The dynamic behavior of short polymers build
from colloids
THESIS
submitted in partial fulfillment of the requirements for the degree of
MASTER OF SCIENCE in
PHYSICS
Author : Loes Huijnen
Student ID : 1383566
Supervisor : MSc. Ruben Verweij
Dr. Daniela Kraft
2ndcorrector : Dr. Luca Giomi
The dynamic behavior of short polymers
build from colloids
Loes Huijnen
Huygens-Kamerlingh Onnes Laboratory, Leiden University P.O. Box 9500, 2300 RA Leiden, The Netherlands
November 11, 2018
Abstract
An important goal in soft matter physics is to steer microscale self-assembly processes. Here we show linear structures made of colloids and the energy landscape that describes the angular mobility. It was done by functionalizing isotropic and anisotropic colloids with a lipid bilayer and insert DNA linkers that have a specific binding group. The DNA linkers are fully mobile along the particle surface and colloids functionalized with DNA linkers are able to form flexible polymers. Specifically, we looked at polymers consisting of four monomers:
tetramers and found very mobile clusters that had an averaged joint mobility of 154±3 deg2/s. In the energy landscape that we experimentally and theoretically found, we can conclude the preferred angles are 180/180 degrees. For polymers made of two dumbbell
particles we found a mobile bilayer, but no mobile clusters were found. Our tunable tetramers could be the design for a model of controlled self-assembly in even larger structures
Contents
1 Introduction 7 1.1 Colloidal systems 7 1.2 Stability 8 1.3 Colloidal self-assembly 10 1.4 Approach 11 2 Experimental methods 13 2.1 Reagents 132.2 DNA functionalization of silica colloids 15
2.3 Silica coating of polystyrene dumbbells 15
2.4 Cover glass treatment 16
2.5 Imaging 17
2.6 Analysis 18
3 Results and discussion 23
3.1 Properties of isotropic colloids 23
3.1.1 Diffusion of colloids 23
3.1.2 Cluster forming 26
3.1.3 Cluster properties 29
3.2 Silica coating on anisotropic colloids 31
3.2.1 Silica coating of PS particles 31
3.2.2 More silica layers 34
3.3 Properties of anisotropic colloids 37
Chapter
1
Introduction
1.1
Colloidal systems
Numerous products that we use, drink or eat in the daily life contain colloids. Colloids can be described as consisting of a dispersed phase distributed uniformly in a dispersion medium. Different states in which the dispersed phase and the dispersion medium exist are gas, liquid or solid [1]. One of the possibilities is if both the dispersed phase and the dispersion medium are liquid. We encounter this example in milk, butter and mayonnaise where the dispersed phase are fat-globules in a aqueous medium. When the dispersed phase are liquid or solid particles in a gaseous medium, we find examples in fog, mist and smoke. Also toothpaste, coffee, make-up, paint and chocolate are examples of colloids, see figure 1.1. Here the nature of the colloidal system are solid particles in a liquid medium.
Figure 1.1: Examples of materials that contain colloids. Solid particles in a liquid medium. From
left to right: paint, chocolate, toothpaste and coffee.
In some of these examples, colloids are naturally present. But in toothpaste, colloids are added by a manufacturer for a practical use, since they make the viscosity of the medium higher. In paint, colloids are essential. When the medium dries, the pigmented colloids will give the color in a painting or on a wall. Industrial use of colloids stretches even further, magnetic colloids can be used as selective probes and for example detect hazardous heavy metal ions in various water samples, such as surface water, drinking water, industrial waste, and food [2]. In the biomedical field colloids are used for its potential of the delivery of drugs and other bio active materials [3].
1. Introduction A typical property of colloidal particles is that they have a diameter of 1 nanometer to 1 mi-crometer. Particles in this size range show Brownian motion that is produced by thermal fluc-tuations. Because of this Brownian motion particles can diffuse. The rate at which particles diffuse is given by the diffusion coefficient and is described by the formula:
D= kBT
6πηR (1.1)
Where kB is the Boltzmann constant, T is the temperature, η is the dynamic viscosity of the
medium and R is the radius of the dispersed particle. The larger the particles, the less they will be influenced by these fluctuations.
1.2
Stability
Particles driven by diffusion typically are in a stable state. Here we explain why a colloidal system is stable and under what conditions the particles aggregate or sediment. In industrial uses the stability of a colloidal dispersion is crucial. If colloids in toothpaste, paint or probes would cluster, the objects are useless. There are attractive and repulsive forces working on the colloidal particles and there are several factors which may influence the attraction and repul-sion. Firstly, van der Waals-London forces play an important role. They are due to interaction between two or more dipoles. Even if the particles do not have a permanent dipole, fluctua-tions of the electron density give rise to a temporary dipole in a particle. This in turn induces dipoles in neighbor particles. The interaction energy due to the van der Waals force between two spheres of radius R, that have a separation distance r, is given by:
U(r, R)vdW = −AR
12r (1.2)
Here A is the Hamaker constant, named after Hamaker, who first approximated the interaction energy by van der Waals forces [4]. Note that the van der Waals interaction energy is negative due to its attractive nature and therefore influences aggregation.
