfer, as shown in figure 27 and described in paragraph 4.3.1. The

structural parameter SP, however, correlates well with the order of colloidal stability,

presented in table 8. Thus it is supposed, that the presence of non encapsulated liberated

nanoparticle agglomerates dominates this value. The more nanoparticles are not

encapsulated in the matrix polymer, the lower the value is, especially since the morphology

of the microparticles is similar, cf. figure 85.

PVB-based Composites

In figure 86 the BSE-SEM images of microparticle composites with F = 0.3 in the stabilizing

polymer PVB for two magnifications are presented in a comparable way to the PMMA based

composites above, cf. figure 85.

figure 86: BSE-SEM images of spray-dried composites of (from left to right) RA, LA, OA, MA and CA coated

magnetite (appearing light for high back scatter electron densities of iron) in PVB, magnification of

2000x and 10000x in the upper and lower row, respectively; to be compared to the images in figure

85 [23]

Two main differences are observed comparing PMMA composites with PVB composites. The

particles are larger and the phase separation is reduced. Only for OA (third images from the

left in figure 86) bright spots depicting dense agglomerates are visible. For MA,

microparticles with different filler contents are observed. However, no dense agglomerates

occur, which means that the distribution is still poor but the deagglomeration due to PVB

adsorption is successful. Both RA and LA do not show any signs of phase separation.

Certainly the size of the particles is determined by the viscosity of the solvent based mixture

which influences droplet formation in the external mixing two fluid pressurized nozzle. As

the viscosity of PVB in DCM is indeed larger than for PMMA, larger droplets are resulting

independently from the fatty acid. The reason why the nanoparticle fillers are distributed

more homogeneously with less agglomeration is the stabilization which is found in figure 52.

In comparison to the discussion for the PMMA composites, above, the granulometric data of

the PVB composite microparticles are presented in table 10.

113

table 10: Granulometric data: median microparticle size, specific surface area calculated from the particle size

distribution, BET surface and structural parameter SP of the spray-dried samples with PVB as the matrix

polymer [23]

fatty acid x50,3 in µm SPSD in m2/g SBET in m2/g SP

RA 5.18 1.24 1.57 78.9

LA 5.81 1.13 1.78 63.5

OA 6.56 1.02 4.60 22.2

MA 6.71 1.00 5.81 17.2

CA 7.18 0.98 3.54 27.7

Just as mentioned before, the particles are larger than the PMMA composite microparticles,

which goes along with smaller calculated specific surfaces SPSD. The structural parameter SP,

however, is larger than for the PMMA composites, except for OA which will be discussed

below. If the SP was independent from the morphology and the polymer, the results would

show an improved encapsulation of the nanoparticles and less liberated agglomerates. For

OA grafted magnetite in PVB solution the following could occur. In figure 52 it can be seen,

that for the highest PVB concentration OA has the lowest content of primary particles, so

PVB is less efficient in stabilizing OA. As for the hypothesis of adsorption of PVB at

unoccupied magnetite surface sites, discussed in paragraph 5.3.3, it would mean that OA

might be less efficient in this mechanism as compared to the other fatty acids. In other

words, the surface of magnetite with OA is less accessible for the PVB to adsorb. Since the

BET surfaces are similar for PMMA and PVB samples the difference in SP for OA must in this

case be due to the low calculated SPSD (larger particle size) and dominated by morphological

differences of the polymers.

6.2.5 Increasing Filler Concentration F

In the final paragraph dedicated to the microparticles prepared by spray drying, the

influence on the particle morphology by increasing filler concentration F is discussed. Three

individual particles of the cyclone fraction of PMMA-based composites with filler

concentrations of 30 %, 50 % and 80% are depicted in the BSE-SEM micrographs of figure 87.

The tremendous backscattering of electrons by the iron atoms in the magnetite

a opa ti les e o es o ious, looki g at the i easi g da k ess of the i opa ti les with higher magnetite concentration, making it hard to judge on the nanoparticle

distribution within the composite microparticles for very high filler concentrations. This is

the reason, why the previously discussed investigations are based on composites with

F = 0.3.

114

figure 87: Inverted BSE-SEM images of individual spray-dried microparticles with similar size from the cyclone

fraction with RA-Fe3O4 at DRA = 0.2 at (from left to right) F = (0.3, 0.5, 0.8) with comparable contrast

to visualize the impact of the increasing magnetite concentration, magnification 20,000x

Looking at the morphology it can be noted, that generally the particles with filler

concentrations above 30 % are more compact and less truncated. However, broken particles

still reveal the shell formation whilst drying. The particle size distributions of the three spray

dryer fractions of the samples with F = (0.3, 0.5, 0.8) determined with laser diffraction are

presented in figure 88.

figure 88: Volume weighted PSD of the three spray dryer fractions at the cylinder, cyclone and filter for the

PMMA-based composites of RA-Fe3O4 with DRA = 0.2 and the filler concentrations F = (0.3, 0.5, 0.8)

For all filler concentrations the largest particle sizes are occurring for the cylinder fraction,

followed by the cyclone fraction and the filter fraction, which is expected when compared

with the results in figure 79. There is a clear trend for the PSDs of the product fraction at the

cyclone and the filler concentration. For an increasing filler concentration the particles are

smaller. No such trend is occurring at the other two fractions. The cylinder material shows

larger particles for the highest filler concentration and comparable PSDs for the other two.

The filter fraction is smallest for F = 0.5 and for F = 0.8 not much different from the cyclone

fraction. A reason for a missing trend at the filter fraction is the deviation of the cyclone

performance and cut-off size which also depends on the specific weight of the particles.

In table 11 the granulometric data of the cyclone particle fractions for the different filler

concentrations presented in figure 88 is summarized. The composite specific weight is

calculated and all the other values are based on measurements.

115

table 11: Specific Weight and dynamic viscosity (cf. A.9) of the spray-dried dispersion and granulometric data of

the cyclone particle fraction for different filler concentrations of PMMA-based RA-Fe3O4 composites with

DRA = 0.2, calculated specifi eight , edia parti le size 0, , spe ifi surfa e areas al ulated ith the PSD SPSD (cf. eq. (76)) and measured with BET SBET, SAUTER diameter x3,2 calculated with PSD and structural parameter

SP (cf. eq. (76))

F

in -

dispersion

in g/cm3

dispersion

in mPa·s

composite

in g/cm3

x50,3

in µm

SPSD

in m2/g

x3,2

in µm

SBET

in m2/g

SP

in -

0.3 1.34 2.6 1.51 4.1 (2.3)† 1.41 2.8 (1.9)† 2.11 66.8

0.5 1.35 1.6 1.87 3.0 (1.7)† 1.39 2.3 (1.5)† 2.26 61.5

0.8 1.37 0.5 2.74 1.8 (1.0)† 1.46 1.5 (1.0)† 1.86 78.5

†multiplicity of the particle size at F = 0.8

The SP value results in the lowest structuring for the highest filler concentration of F = 0.8

which can be explained both by a less truncated spherical shape and a denser less structured

shell. However, the differences must be critically observed, since the composites are based

on different precipitation and phase transfer batches with different initial primary particle

concentrations which have not been assessed. This could also be a reason, why the SP value

of the F = 0.3 sample is higher than the one reported for the same composition in table 9.

A more profound comparison must be considered looking at the resulting particle sizes x50,3

and x3,2. One can see that the particles prepared with a filler concentration of 30 % are 2.3

and 1.9 times larger than the for a higher filler concentration of 80 %. The particle sizes of

the compound with 50 % magnetite nanoparticle concentration are 1.7 and 1.5 times larger

than the particles with 80 % magnetite for x50,3 and x3,2, respectively. There are different

causalities for different particle sizes.

If, hypothetically, the droplet size was not affected by the filler concentration and the

particles were solid, then smaller particles are expected for higher filler concentrations and

constant solids concentration csolid in the dispersion, simply due to the higher specific gravity.

Equation (81) shows the relation of droplet diameter xdroplet and composite particle diameter

xcomposite of a solid particle as a function of the components specific gravities , the

compositional parameter F and D and the concentration of solids in the dispersion csolid.

3

1

polymer

lesnanopartic

surfactant

lesnanoparticsolvent

solidlesnanopartic

composite

droplet

1

11

1

DFFDFF

c

x

x

(81)

With DRA = 0.2, csolid = 0.07 and specific gravities of 5.2 g/cm3, 0.9 g/cm3 and 1.2 g/cm3 for

magnetite, surfactant and polymer, respectively, the previous equation results in

116

xdroplet/xcomposite = (2.53, 2.69, 3.05) for F = (0.3, 0.5, 0.8). Assuming equal droplet sizes the

composite diameters of the samples with 30 % and 50 % magnetite would be 1.2 and 1.1

times larger than the particles with F = 0.8, respectively.

The droplet size however is not expected to be the same, since the viscosity as well as the

fluid density are influenced by the polymer content in the dispersion with cpolymer = (64, 40,

4) g/l for F = (0.3, 0.5, 0.8) and csolid = 0.07 resulting in the dispersion specific weight and

dynamic viscosity as presented in table 11. Applying eq. (72) and assuming that only the

liquid viscosity and specific weights are different the droplet sizes of F = 0.3 and F = 0.5

would be 1.11 and 1.08 times larger than for F = 0.8. With respect to eq. (81) this would lead

to particle ratios of 1.34 for F = 0.3 and 1.22 for F = 0.5 to F = 0.8, respectively. These values

are still smaller than the ones reported in table 11, which means that the shell forming

aspect whilst drying of the droplet is influenced by the solids composition. A higher polymer

content could therefore lead to thinner shells and thus more expanded hollow spheres,

which again would have a higher tendency to collapse and lead to the observed truncated

shapes.

6.3 Experimental Results – Injection Molded Composites

One application for the spray-dried microparticle composites discussed above are micro

injection molding processes. For this the fine cohesive powder is agglomerated by a press

and granulation procedure developed as part of a diploma thesis by HORSCHIG, cf. 3.1.5 and

[94]. This previous step is necessary to guarantee sufficient powder flowability for

processing in the injection molding device, where the granules are fed through a hopper, cf.

6.1.2. The final aim is to produce parts with this shaping method. In order to investigate the

ate ial’s e ha i al p ope ties the elt is i je ted i a tool to ge e ate te sile test

specimens. These dog-bone shaped components are used in the present thesis for cross-

section imaging to investigate the distribution of the magnetite nanoparticles within the

polymer matrix. However, the material properties are not discussed as part of this thesis but

to be found e.g. in [22].

The final section of this thesis, dedicated to reveal the effects of mixing a stable nanoparticle

dispersion with a polymer solution in order to prepare nanoparticle-polymer-composites,

shall visualize and quantitatively investigate the impact of the complex particle interactions

in the solvent based dispersion on the final material processing step of injection molding.

Due to the introduced problem of reproducibility of different precipitation batches (cf. 5.3.1)

it proves to be difficult to compare different injection molding batches, which contain these

uncertainties as well. Nevertheless and keeping the reproducibility problem in mind, in the

following four subsections:

117

- composites of PMMA, PC and PVB are compared at two filler concentrations in 6.3.1,

- varying filler concentrations are compared for PMMA based composites in 6.3.2,

- the identity and origin of the agglomerates are briefly discussed in 6.3.3 and finally

- the solution and spray drying method for composite preparation is compared to the

conventional method of melt compounding in 6.3.4.

6.3.1 PMMA vs. PC vs. PVB

In this paragraph injection molded composites of the three polymers PMMA, PC and PVB

with filler concentrations of F = 0.3 and F = 0.5 of ricinoleic acid capped magnetite

nanoparticles with DRA = 0.2 are compared. In paragraph 5.3.1 it is shown that within PMMA

as well as within PC (depletion) flocculation occurred and this leads to the reduction of the

stable primary particle concentration wPrimary with increasing polymer concentration cpolymer.

This destabilizing effect is even stronger for PC. Contrarily, the polymer PVB stabilizes the

dispersion and increases the concentration of stabilized particles. Below, there are three

different sets of cross-sectional images with different magnifications and resolutions. This is

to ensure to include both the agglomerated magnetite nanoparticles as well as the primary

ones for investigation.

The first set of six images (three polymers with two filler concentrations) is presented in

figure 89 and obtained with bright field optical microscopy imaging after image tilt focus

corrections. Magnetite appears in black. The field of view is 76,800 µm2. Bright lines are

attributed to scratches by the polishing method applied, described in [19]. The resolution is

approximately 0.8 µm.

figure 89: Bright field optical microscopy images, (from left to right) PMMA, PC and PVB for (top row) F = 0.3

and (bottom row) F = 0.5, lens magnification: 20x

All images in figure 89 reveal more agglomerates for the filler concentration of 50 % as

compared to the 30 % in the upper row. The background gray scale is getting darker with

higher filler concentration because of more primary particles and smaller agglomerates not

118

resolved by the images. However, the gray scale values should not be compared between

different polymers because of different optical parameters of the pristine polymers. So there

does not have to be a higher content of primary particles for the 50 % PC sample as

compared to PMMA. Only very few agglomerates are visible for PVB. Furthermore the size of

agglomerates is bigger for PMMA and PC as compared to PVB. Continuing discussions are

based on quantitative particle detection and distribution analyses presented in table 12. For

each polymer and filler concentration two images of different spots on the sample are

investigated and processed to obtain information on size and concentration of agglomerates

and calculate the dispersion parameter covVORONOI , cf. 6.1.3.

table 12: Summary of data obtained from the binarized images of figure 89, coefficient of variance of the

VORONOI polygons covVORONOI (cf. 6.1.3), median size of the number weighted distribution of FERET diameters

xFERET 50,0, area fraction of the dete ted agglo erates φAgglomerates (related to the overall volume / area fraction of

the filler nanoparticles in parentheses), agglomerated particle concentration from the primary particle

concentration reported in 5.3.1 with wAgglomerates = 1 - wPrimary and number of agglomerates detected

polymer F

in %

covVORONOI

in -

xFERET 50,0

in µm

φAgglomerates

in %

φAgglomerates

/ φtotal

in %

wAgglomerates

in %

number of

agglomerates

detected

PMMA 30 0.881 1.15 7.1 82 64.8 4,900

PMMA 50 0.880 2.17 12.5 71 32.4 2,837

PC 30* 0.693 1.57 4.1 47 90.9 2,570

PC 50* 0.805 3.30 7.7 44 46.9 992

PVB 30 0.591 0.83 0.3 3 0.8 120

PVB 50 0.595 0.93 0.3 2 3.9 163

* questionable values for there is a high mass loss of larger magnetite agglomerates

The data in table 12 points out, the largest particles at equal filler concentrations are to be

found in the PC samples and the smallest ones in both PVB samples. This approves the

tendencies of colloidal stability, where PC is heavily destabilizing and PVB stabilizes well. The

distribution analysis with VORONOI polygons concludes that the poorest dispersion exists for

both PMMA samples and the best for both PVB samples. This must be due to the mixing of

the agglomerates within the polymer melt in the injection molding device. The fourth

column in table 12 shows the area fraction of particles detected and this is compared to the

area fraction of particles in the composite (relative agglomerate concentration value in the

fifth column). The total agglomerate concentration increases with increasing filler

concentration and is higher for PMMA as compared to PC and PVB. However, the relative

agglomerate concentration decreases with increasing filler concentration. From what is

concluded in 5.3.1 the agglomerate concentration should decrease with decreasing polymer

concentration in the dispersion and thus increasing filler concentration in the prepared

119

composite for equal csolid. So the decrease of relative agglomerate concentration in column

six is conclusive and should follow the data given in the seventh column of table 12. This

phenomenon of increasing agglomerate concentration with increasing filler concentration

and decreasing relative agglomerate concentration is discussed in 6.1.4. Since PVB is

stabilizing, the relative agglomerate concentration is expected to increase with increasing

filler concentration.

One would expect a higher relative agglomerate concentration for the PC samples when

compared to PMMA, for the destabilization with PC is stronger, cf. figure 76. The reasons for

this discrepancy may be:

- a higher number of undetected unresolved agglomerates for PC, which also would

explain the darker grey scale of the background or

- a loss of magnetite during the processing of the dispersion and thus an

overestimation of the actual calculated filler concentration F.

For the PC sample with a set filler concentration of F = 0.5 a magnetite concentration of

wTGA,600 °C = 0.38 is determined for the molded samples and wTGA,600 °C = 0.28 for the cyclone

fraction of one of the two spray-dried samples as part of the injection molded granules. The

same spray-dried sample showed a measured magnetite concentration of 0.69 for the

cylinder fraction of the spray dryer.

Furthermore the relative agglomerate concentration for the PMMA sample is higher than

the determined values from 5.3.1 for both filler concentrations with a high initial primary

particle concentration of about 90 %. There are also several reasons for this happening. On

the one hand it could be that the precipitation batches used for preparing the injection

molded PMMA-based samples led to lower initial primary particle concentrations and thus

higher relative agglomerate concentrations, cf. 6.1.4. Another possible explanation is, that

the large agglomerates are taken as a compact area of magnetite phase, assuming the

porosity of these agglomerates to be zero, which certainly cannot be the case. Furthermore

the optical method is not limited to the uppermost part of the samples and picks up

agglomerate information from below the actual surface. This will increase the determined

filler concentration as well.

Unresolved agglomerates, not detectable by optical microscopy based on relatively high

wavelength electro-magnetic radiation, should become visible by higher resolution BSE-SEM

imaging. Hence, the six samples of figure 89 are reinvestigated at different spots on the

sample. This time the cross-section is generated by cracking the tensile specimen from the

injection molding procedure, no polishing is applied and hence the sample surface is not as

smooth as before and shows cracks.

120

In figure 90 images with a magnification of 5,000x are displayed with a field of view of

400 µm2 and an approximate resolution of 100 nm.

figure 90: Inverted BSE-SEM images (from left to right) PMMA, PC and PVB for (top row) F = 0.3 and (bottom

row) F = 0.5, magnification: 5,000x

With this resolution the individual primary nanoparticles cannot be resolved either so that

all particles detected are taken to be agglomerates. These images allow to visualize smaller

agglomerated structures especially for the PC sample with F = 0.5. However, the individual

primary particles are still not visible, especially for the PVB samples not more individual

particles than before can be seen on the images. Image processing and particle detection for

the 5,000x magnification BSE-SEM samples is applied just like for the optical micrographs

above and the data is presented in table 13.

The median FERET diameters are all smaller than the once presented in table 12. The reason

for this is the higher resolution of SEM compared to optical microscopy. Contrary to the

lower resolution images the number of agglomerates detected is increasing for the higher

filler concentration. This together with decreasing size means, that more small agglomerates

are detected for higher filler concentrations. Again, for the higher filler concentrations the

lower particle sizes are detected for the destabilizing polymer PMMA and PC. All quantified

states of dispersions are poorer than for the optical microscopy investigations with a higher

field of view with cov values above 1. This could mean that the microscopic mixing is not as

thorough as the macroscopic mixing.

121

table 13: Summary of data obtained from the binarized images of figure 90, coefficient of variance of the

VORONOI polygons covVORONOI (cf. 6.1.3), median size of the number weighted distribution of FERET diameters

xFERET 50,0, area fra tio of the dete ted agglo erates φAgglomerates (related to the overall volume / area fraction of

the filler nanoparticles in parentheses), agglomerated particle concentration from the primary particle

concentration reported in 5.3.1 with wAgglomerates = 1 - wPrimary and number of agglomerates detected

polymer F

in %

covVORONOI

in -

xFERET 50,0

in µm

φAgglomerates

in %

φAgglomerates

/ φtotal

in %

wAgglomerates

in %

number of

agglomerates

detected

PMMA 30 2.865 0.28 20.7 238 64.8 689

PMMA 50 1.556 0.14 14.8 84 32.4 2,739

PC 30* 1.159 0.30 24.4 280 90.9 972

PC 50* 0.893 0.24 16.7 94 46.9 1,980

PVB 30 1.012 0.12 2.4 28 0.8 1,339

PVB 50 1.289 0.24 4.4 25 3.9 348

* questionable values for there is a high mass loss of larger magnetite agglomerates

Another distinguishing difference between table 12 and table 13 is the relative agglomerate

concentration in the sixth column. Some of these values are higher in the latter and exceed

100 % which is impossible. This could be due to the following three reasons.

1) The porosity Agglomerates of the agglomerates is not considered and with this taken to

be zero, whereas for closest packing of the nanoparticles within the agglomerate, the

porosity would be 0.26 resulting in a correction factor of 0.74 (1 – ) for the

determined area fractions. This correction factor would be even lower for more

loosely packed agglomerates.

2) The back scatter electron information reaches down below the investigated surface

area several 100 nm in depth, cf. A.2.2. As a consequence of the penetration of the

electron beam in the submicron scale agglomerates not located right at the surface

are detected as well and increase the agglomerate concentration.

3) The agglomerate concentration could be higher because the magnetite precipitation

batches used to produce the spray-dried dispersions led to a lower initial primary

particle concentration, lower than the one present for the samples the data in

column seven of table 13 is extracted from.

Finally, a closer look has to be taken on the bottom row of figure 90 for the composites with

F = 0.5. There is a grayish background for PVB, a homogeneously dotted background for

PMMA and a clustered dotted background for PC. For PVB and PMMA this concludes that

the primary particles must be well dispersed. For the PC sample it seems like there are

fractal loosely agglomerated structures of the nanoparticles. A higher magnification will shed

light on this.