Second, electrostatic interactions have a large influence on the stability and have a repulsive or attractive nature, dependent on the charge. Colloidal particles often carry an electrical charge causing free ions of the dispersion medium to balance out the colloidal charge. This results in an electrical double layer around the colloid. In figure 1.2a, the electrostatic interaction po-tential is shown. The popo-tential near the particle surface is maximal, decreasing with distance. The value of the potential given at the distance of the first electrical layer, the Stern layer, is called the Stern potential. At further distance, the potential is less affected by the electrical double layer. This point is the slipping plane and gives rise to the Zeta potential. The interac-tion energy driven by the electrostatic double layer of two spheres with radius R at separainterac-tion distance r is given by:
1.2 Stability
U(r, R)el = −ZR exp(−κr)
2 (1.3)
where Z is a constant, 1/κ is the Debye length which describes the characteristic length of the diffuse electric double layer. Next to electrostatic forces, steric forces also increase stability. This sort of stabilization is a result of entropic forces of polymers on the particle surface. They want to maximize their number of configurations and therefore cannot overlay polymers of another particle, which results in repulsion between particles [5].
(a)Double layer giving rise to electrostatic potential
(b)Charge stabilization (c)Steric stabilization
Figure 1.2: Colloidal stability. In figure 1.2a a particle with an electrical double layer and its effect on
the electrostatic potential is shown. The potential near the particle surface is maximal, decreasing with distance. In figure 1.2b and1.2c two stabilization methods are shown. Charge stabilization is caused the electrical double layer and steric stabilization is a result of entropic forces of polymers on the particle surface. Figures adopted from [6]
Also, hydrophobic forces may play a role in the stability of colloidal particles. Hydrophobic forces are attractive of nature when two hydrophobic elements are suspended in an aqueous medium. This type of attraction is very strong and long-ranged [7].
1. Introduction given by:
U(r)net =U(r)att+U(r)rep (1.4)
where U(r)attis the attractive interaction energy and U(r)repis the repulsive interaction energy
at distance r. The net energy defines if a colloidal dispersion is stable when U(r)netis positive or aggregated when U(r)net is negative. Typically the attraction energy is given by the van der Waals interaction force and the repulsion energy by the electrostatic interaction force. The interaction potentials are illustrated in figure 1.3. Figure 1.3a shows that the net interaction potential has a local minima at large distances. Because of the energy barrier at intermediate distances, in general, the system is stable. In figure 1.3b, the net interaction potential has its minimum at short separation distance leading to aggregation.
(a) Stable disperion (b) Aggregation
Figure 1.3: Interaction potentials. The van der Waals and the electric double layer interaction
potential result in a net potential, leading to a stable dispersion (a) or to aggregation (b)
1.3
Colloidal self-assembly
Due to a combination of the forces acting on colloidal particles and the particles diffusive be-havior, we can introduce specific attractions between colloidal particles. This phenomenon is called self-assmembly and can be controlled externally by tuning physical properties. Materi-als made of colloidal building blocks that change their structures to an external sensor, such as light, temperature or pH, may be an ultimate result of controllable self-assembly.
A promising method to achieve self-assembly is by functionalization of specific binding groups to colloidal particles. One way to achieve this is with DNA linkers. DNA consists of a sequence of molecules called nucleobases and together they form a strand. There are four types of nu-cleobases, adanine (A), thymine (T), cytosine (C) and guanine (G), of which A can bind to T and C to G. Followingly, the DNA strand can bind to its complementary strand and form dou-ble stranded DNA. The DNA linkers that can be used for self-assembly consist of a doudou-ble
1.4 Approach
stranded spacer and a single stranded ”sticky end” that serves as the specific binding group. DNA as specific binding groups leads to controllable aggregation and many researchers have been able to attach linkers to the colloidal particles [8–10]. Kim et al. shows that with this tech-nique, they can made ordered crystals made from colloids and the bonds created by the DNA linkers are reversible by heating above the melting temperature of the DNA [10]. But due to sharp association/dissociation transitions and the slow rearrangement kinetics, fabrication of well-organized structures stays complicated. [11].
A different approach is by introducing surface mobile DNA linkers that can be achieved by coating the particles with a lipid bilayer [12–15]. For schematic representation, see figure 2.2. Lipid bilayers consist of lipids that have a hydrophilic head and hydrophobic tails. Due to the hydrophobic forces, the lipids configure such that the tails draw together and form a bilayer. DNA linkers with a hydrophobic anchor insert spontaneously in the lipid bilayer because of hydrophobic forces. A feature of the lipid layer, is that it is mobile and thus DNA linkers can diffuse freely around the particles surface. With this feature many interesting models can be investigated, two examples are flexible molecules and polymers consisting of colloids, see figure 1.4. These structures are promising building blocks for the development of complex materials and nano or micro robots.
Figure 1.4: DNA-mediated polymers and molecules. From left to right: polymers made from 4, 5
and 6 colloids and molecules made with 4 and 6 colloids. Scale bar corresponds to 5 µm.
1.4
Approach
In this work, colloidal particles are functionalized with a lipid bilayer and DNA linkers to be able to study the self-assembly of clusters, specifically short polymer chains build from colloids. In chapter 1, spherical particles are used. We discuss the presence of DNA linkers on the particle surface and their ability to bind to the complementary particle. In order to self-assemble, particles need to be able to move freely and, as to be discussed in this chapter, the substrate has a large influence on this. When clusters have formed, we will discuss the joint mobility and the free energy landscape as function of opening angles. In chapter 2, we will introduce anisotropic particles that have a dumbbell shape. Because the dumbbell particles we use are made of polystyrene (PS), we need to coat them with silica in order for them to be heavy enough to stay in the 2D plane. Secondly, the silica coating is needed for a lipid
1. Introduction bilayer functionalization. We will discuss silica coating under various conditions and which of these conditions result in a successful coating. Chapter 3 will continue with dumbbell shaped colloids and discuss short polymers made from dumbbells. Here as well, the joint mobility of these polymers is discussed. Polymers build from dumbbells are interesting, because they have a rigid part in the center that might influence the dynamic behavior of polymers. Besides, due to this rigid part between the lobes of the dumbbell, we can analyze the sliding and rolling behavior in a joint of a polymer.