122

In figure 91 the six composite samples with the matrix polymer PMMA, PC and PVB and the

filler concentrations of 30 % and 50 % are presented in inverted BSE-SEM images with a

magnification of 50,000x. The field of view and approximate resolution are 4 µm2 and 10 nm,

respectively. These images are not processed for particle detection, for the resolution

difference caused by the different charging behaviors of the matrix polymers does not allow

for a thorough image processing as achieved before.

figure 91: Inverted BSE-SEM images (from left to right) PMMA, PC and PVB for (top row) F = 0.3 and (bottom

row) F = 0.5, magnification: 50,000x

All six images allow for a good investigation of the state of dispersion of the primary particles

as well as smallest agglomerates. The differences between the three matrix polymers are

outstanding and as expected from the colloidal stability investigations of chapter 5. For both

PVB samples, on the right of figure 91, the very well dispersed primary particle state, as

expected from the investigations of the stability in solvent based dispersion in chapter 5, is

presented. For the PMMA samples on the left of figure 91 it seems like there is a very loose

structure between the primary particles and as if they were not as dispersed as for PVB.

With the higher filler concentration of F = 0.5 the image resolution is poor, which could be

due to charging effects similar to this filler concentration and PC. In the PC sample with 30 %

magnetite only very few primary particles are recognizable and located close to the ten

visible agglomerates. At the higher filler concentration of F = 0.5 the state of dispersion

seems to have improved over the lower magnetite concentration, yet the poor resolution

does not allow for judging whether primary particles or tightly bound agglomerates occur.

123

6.3.2 Increasing Filler Concentration – PMMA composite

In this paragraph PMMA-based composites with RA-Fe3O4 and DRA = 0.2 are compared with

increasing filler concentrations of 30 %, 40 %, 50 % and 60 %. The polished sample cross-

sections are investigated with the lower resolution FEI Phenom BSE-SEM device and

magnifications of 2,000x in figure 92 and 24,000x in figure 93. The fields of view are

14,400 µm2 and 100 µm2 with resolutions of 0.5 µm and about 40 nm, respectively. The

PMMA samples with F = 0.3 and F = 0.5 in this section are not the same like the ones

investigated in paragraph 6.3.2, above.

figure 92: (top row) inverted BSE-SEM images of PMMA-based RA-Fe3O4 composites with DRA = 0.2 and (from

left to right) F = (0.3, 0.4, 0.5, 0.6), (bottom row) binary image of the SEM images above for particle

detection, magnification: 2,000x

The binarized images in the bottom row of figure 92 are evaluated by image processing

methods described in 6.1.3, as it was done before in table 12 and table 13 and the data is

presented in table 14. The state of dispersion is about constant for the samples with 30 % up

to 50 % magnetite with cov values between 0.715 and 0.759. Only the highest filled sample

investigated stands out with a value of 1.371 characterizing a poorer dispersion. The median

FERET diameters are smallest for the lowest filled sample. Supposing the agglomerates are

formed in the dispersion before spray drying this infers that with higher stability for lower

polymer concentrations the agglomerates are bigger. It is more probable that additional

agglomeration occurs for the melt mixing in the injection molding device. This could also

explain the high number of agglomerates larger than about 10 µm in the F = 0.6 sample,

which is larger than most microparticles produced in the spray dryer.

The absolute agglomerate fraction is smallest for 30 % and has a maximum at 40 %. It seems

like the 40 % value is a bit off, which can also be concluded from the relative agglomerate

concentration which should decrease with increasing filler concentration, cf. 6.1.4. All

relative agglomerate concentrations are too high and must be questioned, especially for

values higher than 100 %. There are three reasons for this presented in 6.3.1, above.

124

table 14: Summary of data obtained from the binarized images in figure 92, coefficient of variance of the

VORONOI polygons covVORONOI (cf. 6.1.3), median size of the number weighted distribution of FERET diameters

xFERET 50,0, area fra tio of the dete ted agglo erates φAgglomerates (related to the overall volume / area fraction of

the filler nanoparticles in parentheses), agglomerated particle concentration from the primary particle

concentration reported in 5.3.1 with wAgglomerates = 1 - wPrimary and number of agglomerates detected

F

in %

covVORONOI

in -

xFERET 50,0

in µm

φAgglomerates

in %

φAgglomerates

/ φtotal

in %

wAgglomerates

in %

number of

agglomerates

detected

30 0.741 4.96 11.4 131 64.8 4,932

40 0.715 6.16 23.0 180 51.4 4,539

50 0.759 6.96 19.6 111 32.4 2,653

60 1.371 6.59 20.5 86 21.4 3,473

Assuming only the underestimation of the porosity of the agglomerates was the reason the

values in column five of table 14 do not comply with the relative agglomerate concentrations

presented in column five of table 14. Additionally inferring, the colloidal stability was the

same with equal initial primary particle concentrations. Then the porosity may be calculated

by 1 – φAgglomerates/φtotal/wAgglomerates. This results in porosities of 0.50, 0.71, 0.71 and 0.75 for

the samples in increasing order of filler concentration. Concluding, the packing of the

agglomerates is densest for the lowest filler concentration, which is the least stable system

with the highest polymer concentration in the processed dispersion. However, the

presumptions made are too numerous and the uncertainties too high to overemphasize

these calculations. One problem not mentioned in this regard is that there might still be

smaller agglomerates not resolved by the 2,000x magnification in figure 92. Therefore figure

93 displays the samples with 24,000x magnification but poorer image resolution quality

when compared to the BSE-SEM investigations in 6.3.1, which is due to the different

microscopes used, cf. A.2.2.

figure 93: Inverted BSE-SEM images of PMMA-RA-Fe3O4 composites with DRA = 0.2 and (from left to right)

F = (0.3, 0.4, 0.5, 0.6), magnification: 24,000x

The difference between the images is marginal. However, besides the obvious agglomerates

with sizes around 1 µm, the background where the primary nanoparticles are assumed is

increasing in filler concentration (dark spots) FPrimary from left to right in figure 93 with

125

increasing F and thus expected increasing relative concentration of primary particles

wPrimary = 1 – wAgglomerates = f(1 - φAgglomerates/φtotal).

6.3.3 Identity of Agglomerates

So far it is shown that destabilizing polymers lead to agglomeration in the organic solvent

based dispersion (cf. chapter 5). Furthermore agglomerates are identified in the spray-dried

microparticles (cf. 6.2) and the injection molded composites (cf. 6.3). These three occurring

agglomerates have however not yet been put in context and compared, in order to show the

identity of the agglomerates and whether agglomerates found in the final injection molded

material possibly originate from the destabilized dispersion.

This section is to present the sizes of the agglomerates at different unit operations of the

process for the preparation of PMMA-based composite materials with RA-Fe3O4 and 30 %

magnetite concentration. This investigation is attributed to answer the question whether big

agglomerates found in the molded materials are only because of (depletion) flocculation

before spray drying or mainly due to the melt processing of the composite microparticles. In

figure 94 (a), (b) and (c) there are micrographic visualizations of the agglomerates at the

three process steps: DCM-based PMMA and RA-Fe3O4 mixture, spray-dried microparticle

composite and cross-section of an injection molded part, respectively. Images similar to

figure 94 (b) and (c) are presented in 6.2 and 6.3.2, respectively. For generating the optical

micrograph in figure 94 (a) a stable dispersion of RA-Fe3O4 in DCM carrying only primary

nanoparticles with cnanoparticles = 1.5 g/l is placed under an optical microscope between a

sample carrier and a cover slip. Only a homogenous grey liquid is visual in this case.

Subsequently a mixture with PMMA with cpolymer = 58 g/l is put under the microscope and

within a few minutes larger agglomerates appear, just like it is the case for the investigations

presented in figure 63 of paragraph 5.3.5. By image analysis 314 of these agglomerates are

evaluated and the number weighted sum frequency of the FERET diameters are presented in

the graph in figure 94 with a median size of 1.8 µm.

Based on the findings presented in figure 94, the agglomerates after depletion flocculation

visible under the light optical microscope (Carl Zeiss) in figure 94 (a) inspecting a thin film of

the destabilized dispersion, are of the same order like the agglomerates in the inverted BSE-

SEM of an injection molded sample cross-section in figure 94 (c). The reason that in the

diagram of the size distributions in figure 94, the agglomerates in the molded sample cross-

section seem to be smaller, is that agglomerates with sizes below 1 a ot e esol ed with the light optical microscope.

126

figure 94: Visualization of agglomerates of flocculated magnetite nanoparticles (a) under optical microscope of

a flocculated suspension, (b) as dark spots in composite microparticles and (c) in a cross-section of an

injection molded composite material with inverted BSE-SEM; al images with the same scaling; the

diagram depicts the size distribution of the agglomerates in the suspension compared to the cross-

section showing very similar sizes [27]

The agglomerate sizes in the spray-dried microparticle composites are difficult to assess due

to the morphology influence of the microparticles on the image. However, in the center of

the image in figure 94 (b), one can notice two microparticles with a diameter of about 5 and larger dark spots which are identified as magnetite agglomerates that appear on one

side of the microparticle in a ring shape resulting in Janus type particles. This is often to be

found looking at the spray-dried particles and could be related to the flow dynamics in the

atomized singular droplet pushing the high density agglomerates (specific weight of

magnetite is 5.2 g/cm3 compared to the polymer which is about 1.2 g/cm3) to one side of

the microparticle, cf. 6.2.1.

6.3.4 Solution and Spray Drying Process vs. Melt Mixing

In polymer composites processing, the conventional method of preparation is mixing the dry

disperse pa ti le ediu i a pol e elt usi g a so alled o pou de hi h is a extruder device with counter-rotating twin-screws that cause a high shear in the melt to

sufficiently mix the particles in the melt, cf. 2.2.1. Yet, it is argued that the shear is not strong

enough to disperse nanoparticle agglomerates, which is one reason for the solution and

spray drying method investigated in this thesis.

In this final paragraph a comparison of the dispersion of ricinoleic acid coated magnetite

nanoparticles in PMMA using both methods, is presented. In figure 95, the inverted BSE-SEM

images of sample cross-sections for parts prepared with either method are depicted. The

127

d pa ti les o pou ded ith the t aditio al ethod a e p e ipitated a d i i oleic

grafted magnetite nanoparticles using the phase transfer procedure and spray drying

without polymer addition.

figure 95: Comparison of Inverted BSE-SEM of cross-sections of (left) a melt compounded sample of RA-Fe3O4

compounded with PMMA and (right) one that was spray-dried and injection molded with PMMA and

RA-Fe3O4; for both samples F = 0.3 and DRA = 0.2

The images show very clearly, that both the homogeneity is better for the solution and spray

drying method and the sizes of the agglomerates are much smaller as well. Attention has to

be paid that the scaling on the left image is even bigger than on the right hand side to show

the bigger agglomerates. In order to follow the descriptive procedures of paragraphs 6.3.1

and 6.3.2 the images are processed and evaluated to judge on the state of dispersion and

the agglomerate size. Even though in such a case a visual description is sufficient for the

differences are very strong. However, the data is presented in table 15.

table 15: Summary of data obtained from the binarized images in figure 95, coefficient of variance of the

VORONOI polygons covVORONOI (cf. 6.1.3), median size of the number weighted distribution of FERET diameters

xFERET 50,0, area fra tio of the dete ted agglo erates φAgglomerates (related to the overall volume / area fraction of

the filler nanoparticles), and number of agglomerates detected

method covVORONOI

in -

xFERET 50,0

in µm

φAgglomerates

in %

φAgglomerates /

φtotal

in %

number of

agglomerates

detected

classic

compounding 4.207 1.33 14.4 166 4,735

solution and spray

drying 0.836 0.72 15.6 179 3,810

The state of dispersion for the melt compounded composite is very poor with a high

coefficient of variance of the VORNOI polygon areas of 4.2. The FERET diameter is larger for the

melt compounded system however very low, when comparing with the micrograph in figure

95. This is due to the number weighted distribution parameter x50,0. Since the material

128

composition is the same for both methods it is not surprising that the absolute as well as

relative agglomerate concentrations are similar.

This comparison clearly shows, that the classic melt mixing with high shear forces introduced

by the compounder screw geometries is not capable of dispersing nanoparticle

agglomerates as well as the solution and spray drying method both due to a poor

deagglomeration and furthermore a poorer homogenization. The deagglomeration for

PMMA-based solution and spray drying prepared composites is poor due to the colloidal

destabilization by the non-adsorbing polymer. However, the homogeneity proofs to be

rather well, which is probably due to the small size of the composite microparticles fed to

the extruder.

In conclusion, even if colloidal destabilization causes nanoparticle agglomeration the

solution and spray drying method may be the better choice, when wanting to prepare

nanoparticle-polymer composites which are highly filled and where the nanoparticles and

the polymers are synthesized separately.

129

7 General Conclusions and Outlook

As part of this thesis a process chain has been developed to prepare highly filled composites

of fatty acid stabilized magnetite nanoparticles dispersed in different thermoplastic

polymers. This process chain consists of the following individual process units: nanoparticle

synthesis with a co-precipitation reaction, liquid-liquid phase transfer from water to an

immiscible organic solvent phase with the help of amphiphilic molecules, mixing the

stabilized colloidal dispersion with a polymer solution and spray drying the solution to

withdraw the solvent and prepare microparticle composites. The following individual

process units have been investigated in detail: liquid-liquid phase transfer for the

hydrophobization and stabilization of the nanoparticles, mixing the stabilized organic solvent

based dispersion with a dissolved polymer and spray drying of the solution to prepare

composite microparticles to be used for injection molding applications. The phenomenon of

physical-chemical deagglomeration upon liquid-liquid phase transfer of nanoparticle

agglomerates by chemically adsorbing fatty acid molecules on co-precipitated magnetite is

described with a simple physical model based on the ALEXANDER-DE-GENNES theory and

experiments using five different types of fatty acids: ricinoleic, linoleic, oleic, myristic and

caprylic acid. The primary particle concentration of the transferred particles correlates with

the solubility of adsorbed fatty acid chains in the solvent dichloromethane. Repulsive forces

are introduced as a consequence of the physical model depending on the length of the

adsorbed fatty acids and the distance between adsorbed molecules on the nanoparticle

surface. The numerically determined repulsive forces are in the order of magnitude

compared to the attractive VAN DER WAALS forces. Once the repulsive forces overcome the

attractive VAN DER WAALS interactions deagglomeration occurs. In experiments ricinoleic acid

is found to be most effective in deagglomerating and it is capable of leading to

dichloromethane based dispersions with primary particle concentrations (crystallite size of

about 15 nm) of more than 90 % by weight. The decomposition of chemically adsorbed

ricinoleic acid on magnetite nanoparticles is described and a significant step of magnetite

reduction is found when heating the sample in inert atmosphere between 600 °C and 900 °C.

This is an important finding when discussing thermo gravimetric analyses for investigation of

the stabilized particles and the composites as well. Magnetite reduction is most probably

caused by residual carbonaceous species after detachment and decomposition of physically

adsorbed fatty acid and dehydrogenation of adsorbed fatty acids. Dichloromethane proves

to be a good solvent for the technical polymers poly(methyl methacrylate) PMMA,

poly(bisphenol A carbonate) PC, poly(styrene) PS and poly(vinyl butyral) PVB for preparing

composites based on a solution of the polymers in dichloromethane with fatty acid stabilized

magnetite nanoparticles. PMMA, PC and PS in technical quality cause drastic and rapid

130

flocculation of primary fatty acid stabilized magnetite nanoparticles in the solvents

dichloromethane, ethyl acetate, methyl methacrylate and styrene. The primary particle

concentration decreases exponentially with increasing polymer concentrations. This

tendency is not influenced by the type of fatty acids. It is found that PMMA and PC do not

adsorb on the fatty acid coated nanoparticle surface and the destabilization can therefore

best be described with depletion effects. The governing attractive and repulsive forces in a

dispersion of nanoparticles with non-adsorbing polymers and under the absence of

electrostatic double layer forces are added in a DLVO-like treatment. The analysis of this

superposition of the interactions shows that the polymer concentration and coil size as

compared to the nanoparticles size are the most important values determining stability.

Experiments on colloidal stability are based on gravimetric determination of the primary

particle concentration. This can furthermore be evaluated with light extinction

measurements of diluted samples, for the light extinction follows LAMBERT-BEER’s la fo the investigated particle concentrations. MIE theory calculations demonstrate the strong light

absorption of magnetite nanoparticles explaining the deep black color of stable dispersions.

As part of the stability experiments it is found that the polymer PVB leads to stabilization of

all dichloromethane based fatty acid coated magnetite nanoparticle dispersions

independent from the fatty acid used. PVB is observed to adsorb on the nanoparticle

surface. In addition the hydrodynamic size of the primary particles increases with increasing

PVB concentration and the light extinction versus primary particle concentration is different

from the destabilizing polymers PMMA and PC. The apparent p i a pa ti le concentration increases with increasing PVB concentration. Spray Drying is possible for any

composition of the organic solvent based mixture of fatty acid stabilized nanoparticles and

dissolved polymers. A maximum of 10 % by weight of solids in the solution is possible to

process, for higher viscosities lead to fine web-interconnected composite microparticles.

Due to flocculation the smaller spray dried particles found in the filter of the spray drying

device have a lower content of nanoparticle fillers. The largest particles exhibit a higher

content of nanoparticle fillers. More than 80 % of the microparticles make up the product

and are situated in the coarse outlet of an aero cyclone. Their composition is approximately

equal to the composition of disperse and continuous phase in the solution. Depending on

the primary particle concentration the morphology and homogeneity of the spray dried

microparticles is affected. A structural parameter is introduced to account for the

morphology of the spray dried composite microparticles with higher values for better

dispersed composite microparticles. For characterization of the composites both spray dried

microparticles and cross-sections of injection molded composites, the detection of back

scattered electrons is best to evaluate the distribution of the magnetite filers in the

continuous polymeric matrix. The state of dispersion in the organic solvent based

nanoparticle polymer mixtures correlates to the distribution of the nanoparticle fillers in the

final injection molded composites. PMMA and PC based composite exhibit a high content of

131

agglomerates whereas for PVB the nanoparticulate fillers are well dispersed in correlation to

the colloidal stability in organic solvent suspensions. PC shows a stronger effect in

destabilizing the suspensions and has a worse dispersion in the final composite. Finally it is

proven, that even for flocculated systems, the solution and spray drying method leads to a

better dispersed injection molded composite when compared to a classically melt-

compounded composite of the same composition for ricinoleic acid coated magnetite

nanoparticles in PMMA with a filler concentration of 30 % by weight.

There are many possible continuations of the findings of this thesis. As part of the liquid-

liquid phase transfer various influencing parameters are to be investigated with the aim to

increase the initial primary particle concentration. This is in part investigated most recently

in other PhD thesis at the institute of mechanical process engineering and minerals

processing at the TU Bergakademie Freiberg supported by a DFG priority program SPP 1273

Kolloid e fah e ste h ik . I o e tio to these a ti ities colloidal probe AFM

measurements for the phase transfer of hydrophilic particles under the presence of

adsorbing amphiphilic molecules have to be conducted. Furthermore it is essential to

develop a reproducible synthesis of magnetite nanoparticles. In order to understand and

support the stabilization mechanism of PVB it would be interesting to study the effect of

hydroxyl group content in the PVB structure and its impact on the stabilization of fatty acid

coated magnetite nanoparticle dispersions. Studies on the kinetics of flocculation for

destabilizing polymers will be necessary to understand the destabilization mechanism. The

injection molded composites need to be tested for their mechanical and magnetic properties

especially depending on the filler concentration of highly filled composites. For

characterization of the composites the novel pc-AFM (phase contrast atomic force

microscopy) technique for investigation of the state of dispersion needs to be developed to

guarantee reproducible results. The state of dispersion should be evaluated for different

injection molding parameters. Finally the process chain ought to be tested and further

developed for other nanoparticle and polymer systems.

A1

Appendix

A.1 Materials A3

A.1.1 Materials for Nanoparticle Synthesis A3

A.1.2 Fatty Acids A3

A.1.3 Polymers A4

A.1.4 Organic Solvents A10

A.2 Lab Analytics A13

A.2.1 Particle Size Analysis A13

A.2.2 Microscopy A14

A.2.3 Spectroscopy A17

A.2.4 Thermal Gravimetric Analysis – TGA A17

A.2.5 Powder X-Ray Diffraction – XRD A18

A.2.6 Viscosimetry A18

A.3 Iron-Oxide (Magnetite) Nanoparticle Synthesis A19

A.3.1 XRD Analysis A19

A.3.2 TEM Analysis A20

A.3.3 Particle Size and Charge Analyses A21

A.4 Nanoparticle Phase Transfer Procedure A23

A.5 Investigation of the Colloidal Stability – Primary Particle Concentration A25

A.5.1 Colloid-Polymer-Mixtures A25

A.5.2 Gravimetric Determination of the Primary Particle Concentration A25

A.5.3 UV/VIS Extinction Based Determination of the Primary Particle Concentration A28

A.5.4 Determination of Polymer Adsorption A29

A.6 Mechanical Dispersing Methods A31

A.6.1 Planetary Ball Mill – PM A31

A.6.2 Sonotrode Ultrasound – US A31

A.6.3 Rotor-Stator Mixing A32

A.7 Spray Drying A33

A2

A.8 Optical Properties of Magnetite A35

A.9 Viscosity of Polymer Solutions and Solubility A38

A.9.1 Intrinsic Viscosity A38

A.9.2 Polymer Solubility A39

A.10 Limitations of Dynamic Light Scattering for Dispersions with Primary Particles

and Agglomerates A41

A.11 Calculation of HANSEN Solubility Parameters with the Group Contribution Method

A43

A3

A.1 Materials

A.1.1 Materials for Nanoparticle Synthesis

The following table 16 lists all the chemicals used for synthesis of the magnetite

nanoparticles, which is described in A.3. All material is used as received. The de-ionized

water may not be reproducible in specifications, cf. 5.3.1.

table 16: Chemicals used for magnetite nanoparticle synthesis

substance formula supplier specification

water H2O (lab production from tabbed water with ion-

exchanger Seradest SD 2800)

ammonium

hydroxide NH4OH Sigma Aldrich

ACS reagent

26 %

ferric chloride

hexahydrate FeCl3·6H2O Carl Roth GmbH

Iron(III) chloride hexahydrate

≥ %, p.a., AC“

ferrous

sulphate

heptahydrate

FeSO4·7H2O Carl Roth GmbH Iron(II) sulphate heptahydrate

≥ . %, Ph.Eu ., U“P

hydrochloric

acid HCl Carl Roth GmbH

hydrochloric acid 1 mol/l - 1 N

volumetric standard solution

A.1.2 Fatty Acids

Fatty acids are surfactant molecules used in this thesis for phase transfer and

hydrophobization of magnetite nanoparticles. The following list in table 17 presents the

supply of the materials, which are all used as received.

table 17: List of fatty acids used for the experiments in this thesis

fatty acid abbreviation formula supplier specifications

caprylic acid CA C7H16O2 Sigma Aldrich ≥ %

myristic acid MA C14H28O2 Sigma Aldrich “ig a G ade, ≥ %

oleic acid OA C18H34O2 Alfa Aesar technical, ~90% (GC)

linoleic acid LA C18H32O2 Sigma Aldrich ≥ % GC

ricinoleic acid RA C18H34O3 Sigma Aldrich technical, ~80% (GC)

A4

The chemical structures of the organic acids are displayed in figure 96. It clearly shows the

difference in length depending on the number of C18 atoms. One can furthermore locate

the double bonds for OA and RA (at C9) as well as LA (at C9 and C12). The special feature of

ricinoleic acid is the hydroxyl group at the C12 position. There is another list of chain

properties for the five fatty acids in table 2 of paragraph 4.1.2.

figure 96: Chemical structures of the fatty acids: (a) caprylic acid CA, (b) myristic acid MA, (c) oleic acid OA, (d)

linoleic acid LA, (e) ricinoleic acid RA

All fatty acids used are derived from natural mixtures of different fatty acids. This is

especially notable for RA, which only exhibits a GC determined purity of about 80 % at

technical quality. The other 20 % is made up of other fatty acids such as OA and LA. This has

to be kept in mind when interpreting the experimental findings.

A.1.3 Polymers

The polymers used in this thesis need to be soluble in certain organic solvents, which must

also be suitable carrier liquids for the fatty acid stabilized iron oxide nanoparticles. There is a

choice of four widely applied polymers, namely poly(methyl methacrylate), poly(bisphenol A

carbonate), poly(vinyl butyral) and poly(styrene). All of them exhibit a high transmission for

visible light which in the case of PMMA, PC and PS is even better than for general window

glass material. Especially PMMA and PC are therefore used in the optics industry and

generally as a substitute for glass. PVB has many applications from binders in paints and

ceramics to lamination material for safety glass.

The specific weights are similar with the lightest for PS and the densest for PC and PMMA.

The thermal properties differ very much. Since all polymers listed are amorphous there is

not a distinct melting point but a glass transition θg temperature, which is the lowest for PVB

A5

and the highest for PC. This temperature marks a point where many physical properties such

as strength change, even though it does not stand for a phase transition point. The close

order changes from a glassy to a rubbery state.

table 18: General physical properties of the polymers used in this thesis[154, 247]

polymer acro-

nym

in g/cm3

θg

in °C

nD

in -

light

trans-

mittance

tensile

strength

in MPa

izod

impact

strength

in kJ/m2

poly(methyl

methacrylate) PMMA . … . 106 1.49 92 % … …

poly(bisphenol

A carbonate) PC 1.20 150 1.59 85 % 65.5 850

poly(vinyl

butyral) PVB 1.10 … 1.49 - … 58.7

poly(styrene) PS . … . 100 1.58 90 % … 19.7

In figure 97 (left) the thermal decomposition graphs are displayed showing another thermal

characterization to differentiate between the polymers. The most thermal stable material is

PC. This polymer also shows an incomplete decomposition, which is e.g. described in [200].

Optical properties are displayed in column five and six in table 18 with the refractive index

nD (at 20 °C) and the visible light transmittance (through a 3 mm sample for 600 nm),

respectively. The differences are interesting when optical applications demanding special

properties are of interest. Even more important certainly are the mechanical properties, for

most of the material is chosen because of these values. Both the maximum tensile strength

from tensile tests as well as the izod impact strength are reported in table 18. Again,

especially PC stands out with very high impact strength. This high value means, the material

can absorb a lot of energy contrary to the brittle PS. The high value for PVB justifies the

application in safety glass. Since PC might be more expensive and there are health concerns

about not reacted harmful bisphenol A, PMMA often is an alternative. However, the

strength of PC is usually a justification of using this type of polymer and one reason why it is

tested as a matrix polymer for highly filled composites in this thesis.

Certainly also interesting for certain applications is the wettability with water represented by

the contact angle presented in figure 97 on the right. PS is the most hydrophobic material

with an angle around 90° and PMMA similarly hydrophilic like PVB. Those values have been

experimentally determined on thin layers deposited on glass slides with the sessile drop

method.

A6

figure 97: (left) Thermal decomposition in N2 atmosphere with a heating rate of 20 K/min (from left to right, red

to blue) PMMA, PVB, PS, PC; (right) contact angle with water on a thin film using sessile drop analysis

Connected to the wettability are the HSP values, cf. 5.1.2, presented in table 19.

table 19: HSP values of the polymers used in this thesis

polymer d

in MPa1/2

p

in MPa1/2

h

in MPa1/2 source

PMMA 18.1 10.5 5.1 [248]

PC 18.1 5.9 6.9 [248]

PVB 18.6 4.4 13.0 [154]

PS 18.5 4.5 2.9 [248]

The high hydrogen bonding HSP for PVB is most probably due to the hydroxyl groups in its

structure, as presented below when presenting individual chain properties and supplier

information for all four polymers.

Poly(methyl methacrylate) – PMMA

trade name: Diakon CLG 902

supplier: Lucite

The PMMA used in this thesis is a granular batch for injection molding applications. The basic

structure of the repeating unit of PMMA is presented in figure 98. It shows a vinylic polymer

with a functional methyl opposed by a methacrylate group. The polymer is synthesized by

radical polymerization of the monomer methyl methacrylate.

figure 98: Chemical structure of the repeating unit of poly(methyl methacrylate)

A7

The following table 20 summarizes the chain properties of the PMMA sample. The left part

of the table lists the weight and number averaged molar weights Mw and Mn, the repeating

units’ molar weight M0, the number of repeating units N and the polydispersity index PI. On

the right hand side there are: the length of a repeating unit l0, the segment (or KUHN) length

ls, the segment to repeating unit ratio ls/l0, the number of segments Ns and the radius of

g atio i the θ-state RG.

table 20: Polymer chain properties of the PMMA batch Diakon CLG 902

physical

quantity

unit value source physical

quantity

unit value source

Mw g/mol 88,500 [60] l0 nm 0.25 [118]

Mn g/mol 41,305 [60] ls nm 2.175 [118]

M0 g/mol 100.117 - ls/l0 - 8.7 [118]

N g/mol 413 Mn/M0 Ns - 47 N·l0/ls

PI - 2.14 Mw/Mn RG nm 6.09 eq. (38)

Using eq. (40) the swollen radius of gyration for FLORY interaction parameters of χ = 0.10 and

χ = 0.31 (cf. A.9.2) can be calculated resulting in RG χ = 0.10) = 10.1 nm and

RG χ = 0.31) = 8.9 nm.

Poly(bisphenol A carbonate) – PC

trade name: Makrolon 2407

supplier: Bayer Material Science

The poly(bisphenol A carbonate) batch used in this thesis is a granular substance which is

used for injection molding applications as well as for sheet plastics. In figure 99 the structure

of the repeating unit is presented. Compared to the simple structure of PMMA the repeating

unit of PC is not vinylic and more complex. This is due to the fact that this polymer is

synthesized by the polycondensation reaction of bisphenol A and phosgene (COCl2) resulting

in ester bonds. Hence, PC is a typical representative of synthetic polyesters.

figure 99: Chemical structure of the repeating unit of poly(bisphenol A carbonate) [247]

According to table 20 for PMMA the most important chain properties of the PC batch used

are presented in table 21. The physical quantities are the same like in table 20.

A8

table 21: Polymer chain properties of the PC batch Makrolon 2407

physical

quantity

unit value source physical

quantity

unit value source

Mw g/mol 25,900 [249] l0 nm 1.065 [250]

Mn g/mol 10,100 [249] ls nm 3.000 [251]

M0 g/mol 254.285 - ls/l0 - 2.8 -

N g/mol 40 Mn/M0 Ns - 14 N·l0/ls

PI - 2.56 Mw/Mn RG nm 4.58 eq. (38)

Using eq. (40) the swollen radius of gyration for a FLORY interaction parameter of χ = 0.00 (cf.

A.9.2) can be calculated resulting in RG χ = 0.00) = 7.9 nm.

Poly(vinyl butyral) – PVB

trade name: Mowital B 30 T

supplier: Kuraray Europe

The polymer PVB is applied in this thesis, since it is used as a matrix polymer for composite

synthesis of magnetic beads for bioseparation purposes [92]. It is not a typical polymer for

injection molding of parts like PMMA and PC but can certainly be used therefore. Its rather

complex structure is made up of three different repeating units as presented in figure 100.

Yet it is not considered a co- or terpolymer.

figure 100: Chemical structure of the three repeating units of poly(vinyl butyral) with n butyral, m alcohol and p

acetate subunits [247]

The reason for the structure of commercial PVB is its synthesis pathway. Poly(vinyl acetate)

is the starting material which is synthesized to poly(vinyl alcohol) by hydrolysis [247]. Due to

statistical reasons there are few acetate groups as well as several alcohol subunits left in the

PVB structure. This composition is important for the use of PVB. The polymer is only soluble

in organic solvents and not in water like PVA and PVAc. However, the hydroxyl functional

groups are responsible for desired adhesive interactions necessary for the superior binder

properties and for glass lamination. These functionalities are also most probably the reason

for stabilization of fatty acid coated magnetite nanoparticles, cf. paragraph 5.3.3. Unlike the

two previously described polymers the polymer chain is not characterized by repeating unit

A9

and segment lengths, for which no information is found. Instead in table 22 besides the

chain lengths there is the approximate composition with weight fractions wPVB, wPVA and

wPVAc of butyral, alcohol and acetate subunits.

table 22: Polymer chain properties of the PVB batch used, including the composition of the three functional

units displayed in figure 100

physical

quantity

unit value physical

quantity

unit value

Mw g/mol 32,0007 wPVB - . … . 6

Mn g/mol 30,700 wPVA - . … . 6

M0 g/mol 61.43* wPVAc - 0. … . 6

N g/mol ≈ 5007

PI - 1.04

* average value for a hypothetic averaged repeating unit calculated with the average composition by weight of 59.5 % vinyl butyral, 25.5 % vinyl alcohol and 2.5 % vinyl acetate

Due to the lack of information of the segment lengths it is not possible to calculate

theoretical radii of gyration.

Poly(styrene) – PS

trade name: unspecified

supplier: unspecified

The fourth polymer used only for the preliminary experimental investigations in paragraph

5.3.6 is poly(styrene). Similar to PMMA it is a vinylic polymer typically derived from radical

polymerization of the monomer styrene. Its structure is displayed in figure 101. The special

functionality in the structure is the phenolic side group. It causes the small hydrogen

bonding HSP in table 19 as well as the high contact angle with water in figure 97.

figure 101: Chemical structure of the repeating unit of poly(styrene)

ATR-FTIR and TGA/FTIR investigations show that the material is PS with no additives found.

6 http://www.kuraray-am.com/pvoh-pvb/downloads/Mowital_Technical_Data_Sheet.pdf, march 29th 2012 7 http://www.kuraray-am.com/pvoh-pvb/downloads/Mowital_brochure.pdf, march 29th 2012

A10

A.1.4 Organic Solvents

Organic solvents are liquid carbon containing molecular substances which can dissolve gases,

other liquids or solids without a chemical reaction involved [140]. The organic solvents in this

thesis shall both act as good carrier liquids for fatty acid grafted magnetite nanoparticles as

well as be able to dissolve the polymers introduced above. In the following table 23 the four

solvents used in this thesis are presented.

table 23: List of solvents used in the experiments of this thesis

solvent acronym formula supplier specifications

dichloro-

methane DCM CH2Cl2 Carl Roth GmbH

ROTIPURAN® ≥ 99.5 %,

p.a., ACS, ISO

ethyl

acetate EA C4H8O2 Carl Roth GmbH

ROTIPURAN® ≥ 99.5 %,

p.a., ACS, ISO

methyl

methacrylate MMA C5H8O2 Carl Roth GmbH ≥ 99 %, extra pure

styrene ST C8H8 VWR International Merck KGaA, synthesis

quality

A summary of the general physical properties, specific weight and dynamic viscosity ,

refractive index nD, the HSP and the molar volume v, which is M/ , are to be found in table

24.

table 24: General physical properties of the solvents

solvent (20°C)

[252]

/g/cm3

(20°C)[252]

/mPa·s

nD

/-

d[140]

/MPa1/2

p[140]

/MPa1/2

h[140]

/MPa1/2

v

/cm3/mol

dichloro-

methane 1.331 0.436 1.42 18.2 6.3 7.8 63.86

ethyl

acetate 0.900 0.450 1.37 15.8 5.3 7.2 98.56

methyl

methacrylate 0.940 0.600 1.41 15.8 6.5 5.4 106.51

styrene 0.904 0.754 1.55 17.8 1.0 3.1 114.45

Dichloromethane – DCM

The chemical structure of the simple molecule dichloromethane is presented in figure 102. It

is the ostl used sol e t i this thesis’ e pe i e ts.

A11

figure 102: Chemical structure of dichloromethane

DCM is a polar aprotic solvent which is immiscible with water having a water solubility of

13 g/l at 20 °C (2 g/l at 30 °C). It is a very commonly used solvent e.g. as paint stripper.

Furthermore it can be found in spray painting operations, for decreasing of automotives and

leather products as well as in household products [140]. Since it may be carcinogenic its use

is limited and strongly regulated by EU REACH8. This is also one reason why as part of the

present research it is tested to substitute DCM with the following solvents.

Ethyl Acetate – EA

The chemical structure of the organic solvent ethyl acetate is depicted in figure 103. In this

thesis it is applied in paragraph 5.3.6.

figure 103: Chemical structure of ethyl acetate

Just like DCM it is a polar aprotic solvent. This ester is very common in use, e.g. in glues and

strippers but has the disadvantage over DCM that it can result in explosible atmospheres. Its

solubility in water is higher than for DCM with 83 g/l at 20 °C.

Methyl Methacrylate – MMA

The chemical structure of the solvent and vinylic monomer methyl methacrylate is shown in

figure 104. It is applied in the investigations of paragraph 5.3.6.

figure 104: Chemical structure of methyl methacrylate

8 Official Journal of the European Union L 86/7, Commission Regulation (EU) No 276/2010

A12

The structure is similar to ethyl acetate and it is aprotic. It is the base material for the

polymer PMMA. The reason for application in this thesis is to test its ability to transfer and

stabilize magnetite nanoparticles with the surfactant ricinoleic acid as well as check the

colloidal stability of the particles with dissolved PMMA. The solubility in water is 15 g/l at

20 °C similar to DCM.

Styrene – ST

The chemical structure of the solvent and vinylic monomer styrene is shown in figure 105. It

is applied in the investigations of paragraph 5.3.6 together with MMA.

figure 105: Chemical structure of styrene

The structure shows that styrene is the only non-polar solvent of the present thesis. Styrene

is the base material for the polymer PS and a number of common co-polymers such as ABS.

The reason for application in this thesis is to test its ability to transfer and stabilize magnetite

nanoparticles with the surfactant ricinoleic acid as well as check the colloidal stability of the

particles with dissolved PS. In water it is only soluble with 0.3 g/l at 20 °C.

A13

A.2 Lab Analytics

A.2.1 Particle Size Analysis

For details on the methods of particle size analysis, the following literature can be adviced

[198, 199, 253, 254].

Laser Diffraction

device: HELOS/KR with QUIXEL

company: Sympatec

Laser diffraction is a static particle sizing method of a bulk of particles which relies on the

evaluation of FRAUNHOFER diffraction. The measurement range is 0.1 µm to 8750 µm,

therefore it is applied for the spray dried particles in the present thesis. It is important to

disperse the particles, which is achieved by mixing of the powder in low viscosity silicon oil

M3 as the carrier liquid. Then ultrasonic treatment is applied in steps of 1 min until the PSD

is constant. The silicon oil guarantees good wettability without swelling of the composite

microparticles.

Analytical Centrifugation with Photo extinction

device: SA-CP 3

company: Shimadzu

Analytical centrifugation is a cumulative size method for fine particles in suspension. The

particle concentration in a certain plane is monitored with light extinction. The

measurement range is depending on the differences in specific weight and the liquid

viscosity and reaches down to 20 nm. The speed of rotation is accelerating with 120 min-2 up

to 5000 min-1 which corresponds to 2,040·g with respect to the radius of measurement

[255]. Dilution of the samples occurs until a certain extinction of the mixed suspension has a

distinct value (between 70 % and 120 % of an arbitrary unit). In the present work the

method is preferred for agglomerated nanoparticles with a broad PSD. Due to the

disregarded extinction intensity over size for particles smaller the illumination wavelength

(MIE scattering) the PSD is presented as intensity weighted.

Dynamic Light Scattering

device: Zetasizer ZS Nano

company: Malvern Instruments

The method of dynamic light scattering – DLS (synonyms: photon cross correlation – PCS,

quasi elastic lights scattering – QELS) is based on the free diffusion of particles due to

A14

BROWNian motion of the carrier liquid molecules [112]. Autocorrelation of time dependent

intensity fluctuations of scattered laser comprises the diffusion coefficient D. The well

established STOKES-EINSTEIN relation brings together this diffusion coefficient D and the

hydrodynamic particle size x. By definition of the ISO13321 the so called cumulant method

extracts a single mean size (z-average or xDLS) and a polydispersity index PDI as a result of

DLS. Since for a particle sizing method one is expecting to get a PSD there are mathematical

tools, such as the so called NNLS method (non-negative least square) to derive intensity

weighted PSD from the autocorrelation function. One questionable feature is the

assumption of GAUSSian profiles of the size fractions and the limitation to a certain limited

number of size fractions often creating size ranges with no particle content, which may not

meet reality. Therefore and by experience, the results of a PSD from a DLS method must be

carefully interpreted, cf. A.10. This is even more problematic when wanting to extract

volume or number weighted distributions since good knowledge of the complex optical

parameters must be given, especially when there are particles with sizes roughly as big as

the wavelength of the laser at about 600 nm (MIE scattering). Certainly when measuring

monomodal distributions of particles smaller 60 nm (RAYLEIGH scatterers) DLS is a perfectly

suitable method. Care must be taken for preventing dust or bigger particles in the sample.

Therefore the diluting solvent is typically filtered with a 200 nm PTFE membrane filter. The

Quartz glass cuvettes are also rinsed several times prior to measurements.

Typical measurement settings are:

- wavelength 632.8 nm

- 173° backscattering

- 15 times 30 second measurement duration and averaging

- three measurement repetitions per sample

Furthermore the -potential or more precisely the electrophoretic mobility UE is determined

with this device using special electrophoretic cells and determining the velocity not by DLS

but by a laser Doppler method.

A.2.2 Microscopy

Optical Microscopy

device: Axiolab.A1

company: Carl Zeiss

The optical microscope works with three contrast modes: bright field, dark field and

differential interference contrast. Four objective lenses are installed: 5x, 10x, 20x and 50x. A

A15

digital camera AxioCam with five megapixels is attached. The images are digitally processed

using the firmware AxioVision.

Back Scattering Scanning Electron Microscopy – BSE-SEM

devices: Phenom Nanolab 600

company: FEI FEI

In electron microscopy there are several detection modes all based on different emissions

from the sample. Most widely used in imaging is detecting the so called secondary electrons

(SE-SEM) emitted from the surface with only few nm depth [256]. Their detection leads to

the common plastic images with good depth of field, ideal for imaging morphologies or e.g.

particles. Another, for composite analysis more interesting, imaging mode is based on the

electrons backscattered by the atoms they hit (BSE-SEM). Their detection leads to phase

sensitive images since heavier atoms are more efficient in back scattering electrons so they

appear lighter in the images. However, depending on the energy of the primary beam

electrons and the atomic number the depth information of backscattered electrons is in the

order of several 100 nm [257, 258]. Higher beam energies and lighter atoms cause a deeper

electron penetration.