Chapter
2
Experimental methods
2.1
Reagents
Chemicals:
We use ammonium hydroxyde solution, which is 28 - 30 % NH3 basis, bought from
Honey-well. In the rest of this thesis I will address it as ammonia. Ethanol (≥99.8%) was bought from Honeywell and tetraethyl orthosilicate (TEOS), which is ≥ 99.0% was bought from Sigma-Aldrich. Styrene, 3-(Trimethoxysilyl) Propyl Methacrylate (TPM), divinylbenzeen (DVB) and azobisisobutyronitril (AIBN) were used for the synthesis of the dumbbells. For the PAA coat-ing of coverglasses the followcoat-ing chemicals are used: Hellmanex, 35% w/w hydrogen peroxide (H2O2), 20% w/w acrylamide in water, N,N,N’,N’ - Tetramethylethylenediamine (TEMED),
ammonium persulfate (APS) and acetic acid. For the PEGylation of cover glass also PEG2000-SVA of Lasan Bio, and aminopropyl triethoxysilane (APTS) were used and casein was used for the casein coating of coverglass. DOPC, DOPE-PEG2000, DOPE-Rhodamine and Cholesterol-TopFluor were bought from Avanti.
Buffers:
HEPES buffer type 1 was made with 10 mM HEPES, 40 mM NaCl and 3 mM NaN3 dissolved
in miliQ water of 18.2 MΩcm (further addressed as water). The final pH was set to 7.35, using NaOH. For a higher salt concentration buffer, buffer type 2, 10mM HEPES and 190 mM NaCl dissolved in water, were used. Here, the final pH was set to 7.4, using NaOH. In order to pre-vent the growth of bacteria, the buffer was either sterilized using an autoclave or the buffer was filtered with a filter of pore size 0.1 µm. A PBS buffer for the PEGylation of cover glass was made with 50 mM NaCl, 10 mM PBS and the final pH was 7.5.
Particles:
The spherical silica colloids were purchased from Microparticles GmbH and are 2.12± 0.06 µm in diameter. They were washed three times in water. The dumbbell particles that are used in chapter 3.2 are fabricated from uncrosslinked polystyrene spheres (16% wt.) that are
stabi-2. Experimental methods lized in PVA solution by the method revised by Kim et al. [16]. A swelling solution is made of styrene, 1.5% v/v DVB and 2% wt. AIBN. To make heavier dumbbells we make a second type swelling solution where 10% v/v TPM is added. TPM has a higher density than polystyrene and its chemical structure is similar to silica. Afterwards, the swelling solution is mixed with a 0.5% hydroquinone and a 1% PVA solution both in water. This is added to the polystyrene spherical colloids and causes them to swell and form a protrusion that we polymerize in an oil bath for 24 hours [17]. This resulted in type 1 dumbbell particles (figure 2.1a), that have a total length of 2.72±0.12 µm and the radius of the lobes is 0.85±0.06 µm. A second type of dumbbells are used in chapter 3.3 and have a total length of 5.26±0.12 µm and the radius of the lobes was measured to be 1.38±0.06 µm (figure 2.1b).
(a)Type 1 dumbbell particles (b)Type 2 dumbbell particles
Figure 2.1: SEM images of polystyrene dumbbell colloids. DNA strands:
To induce self-assembly, the particles are divided into two groups, one coated with linker type A and the other with linker type B, described in [8]. We use the model where the DNA linkers are double stranded with a flexible single stranded DNA spacer and complementary sticky ends. To produce this, the linkers are hybridized with a backbone and have a sticky end that can bind to the other DNA linker type. Hybridization happens at room temperature and to en-sure all linkers are single stranded before hybridization, we heated the DNA strands to 90°C for 10 minutes. The sequence of the DNA linker type A is: Double stearyl-Heg(18)spacer-5’-TT- TAT-CGC-TAC-CCT-AGT-CAC-CTT-CGC-ACA-GTC-ACA-TTC-AGA-GAG-CCC-TGT-CTA-GAG-AGC-CCT-GCC-TTA-CGA-GTA-GAA-GTA-GG-3’-6-FAM. The sequence of DNA linker type B is: Double stearyl-HEG(18)spacer-5’-TT-TAT-CGC-TAC-CCT-AGT-CAC-CTT-CGC-ACA-GTC- ACA-TTC-AGA-GAG-CCC-TGT-CTA-GAG-AGC-CCT-GCC-TTA-CGA-CCT-ACT-TCT-AC-3’-Cy3. The sequence of the backbone is TCG-TAA-GGC-AGG-GCT-CTC-TAG-ACA-GGG-CTC-TCT-GAA-TGT-GAC-TGT-GCG-AAG-GTG-ACT-GTG-CGA-AGG-GTA-GCG-ATT-TT. The sticky end is 11 basepares long and has a melting temperature of 45°C. DNA linkers were made by mixing 10 µL, 20 µM of linker type A or B with 10 µL, 20 µM of backbone and 90 µL type 1
2.2 DNA functionalization of silica colloids
HEPES buffer. The final molarity of 1.82 µM was diluted to 0.02 µM.