The difference between SE- and BSE-SEM is visualized in figure 106 for the same sample area

of a fractured cross-section of a PMMA based RA-Fe3O4 composite with 30 % by weight

magnetite.

figure 106: Comparison of a micrograph of a fractured composite surface of RA-Fe3O4 in PMMA with F = 0.3 and

DRA = 0.2 (left) with detection of the secondary electrons and (right) when detecting the back

scattered electrons with much better phase contrast (the lighter sections are due to strong back

scattering of the heavy iron atoms in the magnetite nanoparticles)

The left image (secondary electron detection) only shows the morphology of the fractured

surface. The bright spots in the right image (back scattered electrons) show the location of

magnetite phases because the iron atoms are the heaviest atoms (Z = 26) in the composite

where only hydrogen, carbon and oxygen is found with atom numbers of 1, 6 and 8

respectively. Even sputter coating with an ultrathin layer of AgPd for suppressing sample

charging preserves the phase contrast.

A16

The two applied microscopes have the following specifications:

Phenom FEI: constant 5kV acceleration voltage, only back scattering information

available, low resolution and maximum 24,000x magnification, desktop

device

Nanolab 600 FEI: variable acceleration voltage, secondary as well as back scattering

detection, high resolution even at low kV, up to 50,000x magnifications

in BSE-mode, high-end device

The reason working with these two instruments is the limited availability of the superior

Nanolab 600.

Transmission Electron Microscopy – TEM

device: CM 30

company: Philips

The TEM images are acquired with a CM30 from Philips with an acceleration voltage of

300kV. The samples are prepared by embedding in an epoxy resin and sectioning in an

ultramicrotome with a layer thickness of approximately 80 nm. The sample slices are put on

a carbon grid.

phase contrast Atomic Force Microscopy – pc-AFM

device: XE 100, true non contact mode

company: Park Systems

As part of the efforts to characterize nanoparticle-polymer composites a method based on

non-contact atomic force microscopy (AFM) has been developed, namely phase contrast

atomic force microscopy (pc-AFM) [19]. However, the results of this method are not yet

reproducible and too many influencing parameters are disturbing the evaluation. Hence, it is

not used for the investigation of sample cross-sections in this thesis. Nevertheless it has

been giving the first hint of agglomerates in cross-sections because it enabled the resolution

of individual nanoparticles within the large agglomerate phases, presented in Figure 107.

A17

Figure 107: pc-AFM image of a PMMA-RA-Fe3O4 composite with F = 0.3

A.2.3 Spectroscopy

Photo spectrometry – UV/VIS

devices: Lambda 3B, dual beam Cary 60, single beam

company: Perkin Elmer Agilent

UV/VIS measurements are carried out in a Lambda 3B device from Perkin Elmer with dual

beam set-up using Quartz cuvettes and the carrier liquid in reference position. Only for the

investigations in 5.3.6 the new device Cary 60 from Agilent is used. Both devices work with

CZERNY-TURNER spectrometers. There are two light sources for the Lambda 3B (mercury for

UV and Halogen for VIS) and a single Xenon pulse lamp source for the Cary 60.

Fourier Transform Infrared Spectroscopy – FTIR

device: Tensor 27, equipped with ATR

company: Bruker Optics

For vibrational spectroscopic purposes FOURIER transform infrared spectroscopy (FTIR) is

applied. The IR source is in the MIR range. Solid and liquid samples are investigated using the

attenuated total reflection (ATR) sitting on a ZnSe crystal. The detector for these samples is

PELTIER cooled. An additional gas analysis chamber with nitrogen cooled detector and high

sensitivity is attached to the Tensor 27 for analyzing evolving gas from the TGA, cf. A.2.4.

A.2.4 Thermal Gravimetric Analysis – TGA

device: STA 449 F3 Jupiter

company: Netzsch

Thermal gravimetric analysis is realized with a simultaneous thermal analyzer STA 449 F3

Jupiter from Netzsch which is capable of simultaneously conducting differential scanning

calorimetry (DSC) as well as thermal gravimetric analysis (TGA). In this thesis only the TGA

A18

signals are used. The oven is made of SiC and has a temperature range from room

temperature to 1550 °C. Al2O3 crucibles are used. The purge and protection gas of the scale

is nitrogen N2 with 99.999 % purity. If not mentioned otherwise the samples are heated from

room temperature to 900 °C at a heating rate of 20 K/min with a purge gas flow of

20 ml/min. This leads to sufficiently dense evolving gases for a thorough evaluation with

coupled FTIR.

A.2.5 Powder X-Ray Diffraction – XRD

device: X’Pe t PRO MPD

company: PANalytical

Powder X-Ray diffraction is conducted with a X’Pe t PRO MPD f o PANal ti al usi g CuKα

radiation at a wavelength of 1.5406 Å. This method is applied for phase and crystallite size

analysis [259].

A.2.6 Viscosimetry

The kinetic viscosities of solvents and polymer solutions are determined with UBBELOHDE

viscosimetry [49]. The principle of this method is measuring the time a tested fluid needs to

free fall flow through a defined capillary. With respect to HAGEN-POISEUILLE’s la the ti e it takes to flow through this capillary is proportional to the kinetic viscosity. The capillaries

used are geometry Oc (0.5...3 mm2/s) a d geo et I . … mm2/s). The time is measured

automatically while the capillary filled with the tested liquid rests in a thermostatic bath.

A19

A.3 Iron-Oxide (Magnetite) Nanoparticle Synthesis

The standard procedure for co-precipitation of magnetite nanoparticles is presented here.

The following recipe is for the synthesis of 1 l suspension with 20 g Fe3O4.

- 47.03 g FeCl3∙6H2O (0.174 mol) and 24.19 g FeSO4∙7H2O (0.087 mol) are dissolved in

1 l deionized water using a 1 l round bottom flask

- The solution is stirred at 4000 min-1 for 10 min

- While stirring the solution is heated to 70 °C

- The stirring speed in raised to 9000 min-1 and quickly 60 ml of a 26 % NH4OH solution

are added at the tip of the rotor using a syringe

- After 2 min the stirrer is reduced to 4000 min-1 again and at 70 °C the dispersion is

stirred for another 20 min

- At the end of the process step the dispersion is spilled in a storage container and

cooled down to room temperature

The reaction steps are as follows:

OHOFeOHFeOHFeOOH

ClNHOHFeOOHOHNHFeCl

SONHOHFeOHNHFeSO

24322

4243

424244

42

6262

2

(82)

A.3.1 XRD Analysis

Using powder x-ray diffraction one can determine the phase and size of the crystallites by

investigating the peak positions and line broadening of the angle dependent x-ray

interferences, which can be defined as the primary particle size [260]. Using a GAUSSian

profile for the diffraction peaks one can find the peak location Θ and the width which is the

full width at half of the maximum FWHM. By plotting FWHM∙cos(Θ o e ∙si Θ) the

crystallite size is found in the y-intercept.

interceptecrystallit

y

Kx CuK

(83)

The SCHERRER correction K is often defined as unity, the wavelength of a CuKα radiation

source is CuKα = 1.5406 Å.

A20

figure 108: (left) powder diffractogram (XRD) of the washed and dried co-precipitated magnetite, all peaks

correspond to the magnetite crystal system, the four major peaks are used for calculation of the

crystallite size (right) using the Williamson-hall plot [260]

Using eq. (92) and the WILLIAMSON-HALL plot in figure 108 one obtains a crystallite size of the

magnetite crystals of xcrystallite = 14.8 nm. The peak positions correspond to either magnetite

or maghemite. Another study based on MÖSSBAUER spectroscopy identifies the magnetite

phase of freshly synthesized nanocrystals using the same co-precipitation procedure [153].

A.3.2 TEM Analysis

The TEM in figure 109 shows embedded magnetite nanoparticles. All individual particles are

evaluated by image processing with the FERET diameter (maximum FERET diameter at various

angles).

figure 109: (left) transmission electron micrograph of the precipitated magnetite nanoparticles in a PMMA

matrix with F = 0.3, (right) particle size distribution (number frequency) as obtained from image

analysis of the TEM image

In figure 109 on the right the graph shows the size distribution of the primary particles with

a maximum mode between 12.5 nm and 15 nm which corresponds to the crystallite size

determined with XRD in A.3.1. The largest detected particles are 35 nm and the smallest

roughly 5 nm.

A21

A.3.3 Particle Size and Charge Analyses

The phase transferred functionalized magnetite nanoparticles in DCM present a colloidal

system of a large fraction of stable primary particles of a mean size of about 20 nm and a

smaller fraction of sedimented agglomerates in the size range of about 100 nm up to 4 µm.

This pristine dispersion is characterized with analytical centrifugation and the intensity

weighted cumulative size distribution is depicted in figure 110 together with the volume

weighted frequency distribution of the primary particle fraction measured with DLS. The size

of the primary particles coincides well with the TEM and XRD results presented above. A

small deviation to larger sizes for the DLS measurement can be attributed to the higher

hydrodynamic diameter caused by the adsorbed layer of ricinoleic acid forming a brush-like

structure. Because of the lack of light scattering correction for the sedimentation analysis (cf.

figure 110 top) the broad distribution, spanning from about 20 nm to 4 µm, cannot be

evaluated with a mass weighted particle size distribution.

figure 110: (top) Intensity weighted cumulative particle size distribution of phase transferred magnetite

particles as determined with the cuvette centrifuge and (bottom) volume weighted particle size

distribution of the supernatant after centrifugation without polymer as measured with DLS compared

to the TEM investigation of encapsulated magnetite nanoparticles (inset) [20]

Not functionalized and phase transferred particles co-precipitated are agglomerated due to

the low absolute -potential of -5.0 mV. Their PSD is presented in figure 111. The median size

is approximately 1 µm. When washing the suspension after the precipitation reaction with

de-ionized water keeping the pH at 9.0, the absolute -potential rises to -29.7 mV and the

particles deagglomerate with median sizes of approximately 50 nm.

A22

figure 111: Particle size distributions of precipitated magnetite in water with full ion strength after the reaction

with given zeta-potential (blue line measured with analytic centrifugation) and after washing

reducing the ion concentration increasing the absolute zeta-potential (green line measured with DLS)

The net charge of the magnetite nanoparticles in water at pH of 9.0 is negative.

A23

A.4 Nanoparticle Phase Transfer Procedure

The experimental procedure to transfer aqueous magnetite nanoparticles to the heavier

fatty acid carrying solvent phase of dichloromethane is schematically visualized in four steps

in figure 112.

figure 112: Steps of a gravity driven phase transfer using a organic solvent which is heavier than water, e.g.

DCM

The steps for the gravity driven phase transfer in figure 112 are as follows:

(1) Mixing fatty acid with dichloromethane (under stirring for 5 minutes) and

placing the mixture in a sufficiently large beaker.

The composition of the mixture is defined by the surfactant ratio on magnetite nanoparticles

D and surfactant concentration in the solvent as shown in eq. (84). Furthermore the

nanoparticle concentration cnanoparticles in the solvent after completed phase transfer is given.

DDV

mc

mm

m

m

mD

solvent

solvent

lesnanoparticlesnanopartic

solventsurfactant

surfactant

lesnanopartic

surfactant

(84)

Typical values for D and are 0.2 and 0.02, respectively, cf. 3.2 and [88, 99]. These will

guarantee the complete transfer of the particles and a minimum amount of fatty acid

surfactants.

(2) The synthesized magnetite nanoparticle suspension is carefully added with

the properties: pH 9.0, 20 g/l Fe3O4 concentration and = -5 mV.

(3) The aqueous agglomerated nanoparticles quickly settle to the liquid-liquid

interface and by adsorption of the fatty acids enter the organic solvent phase.

Depending on the fatty acid and the quality of the precipitation batch after

less than 6 hours the entire particle mass is phase transferred.

A24

(4) A clear reflecting interface with absence of black particles in the upper water

phase marks the completed phase transfer. In order to reduce the amount of

ammonia salts at the interface and reduce the pH the magnetite free water is

replaced three times with an equal amount of deionized water. After the last

washing step the entire water is carefully removed.

In paragraph 5.3.6 the phase transfer of magnetite to a lighter organic solvent phase (methyl

methacrylate or styrene) is realized in cylindrical separation funnels. There the steps

(highlighted with *) are as follows.

(1)* Placing the precipitated nanoparticle suspension in the funnel.

(2)* Mixing the fatty acids with the solvent, with respect to eq. (84).

(3)* Adding the fatty acid solvent mixture to the separation funnel and emulsifying

by shaking. Next a repeated procedure of shaking for about 10 seconds and

letting the emulsion separate for 10 minutes six times. With D = 0.2 and

= 0.02 this guarantees a completed phase transfer for ricinoleic acid.

(4)* Addition of 1N HCl to reach a pH of 6.0 in the water phase and thus breaking

the emulsion leading to the clear and reflective interface.

(5)* Washing step of the particle free water phase three times with an equal

amount of deionized water and removal of the lighter water phase at the

bottom of the separation funnel.

The use of separation funnels for the gravity driven phase transfer to a heavier solvent phase

fails due to the choking of the lower outlet by not deagglomerated particle sediments, cf.

chapter 4.

A25

A.5 Investigation of the Colloidal Stability – Primary

Particle Concentration

A.5.1 Colloid-Polymer-Mixtures

Organic solvent based nanoparticle dispersions and polymer solutions are thoroughly mixed

using a vortexer (VORTEX 3 from IKA) for 15 min. The compositions and concentrations of

the experimentally investigated dispersions are presented in table 25.

table 25: Concentrations cpolymer and cnanoparticles of the investigated dispersions of polymers, nanoparticles in

DCM as an organic solvent with a surfactant (fatty acid) to nanoparticle mass ratio D of 0.2, as well as the

resulting filler concentration F of the particles in a composite synthesized with the dispersion withdrawing the

solvent assuming specific weights of 5.2 g/cm3 and 1.2 g/cm

3 for the magnetite nanoparticles and the polymer

as well as the surfactant layer, respectively.

# cpolymer

in g/l

cnanoparticles

in g/l

F

in % by weight

F

in % by volume

1 52.0 24.4 30.0 8.7

2 31.7 24.4 40.0 12.8

3 19.5 24.4 50.0 17.7

4 11.4 24.4 60.0 23.8

5 5.6 24.4 70.0 31.7

6 1.2 24.4 80.0 42.1

7 0.0 24.4 83.3 46.3

The values are chosen so that they meet economic and ecologic specifications for a relevant

technical process of material preparation, cf. paragraph 3.2. The solids mass concentration

csolid in the solvent is set constant at 0.07 which results in a well processable feed for the

spray dryer. The filling degrees F are chosen to meet the specifications for a highly filled

composite material exceeding volume filler concentrations of 10 %.

A.5.2 Gravimetric Determination of the Primary Particle

Concentration

In order to evaluate coagulation and colloidal stability, the weight fraction of primary

particles is quantified as this is a suitable criterion for the desired composite material. For a

broad distributed system with particles and agglomerates spanning a size range of about

10 nm to 10 µm classic particle sizing methods do not give reliable information on volume

weighted distributions, cf. A.2.1 and A.10. The new approach developed here is to

gravimetrically compare the samples of both a mixed system with all particle sizes and the

A26

supernatant of a centrifuged system with only primary particles. For this the nanoparticle

dispersion is mixed with the dissolved polymer at the desired compositions in a 10 ml

centrifugal cuvette. After intensive mixing using a vortexer (VORTEX 3 from IKA) for 15 min, a

sample of 2 ml is taken. Subsequently, the cuvettes are introduced in a lab centrifuge

(Hettich Universal 30F) at 2,800·g for 20 min. After centrifugation a sample b is taken of the

supernatant. The size of the particles x in the supernatant taken from a distance R1 = 0.01 m

below the surface of the dispersion which rotates at a radius of R2 = 0.08 m apart form the

rotation axis is calculated following STOKES law of settling particles in a centrifugal field

presented in eq. (85).

2

1

2

21fluid

2fluidlenanopartic

ln18

2

R

RR

ntx

(85)

The specific weights are 1.33 g/cm3 and 5.2 g/cm3 for the fluid DCM and the particles Fe3O4,

respectively. This results in particles smaller than 26 nm in the supernatant with a fluid

viscosity of 0.41 mPas without addition of a polymer. The sample 7 of table 25 has a the

highest fluid viscosity of 2.2 mPas, 2.8 mPas and 8.4 mPas for PMMA, PC and PVB,

respectively. This leads to particles smaller than 60 nm, 68 nm and 117 nm in the

supernatant for PMMA, PC and PVB, respectively.

Both samples a and b are oven dried for 3 h at 50 °C to evaporate the solvent. The

compositions are then characterized with thermo gravimetric analysis (TGA) using a Netzsch

STA 449 from room temperature to 900 °C with a heating rate of 20 K/min under a nitrogen

atmosphere. Figure 2 shows exemplarily the decomposition curves for the PMMA system

and the samples # 1, 3 and 5 with polymer concentrations cpolymer of 52.0 g/l, 19.5 g/l and

5.6 g/l, respectively. The pristine nanoparticle dispersion (sample # 7) is characterized as well

revealing the amount of fatty acid and a nanoparticle residual mass of about 83 %, as

expected.

The weight concentration of the primary particles wPrimary is simply the ratio of the mass

residues (at 600 °C, cf. 4.3.4 and [21]) of sample b wb to sample a wa, as presented in

eq. (86).

a

bPrimary w

ww (86)

The quantity w generally stands for mass concentrations. Indices a and b are the samples

before and after centrifugation, respectively, as explained in the graphical insets in figure

113.

A27

For polymers which do not decompose but have residual ash masses without any fillers like

for PC, the residual mass of the functionalized nanoparticles as described in eq. (86) has to

be corrected defined by the following eq. (87).

polymerr,nanor,

polymerr,*a/bnanor,

a/b ww

wwww

(87)

With wa/b* being the corrected mass concentrations using the residual mass of the fatty acid

coated nanoparticles wr,nano which are attributed to the decomposition of organic material

like surfactants used for stabilization (wr,nano = 0.833 for D = 0.2) and wr,polymer which accounts

for the residual mass due to ash when decomposing the pure polymer (wr,polymer ≈ 0.2 for

PC).

figure 113: Three representative sets of TGA analyses of the composition of the solids in the pristine dispersion -

samples (a) and in the supernatant - samples (b) to determine the primary particle concentration

wPrimary (numbers explained in table 25) [20]

The error for the primary particle concentration u(wPrimary) is given in eq. (88) with respect to

the errors of the TGA determined residual mass fractions u(wa) and u(wb).

2a

b

aba,Primary

1

w

w

wwuwu (88)

The errors from TGA measurements are taken to be u(wa) = u(wb) = u(wa,b) = 0.02, by

experience regarding sample preparation and few repetitive measurements of selected

samples.

The uncertainties and consequent errors of the polymer concentration cpolymer are caused by

the several gravimetric mixing steps of polymer solutions, nanoparticle dispersions and the

pure solvent. The following calculations in eq. (89) obtain the indices 1 for the polymer

solution, 2 for the nanoparticle dispersion and 3 for the solvent.

A28

3311

31122

23112

23112

1312131

1312131

2polymer

311

32211

311polymer

1

111

111

muwm

Duwmwm

wuwmDm

muwmwD

wuwmBm

mumwBw

Bcu

B

wm

mwDmwm

wmc

(89)

The following values shall be given: w1 = 0.15, u(w1) = 0.0015, w2 = 0.08, u(w1) = 0.0063,

D = 0.2, u(D) = 0.0006 and u(m1) = u(m2) = u(m3) = 0.01.

A.5.3 UV/VIS Extinction Based Determination of the Primary Particle

Concentration

Besides gravimetric determination of the primary particle concentration, it is of interest to

correlate wPrimary to the concentration dependent photo metrically determined extinction E

of the diluted supernatant at a fixed wavelength . This method is quicker than the time-

consuming TGA and could be used instead, if a linear dependency of E versus wPrimary is

noticed. This linear relation follows the assumption of a monodisperse distribution of

particles in the diluted supernatant and thus elimination of the size contribution in the

extinction E as described by MALYNYCH et al. [261] combining MIE’s scattering theory and

LAMBERT-BEER’s la in eq. (90):

edV

NxCxE log,, ext

, (90)

where x is the assumed constant particle diameter, NP is the particle number in the sample

Volume V and d is a fixed inner thickness of the cuvette (pathway of light through the

sample). Using the definition of the volume concentration of the particles in eq. (91):

V

xN

V

VN

6

3lenanopartic

, (91)

with the volume of a spherical particle VP with particle diameter x, eq. (90) can be

transformed into eq. (92).

edx

xCxE log6

,,3ext

(92)

For each sample b 20 µl are diluted with 1 ml of solvent resulting in a maximum volume

concentration of 9.3·10-5 if all nanoparticles are in the supernatant. This is the case for

wPrimary = 1 because the particle concentration of the samples a (before centrifugation) are

A29

constant and should decrease with wPrimary in the diluted samples b, so that the particle

volume concentration φ after dilution is (eq. (93)):

µl

µlcw

1000

20

lesnanopartic

lesnanoparticPrimarylesnanopartic

,

(93)

where nanoparticles is the specific weight of the nanoparticles, which is presumed to be

5,200 g/l for magnetite without considering the adsorbed fatty acid layer that also is not

apparent in wPrimary.

The dilutions are analyzed using a photospectrometer (Perkin Elmer Lambda 3B) with the

pure solvent as the reference sample in quartz cuvettes with d = 10 mm.

The primary particle concentration can also be defined with eq. (94).

0Primary

w, (94)

with the volume fraction of particles φ in the sample b and φ0 in the sample a.

Under the assumption of Rayleigh scattering (small particles with x < 0.1· which have a

refractive index close to 1) the scattering part of extinction is defined by [33] in eq. (95).

1

32

3,

solvent

lesnanopartic

4

3

n

nd

xxE

(95)

Here nnanoparticles and nsolvent are the real terms of the refractive indices of the nanoparticles

and the solvent, respectively. It shows that scattering is hyperbolic proportional to

wavelength with -4. Appendix A.8 presents optical properties and theoretical extinction

behavior of magnetite nanoparticles.

A.5.4 Determination of Polymer Adsorption

The adsorption of polymers onto the nanoparticle surface, here defined as the mass of

polymer mpolymer per mass of nanoparticles mnanoparticles, is determined by solvent evaporation

of samples b above a strong neodymium magnet and subsequent thorough washing with

DCM, holding back the super-paramagnetic nanoparticles and adsorbed matter. The washed

and dried residue is then analyzed with TGA to ascertain with the following relation:

1polymenor,

1polymerr,

lesnanopartic

polymer rwwm

m,

(96)

where wr,polymer and wr, no polymer are the residual mass fractions derived from TGA with and

without polymer, respectively. For PVB several polymer concentrations from 0 g/l to 52 g/l

A30

are considered, whereas for PC and PMMA only the adsorption at one polymer

concentration of 24.9 g/l is investigated. This procedure is necessary for the investigations in

paragraph 5.3.3.

A31

A.6 Mechanical Dispersing Methods

A.6.1 Planetary Ball Mill – PM

device: Pulverisette 5/4 type 05.102

company: Fritsch

Media mills, such as agitated ball mills are widely used for dispersing agglomerated

submicron particles and comminution down to nanoparticles sizes [52, 202, 262-264].

Planetary ball mills are non-agitated media mills which are also used for nanoscale

production of particles [265]. The special motion of the balls has recently been investigated

numerically [266].

For the experiments, balls with 0.1 mm diameter made of yttrium stabilized zirconium oxide

YSZ are applied (Alpine Power Beads). The batch grinding vessel has a volume of 80 ml and is

also made of YSZ. It is filled up to 30 % by volume with the balls including voids which equals

a mass of 20 g of balls. The suspension volume is 20 ml which corresponds to 1.5-times the

volume of the voids of the ball packing. The sun wheel is rotating at 37.7 s-1 with faster

counter-rotating planets with -82.6 s-1. The sun wheel and planet vessel diameters are

200 mm and 65 mm, respectively. This results in an acceleration of the dispersing vessel of

188.3 m/s2 (19·g).

A.6.2 Sonotrode Ultrasound – US

device: SONOPLUS HD 2200

company: Bandelin

Sonotrode ultrasonic treatment is very common for dispersing submicron particle

agglomerates [51]. The mechanical dispersing stress is caused by cavitation events which is

an instable grow of gas bubbles until collapse with subsequent high local shear stress [267].

Adviceable literature on effects of ultrasound cavitation, so called sonochemistry in general

is offered by SUSLICK [268, 269].

The sonotrode used is a 200 W SONOPLUS HD 2200 from Bandelin with the ultrasound

transducer UW 2200 and controller GM 2200. The sonotrode type is VS 70 T with a diameter

of 13 mm made of a titanium alloy. The maximum amplitude is reported as 153 µm9 at a

frequency of 20 kHz. For the experiments a suspension volume of 20 ml is dispersed and

cooled on the mantle of the batch vessel at 20 °C. To release larger not cavitating but

9 http://www.bandelin.com/produkte/p4/p41.htm, June 14th 2012

A32

ultrasound damping gas bubbles the US is pulsed with 50 %, which means ½ a second

ultrasound emission every second.

A.6.3 Rotor-Stator Mixing

device: Ultra-Turrax T18 basic

company: IKA

Dispersing with rotor stator systems is also a common method for submicron particles [50,

204]. The mechanism of deagglomeration is a very high shear in a small gap which is

constantly flushed with a circulating suspension. The device used in this study is an Ultra-

Turrax T18 basic from IKA equipped with the dispersing element S 10 N 10 G. Just like when

dispersing with US a suspension volume of 20 ml is stressed and cooled on the mantle of the

batch vessel at 20 °C. The rotor diameter is 7.6 mm10 with a 200 µm gap. The rotational

speed is set at 200 s-1 resulting in a mantle speed of the rotor of 9.6 m/s which is equal to

the speed of liquid at the rotor resulting in a shear rate of approximately 47,750 s-1 within

the gap.

10 http://www.ika.de/Products-Lab-Eq/Dispersers-Homogenizer-csp-177/, June 14th 2012

A33

A.7 Spray Drying

device: Mini Spray Dryer B-290

company: Büchi

Microparticle composite formation is achieved by quick solvent evaporation using a

commercial lab-scale co-current spray dryer Büchi B-290 with an inert nitrogen atmosphere

equipped with an external mixing two fluid nozzle (cf. figure 70) and a condenser Büchi B-

295. A photograph and a principle scheme of the system are presented in figure 114.

figure 114: Configuration of the lab-scale spray dryer with inert gas flow (left) photograph after spray drying a

PMMA-based composite with RA-Fe3O4 and F = 0.3, (right) schematic drawing with: a) external

mixing two fluid nozzle, b) ventilator for drying gas circulation, c) heater, d) condenser; c)-d) cannot

be seen in the photograph on the left

The important setting parameters are presented in table 26.

table 26: Important setting parameters of the lab scale spray dryer

inlet temperature (control parameter) 60 °C

temperature at the condenser -20 °C

flow rate of drying gas 3.50·101 m3/h

nozzle atomizing gas flow rate 1.05·100 m3/h

feed flow rate (squeeze pump) 7.20·10-4 m3/h

nozzle diameter (feed) 0.7 mm

nozzle diameter (atomizing gas) 1.4 mm

The cyclone cut-off size where 50 % of the feed is transferred to the coarse product is

approximately 1 µm, as determined experimentally.

A34

The experimental protocol is as follows:

(1) turning on the system and setting the condenser at a temperature of 10 °C

before flushing with dry nitrogen

(2) start the ventilator at maximum power (3.50·101 m3/h), the overpressure

behind the filter should read +30 mbar, a lower value may be due to a clocked

filter or a problem with a valve; overpressure is necessary for maintaining an

explosive safe state within the dryer

(3) 10 min purging the entire system with dry technical quality nitrogen at 283 l/h

through the nozzle, the oxygen sensor drops to a value of 0.0 %.

(4) turning on the heating element and setting the condenser to -20 °C

(5) waiting 10 min for a static process with constant temperatures at inlet and

outlet of the cylinder

(6) setting the purge and atomizing gas to 1.05·100 m3/h and pumping pure

solvent with a flow rate of 7.20·10-4 m3/h for five minutes until the system is

static again

(7) feeding the entire dispersion at a flow rate of 7.20·10-4 m3/h

(8) finishing with another five minutes atomizing pure solvent through the feed

(9) turning off the heating device and continuing drying gas circulation through

the system for 10 minutes, then turning off the ventilator and opening the

device to take out the material

A35

A.8 Optical Properties of Magnetite

This excursus is to introduce to the optical properties of the material magnetite and with this

explain the appearance of colloidal dispersions of magnetite nanoparticles. The strong

absorption behavior of magnetite is the bottleneck of characterizing the colloid. This is one

reason why e.g. the gravimetric method in A.5 had to be introduced instead of an optical

method.

A very important optical parameter is the wavelength dependent complex refractive index

which is given in eq. (97).

iknn ~ (97)

This physical quantity depends on the electromagnetic wavelength and is for example

important for MIE evaluation of particle sizes with the methods of dynamic light scattering or

laser diffraction. However, it turns out that it is hard to obtain this data in the literature.

One way to extract the complex indices is by measuring the dielectric properties, because of

the relation in eq. (98) [270].

2,

2

22

211

22

211 kn (98)

Dielectric constants 1 and 2 can be obtained e.g. by reflectance measurements [271]. The

data for magnetite is extracted from SCHLEGEL et al. [271] and presented as a function of the

energy E in figure 115 on the left. This energy can be transformed into wavelengths by

PLANCK’s fa ous elatio E = h·f with the PLANCK constant h and the frequency f with f = c / ,

where c is the speed of light and the wavelength.

figure 115: (left) dielectric properties of magnetite, data extracted from [271] as a function of the energy E,

which is (right) calculated for wavelengths

A36

With the data in figure 115 on the right and eq. (98) the complex refractive index can be

determined as a function of wavelength, which is graphically presented in figure 116 for the

visible electromagnetic spectrum in the center.

figure 116: real and imaginary part of the refractive index of magnetite as calculated from the data in figure

115 using eq. (90)

The imaginary part of the refractive index is responsible for the light absorption and the high

values in figure 116 justify why magnetite appears black. Magnetite is a highly absorbing

medium, which also corresponds to the deep black color of magnetite dispersion already at

low particle concentrations.

The light extinction in particulate dispersionsis introduced in (90) showing to be proportional

to the extinction cross section Cext which is determined by MIE theory. With the help of the

complex refractive index over a wide wavelength range in figure 116 and the free MIE

calculation program MiePlot11 one can calculate the extinction cross-section Cext as well as

the absorption and scattering cross-sections which make up Cext of magnetite nanoparticles

in any solvent. This value is necessary for calculating expected extinction values in UV/VIS

analyses.

The results of the Mie calculation are presented in figure 117. Since the extinction is

dominated by absorption by 99.50 % for 300 nm and 99.98 % for 900 nm the scattering cross

section in the graph is multiplied by 100 and the absorption is divided by two for

visualization purposes.

11 Version 4.2, http://www.philiplaven.com/mieplot.htm (march 22nd 2012)

A37

figure 117: Extinction, scattering and absorption cross-sections Cext, Cscat and Cabs of magnetite nanoparticles

with a diameter of 15 nm and optical properties defined in figure 116 in dichloromethane with a

refractive index of 1.4242 as a function of the wavelength, notice that scattering and absorption

values are 100- and 0.5-fold, i.e. extinction is mainly due to absorption

A38

A.9 Viscosity of Polymer Solutions and Solubility

Using capillary viscosimetry by UBBELOHDE (cf. A.2.6) the dynamic viscosities of polymer

solutions are determined and the results for PMMA, PC, PVB and PS in DCM as well as

PMMA in EA are plotted in figure 118.

figure 118: dynamic viscosities of the polymers of this thesis in DCM as well as PMMA in EA as a function of the

polymer concentration, determined with UBBELOHDE viscosimetry

In a linear plot of ordinate and abscissa the viscosity increases progressively for all solutions.

The least increase is noticed for PMMA in EA as well as in DCM. Most progressive of all is the

solution of PVB in DCM. Following the theory introduced in 5.1.1 it is more useful plotting

the reduced specific viscosity to evaluate the intrinsic viscosity [ ] in A.9.1.

A.9.1 Intrinsic Viscosity

Plotting the reduced viscosity sp/cpolymer (cf. eq. (44)) over the polymer concentration

enables to estimate the intrinsic viscosity [ ] which is related to the molar mass in eq. (45).

This is done for all combinations in figure 118 but PVB, for which a linear fit seems

unjustified. The results are plotted in figure 119 and quantitatively evaluated in table 27.

figure 119: Reduced viscosity over polymer concentration for various polymers in DCM and PMMA in EA. The

lines show the linear fit to evaluate the intrinsic viscosity following eq. (46)

A39

In table 27 the intrinsic viscosities are presented with the error evaluated by linear fitting the

data in figure 119 considering the data uncertainties. The fit quality is presented with R2.

table 27: Evaluation of the intrinsic viscosities [ ], the overlap concentration c*

polymer using eq. (42), the radius of

gyration RG using eq. (46), the hydrodynamic radius determined with DLS and the ratio of hydrodynamic radius

to radius of gyration X using eq. (41)

[ ] in l/g R2 in - c*

polymer in g/l RG in nm RH in nm X = RH/RG

PMMA

in DCM 0.031 ± 0.035 0.900 32.2 ± 36.4 5.9 ± 2.2 4.0 ± 0.7 0.68 ± 0.37

PC

in DCM 0.019 ± 0.038 0.980 52.6 ± 105.3 3.1 ± 2.1 3.0 ± 0.9 0.97 ± 0.95

PS

in DCM 0.036 ± 0.010 0.995 27.8 ± 7.7 - - -

PMMA

in EA 0.023 ± 0.003 0.962 43.5 ± 5.7 5.3 ± 0.2 - -

A.9.2 Polymer Solubility

The polymer solubilities in this paragraph are all based on HSP and evaluation of solubility

distance D1,2 using eq. (50) as well as the FLORY interaction parameter with eq. (52). The

theory behind this approach is introduced in 5.1.2.

The solubilities of PMMA, PC and PVB in DCM are presented in table 28. These information

are used for investigations and discussions in 5.3.1. A comparison of the solubility of the

polymer PMMA in the three solvents DCM, EA and MMA is needed for the discussions in

5.3.6. The results are presented in table 29. A comparison of the solubility of the polymer PS

in the two solvents DCM and ST is needed for the discussions in 5.3.6 as well. The results are

presented in table 30.

table 28: Polymer solubility in DCM for PMMA, PC and PVB with HSP values from table 19 and table 24,

solubility distance D1,2 using eq. (50), FLORY interaction parameter χ using eq. (52)

RH

in nm

D1,2

in MPa1/2

χ

in -

PMMA 4.0 ± 0.7 5.00 0.10

PC 3.0 ± 0.9 1.00 0.00

PVB 8.5 ± 0.4 5.59 0.12

A40

table 29: Solubility of PMMA in the solvents dichloromethane, ethyl acetate and methyl methacrylate with HSP

values from table 19 and table 24, solubility distance D1,2 using eq. (50), FLORY interaction parameter χ using

eq. (52)

D1,2

in MPa1/2

χ

in -

DCM 5.00 0.10

EA 7.25 0.31

MMA 6.10 0.24

table 30: Solubility of PS in dichloromethane and styrene with HSP values from table 19 and table 24, solubility

distance D1,2 using eq. (50), FLORY interaction parameter χ using eq. (52)

D1,2

in MPa1/2

χ

in -

DCM 5.25 0.11

ST 3.77 0.10

A41

A.10 Limitations of Dynamic Light Scattering for Dispersions

with Primary Particles and Agglomerates

Certainly dynamic light scattering is a straight forward method when it comes to

determining the size of nanoparticles. Yet, it shall be emphasized here, that, based on

numerous experiences with different types of nanoparticles, this method is not reliable for

dispersions with both small primary particles and large agglomerates tending to settle

rapidly. Generally, for a broad particle size distribution the volume or number weighted

particle size distribution cannot be assessed quantitatively, i.e. the fractions of the sizes are

difficult to be determined. A certain fraction of agglomerated nanoparticles can even make it

impossible to identify the primary particles.

In the following graphs in figure 120 correlation coefficients of a single dynamic light

scattering experiment are presented at different time steps of ricinoleic acid transferred

particles containing agglomerates. Furthermore the intensity weighted frequency

distributions are depicted.

figure 120: (left) correlation coefficient and (right) frequency distribution intensity weighted of one single DLS

experiment for ricinoleic acid transferred particles containing agglomerates at time steps t1 through

t4 which are 2 minutes apart each; additionally the result for the sample after centrifugation

containing only primary particles

There are three fractions found where the smallest clearly represents the primary particles.

Two larger fractions appear at about 100 nm and between 400 nm and 1 µm. The quantity of

the fractions changes with measurement time. This is even more intriguing for the

correlation coefficient which can be regarded as the data of DLS which has the least impact

by mathematical artifacts. The correlation coefficient for longer correlation times (plotted on

the abscissa) which correspond to the larger particles (agglomerates) decreases with

increasing measuring time. This concludes that the sample is losi g la ge pa ti les due to sedimentation and thus withdrawal from the measurement zone. For the purpose of a

statistically reliable sample the measurement time should even be longer when larger

particles occur, due to longer correlation times.

A42

Finally there are at least two reasons why DLS should not be used for such samples. First of

all the size range is too broad and secondly and most importantly the sample is not stable

due to sedimentation of the agglomerates.

A43

A.11 Calculation of HANSEN Solubility Parameters with the

Group Contribution Method

The concept of HSP is introduced in 5.1.2. Details on the theory and usage of HANSEN

Solubility Parameters (HSP) can furthermore be found in [137, 154]. For the fatty acids

capped onto the magnetite nanoparticle surface HOY’s ethod as des i ed i [154] is

applied with the input parameters Ft,i, Fp,i, Vi, T,i(P), B = 277 of the groups i which are defined

below in table 32 and table 33. The parameter ni stands for the number of each group i

within the substance.

The first parameter is the molar attraction function with contributes to the total solubility

parameter (HILDEBRANDT parameter) Ft in eq. (99).

i

FnF it,it (99)

Next is the polar component of the molar attraction function corresponding to the polar HSP

Fp in eq. (100).

i

FnF ip,ip (100)

The molar volume V is defined in eq. (101).

i

VnV ii (101)

The parameter T(P) in eq. (102) is the LYDERSON correction for polymer non-ideality.

i

PiT

PT n )(

,i)( (102)

Eqs. (99) - (102) are so called additive molar functions. The next two eqs. (103) and (104) are

named auxiliary equations necessary to finally evaluate the HSP.

)(

5.0~P

T

n (103)

V

Δ pTP

)()( 777 (104)

The HILDEBRANDT parameter t is defined in eq. (105).

A44

V

nBFthpdt

~222 (105)

The definition of the polar HSP p is given in eq. (106).

2

1

)( ~1

nBF

F

t

p

Ptp (106)

The HSP due to hydrogen bonding h is defined in eq. (107).

2

1

)(

)( 1

P

P

th (107)

With eq. (108) the dispersive HSP is to calculate using eqs. (105) - (107).

2h

2p

2td (108)

In table 31 the group contribution parameters are listed for the groups that make up the

structures of the fatty acids.

table 31: Molar group parameters of the group contribution method of the groups relevant for the structure of

fatty acids

group Ft,i in

J∙ 3)1/2/mol

Fp,i in

J∙ 3)1/2/mol

Vi

cm3/mol T,i

(P)

-CH3 303.5 0 21.55 0.0220

-CH2- 269.0 0 15.55 0.0200

=CH- 249.0 59.5 13.18 0.0185

-CHOH- 591.0 591 12.45 0.0490

-COOH 565.0 415 17.30 0.0400

The numbers of the groups for the fatty acids RA, LA, OA, MA and CA are presented in table

32.

table 32: Number of specific groups in the fatty acids used

Fatty Acid -CH3 -CH2- =CH- -CHOH- -COOH

Ricinoleic Acid (RA) 1 13 2 1 1

Linoleic Acid (LA) 1 12 4 0 1

Oleic Acid (OA) 1 14 2 0 1

Myristic Acid (MA) 1 12 0 0 1

Caprylic Acid (CA) 1 6 0 0 1

In case the fatty acids are grafted chemically to the magnetite surface the carboxyl group is

left out and the numbers of the individual groups are given in table 33.

A45

table 33: Number of specific groups in the grafted fatty acids, neglecting influence from the chemically bound

complex

Fatty Acid

capped Fe3O4 -CH3 -CH2- =CH- -CHOH- -COOH

RA-Fe3O4 1 13 2 1 0

LA-Fe3O4 1 12 4 0 0

OA-Fe3O4 1 14 2 0 0

MA-Fe3O4 1 12 0 0 0

CA-Fe3O4 1 6 0 0 0

Using eqs. (99) - (108) as well as the information in table 32, the HILDEBRANDT and HSP for the

fatty acids are calculated and presented together with the solubility distance to DCM in table

34.

table 34: HSP of FA calculated with the group contribution method and the number of groups from table 32 and

solubility distance in DCM DFA-Fe3O4-DCM with eq. (50) and HSP for DCM in A.1.4

Fatty Acid t in

MPa1/2 d in

MPa1/2 p in

MPa1/2 h in

MPa1/2

DFA-DCM

in MPa1/2

RA 19.68 16.84 8.32 5.89 3.89

LA 18.47 17.16 6.38 2.46 5.73

OA 18.31 17.30 5.73 1.78 6.31

MA 18.20 17.30 5.60 0.76 7.30

CA 18.33 16.87 7.10 1.05 7.30

Using eqs. (99) - (108) as well as the information in table 33, the HILDEBRANDT and HSP for the

fatty acids chemically bound to the Fe3O4 surface are calculated and presented together with

the solubility distance to DCM in table 35.

table 35: HSP of FA-Fe3O4 calculated with the group contribution method and the number of

groups from table 33 and solubility distance in DCM DFA-Fe3O4-DCM with eq. (50) and HSP for

DCM in A.1.4

Fatty Acid

capped Fe3O4 t in

MPa1/2 d in

MPa1/2 p in

MPa1/2 h in

MPa1/2

DFA-Fe3O4-DCM

in MPa1/2

RA-Fe3O4 19.40 17.27 6.89 5.55 2.98

LA-Fe3O4 18.07 17.61 4.02 0.50 7.74

OA-Fe3O4 17.90 17.68 2.81 0 8.61

MA-Fe3O4 17.66 17.66 0 0 10.08

CA-Fe3O4 17.38 17.38 0 0 10.16

The final table 36 compares the solubility distances and FLORY interaction parameters of end-

grafted fatty acids in the solvents dichloromethane, ethyl acetate, styrene and methyl

methacrylate.