Inert DNA is added to the particles for steric stabilization to exclude nonspecific binding by van der Waals attraction. Inert DNA is made by hybridization of the sequences cholesterol- TEG-3’-TTT-TAG-CGA-TGG-GAA-GCG-TGT-CAG-TTA-GAT-CTC-TCG-GGA-CGG-ATT-GC-5’ (strand A) and cholesterol-TEG-TEG-3’-TTT-TAG-CGA-TGG-GAA-GCG-TGT-CAG-TTA-GAT-CTC-TCG-GGA-CGG-ATT-GC-5’TTT-ATC-GCT-ACC-CTT-CGC-ACA-GTC-AAT-CTA-GAG- cholesterol-TEG-5’TTT-ATC-GCT-ACC-CTT-CGC-ACA-GTC-AAT-CTA-GAG-AGC-CCT-GCC-TTA-CGA-3’ (strand B). Here 10 µL, 20 µM of strand A was mixed with 10 µL, 20 µM of strand B and 90 µL type 1 HEPES buffer and the final molarity was 1.82 µM.
2.2
DNA functionalization of silica colloids
The colloidal particles are coated with a lipid bilayer first. This is done by rupture of small unilamellar vesicles (SUVs) that spread on the surface of the colloids, see figure 2.2 a). SUVs were prepared by drying DOPC (98.9 mole%), DOPE-PEG2000 (1 mole%) and fluorescent la-beled lipids (0.1 mole%) using a vacuum desiccator. A mixture of 77 µL DOPC, 7.34 µL DOPE-PEG2000 and 2 µL DOPE-Rhodamine or Cholesterol-TopFluor was typically used. After the chloroform has evaporated from the mixture, the lipids are resuspended with 1 mL of HEPES buffer type 1. The SUVs are then made by sonication on ice for 9 min (18 seconds on / 42 seconds off) on 30% load and 30% amplitude using a tip sonicator or 30 min (30 seconds on, 15 seconds off) on 90% amplitude using a cup horn. Titanium particles from the sonicator tip were removed by centrifugation for 45 minutes on 906 g. After coating the colloids with a lipid bilayer, excess SUVs were removed by centrifugation for 2 min of 28.6 g.
The DNA linkers were inserted in the lipid bilayer by incubation for one hour. A hydrophobic double stearyl group at the end of the DNA linkers links the DNA linkers to the lipid bilayer via hydrophobic forces. Inert DNA is added here as well to increase stability. Due to the mo-bility of the bilayer, the DNA can move around the surface of the colloidal particle, see figure 2.2 b). The typical DNA linker concentration was 650 µm−2 and inert DNA concentration 1.7x105µm−2. After functionalization of the colloids with DNA, excess DNA was removed by 2 washing cycles, which consisted of centrifugation for 2 min of 28.6 g.
2.3
Silica coating of polystyrene dumbbells
Because the lipid bilayer coating is not mobile on polystyrene dumbbells [18], we coat them with a silica layer. A second reason for a silica coating is to make the dumbbells heavier. Polystyrene is a material with a density of 1.04 g/mL and colloids made of polystyrene go out of the 2D focus plane very easy. Because we want to measure our system in 2D, we add silica to confine the dumbbells to the 2D surface. Different procedures with silica coatings are performed on dumbbell particles type 1. Here we describe the standard procedure: Firstly, polystyrene dumbbell particles that had formed aggregates were broken up by sonication in an ultrasonic bath for 20 minutes. Then, the dumbbell dispersion is added to a mixture of
2. Experimental methods
(a)Lipid bilayer coating by rupture of SUVs
(b) DNA functionalized colloids and their joint flexibility
Figure 2.2: Overview of the experimental system. First colloidal particles are coated with a lipid
bilayer. Due to the mobility of the bilayer, DNA linkers can diffuse freely around the surface. Figures adapted from [18]
ethanol, ammonia and water, while mixed with a stirring bean on 400 rpm. TEOS was slowly added from a syringe.
Seeded growth
Here, the first silica coating was prepared by an addition of 42.2 mL ethanol, 3.34 mL ammonia, 3.16 mL dumbbell dispersion and 1.11 mL TEOS. Concentrations were calculated to be [H2O]
= 5.0 M, [NH3] = 1.0 M and [TEOS] = 0.1 M. The reaction was left for 8 or more hours and
afterwards a small sample was taken out and washed three times at 179 g for 30 minutes, followed by 30 minutes sonication. Further silica coatings were created by seeded growth, which means that the silica coated colloids are left to react with additions of TEOS and water in a 1:2 molar ratio to increase the silica coating. 1.0 mL TEOS and 161 µL water were added and the reaction was left for 8 or more hours. Ammonia evaporates very fast, so during the seeded growth reactions, it was possible that ammonia was evaporated. Therefore we also add 1.67 mL of ammonia to the seeded growth reaction. The pH was measured with pH indicator strips 7.5-14 of MColoPhast.
2.4
Cover glass treatment
Casein coating:
dif-2.5 Imaging
ferent substrates were used. Casein cover glass was prepared by first cleaning the glass by 30 minutes of sonication in a 2% Hellmanex in MiliQ solution. We replaced the Hellmanex solution with water and the glass was sonicated for another 30 minutes. Again the water was replaced. The cleaned glass cover slip is then passivated for 30 minutes in a 5 g/mL casein solution and rinsed twice with water.
PEG coating:
For the polyethylene glycol (PEG) coating of cover glass, we cleaned the glasses with a Piranha base: Cover glass was immersed in Hellmanex 2% for 30 minutes while stirring. Afterwards the cover glass was rinsed three times with water. The cover glass was then immersed in 5:1:1 volume ratio H2O : NH4OH (28-30% w/w) : H2O2 (35% w/w) and this was heated to 70 °C
for 30 minutes, while stirring. The glass was rinsed with water for three times.