A46

table 36: solubility distances of fatty acid caped Fe3O4 in different solvents using eq. (50) as well as FLORY

interaction parameter using eq. (52) and the HSP of the solvents reported in A.1.4 and of the FA-Fe3O4 listed

above

dichloromethane ethyl acetate styrene methyl methacrylate

D12

in MPa1/2 χ

in - D12

in MPa1/2 χ

in - D12

in MPa1/2 χ

in - D12

in MPa1/2 χ

in -

RA-Fe3O4 2.98 0.03 3.72 0.08 6.47 0.29 2.96 0.06

LA-Fe3O4 7.74 0.23 7.72 0.36 4.00 0.11 6.57 0.28

OA-Fe3O4 8.61 0.29 8.50 0.43 3.60 0.09 7.55 0.37

MA-Fe3O4 10.08 0.39 9.69 0.56 3.27 0.07 9.24 0.55

CA-Fe3O4 10.16 0.40 9.48 0.54 3.36 0.08 9.02 0.52

R1

References

[1] V. DUSASTRE, S. TOMLIN, M. BELLANTONE, C. LOBO: Our changing nature. In: Nat Mater 1 (2002), p. 1-1.

[2] Y. BRÉCHET, J.Y. CAVAILLÉ, E. CHABERT, L. CHAZEAU, R. DENDIEVEL, L. FLANDIN, C. GAUTHIER: Polymer Based Nanocomposites: Effect of Filler-Filler and Filler-Matrix Interactions. In: Advanced Engineering Materials 3 (2001), p. 571-577.

[3] G. SCHMIDT, M.M. MALWITZ: Properties of polymer-nanoparticle composites. In: Current Opinion in Colloid & Interface Science 8 (2003), p. 103-108.

[4] B. HICKSTEIN, U.A. PEUKER: Modular process for the flexible synthesis of magnetic beads—Process and product validation. In: J. Appl. Polym. Sci. 112 (2009), p. 2366-2373.

[5] F. HUSSAIN, M. HOJJATI, M. OKAMOTO, R.E. GORGA: Review article: Polymer-matrix nanocomposites, processing, manufacturing, and application: An overview. In: J. Compos Mater. 40 (2006), p. 1511-1575.

[6] S.J. AHMADI, Y.D. HUANG, W. LI: Synthetic routes, properties and future applications of polymer-layered silicate nanocomposites. In: Journal of Materials Science 39 (2004), p. 1919-1925.

[7] T. BANERT, U.A. PEUKER: Preparation of highly filled super-paramagnetic PMMA-magnetite nano composites using the solution method. In: Journal of Materials Science 41 (2006), p. 3051-3056.

[8] S. LI, M.M. LIN, M.S. TOPRAK, D.K. KIM, M. MUHAMMED: Nanocomposites of polymer and inorganic nanoparticles for optical and magnetic applications. In: Nano Reviews 1 (2010), p. 5214.

[9] A. CAMENZIND, W.R. CASERI, S.E. PRATSINIS: Flame-made nanoparticles for nanocomposites. In: Nano Today 5 (2010), p. 48-65.

[10] S. BEHRENS: Preparation of functional magnetic nanocomposites and hybrid materials: recent progress and future directions. In: Nanoscale 3 (2011).

[11] P. DALLAS, V. GEORGAKILAS, D. NIARCHOS, P. KOMNINOU, T. KEHAGIAS, D. PETRIDIS: Synthesis, characterization and thermal properties of polymer/magnetite nanocomposites. In: Nanotechnology 17 (2006), p. 2046-2053.

[12] B.A. ROZENBERG, R. TENNE: Polymer-assisted fabrication of nanoparticles and nanocomposites. In: Progress in Polymer Science 33 (2008), p. 40-112.

[13] A.C. BALAZS, T. EMRICK, T.P. RUSSELL: Nanoparticle polymer composites: Where two small worlds meet. In: Science 314 (2006), p. 1107-1110.

[14] R.H. FRENCH, V.A. PARSEGIAN, R. PODGORNIK, R.F. RAJTER, A. JAGOTA, J. LUO, D. ASTHAGIRI, M.K. CHAUDHURY, Y.M. CHIANG, S. GRANICK, S. KALININ, M. KARDAR, R. KJELLANDER, D.C. LANGRETH, J. LEWIS, S. LUSTIG, D. WESOLOWSKI, J.S. WETTLAUFER, W.Y. CHING, M. FINNIS, F. HOULIHAN, O.A. VON LILIENFELD, C.J. VAN OSS, T. ZEMB: Long range interactions in nanoscale science. In: Reviews of Modern Physics 82 (2010), p. 1887-1944.

R2

[15] K.J.M. BISHOP, C.E. WILMER, S. SOH, B.A. GRZYBOWSKI: Nanoscale forces and their uses in self-assembly. In: Small 5 (2009), p. 1600-1630.

[16] T. BANERT: Synthese funktionaler Nanokomposit-Partikeln für die Bioseparation durch Sprühtrocknung - Auswahl und Eignung von Stoffsystemen. TU Clausthal, Fakultät für Mathematik/Informatik und Maschinenbau, Dissertation. 2008

[17] T. BANERT, U.A. PEUKER: Herstellung hochgefüllter Bauteile aus Nanokompositmaterial. In: Chemie Ingenieur Technik 77 (2005), p. 1232.

[18] X.D. ZHOU, H.C. GU: Synthesis of PMMA-ceramics nanocomposites by spray process. In: Journal of Materials Science Letters 21 (2002), p. 577-580.

[19] M. RUDOLPH, K. BAUER, U.A. PEUKER: Phase-contrast atomic force microscopy for the characterization of the distribution of nanoparticles in composite materials. In: Chemie Ingenieur Technik 82 (2010), p. 2189-2195.

[20] M. RUDOLPH, U.A. PEUKER: Coagulation and stabilization of sterically functionalized magnetite nanoparticles in an organic solvent with different technical polymers. In: Journal of Colloid and Interface Science 357 (2011), p. 292-299.

[21] M. RUDOLPH, J. ERLER, U.A. PEUKER: A TGA/FTIR perspective of fatty acid adsorbed on magnetite nanoparticles - Decomposition steps and magnetite reduction. In: Colloids and Surfaces A: Physicochemical and Engineering Aspects 397 (2012), p. 16-23.

[22] S. KIRCHBERG, M. RUDOLPH, G. ZIEGMANN, U.A. PEUKER: Nanocomposites Based on Technical Polymers and Sterically Functionalized Soft Magnetic Magnetite Nanoparticles: Synthesis, Processing, and Characterization. In: Journal of Nanomaterials 2012 (2012), p. Article ID 670531.

[23] M. RUDOLPH, U.A. PEUKER: Phase Transfer of Agglomerated Nanoparticles - Deagglomeration by Adsorbing Grafted Molecules and Colloidal Stability in Polymer Solutions. In: Journal of Nanoparticle Research 14 (2012), p. 990.

[24] M. RUDOLPH, S. KIRCHBERG, C. TURAN, G. ZIEGMANN, U.A. PEUKER, Synthesis of Highly Filled Nanomagnetite Polymeric Composites via Sterically Stabilized Organosols and the Spray Drying Process, WCPT6, Nuremberg, 2010.

[25] M. RUDOLPH, C. TURAN, S. KIRCHBERG, G. ZIEGMANN, U.A. PEUKER, Processing and Characterization of Highly Filled Polymer Nanoparticle Composites for Micro Injection Molding Applications, ECCM14, Budapest, 2010.

[26] M. RUDOLPH, U.A. PEUKER, On the significance of nanoparticle interactions for the synthesis of highly filled polymer-nanoparticle-composites with the solution and spray-drying process, in: V.S. Litvinenko (Ed.), International Forum-Competition of Young Researchers - Topical Issues of Subsoil Usage, Ministry of Education and Science Russian Federation, St. Petersburg, Russia, 2011, p. 278.

[27] M. RUDOLPH, C. TURAN, S. KIRCHBERG, G. ZIEGMANN, U.A. PEUKER, in: G. Tiddy, R.B.H. Tan (Eds.), NanoFormulation, The Royal Society of Chemistry, 2012, pp. 177-187.

[28] nano.DE-Report 2011 - Status quo der Nanotechnologie in Deutschland, nano.DE-Report, Bundesministerium für Bildung und Forschung, Berlin, 2011.

[29] R.K. GUPTA, E. KENNEL, K.-J. KIM (Ed.) Polymer Nanocomposites Handbook. CRC Press: Boca Raton, 2010.

R3

[30] M. ALEXANDRE, P. DUBOIS: Polymer-layered silicate nanocomposites: preparation, properties and uses of a new class of materials. In: Materials Science and Engineering: R: Reports 28 (2000), p. 1-63.

[31] R. KRISHNAMOORTI, R.A. VAIA: Polymer Nanocomposites. 260. American Chemical Society, 2001. - ISBN 0-8412-3768-9

[32] T. HANEMANN, D.V. SZABO: Polymer-Nanoparticle Composites: From Synthesis to Modern Applications. In: Materials 3 (2010), p. 3468-3517.

[33] H. ALTHUES, J. HENLE, S. KASKEL: Functional inorganic nanofillers for transparent polymers. In: Chem. Soc. Rev. 36 (2007), p. 1454-1465.

[34] G. ALLEGRA, G. RAOS, M. VACATELLO: Theories and simulations of polymer-based nanocomposites: From chain statistics to reinforcement. In: Progress in Polymer Science 33 (2008), p. 683-731.

[35] S. GYERGYEK, M. HUSKIC, D. MAKOVEC, M. DROFENIK: Superparamagnetic nanocomposites of iron oxide in a polymethyl methacrylate matrix synthesized by in situ polymerization. In: Colloid Surf. A-Physicochem. Eng. Asp. 317 (2008), p. 49-55.

[36] S. GYERGYEK, D. MAKOVEC, A. MERTELJ, M. HUSKIC, M. DROFENIK: Superparamagnetic nanocomposite particles synthesized using the mini-emulsion technique. In: Colloid Surf. A-Physicochem. Eng. Asp. 366 (2010), p. 113-119.

[37] R.Y. HONG, B. FENG, X. CAI, G. LIU, H.Z. LI, J. DING, Y. ZHENG, D.G. WEI: Double-Miniemulsion Preparation of Fe3O4/Poly(methyl methacrylate) Magnetic Latex. In: J. Appl. Polym. Sci. 112 (2009), p. 89-98.

[38] D. HORÁK, N. CHEKINA: Preparation of magnetic poly(glycidyl methacrylate) microspheres by emulsion polymerization in the presence of sterically stabilized iron oxide nanoparticles. In: J. Appl. Polym. Sci. 102 (2006), p. 4348-4357.

[39] J.S. HUANG, S.R. WAN, M. GUO, H.S. YAN: Preparation of narrow or mono-disperse crosslinked poly((meth)acrylic acid)/iron oxide magnetic microspheres. In: Journal of Materials Chemistry 16 (2006), p. 4535-4541.

[40] N.S. KOMMAREDDI, M. TATA, V.T. JOHN, G.L. MCPHERSON, M.F. HERMAN, Y.-S. LEE, C.J. O'CONNOR, J.A. AKKARA, D.L. KAPLAN: Synthesis of Superparamagnetic Polymer - Ferrite Composites Using Surfactant Microstructures. In: Chemistry of Materials 8 (1996), p. 801-809.

[41] M. KRYSZEWSKI, J.K. JESZKA: Nanostructured conducting polymer composites -- superparamagnetic particles in conducting polymers. In: Synthetic Metals 94 (1998), p. 99-104.

[42] R. MOHR, K. KRATZ, T. WEIGEL, M. LUCKA-GABOR, M. MONEKE, A. LENDLEIN: Initiation of shape-memory effect by inductive heating of magnetic nanoparticles in thermoplastic polymers. In: Proceedings of the National Academy of Sciences of the United States of America 103 (2006), p. 3540-3545.

[43] J.L. WILSON, P. PODDAR, N.A. FREY, H. SRIKANTH, K. MOHOMED, J.P. HARMON, S. KOTHA, J. WACHSMUTH: Synthesis and magnetic properties of polymer nanocomposites with embedded iron nanoparticles. In: Journal of Applied Physics 95 (2004), p. 1439-1443.

[44] M. YAMAGATA, M. ABE, H. HANDA, H. KAWAGUCHI, Magnetite/polymer composite particles prepared by molecular assembling followed by in-situ magnetite formation, 41st

R4

International Symposium on Macromolecules, Wiley-V C H Verlag Gmbh, Rio de Janeiro, BRAZIL, 2006, pp. 363-370.

[45] R.H. BAUGHMAN, A.A. ZAKHIDOV, W.A. DE HEER: Carbon nanotubes - the route toward applications. In: Science 297 (2002), p. 787-792.

[46] S. SINHA RAY, M. OKAMOTO: Polymer/layered silicate nanocomposites: a review from preparation to processing. In: Progress in Polymer Science 28 (2003), p. 1539-1641.

[47] A.K. GEIM, K.S. NOVOSELOV: The rise of graphene. In: Nat. Mater. 6 (2007), p. 183-191.

[48] P. EYERER, in: P. Eyerer (Ed.), Polymers - Opportunities and Risks I, Springer, Berlin, Heidelberg, 2010, pp. 1-17.

[49] B. TIEKE: Makromolekulare Chemie. WILEY-VCH Verlag GmbH & Co. KgaA: Weinheim, 2005.

[50] J. BALDYGA, W. ORCIUCH, L. MAKOWSKI, K. MALIK, G. OZCAN-TASKIN, W. EAGLES, G. PADRON: Dispersion of Nanoparticle Clusters in a Rotor-Stator Mixer. In: Ind. Eng. Chem. Res. 47 (2008), p. 3652-3663.

[51] C. SAUTER, M.A. EMIN, H.P. SCHUCHMANN, S. TAVMAN: Influence of hydrostatic pressure and sound amplitude on the ultrasound induced dispersion and de-agglomeration of nanoparticles. In: Ultrasonics Sonochemistry 15 (2008), p. 517-523.

[52] C. SCHILDE, I. KAMPEN, A. KWADE: Dispersion kinetics of nano-sized particles for different dispersing machines. In: Chemical Engineering Science 65 (2010), p. 3518-3527.

[53] L. XIE, C.D. RIELLY, G. ÖZCAN-TASKIN: Break-Up of Nanoparticle Agglomerates by Hydrodynamically Limited Processes. In: Journal of Dispersion Science & Technology 29 (2008), p. 573-579.

[54] J.B. PARSE, J.A. WERT: A Geometrical Description of Particle Distributions in Materials. In: Model. Simul. Mater. Sci. Eng. 1 (1993), p. 275-296.

[55] N. YANG, J. BOSELLI, I. SINCLAIR: Simulation and quantitative assessment of homogeneous and inhomogeneous particle distributions in particulate metal matrix composites. In: Journal of Microscopy 201 (2000), p. 189-200.

[56] Z.P. LUO, J.H. KOOT: Quantifying the dispersion of mixture microstructures. In: Journal of Microscopy 225 (2007), p. 118-125.

[57] P. GANGULY, W.J. POOLE: Characterization of reinforcement distribution inhomogeneity in metal matrix composites. In: Mater. Sci. Eng. A-Struct. Mater. Prop. Microstruct. Process. 332 (2002), p. 301-310.

[58] A. TSCHESCHEL: Räumliche Statistik zur Charakterisierung gefüllter Elastomere. TU Bergakademie Freiberg, Fakultät für Mathematik und Informatik, Dissertation. 2005

[59] S. KIRCHBERG, G. ZIEGMANN: Thermogravimetry and dynamic mechanical analysis of iron silicon particle filled polypropylene. In: J. Compos Mater. 43 (2009), p. 1323-1334.

[60] S. KIRCHBERG: Einfluss von Füllgrad und Geometrie weichmagnetischer Partikel auf die Verarbeitungs- und Materialeigenschaften ausgewählter Thermoplaste. TU Clausthal, 2011

R5

[61] J.W. CHO, D.R. PAUL: Nylon 6 nanocomposites by melt compounding. In: Polymer 42 (2001), p. 1083-1094.

[62] H.R. DENNIS, D.L. HUNTER, D. CHANG, S. KIM, J.L. WHITE, J.W. CHO, D.R. PAUL: Effect of melt processing conditions on the extent of exfoliation in organoclay-based nanocomposites. In: Polymer 42 (2001), p. 9513-9522.

[63] S. FILIPPI, E. MAMELI, C. MARAZZATO, P. MAGAGNINI: Comparison of solution-blending and melt-intercalation for the preparation of poly(ethylene-co-acrylic acid)/organoclay nanocomposites. In: Eur. Polym. J. 43 (2007), p. 1645-1659.

[64] Z. SHEN, G.P. SIMON, Y.-B. CHENG: Comparison of solution intercalation and melt intercalation of polymer-clay nanocomposites. In: Polymer 43 (2002), p. 4251-4260.

[65] D.A. SHIPP, Comprehensive Nanoscience and Technology, Academic Press, Amsterdam, pp. 265-276.

[66] S.I. YUN, D. ATTARD, V. LO, J. DAVIS, H. LI, B. LATELLA, F. TSVETKOV, H. NOORMAN, S. MORICCA, R. KNOTT, H. HANLEY, M. MORCOM, G.P. SIMON, G.E. GADD: Spray-dried microspheres as a route to clay/polymer nanocomposites. In: J. Appl. Polym. Sci. 108 (2008), p. 1550-1556.

[67] T. BANERT, U.A. PEUKER: Herstellung von hochgefüllten Fe3O4/PMMA-Nanokompositen im Sprühprozess. In: Chemie Ingenieur Technik 77 (2005), p. 224-227.

[68] Y. IMAI, A. TERAHARA, Y. HAKUTA, K. MATSUI, H. HAYASHI, N. UENO: Transparent poly(bisphenol A carbonate)-based nanocomposites with high refractive index nanoparticles. In: Eur. Polym. J. 45 (2009), p. 630-638.

[69] A.F. MANSOUR: Optical Properties of Fullerene/PMMA. In: International Journal of Polymeric Materials 54 (2005), p. 227-235.

[70] J. LIU, T. LIU, S. KUMAR: Effect of solvent solubility parameter on SWNT dispersion in PMMA. In: Polymer 46 (2005), p. 3419-3424.

[71] M. OEHRING, R. BORMANN: Nanocrystalline alloys prepared by mechanical alloying and ball milling. In: Materials Science and Engineering: A 134 (1991), p. 1330-1333.

[72] P.D. CASTRILLO, D. OLMOS, D.R. AMADOR, J. GONZÁLEZ-BENITO: Real dispersion of isolated fumed silica nanoparticles in highly filled PMMA prepared by high energy ball milling. In: Journal of Colloid and Interface Science 308 (2007), p. 318-324.

[73] P.D. CASTRILLO, D. OLMOS, J. GONZÁLEZ-BENITO: Novel polymer composites based on a mixture of preformed nanosilica-filled poly(methyl methacrylate) particles and a diepoxy/diamine thermoset system. In: J. Appl. Polym. Sci. 111 (2009), p. 2062-2070.

[74] J. GONZÁLEZ-BENITO, G. GONZÁLEZ-GAITANO: Interfacial Conformations and Molecular Structure of PMMA in PMMA/Silica Nanocomposites. Effect of High-Energy Ball Milling. In: Macromolecules 41 (2008), p. 4777-4785.

[75] J. GONZÁLEZ-BENITO, D. OLMOS: Efficient dispersion of nanoparticles in thermoplastic polymers. In: Society of Plastic Engineers - Plastics Research Online DOI:10.1002/spepro.002566 (2010).

[76] M. HEDAYATI, M. SALEHI, R. BAGHERI, M. PANJEPOUR, A. MAGHZIAN: Ball milling preparation and characterization of poly (ether ether ketone)/surface modified silica nanocomposite. In: Powder Technology 207 (2011), p. 296-303.

R6

[77] G. GORRASI, M. SARNO, A. DI BARTOLOMEO, D. SANNINO, P. CIAMBELLI, V. VITTORIA: Incorporation of carbon nanotubes into polyethylene by high energy ball milling: Morphology and physical properties. In: Journal of Polymer Science Part B: Polymer Physics 45 (2007), p. 597-606.

[78] Y. MORI, Y. YAMAOKA, S. HORI, T. TAGUCHI, K. TSUCHIYA, Preparation of Inorganic Nanoparticles - Polymer Nanocomposite through Hetero-Coagulation and its Flame Retardancy, WCPT6, Nuremberg, 2010.

[79] R. SENGUPTA, A. BANDYOPADHYAY, S. SABHARWAL, T.K. CHAKI, A.K. BHOWMICK: Polyamide-6,6/in situ silica hybrid nanocomposites by sol-gel technique: synthesis, characterization and properties. In: Polymer 46 (2005), p. 3343-3354.

[80] R.A. ALI-ZADE: Physical Characteristics of Polymer Magnetic Microspheres. In: Turkish Journal of Physics 28 (2004), p. 359-368.