The glass was rinsed further one time with ethanol and two times with methanol. Subse-quently, the cover glass was immersed in a 5% acetic acid in methanol and 1% APTS and was stirred for 30 minutes. The glass was rinsed five times with methanol and was desiccated overnight in vacuum. Lastly the glass was immersed in a 5 g/L PEG2000-SVA in PBS buffer and the reaction was left for 8 to 10 hours at 4°C. Afterwards, the glass was washed 4 times with water and was dried under a nitrogen flow.
PAA coating:
For the coating of cover glass with PAA, the glass was cleaned with a Piranha base described above in the PEG coating. Afterwards, the glass was cleaned two times with ethanol and im-mersed in a 1% v/v acetic acid and 0.5% v/v TPM in ethanol and stirred for 10 to 15 minutes. A 0.035% v/v TEMED solution and a 0.070% w/v ammonium persulfate solution was added and the solution and was stirred for another 1 to 2 hours for polymerization. For the PAA coated cover glasses of type 2 and type 3, this step was done under a nitrogen atmosphere and for type 2 cover glass 0.012% bis-acrylamide was added and for type 3, 0.016% bis-acrylamide was added as a cross linker.
2.5
Imaging
Both DNA linkers and SUVs are fluorescently labeled. Colloidal particles coated with DNA linker type 1 are labeled with 6-FAM which has a maximum excitation at 492 nm and particles coated with DNA linker type 2 are labeled with Cy3 which has a maximum excitation at 554 nm. The SUVs were fluorescently marked with DOPE-Rhodamine, which has an excitation maximum at 550 nm and Cholestol-TopFluor which has an excitation maximum at 495 nm. We use a 488 nm laser, which excites 6-FAM and Cholesterol-TopFluor and a 561 nm laser, that excites Cy3 and DOPE-Rhodamine. A NIKON Ti-E microscope with a Nikon A1R confocal scanhead that has galvano and resonant scanning mirrors, was used to image the particles
2. Experimental methods with a 100 X oil objective.
Dumbbell particles and its silica coating was imaged with an FEI nanoSEM 200 scanning elec-tron microscope (SEM). The sample was prepared on a SEM stub and sputter coated with a Cressington 208HR sputter coater with a thin layer of 80% platina and 20% palladium to pre-vent charging, caused by the electron beam. It is coated for 200 seconds and was presputtered for 10 seconds on 20 mA, while tilted at 20° and rotating.
2.6
Analysis
Fluorescence intensity:
For the fluorescence intensity the same amount of fluorescently labeled DNA linkers were used for each sample. No other materials were fluorescently labeled in order to compare the amount of DNA linkers present on the particles by comparing the intensity. Three samples were pre-pared with the before described amount of SUVs and 20.7 µL DNA linkers of 4.5 µM to obtain a linker density of 3x105 µm−2. The first sample served as control sample and was exposed to the same amount of light, but was not washed. The second sample was washed after the DNA linkers were incubated for one hour. There was a third sample that we washed with 1.3 µM inert DNA. Verweij and Rinaldin [18] found that the stability greatly improved when inert DNA was added after washing as well and therefore DNA might get washed away. During imaging, all relevant settings were kept constant for each sample. Intensities were measured for each separate particle in ImageJ. The pixel intensities for the two different colored particles were averaged and from this standard deviation was calculated.
Tracking:
We adapted our own code to track particles in a cluster of 4 spheres or 2 dumbbells. First, we selected the particles in the cluster by clicking in its center. The frame is inverted in intensity, such that the intensity of the outer edge of the particle is maximal. Then the frame is interpo-lated and converted to polar coordinates. For each selected particle the position is found by measuring the maximum intensity with the help of a predefined particle radius. These coor-dinates are converted to the original Cartesian coorcoor-dinates and a circle is fitted to the particle, using the least squares method. Once the position of the particles are found, we iterate over all frames and for each frame, the location is refined. Afterwards, we checked the positions of the selected particles with the AnnotatePlugin from pimsviewer.
Diffusivity:
To be able to conclude which substrate was best to use for diffusion of particles, we analyzed the mean diffusion constant for each substrate. For each movie, which typically consisted of five minutes with a frame rate 10 frames per second, the particles are located and tracked using the python module trackpy [19]. Based on the mean squared displacement (MSD) as a function
2.6 Analysis
of lag time the diffusion constant is fitted by the power law Atn where A is the MSD, t is the lag time and n is the power law exponent, which is 1 for diffusive particles. The diffusion con-stant in 2 dimensions is given by D = MSD / 4t, where t is the time. A histogram is plotted for all diffusion constants for the particles and a threshold of D = 7.5x10−3 µm2/s was set, where particles that have a diffusion constant below this threshold are stuck to the substrate.
Cluster forming:
The number of clusters and total particles was counted by hand. Because we end up with a total number of particles and clusters, we cannot take the standard deviation and for this reason there is no error given in the graphs about the cluster forming. The minimum num-ber of particles used to derive the percentage of clusters, was 100 unless stated otherwise, and was typically between 100 and 400. A cluster is defined when it is bound for at least 10 seconds.
Joint mobility:
For the analysis of the trimers, the opening angle was defined as the angle between the three particles and is found by using trigonometric functions on the particles positions. The joint mobility was defined as the slope of the mean squared displacement of the opening angles of all frames [12]. It was measured with the use of trackpy and a least squares fit was applied to find the joint mobility for lag-times up to 1.5 seconds. Values above these lagtimes, become less trustworthy because there are less data points. Clusters composed of four spherical par-ticles or two dumbbell parpar-ticles have two opening angles and are defined as the two opening angles between three particles in the cluster, see the inset in figure 3.8a. The analysis done on these two opening angles is exactly the same as for joint flexibility of the the opening angle of the trimer.
Free energy of a trimer:
Only clusters that have a joint mobility larger than 70 degrees2/s are taken into account for the energy calculation as a function of the opening angles. The local force field of a trimer is calculated by using the maximum likelihood estimation method. This allows us to estimate the error in our measurement [18, 20]. The force that best fits the observed displacements is found by the probability of observing a single transition from opening angle θ0to θ1, which is
given by a Gaussian probability density function:
Pτ(θ1|θ0) = 1 2√πDτ exp(−(θ1−θ0−βF(θ0)Dτ) 2 4Dτ ) (2.1)
where Pτis the transition probability at sampling time τ, D is the diffusion coefficient, β is one
over the Boltzmann constant times the temperature, and F is the force at θ0. The probability of
observing more transitions is the product of the individual ones. By taking the logarithm, the product becomes a sum and defines the log-likelihood:
2. Experimental methods L({θ}|D, Fj) = N
∑
j=1 log Pτ(θj|θj+1, D, Fj) (2.2)here the best fit of the diffusion constant and the local force maximize the probability and are therefore called the maximum likelihood estimates (MLEs). To find the MLEs, a Bayesian esti-mator was implemented and makes use of the python package EMCEE. The error of the MLEs was derived as the standard deviation of the chain of Markov Chain Monte Carlo (MCMC) samples. The energy is then derived by integration over the force extracted from the joint mo-bility of the particles. Because the resulting free energy is symmetric around 180°, the energy is mirrored and averaged over this angle and only values between 60° and 180° are shown.
Free energy landscape of a tetramer:
Like the case for the trimer, the energy of a tetramer is derived, only when both the opening angles had a joint flexibility larger than the threshold. Instead of deriving the force from the MLE method, we use the histogram values of the found opening angles to find the probability density function and via Boltzmann weighing the free energy landscape can be calculated:
βF= −ln P+βF0 (2.3)
Here F is the free energy in terms of β. P is the probability density and F0is an arbitrary offset
that can be added to the free energy. The resulting energy landscape is symmetric at the open-ing angles θ1+θ2= 360° and θ1= θ2
Theoretical free energy:
The outline we use to derive the theoretical free energy is inspired by Meng et al. [21] and we summarize the steps that are taken. In the canonical ensemble, the probability of a cluster that is in state s is proportional to the Boltzmann distribution:
Ps∼e−βFs =Zse−βUs (2.4)
Here Zsis the partition function and Fsis the Helmholtz free energy and Usis pair potential.
We consider the latter the same for all particles in the cluster and at constant temperature we find Ps ∼ Zs, and therefore we can find an expression for the free energy as function of the
partition function. The internal partition function Zsis used to calculate the free energy and is
split up into three contributions:
Zs =Zt,sZr,sZv,s (2.5)
Zt,s is the translational partition function, Zr,s is the rotational partition function and Zv,s is
the vibrational partition function. Zt is proportional to the free volume that the cluster can
explore and is approximately equal for all clusters. We can neglect electron and quantum contributions. In our model we assume that the vibrational partition function is constant, because the bond length is determined by the DNA linker length and should only weakly
2.6 Analysis
depend on the opening angles. The rotational entropy is defined as: Zrot=
√
Icl
σ (2.6)
Where I is moment of inertia of the total cluster relative to its center of gravity (cg) and σ is the rotational symmetry number. Because most configurations in our tetramer are not symmetric, except for some, we will use σ= 1. The moment of inertia of the total cluster is the sum of the moment of inertia of each particle in the cluster around the cg and for each particle the parallel axis theorem can be used to find an expression relative to the cg instead of the origin. Our system of clusters is in 2 dimensions and thus we here use the parallel axis theorem in 2 dimensions that describes flat disks instead of spheres:
Icg = I−md2 (2.7)
Here d is the translation from the origin to the cg and I is the is the moment of inertia of a disk with zero height, which is described by I =1/2mR2with R the radius of the disk. To find the positions of the disks relative to the cg, we parameterized the positions of the disks as function of the two opening angles θ1and θ2, see figure 2.3:
r1 = (0, 0) r2 = (2R, 0) r3 = (−2R cos θ1,−2R sin θ1) +r2 r4 = (−2R cos α,−2R sin α) +r3 Figure 2.3: Schematic of particles positions
Here, ri is the position of disk i in the cluster, R is the radius of the disk and α is defined as θ2+θ1−π. With these positions the cg is derived as:
rcg = ∑ mi
ri
∑ mi
(2.8) where m is the mass of particle i. Now we can subtract the positions of the disks from the po-sition of the cg to find the popo-sitions of the disks relative the the cg. Using the relative popo-sitions in the equation that that describes the moment of inertia of a disk and adding these values for all four disks, we find the moment of inertia of the cluster:
Icl(θ1, θ2) =
4 2mR
2+2mR2(5−2 cos θ
2. Experimental methods
Shell thickness:
Dumbbells before and after the silica coating were imaged with scanning electron microscope (SEM). The shell thickness is derived on the basis of two methods. The first method is by mea-suring the length of the dumbbells before and after the silica coating. The mean shell thickness was based on the subtraction of length of the dumbbells without shell from the ones with shell and the mean and standard deviation were calculated. The second method was by direct mea-surement of the shell. For this method the dumbbells with a silica coating were heated for 5 hours in an oven at 450°C. At this temperature polystyrene melts, however silica does not, and in some cases the shell cracks open. This allowed us to measure the shell thickness in ImageJ. Here again, the mean and standard deviation were derived. The resulting shell thickness was equal for both methods, but for the second method the error was smaller. Therefore, we only show results of the thickness based on this method.