[81] R.A. ALI-ZADE: Investigation of polymer magnetic microspheres. In: Colloids and Surfaces A: Physicochemical and Engineering Aspects 255 (2005), p. 111-117.

[82] Y.H. CHEN, Y.Y. LIU, R.H. LIN, F.S. YEN: Characterization of magnetic poly(methyl methacrylate) microspheres prepared by the modified suspension polymerization. In: J. Appl. Polym. Sci. 108 (2008), p. 583-590.

[83] A. KLAUSCH, H. ALTHUES, C. SCHRAGE, P. SIMON, A. SZATKOWSKI, M. BREDOL, D. ADAM, S. KASKEL: Preparation of luminescent ZnS:Cu nanoparticles for the functionalization of transparent acrylate polymers. In: J. Lumin. 130 (2010), p. 692-697.

[84] H. NOLTE, C. SCHILDE, A. KWADE: Production of highly loaded nanocomposites by dispersing nanoparticles in Epoxy Resin. In: Chemical Engineering and Technology 33 (2010), p. 1447-1455.

[85] L.P. RAMÍREZ, K. LANDFESTER: Magnetic Polystyrene Nanoparticles with a High Magnetite Content Obtained by Miniemulsion Processes. In: Macromolecular Chemistry and Physics 204 (2003), p. 22-31.

[86] S. KIRCHBERG, M. RUDOLPH, Abschlussbericht - DFG-Projekt ZI 648/16-1 und PE 1160/7-1 - Entwicklung der Prozesskette für die Synthese und Verarbeitung hochgefüllter Polymer-Nanopartikel-Komposite, TU Clausthal TU Bergakademie Freiberg, 2012, p. 64.

[87] B. HICKSTEIN, R. CECILIA, U. KUNZ, U.A. PEUKER, A. KIRSCHNING: Magnetisierbare Funktionspartikel für die festphasenunterstützte Synthese am Beispiel einer Transferhydrierungsreaktion. In: Chemie Ingenieur Technik 79 (2007), p. 2089-2097.

[88] S. MACHUNSKY, U.A. PEUKER: Liquid-Liquid Interfacial Transport of Nanoparticles. In: Physical Separation in Science and Engineering 2007 (2007), p. 7 pages.

[89] U. PEUKER, Verfahren zur Extraktion von Nano- bis Mikropartikel, in: D.P.-u. Markenamt (Ed.), Deutschland, 2005.

[90] U. PEUKER, Verfahren zur Herstellung eines Nanopartikel-Polymer-Komposits und daraus produzierter optisch transparenter Kunststoffpartikel, in: D.P.-u. Markenamt (Ed.), Deutschland, 2005.

[91] F. HABASHI: Principles of Extractive Metallurgy: General Principles. Gordon & Breach, 1969. - ISBN 9780677017709

R7

[92] U.A. PEUKER, O. THOMAS, T.J. HOBLEY, M. FRANZREB, S. BERENSMEIER, M. SCHÄFER, B. HICKSTEIN, M.C. FLICKINGER, Encyclopedia of Industrial Biotechnology, John Wiley & Sons, Inc., 2009.

[93] H. SCHUBERT: Mechanische Verfahrenstechnik. Dt. Verl. für Grundstoffind.: Leipzig, 1986.

[94] A. HORSCHIG: Konfektionierung von polymeren Mikropartikeln zur Verbesserung der Fließeigenschaften im Mikrospritzguss. TU Bergakademie Freiberg, Institut für Mechanische Verfahrenstechnik und Aufbereitungstechnik, Diplomarbeit. 2010

[95] L. ZHANG, R. HE, H.-C. GU: Oleic acid coating on the monodisperse magnetite nanoparticles. In: Applied Surface Science 253 (2006), p. 2611-2617.

[96] P. ROONASI, A. HOLMGREN: A Fourier transform infrared (FTIR) and thermogravimetric analysis (TGA) study of oleate adsorbed on magnetite nano-particle surface. In: Applied Surface Science 255 (2009), p. 5891-5895.

[97] S. AYYAPPAN, G. GNANAPRAKASH, G. PANNEERSELVAM, M.P. ANTONY, J. PHILIP: Effect of surfactant monolayer on reduction of Fe3O4 nanoparticles under vacuum. In: J. Phys. Chem. C 112 (2008), p. 18376-18383.

[98] Z.X. SUN, F.W. SU, W. FORSLING, P.O. SAMSKOG: Surface characteristics of magnetite in aqueous suspension. In: Journal of Colloid and Interface Science 197 (1998), p. 151-159.

[99] S. MACHUNSKY, P. GRIMM, H.J. SCHMID, U.A. PEUKER: Liquid-liquid phase transfer of magnetite nanoparticles. In: Colloid Surf. A-Physicochem. Eng. Asp. 348 (2009), p. 186-190.

[100] G. MÉRIGUET, E. DUBOIS, R. PERZYNSKI: Liquid-liquid phase-transfer of magnetic nanoparticles in organic solvents. In: Journal of Colloid and Interface Science 267 (2003), p. 78-85.

[101] A. SWAMI, A. KUMAR, M. SASTRY: Formation of Water-Dispersible Gold Nanoparticles Using a Technique Based on Surface-Bound Interdigitated Bilayers. In: Langmuir 19 (2003), p. 1168-1172.

[102] H. YAO, O. MOMOZAWA, T. HAMATANI, K. KIMURA: Stepwise Size-Selective Extraction of Carboxylate-Modified Gold Nanoparticles from an Aqueous Suspension into Toluene with Tetraoctylammonium Cations. In: Chem. Mater. 13 (2001), p. 4692-4697.

[103] H. ZHU, C. TAO, S. ZHENG, S. WU, J. LI: Effect of alkyl chain length on phase transfer of surfactant capped Au nanoparticles across the water/toluene interface. In: Colloids and Surfaces A: Physicochemical and Engineering Aspects 256 (2005), p. 17-20.

[104] W. CHENG, E. WANG: Size-Dependent Phase Transfer of Gold Nanoparticles from Water into Toluene by Tetraoctylammonium Cations: A Wholly Electrostatic Interaction. In: J. Phys. Chem. B 108 (2004), p. 24-26.

[105] A. PRAKASH, H. ZHU, C.J. JONES, D.N. BENOIT, A.Z. ELLSWORTH, E.L. BRYANT, V.L. COLVIN: Bilayers as phase transfer agents for nanocrystals prepared in nonpolar solvents. In: ACS Nano 3 (2009), p. 2139-2146.

[106] D.P. CISTOLA, J.A. HAMILTON, D. JACKSON, D.M. SMALL: Ionization and phase behavior of fatty acids in water: application of the Gibbs phase rule. In: Biochemistry 27 (1988), p. 1881-1888.

R8

[107] J.R. KANICKY, D.O. SHAH: Effect of Degree, Type, and Position of Unsaturation on the pKa of Long-Chain Fatty Acids. In: Journal of Colloid and Interface Science 256 (2002), p. 201-207.

[108] H. SCHUBERT: Aufbereitung fester Stoffe. Bd. 2. Sortierprozesse. Dt. Ver. für Grundstoffindustrie: Stuttgart, 1996. - ISBN 3-342-00555-6

[109] L.H. SPERLING: Introduction to Physical Polymer Science. John Wiley & Sons, Inc.: Hoboken, New Jersey, 2006.

[110] J.N. ISRAELACHVILI: Intermolecular & Surface Forces. Academic Press Ltd.: London, 1992. - ISBN 0-12-375181-0

[111] R.J. HUNTER: Foundations of Colloid Science. Oxford University Press: Oxford et al., 1986.

[112] H.-D. DÖRFLER: Grenzflächen und kolloid-disperse Systeme. Springer: Berlin et al., 2002.

[113] S. GYERGYEK, D. MAKOVEC, M. DROFENIK: Colloidal stability of oleic- and ricinoleic-acid-coated magnetic nanoparticles in organic solvents. In: Journal of Colloid and Interface Science 354 (2011), p. 498-505.

[114] T. BANERT, U.A. PEUKER: Synthesis of magnetic beads for bio-separation using the solution method. In: Chemical Engineering Communications 194 (2007), p. 707-719.

[115] S. AYYAPPAN, G. PANNEERSELVAM, M.P. ANTONY, N.V. RAMA RAO, N. THIRUMURUGAN, A. BHARATHI, J. PHILIP: Effect of initial particle size on phase transformation temperature of surfactant capped Fe3O4 nanoparticles. In: Journal of Applied Physics 109 (2011).

[116] I.V. CHERNYSHOVA, S. PONNURANGAM, P. SOMASUNDARAN: Adsorption of Fatty Acids on Iron (Hydr)oxides from Aqueous Solutions. In: Langmuir 27 (2011), p. 10007-10018.

[117] M. KLOKKENBURG, J. HILHORST, B.H. ERNÉ: Surface analysis of magnetite nanoparticles in cyclohexane solutions of oleic acid and oleylamine. In: Vib. Spectrosc. 43 (2007), p. 243-248.

[118] H.-J. BUTT, M. KAPPL: Surface and Interfacial Forces. WILEY-VCH Verlag GmbH & Co. KGaA: Weinheim, 2010.

[119] S. ALEXANDER: Adsorption of chain molecules with a polar head a scaling description. In: Journal de Physique 38 (1977), p. 983-987.

[120] V. KOROLEV, O. BALMASOVA, A. RAMAZANOVA: The sorption isotherms of oleic, linoleic, and linolenic acids from solutions in cyclohexane and heptane on magnetite. In: Russian Journal of Physical Chemistry A, Focus on Chemistry 83 (2009), p. 1018-1021.

[121] J.T.G. OVERBEEK: Interparticle forces in colloid science. In: Powder Technology 37 (1984), p. 195-208.

[122] C. BARRERA, A.P. HERRERA, C. RINALDI: Colloidal dispersions of monodisperse magnetite nanoparticles modified with poly(ethylene glycol). In: Journal of Colloid and Interface Science 329 (2009), p. 107-113.

[123] A.K. DOLAN, S.F. EDWARDS: Theory of stabilization of colloids by adsorbed polymer. In: Proc. R. Soc. London Ser. A-Math. Phys. Eng. Sci. 337 (1974), p. 509-516.

R9

[124] J.-C. LIU, J.-H. JEAN, C.-C. LI: Dispersion of Nano-Sized Alumina Powder in Non-Polar Solvents. In: J. Am. Ceram. Soc. 89 (2006), p. 882-887.

[125] A. VRIJ: Polymers at Interfaces and the Interactions in Colloidal Dispersions. In: Pure & Applied Chemistry 48 (1976), p. 471-483.

[126] T. COSGROVE: Colloid Science - Principles, Methods and Applications. Blackwell Publishing, 2005. - ISBN 978-1-4051-2673-1

[127] G. BREZESINSKI, H.-J. MÖGEL: Grenzflächen und Kolloide: physikalisch-chemische Grundlagen. Spektrum Akademischer Verlag GmbH: Heidelberg et al., 1993.

[128] D.H. NAPPER: Steric stabilization. In: Journal of Colloid and Interface Science 58 (1977), p. 390-407.

[129] E.L. MACKOR: A theoretical approach of the colloid-chemical stability of dispersions in hydrocarbons. In: Journal of Colloid Science 6 (1951), p. 492-495.

[130] E.L. MACKOR, J.H. VAN DER WAALS: The statistics of the adsorption of rod-shaped molecules in connection with the stability of certain colloidal dispersions. In: Journal of Colloid Science 7 (1952), p. 535-550.

[131] J.T.G. OVERBEEK: Colloid stability in aqueous and non-aqueous media. Introductory paper. In: Discussions of the Faraday Society 42 (1966), p. 7-13.

[132] P.G. DE GENNES: Polymer solutions near an interface. Adsorption and depletion layers. In: Macromolecules 14 (1981), p. 1637-1644.

[133] P.G. DE GENNES: Polymers at an interface; a simplified view. In: Advances in Colloid and Interface Science 27 (1987), p. 189-209.

[134] H.N.W. LEKKERKERKER, R. TUINIER: Colloids and the Depletion Interaction. Springer: Heidelberg, 2011. - ISBN 978-94-007-1222-5

[135] B. FAURE, G. SALAZAR-ALVAREZ, L. BERGSTRÖM: Hamaker Constants of Iron Oxide Nanoparticles. In: Langmuir 27 (2011), p. 8659-8664.

[136] H.-J. BUTT, B. CAPPELLA, M. KAPPL: Force measurements with the atomic force microscope: Technique, interpretation and applications. In: Surface Science Reports 59 (2005), p. 1-152.

[137] C.M. HANSEN: Hansen Solubility Parameters. A User’s Handbook. CRC Press: Boca Raton, 2007. - ISBN 0-8493-7248-8

[138] C.M. HANSEN: The Universality of the Solubility Parameter. In: Product R&D 8 (1969), p. 2-11.

[139] T. LINDVIG, M.L. MICHELSEN, G.M. KONTOGEORGIS: A Flory-Huggins model based on the Hansen solubility parameters. In: Fluid Phase Equilib. 203 (2002), p. 247-260.

[140] G. WYPYCH, Handbook of Solvents, ChemTec Publishing, 2001.

[141] O. CARP, L. PATRON, D. CULITA, P. BUDRUGEAC, M. FEDER, L. DIAMANDESCU: Thermal analysis of two types of dextran-coated magnetite. In: J. Therm. Anal. Calorim. (2009).

[142] L. FU, V.P. DRAVID, D.L. JOHNSON: Self-assembled (SA) bilayer molecular coating on magnetic nanoparticles. In: Applied Surface Science 181 (2001), p. 173-178.

R10

[143] T. GONG, D. YANG, J. HU, W. YANG, C. WANG, J.Q. LU: Preparation of monodispersed hybrid nanospheres with high magnetite content from uniform Fe3O4 clusters. In: Colloids and Surfaces A: Physicochemical and Engineering Aspects 339 (2009), p. 232-239.

[144] W. JIANG, Y. WU, B. HE, X. ZENG, K. LAI, Z. GU: Effect of sodium oleate as a buffer on the synthesis of superparamagnetic magnetite colloids. In: Journal of Colloid and Interface Science 347 (2010), p. 1-7.

[145] Q. LAN, C. LIU, F. YANG, S. LIU, J. XU, D. SUN: Synthesis of bilayer oleic acid-coated Fe3O4 nanoparticles and their application in pH-responsive Pickering emulsions. In: Journal of Colloid and Interface Science 310 (2007), p. 260-269.

[146] Y. SAHOO, H. PIZEM, T. FRIED, D. GOLODNITSKY, L. BURSTEIN, C.N. SUKENIK, G. MARKOVICH: Alkyl Phosphonate/Phosphate Coating on Magnetite Nanoparticles: A Comparison with Fatty Acids. In: Langmuir 17 (2001), p. 7907-7911.

[147] L. SHEN, P.E. LAIBINIS, T. ALAN HATTON: Bilayer Surfactant Stabilized Magnetic Fluids: Synthesis and Interactions at Interfaces. In: Langmuir 15 (1999), p. 447-453.

[148] W.R. VIALI, G.B. ALCANTARA, P.P.C. SARTORATTO, M.A.G. SOLER, E. MOSINIEWICZ-SZABLEWSKA, B. ANDRZEJEWSKI, P.C. MORAIS: Investigation of the molecular surface coating on the stability of insulating magnetic oils. In: J. Phys. Chem. C 114 (2011), p. 179-188.

[149] W. CAI, J. WAN: Facile synthesis of superparamagnetic magnetite nanoparticles in liquid polyols. In: Journal of Colloid and Interface Science 305 (2007), p. 366-370.

[150] H.-D. BELITZ, W. GROSCH: Food Chemistry. 180-215. Springer-Verlag: Berlin Heidelberg, 1999. - ISBN 3-540-64692-2

[151] G. LITWINIENKO: Autooxidation of Unsaturated Fatty Acids and Their Esters. In: J. Therm. Anal. Calorim. 65 (2001), p. 639-646.

[152] V. PÉREZ-DIESTE, O.M. CASTELLINI, J.N. CRAIN, M.A. ERIKSSON, A. KIRAKOSIAN, J.L. LIN, J.L. MCCHESNEY, F.J. HIMPSEL, C.T. BLACK, C.B. MURRAY: Thermal decomposition of surfactant coatings on Co and Ni nanocrystals. 5053-5055. AIP, 2003.

[153] W. CHEN, S. MORUP, M.F. HANSEN, T. BANERT, U.A. PEUKER: A Mössbauer study of the chemical stability of iron oxide nanoparticles in PMMA and PVB beads. In: Journal of Magnetism and Magnetic Materials 320 (2008), p. 2099-2105.

[154] J. BRANDRUP, E.H. IMMERGUT, E.A. GRULKE, A. ABE, D.R. BLOCH: Polymer Handbook (4th Edition). John Wiley & Sons, 1999. - ISBN 978-0-471-16628-3 (Hardcopy); 0-471-47963-5 (Database)

[155] D.H. NAPPER: Polymeric Stabilization of Colloidal Dispersions. Academic Press Inc.: New York, 1983.

[156] G. STROBL: The Physics of Polymers. Springer-Verlag: Berlin, Heidelberg, 2007.

[157] I. TERAOKA: Polymer Solutions - An Introduction to Physical Properties. John Wiley & Sons, Inc.: New York, 2002.

[158] J.W. GOODWIN: Colloids and interfaces with surfactants and polymers: an introduction. John Wiley & Sons Ltd: Chichester, 2004. - ISBN 0-470-94143-5

R11

[159] S.V. TAPAVICZA, J.M. PRAUSNITZ: Thermodynamik von Polymerlösungen: Eine Einführung. In: Chemie Ingenieur Technik 47 (1975), p. 552-562.

[160] G.J. FLEER: Polymers at interfaces and in colloidal dispersions. In: Advances in Colloid and Interface Science 159 (2010), p. 99-116.

[161] G.J. FLEER, R. TUINIER: Analytical phase diagrams for colloids and non-adsorbing polymer. In: Advances in Colloid and Interface Science 143 (2008), p. 1-47.

[162] S. LUZZATI, M. ADAM, M. DELSANTI: Size determination of a chain in a bad solvent by intensity light scattering comparison with a good solvent state. In: Polymer 27 (1986), p. 834-838.

[163] C.M. KOK, A. RUDIN: Relationship between the Hydrodynamic Radius and the Radius of Gyration of a Polymer in Solution. In: Makromolekulare Chemie-Rapid Communications 2 (1981), p. 655-659.

[164] C.M. HANSEN: Polymer additives and solubility parameters. In: Progress in Organic Coatings 51 (2004), p. 109-112.

[165] S. ASAKURA, F. OOSAWA: On Interaction between Two Bodies Immersed in a Solution of Macromolecules. In: The Journal of Chemical Physics 22 (1954), p. 1255-1256.

[166] S. ASAKURA, F. OOSAWA: Interaction Between Particles Suspended in Solutions of Macromolecules. In: Journal of Polymer Science 33 (1958), p. 183-192.

[167] J.-F. BERRET, K. YOKOTA, M. MORVAN, R. SCHWEINS: Polymer-Nanoparticle Complexes: From Dilute Solution to Solid State. In: The Journal of Physical Chemistry B 110 (2006), p. 19140-19146.

[168] H. DE HEK, A. VRIJ: Interactions in mixtures of colloidal silica spheres and polystyrene molecules in cyclohexane : I. Phase separations. In: Journal of Colloid and Interface Science 84 (1981), p. 409-422.

[169] N. DUTTA, D. GREEN: Nanoparticle stability in semidilute and concentrated polymer solutions. In: Langmuir 24 (2008), p. 5260-5269.

[170] R.I. FEIGIN, D.H. NAPPER: Depletion stabilization and depletion flocculation. In: Journal of Colloid and Interface Science 75 (1980), p. 525-541.

[171] A.A. LOUIS, P.G. BOLHUIS, E.J. MEIJER, J.P. HANSEN: Polymer induced depletion potentials in polymer-colloid mixtures. In: The Journal of Chemical Physics 117 (2002), p. 1893-1907.

[172] M.E. MACKAY, A. TUTEJA, P.M. DUXBURY, C.J. HAWKER, B. VAN HORN, Z.B. GUAN, G.H. CHEN, R.S. KRISHNAN: General strategies for nanoparticle dispersion. In: Science 311 (2006), p. 1740-1743.

[173] J.A. ZHOU, J.S. VAN DUIJNEVELDT, B. VINCENT: The Phase Behavior of Dispersions of Silica Particles in Mixtures of Polystyrene and Dimethylformamide. In: Langmuir 26 (2010), p. 9397-9402.

[174] C. COWELL, R. LI-IN-ON, B. VINCENT: Reversible flocculation of sterically-stabilised dispersions. In: Journal of the Chemical Society, Faraday Transactions 1: Physical Chemistry in Condensed Phases 74 (1978), p. 337-347.

R12

[175] R. SHENHAR, T.B. NORSTEN, V.M. ROTELLO: Polymer-Mediated Nanoparticle Assembly: Structural Control and Applications. In: Advanced Materials 17 (2005), p. 657-669.

[176] R. TUINIER, P.A. SMITH, W.C.K. POON, S.U. EGELHAAF, D.G.A.L. AARTS, H.N.W. LEKKERKERKER, G.J. FLEER: Phase diagram for a mixture of colloids and polymers with equal size. In: EPL (Europhysics Letters) (2008), p. 68002.