Fluorescence recovery after photo bleaching:
Flurorescence recovery after photo bleaching (FRAP) allows us to determine if the bilayer is mobile. This technique was done on a confocal microscope were a laser of a wavelength cor-responding to the fluorescent dye in the lipids is absorbed and excited. A region of interest (ROI) is chosen for excitation until the fluorescently labeled lipids are bleached. Another ROI on the particle is chosen as a reference point. Lipids outside the stimulation ROI were not bleached and still emit light. In a mobile bilayer, these lipids can diffuse to the stimulation ROI, where we can measure the recovery of the intensity. The reference ROI shows a decrease of intensity, due to the diffusive behavior of the lipids and by dividing the stimulation ROI by the reference ROI we find a normalized intensity. Lastly the intensity is measured on a ROI that is chosen for the background. We subtracted this from the intensity on both the reference and the stimulation ROI. The normalized fluoresence intensity as a function of time is given by:
f(t) = A(1−e−t/τ1/2) (2.10)
where f is the intensity, A is the amplitude, τ1/2is the half life and t is the time. The mean half life was obtained by fitting this expression with the least squares method to the intensity we found in 8 measurements on dumbbells.
Chapter
3
Results and discussion
3.1
Properties of isotropic colloids
3.1.1 Diffusion of colloids
In this subsection we analyze the diffusion coefficient for different substrates. Also, the num-ber of particles that are stuck to the substrate will be analyzed for different substrates. We discovered that many colloids fully functionalized with a lipid layer and DNA linkers, were fixed on uncoated glass. This unwanted feature could be resolved by using a different sub-strate. Several substrates were studied.
Firstly, we tried a polyethylene glycol (PEG) coated glass substrate. PEG is a hydrophilic poly-mer and has been used before on microscopic glass slides for avoiding non-specific sticking of proteins in single-molecule fluorescence studies [22]. The PEG polymers provide steric stabi-lization, such that the colloids avoid contact with the glass. PEG coating on glass cover slips is described in chapter 2.4. However, we did not obtain usable results for PEG coated glass; 97% of the particles had a diffusion coefficient lower than the threshold defined in chapter 2.6 and the rest of the particles had a diffusion coefficient of 0.02±0.01 µm2/s. In our system the temperature is 20 °C which is 293 K, the dispersion medium is water which has a viscosity co-efficient of 1.002x10−3kg/ms and the particle radius is 1.06 µm. We can substitute these values in equation 1.1, finding a diffusion coefficient of 0.2 µm2/s for our system. The diffusion on PEG coated glass is 10 times lower than the expected diffusion coefficient, therefore PEG did not provide the steric stabilization that was expected.
Secondly, we looked into glass coverslips that are passivated with casein as suggested by van der Meulen [11], who uses a similar substrate. The results for the diffusion were promising. A diffusion coefficient of 0.07±0.03 µm2/s was found. However, 24 hours after preparation of the sample, the colloids were completely clustered and we were unable to extract data on short polymers build from colloids.
3. Results and discussion Lastly, polyacrylic acid (PAA) coated glass cover slips as described in 2.4 were studied. The PAA polymers that cover the glass are anionic, making the glass slightly negative. In contrast to PEG polymers, PAA polymers might provide steric stabilization. Results of coverslips with PAA coating of type 1 are shown in figure 3.1.
(a)Diffusion after preparation (b)Diffusion after 3 hours (c)Diffusion after 6 hours
(d)Diffusion after 24 hours (e)Diffusion after 48 hours (f)Diffusion after 56 hours
Figure 3.1: Diffusion of colloidal spheres on PAA coated cover glass type 1. The influence of
time is studied on a sample with spherical colloids functionalized with 350 DNA linkers/µm2. Time does not influence the average of the diffusion constant of the particles.
As can be seen in the figures of 3.1, many particles have a diffusion coefficient between 0 and 0.0075 µm2/s. These particles are stuck to the substrate. Directly after preparation 63% of the particles is stuck. The sample was imaged again after 3, 6, 24, 48 and 56 hours and in between, the sample was put on a shaker. The percentage of particles sticking to the substrate is in chronological order 45%, 40%, 28%, 35% and 32%, hinting that the shaker decreases the num-ber of particles stuck to the substrate. The histograms in figure 3.1 are bimodal, with the first mode the sticking particles plus particles bound to the sticking particles. These as well, have a lower diffusion constant, since they are in a cluster that is not diffusive. The second mode are the particles and clusters that are diffusive and a diffusion constant was obtained by fitting a Gaussian to the second mode of the histogram. Shaking the sample in time does not have an influence on the diffusion constant, which is 0.09±0.01 µm2/s after preparation of the sample.
A second type of PAA coated cover slips, where 0.012% bis-acrylamide was added as cross linker, was analyzed, see figure 3.2. We measured the diffusion coefficient for three types of samples. The first sample was functionalized with 650 DNA linkers/µm2, the second sample had 420 DNA linkers/µm2and the third sample with 230 DNA linkers/µm2. Figure 3.2 shows
3.1.1. Diffusion of colloids
that on this type of substrate not many particles stick to the substrate. The percentage of par-ticles stuck to the substrate is in order of decreasing linker density 5.2%, 5.6% and 0.9% and thus this type of substrate has a large positive influence .