[177] H.N.W. LEKKERKERKER, W.C.K. POON, P.N. PUSEY, A. STROOBANTS, P.B. WARREN: Phase Behaviour of Colloid + Polymer Mixtures. In: EPL (Europhysics Letters) 20 (1992), p. 559.

[178] A.P. GAST, L. LEIBLER: Effect of polymer solutions on sterically stabilized suspensions. In: The Journal of Physical Chemistry 89 (1985), p. 3947-3949.

[179] B. VINCENT, P.F. LUCKHAM, F.A. WAITE: The effect of free polymer on the stability of sterically stabilized dispersions. In: Journal of Colloid and Interface Science 73 (1980), p. 508-521.

[180] S.M. ILETT, A. ORROCK, W.C.K. POON, P.N. PUSEY: Phase behavior of a model colloid-polymer mixture. In: Physical Review E 51 (1995), p. 1344-1352.

[181] M. SCHMIDT, A.R. DENTON: Demixing of colloid-polymer mixtures in poor solvents. In: Physical Review E - Statistical, Nonlinear, and Soft Matter Physics 65 (2002), p. 061410/061411-061410/061416.

[182] M.S. ROMERO-CANO, A.M. PUERTAS: Phase behaviour of a model colloid-polymer mixture at low colloid concentration. In: Soft Matter 4 (2008), p. 1242-1248.

[183] M.S. ROMERO-CANO, A.M. PUERTAS, F.J. DE LAS NIEVES: Colloidal aggregation under steric interactions: Simulation and experiments. In: J. Chem. Phys. 112 (2000), p. 8654-8659.

[184] E.R. SOULÉ, J. BORRAJO, R.J.J. WILLIAMS: Thermodynamic Analysis of a Polymerization-Induced Phase Separation in Nanoparticle-Monomer-Polymer Blends. In: Macromolecules 40 (2007), p. 8082-8086.

[185] A. PFAU, A. JANKE, W. HECKMANN: Determination of the bulk structure of technical multiphase polymer systems with AFM: comparative AFM and TEM investigation. In: Surface and Interface Analysis 27 (1999), p. 410-417.

[186] R.P. SEAR, D. FRENKEL: Phase behavior of colloid plus polydisperse polymer mixtures. In: Physical Review E - Statistical Physics, Plasmas, Fluids, and Related Interdisciplinary Topics 55 (1997), p. 1677-1681.

[187] Y. LALATONNE, J. RICHARDI, M.P. PILENI: Van der Waals versus dipolar forces controlling mesoscopic organizations of magnetic nanocrystals. In: Nat Mater 3 (2004), p. 121-125.

[188] Y.-D. YAN, J.L. BURNS, G.J. JAMESON, S. BIGGS: The structure and strength of depletion force induced particle aggregates. In: Chemical Engineering Journal 80 (2000), p. 23-30.

[189] J.L. BURNS, Y.-D. YAN, G.J. JAMESON, S. BIGGS: Relationship between interaction forces and the structural compactness of depletion flocculated colloids. In: Colloids and Surfaces A: Physicochemical and Engineering Aspects 162 (2000), p. 265-277.

R13

[190] E.H.A. DE HOOG, W.K. KEGEL, A. VAN BLAADEREN, H.N.W. LEKKERKERKER: Direct observation of crystallization and aggregation in a phase-separating colloid-polymer suspension. In: Physical Review E - Statistical, Nonlinear, and Soft Matter Physics 64 (2001), p. 214071-214079.

[191] H.C. HAMAKER: The London--van der Waals attraction between spherical particles. In: Physica 4 (1937), p. 1058-1072.

[192] E. RUCKENSTEIN, D.C. PRIEVE: Adsorption and desorption of particles and their chromatographic separation. In: AIChE Journal 22 (1976), p. 276-283.

[193] V. TIRTAATMADJA, H.G. MCKINLEY, J.J. COOPER-WHITE: Drop formation and breakup of low viscosity elastic fluids: Effects of molecular weight and concentration. In: Phys. Fluids 18 (2006).

[194] H.-I. HSIANG, C.-C. CHEN, J.-Y. TSAI: Dispersion of nonaqueous Co2Z ferrite powders with titanate coupling agent and poly(vinyl butyral). In: Applied Surface Science 245 (2005), p. 252-259.

[195] J.H. JEAN, S.F. YEH, C.J. CHEN: Adsorption of poly(vinyl butyral) in nonaqueous ferrite suspensions. In: J. Mater. Res. 12 (1997), p. 1062-1068.

[196] E. POTAPOVA: Adsorption of surfactants and polymers on iron oxides: implications for flotation and agglomerations of iron ore. Lulea University of Technology, Hjalmar Lundbohm Research Centre, 2011

[197] P. ROONASI: Adsorption and surface reaction properties of synthesized magnetite Nano-particles. Lulea University of Technology, Hjalmar Lundbohm Research Centre, 2007

[198] R. FINSY: Particle Sizing by Quasi-Elastic Light-Scattering. In: Advances in Colloid and Interface Science 52 (1994), p. 79-143.

[199] M.G. RASTEIRO, C.C. LEMOS, A. VASQUEZ: Nanoparticle characterization by PCS: The analysis of bimodal distributions. In: Part. Sci. Technol. 26 (2008), p. 413-437.

[200] B.N. JANG, C.A. WILKIE: A TGA/FTIR and mass spectral study on the thermal degradation of bisphenol A polycarbonate. In: Polym. Degrad. Stab. 86 (2004), p. 419-430.

[201] K. HIGASHITANI, K. IIMURA, H. SANDA: Simulation of deformation and breakup of large aggregates in flows of viscous fluids. In: Chemical Engineering Science 56 (2001), p. 2927-2938.

[202] C. SCHILDE, H. NOLTE, C. ARLT, A. KWADE: Effect of fluid-particle-interactions on dispersing nano-particles in epoxy resins using stirred-media-mills and three-roll-mills. In: Compos. Sci. Technol. 70 (2009), p. 657-663.

[203] F. MÜLLER, W. PEUKERT, R. POLKE, F. STENGER: Dispersing nanoparticles in liquids. In: International Journal of Mineral Processing 74 (2004), p. S31-S41.

[204] J. BALDYGA, L. MAKOWSKI, W. ORCIUCH, C. SAUTER, H.P. SCHUCHMANN: Deagglomeration processes in high-shear devices. In: Chemical Engineering Research and Design In Press, Corrected Proof (2008).

[205] H. NOLTE, A. KWADE: Effect of kneading process conditions on the particle fineness and the polydispersity in nanocomposites. In: Compos. Sci. Technol. 72 (2012), p. 939-947.

R14

[206] M. POHL, S. HOGEKAMP, N.Q. HOFFMANN, H.P. SCHUCHMANN: Dispergieren und Desagglomerieren von Nanopartikeln mit Ultraschall. In: Chemie Ingenieur Technik 76 (2004), p. 392-396.

[207] A. KRUEGER, F. KATAOKA, M. OZAWA, T. FUJINO, Y. SUZUKI, A.E. ALEKSENSKII, A.Y. VUL, E. OSAWA: Unusually tight aggregation in detonation nanodiamond: Identification and disintegration. In: Carbon 43 (2005), p. 1722-1730.

[208] M. OZAWA, M. INAGUMA, M. TAKAHASHI, F. KATAOKA, A. KRÜGER: Preparation and Behavior of Brownish, Clear Nanodiamond Colloids. In: Advanced Materials 19 (2007), p. 1201-1206.

[209] J. OPITZ, M. MKANDAWIRE, M. SORGE, N. ROSE, M. RUDOLPH, P. KRUEGER, I. HANNSTEIN, V.A. LAPINA, D. APPELHANS, W. POMPE, J. SCHREIBER, G. ROEDEL, Green fluorescent nanodiamond conjugates and their possible applications for biosensing, Proc. SPIE-Int. Soc. Opt. Eng., San Diego, CA, 2010.

[210] M. RUDOLPH: Wässrige Suspensionen von Detonations-Nanodiamant - Untersuchungen zur mechanischen Dispergierung in einer Planetenmühle und mit Ultraschall zur Herstellung eines stabilen Dispersionskolloids. TU Bergakademie Freiberg, Institut für Mechanische Verfahrenstechnik und Aufbereitungstechnik, Fraunhofer IZFP-D, Diploma Thesis. 2008

[211] T. MOKHTARI, A. CHAKRABARTI, C.M. SORENSEN, C.-Y. CHENG, D. VIGIL: The effect of shear on colloidal aggregation and gelation studied using small-angle light scattering. In: Journal of Colloid and Interface Science 327 (2008), p. 216-223.

[212] C. EISERMANN, M.R. MALLEMBAKAM, C. DAMM, W. PEUKERT, S. BREITUNG-FAES, A. KWADE: Polymeric stabilization of fused corundum during nanogrinding in stirred media mills. In: Powder Technology 217 (2012), p. 315-324.

[213] J.W.T. LICHTENBELT, H.J.M.C. RAS, P.H. WIERSEMA: Turbidity of coagulating lyophobic sols. In: Journal of Colloid and Interface Science 46 (1974), p. 522-527.

[214] T. BREMERSTEIN: Ethyl Acetate as a Substitute for Dichloromethane in the Solution and Spray Drying Process for the Synthesis of Highly Filled Poly(methyl methacrylate)-Nanomagnetite-Composites. TU Bergakademie Freiberg, Institut für Mechanische Verfahrenstechnik und Aufbereitungstechnik, Diplomarbeit. 2011

[215] T. BREMERSTEIN, M. RUDOLPH, U.A. PEUKER, Effect of solvent exchange on the stability of sterically functionalized magnetite nanoparticles in poly(methylmethacrylate)-solutions and resultant spray dried composites, to be submitted Chemical Engineering Science, 2012.

[216] R. HARTMANN: Phase Transfer of magnetite nanoparticles to dichloromethane, styrene and methyl methacrylate and colloidal stability. TU Bergakademie Freiberg, Institut für Mechanische Verfahrenstechnik und Aufbereitungstechnik, 2012

[217] R. HARTMANN, M. RUDOLPH, U.A. PEUKER, Phase transfer of agglomerated nanoparticles for the synthesis of polymer nanoparticle composites with the solution and spray drying process, International Forum-Competition of Young Researchers, Gornyi Institute, St. Peterburg, Russia, 2012.

[218] K. MASTERS: Spray drying handbook. Longman Scientific & Technical, 1991. - ISBN 9780470217436

R15

[219] K. OKUYAMA, M. ABDULLAH, I. WULED LENGGORO, F. ISKANDAR: Preparation of functional nanostructured particles by spray drying. In: Advanced Powder Technology 17 (2006), p. 587-611.

[220] B. MULHEM, G.N. SCHULTE, U. FRITSCHING: Solid-liquid separation in suspension atomization. In: Chemical Engineering Science 61 (2006), p. 2582-2589.

[221] F. ISKANDAR, L. GRADON, K. OKUYAMA: Control of the morphology of nanostructured particles prepared by the spray drying of a nanoparticle sol. In: Journal of Colloid and Interface Science 265 (2003), p. 296-303.

[222] P. SCHMIDT, P. WALZEL: Zerstäuben von Flüssigkeiten. In: Physik in unserer Zeit 15 (1984), p. 113-120.

[223] A.H. LEFEBVRE: Atomization and Sprays. Hemisphere Publishing Corporation, 1989. - ISBN 9780891166030

[224] N. ASHGRIZ: Handbook of Atomization and Sprays: Theory and Applications. Springer, 2011. - ISBN 9781441972637

[225] J. SLOTH, K. JORGENSEN, P. BACH, A.D. JENSEN, S. KIIL, K. DAM-JOHANSEN: Spray Drying of Suspensions for Pharma and Bio Products: Drying Kinetics and Morphology. In: Ind. Eng. Chem. Res. 48 (2009), p. 3657-3664.

[226] R. VEHRING: Pharmaceutical particle engineering via spray drying. In: Pharm. Res. 25 (2008), p. 999-1022.

[227] D.E. WALTON, C.J. MUMFORD: The Morphology of Spray-Dried Particles: The Effect of Process Variables upon the Morphology of Spray-Dried Particles. In: Chemical Engineering Research and Design 77 (1999), p. 442-460.

[228] D.E. WALTON: The morphology of spray-dried particles a qualitative view. In: Drying Technology 18 (2000), p. 1943-1986.

[229] X.D. ZHOU, S.C. ZHANG, W. HUEBNER, P.D. OWNBY, H. GU: Effect of the solvent on the particle morphology of spray dried PMMA. In: Journal of Materials Science 36 (2001), p. 3759-3768.

[230] C.S. HANDSCOMB, M. KRAFT, A.E. BAYLY: A new model for the drying of droplets containing suspended solids after shell formation. In: Chemical Engineering Science 64 (2009), p. 228-246.

[231] M. HECKELE, W.K. SCHOMBURG: Review on micro molding of thermoplastic polymers. In: J. Micromech. Microeng. 14 (2004), p. R1-R14.

[232] J. GIBOZ, T. COPPONNEX, P. MÉLÉ: Microinjection molding of thermoplastic polymers: a review. In: J. Micromech. Microeng. 17 (2007), p. R96.

[233] V. PIOTTER, W. BAUER, T. BENZLER, A. EMDE: Injection molding of components for microsystems. In: Microsystem Technologies 7 (2001), p. 99-102.

[234] H. NOLTE, C. SCHILDE, A. KWADE: Determination of particle size distributions and the degree of dispersion in nanocomposites. In: Compos. Sci. Technol. 72 (2012), p. 948-958.

[235] U. WAGENKNECHT, B. KRETZSCHMAR, P. PÖTSCHKE, F.R. COSTA, S. PEGEL, K.W. STÖCKELHUBER, G. HEINRICH: Polymere Nanokomposite mit anorganischen Funktionsfüllstoffen. In: Chemie Ingenieur Technik 80 (2008), p. 1683-1699.

R16

[236] Z.P. LUO, J.H. KOO: Quantitative study of the dispersion degree in carbon nanofiber/polymer and carbon nanotube/polymer nanocomposites. In: Materials Letters 62 (2008), p. 3493-3496.

[237] J.-C. LIN, W. HEESCHEN, J. REFFNER, J. HOOK: Three-Dimensional Characterization of Pigment Dispersion in Dried Paint Films Using Focused Ion Beam-Scanning Electron Microscopy. In: Microsc. Microanal. 18 (2012), p. 266-271.

[238] T. FERREIRA, W. RASBAND: ImageJ User Guide - IJ 1.45m. imagej.nih.gov/ij/docs/guide/, 2011.

[239] R. KLEIN, H. NOLTEMEIER, Springer Berlin / Heidelberg, 1988, pp. 148-157.

[240] K. BAUER, C. ELOO, U. PEUKER: Process optimized and environmentally compliant nanoparticle-wax formulations as additives for thermoplastic polymers. In: Chemie Ingenieur Technik 84 (2012), p. 286-294.

[241] T. GLASKOVA, M. ZARRELLI, A. BORISOVA, K. TIMCHENKO, A. ANISKEVICH, M. GIORDANO: Method of quantitative analysis of filler dispersion in composite systems with spherical inclusions. In: Compos. Sci. Technol. 71, p. 1543-1549.

[242] B.M. TYSON, R.K. ABU AL-RUB, A. YAZDANBAKHSH, Z. GRASLEY: A quantitative method for analyzing the dispersion and agglomeration of nano-particles in composite materials. In: Composites Part B: Engineering 42 (2011), p. 1395-1403.

[243] M. FATNASSI, C. TOURNÉ-PÉTEILH, T. CACCIAGUERRA, P. DIEUDONNÉ, J.M. DEVOISSELLE, B. ALONSO: Tuning nanophase separation and drug delivery kinetics through spray drying and self-assembly. In: New J. Chem. 34 (2010), p. 607-610.

[244] J.O.H. SHAM, Y. ZHANG, W.H. FINLAY, W.H. ROA, R. LÖBENBERG: Formulation and characterization of spray-dried powders containing nanoparticles for aerosol delivery to the lung. In: International Journal of Pharmaceutics 269 (2004), p. 457-467.

[245] K. TOMODA, T. OHKOSHI, T. NAKAJIMA, K. MAKINO: Preparation and properties of inhalable nanocomposite particles: Effects of the size, weight ratio of the primary nanoparticles in nanocomposite particles and temperature at a spray-dryer inlet upon properties of nanocomposite particles. In: Colloids and Surfaces B: Biointerfaces 64 (2008), p. 70-76.

[246] S.I. YUN, V. LO, J. NOORMAN, J. DAVIS, R.A. RUSSELL, P.J. HOLDEN, G.E. GADD: Morphology of composite particles of single wall carbon nanotubes/biodegradable polyhydroxyalkanoates prepared by spray drying. In: Polym. Bull. 64 (2010), p. 99-106.

[247] J.E. MARK, Polymer Data Handbook (2nd Edition), Oxford University Press, 2009.

[248] E. STEFANIS, I. TSIVINTZELIS, C. PANAYIOTOU: The partial solubility parameters: An equation-of-state approach. In: Fluid Phase Equilib. 240 (2006), p. 144-154.

[249] Z. DOBKOWSKI: Dependence of polymer specific volume on molecular characteristics. In: Eur. Polym. J. 20 (1984), p. 399-403.

[250] K.M. ZIMMER, A. LINKE, D.W. HEERMANN, J. BATOULIS, T. BÜRGER: Ellipsoidal potential parameterization for bisphenol-A polycarbonate. In: Macromol. Theory Simul. 5 (1996), p. 1065-1074.

[251] K. AKASHI, Y. NAKAMURA, T. NORISUYE: Small-angle X-ray scattering from bisphenol A polycarbonate in tetrahydrofuran. Molecular characteristics and excluded-volume effects. In: Polymer 39 (1998), p. 5209-5213.

R17

[252] V.-G.V.U. CHEMIEINGENIEURWESEN: VDI-Wärmeatlas. Springer, 2006. - ISBN 9783540255031

[253] N. STANLEY-WOOD, R.W. LINES: Particle Size Analysis. Royal Society of Chemistry, 1992. - ISBN 9780851864877

[254] M. STIEß, S. RIPPERGER: Mechanische Verfahrenstechnik- Partikeltechnologie 1. Springer, 2008. - ISBN 9783540325512

[255] G.E.S. SALAS: Sedimentationsverhalten von Submikrometerpartikeln in wässrigen Suspensionen. TU Dresden, dissertation. 2007

[256] K. KANAYA, S. OKAYAMA: Penetration and energy-loss theory of electrons in solid targets. In: J. Phys. D: Appl. Phys. 5 (1972), p. 43.

[257] A. BENTABET, N. BOUARISSA: Backscattering coefficients and mean penetration depths of 1–4 keV electron scattering in solids. In: Appl. Phys. A: Mater. Sci. Process. 88 (2007), p. 353-358.

[258] M. DAPOR: Mean energy and depth of penetration of electrons backscattered by solid targets. In: Applied Surface Science 70–71, Part 1 (1993), p. 327-331.

[259] L. SPIEß: Moderne Röntgenbeugung. Teubner, 2005. - ISBN 9783519005223

[260] G.K. WILLIAMSON, W.H. HALL: X-ray line broadening from filed aluminium and wolfram. In: Acta Metallurgica 1 (1953), p. 22-31.

[261] S. MALYNYCH, G. CHUMANOV: Extinction spectra of quasi-spherical silver sub-micron particles. In: Journal of Quantitative Spectroscopy and Radiative Transfer 106 (2007), p. 297-303.

[262] C. BERNHARDT, E. REINSCH, K. HUSEMANN: The influence of suspension properties on ultra-fine grinding in stirred ball mills. In: Powder Technology 105 (1999), p. 357-361.

[263] R. STADLER, R. POLKE, J. SCHWEDES, F. VOCK: Wet grinding in agitated ball mills. In: Nassmahlung in Ruehrwerksmuehlen 62 (1990), p. 907-915.

[264] C. KNIEKE, C. STEINBORN, S. ROMEIS, W. PEUKERT, S. BREITUNG-FAES, A. KWADE: Nanoparticle production with stirred-media mills: Opportunities and limits. In: Chemical Engineering and Technology 33 (2010), p. 1401-1411.

[265] F. BATH: Planetary ball mills for the manufacture of nanoparticles. In: Chemie Ingenieur Technik 77 (2005), p. 1276-1278.

[266] S. ROSENKRANZ, S. BREITUNG-FAES, A. KWADE: Experimental investigations and modelling of the ball motion in planetary ball mills. In: Powder Technology 212 (2011), p. 224-230.

[267] W. LAUTERBORN, C.-D. OHL: Cavitation bubble dynamics. In: Ultrasonics Sonochemistry 4 (1997), p. 65-75.

[268] K.S. SUSLICK: Sonochemistry. In: Science 247 (1990), p. 1439-1445.

[269] K.S. SUSLICK, G.J. PRICE: Applications of ultrasound to materials chemistry. 295-326. Annual Reviews Inc, 1999. - ISBN 00846600 (ISSN)

[270] F. WOOTEN: Optical Properties of Solids. Academic Press: New York et al., 1972.

R18

[271] A. SCHLEGEL, S.F. ALVARADO, P. WACHTER: Optical properties of magnetite (Fe3O4). In: Journal of Physics C: Solid State Physics 12 (1979), p. 1157-1164.

View publication statsView publication stats