(a) Diffusion of spheres with 650 DNA linkers/µm2
(b) Diffusion of spheres with 420 DNA linkers/µm2
(c) Diffusion of spheres with 230 DNA linkers/µm2
Figure 3.2: Diffusion of colloidal spheres on PAA coated cover glass type 2. The influence of
DNA linkers is studied on a sample with spherical colloids. Here we see that density of DNA linkers might have an influence on the diffusion coefficient.
We also found that the amount of DNA linkers on the surface of the particle has an influence to the diffusion coefficient of the colloids. Figure 3.2 shows that the histogram of diffusion coefficients shifts to the right for lower DNA linker densities. The value for colloids with 650 linkers/µm2 is 0.03 ± 0.04 µm2/s, 0.02 ± 0.06µm2/s for colloids with 650 linkers/µm2 and increases to 0.08±0.05 µm2/s for colloids with 230 linkers/µm2/s.
Although we see a shift of diffusion coefficient, the error is quite large. We also analysed the diffusion coefficient for a sample with a very high linker density, namely 3x105 linkers /µm2 on PAA coated glass type 1. Here the diffusion coefficient is 8x10−5 ±9x10−5 µm2/s, which indicates all colloids are stuck to the substrate. Comparing this with figure 3.1 also indicates that the DNA linker density has an effect on the diffusion coefficient.
Figure 3.3: Diffusion of colloidal spheres on PAA coated cover glass type 3.
3. Results and discussion most successful results on the diffusion coefficient. The histogram of the diffusion coefficients is shown in figure 3.3 and a mean diffusion coefficient was found at 0.10 ± 0.05 µm2/s. In this sample the diffusion coefficient corresponds most closely to the diffusion coefficient that we theoretically found for our spherical particles and we can conclude that the PAA coated glass type 3 provides the most steric stabilization. We conclude from figure 3.1, 3.2 and 3.3 that crosslinking the PAA polymers has a positive effect of the diffusion coefficient of the colloids. Besides, adding a higher concentration of bis-acrylamide during the cross-linking also might improve the diffusion coefficient of the particles.
3.1.2 Cluster forming
Specific cluster formation can only happen with the right number of DNA linkers [12]. How-ever, Rinaldin and Verweij showed that washing DNA functionalized colloids negatively in-fluenced the stability provided by inert DNA and they showed that adding extra inert DNA during the washing cycle has a positive effect on the stability [18]. Therefore, the following ex-periment is carried out to make sure washing the DNA functionalized particles does not effect the DNA linker concentration. We measured the amount of fluorescently labeled DNA link-ers on two samples, where one of the samples was washed and the other sample was washed with extra inert DNA. We compared this to a control sample that we did not wash, but the sample was exposed to the same amount of light, so we could compare the fluorescence inten-sity. Figure 3.4a shows the mean fluorescence intensity and it shows that for all three samples the intensity has the same value for red and green emission. We can conclude that there is no significant effect on the DNA linker density on the particles surface by washing the sample. Figure 3.4b shows the percentage of small specific and aspecific clusters and thus this figure indicates the effect of washing on the binding. We measured the amount of clusters 24 hours after preparation of the sample. In the sample that is washed, 197 particles were analyzed and of that number, 12% of the colloids were in a specific cluster and 9.1% in an aspecific cluster. In the sample that is washed with inert DNA, we analyzed 202 particles, 1.0% of that number of particles was found in a specific cluster and 8.9% in an aspecific cluster. For the results of the control sample 216 particles were analyzed and none of the particles were found in a specific cluster, but 18.5% of the particles were aspecifically clustered.
Figure 3.4b thus shows that the washed sample is able to form specifically assembled clusters, in contrast to the sample that is washed with extra DNA and the control sample. In the sample that is washed with inert DNA, the extra amount inert DNA provides even more stability to the already stabilized colloidal dispersion. This might cause the absence of specific clusters. In the control sample the amount of DNA linkers is the same as the other two samples, there-fore the DNA linkers on the surface of the particle should not influence the specific clustering. However, because the control sample is not washed, excess DNA linkers might be present in the medium and complementary strands in the medium find the possibility to bind to the
sur-3.1.2. Cluster forming
(a)Pixel intensities (b)Clustering
Figure 3.4: Availability of DNA linkers. Three samples are studied. One sample that is washed,
one is washed with inert DNA and there is a control sample which is not washed. In 3.4a the linker intensities are shown for both red and green emitted linkers. In 3.4b the amount specific and aspecific clusters is presented.
face linkers, making them unavailable for binding to complementary particles. The number of aspecifically bounded clusters is the same for the two samples that are washed. If the inert DNA provides extra stability, then the amount of aspecific clusters also should decrease. How-ever, aspecific clusters might also form before the washing step, during functionalization with SUVs or with DNA and therefore the extra stability will not play a role in this type of clustering.
To prove that the aspecific clusters that have formed might already develop during the func-tionalization process, cluster growth is analyzed over time. Figure 3.5 shows that almost no specific clusters are forming and they are not growing in time due to the stabilization of the inert DNA. The aspecifically formed clusters stay constant in time, and thus we may conclude that the inert DNA stabilizes the colloids and prevents further aspecific clustering. Besides, this indicates that the amount of aspecific clusters that is present, has formed during SUV coating of the colloids.
Before we could invest in the dynamical properties of specific clusters, we needed to increase the specific binding and a medium with higher salt concentration might be the solution. The DNA strands we use are negatively charged, which means that they, besides steric stabiliza-tion, also provide charge stabilization. Biancaniello et al. showed (see figure 3.6a) that specific aggregates only happened with a salt concentration of 150 mM or more if there is a low per-centage of hybridizing DNA [23]. Salt in the solution decreases the distance to the slipping plane (figure 1.2b) and therefore charge stabilization will be neutralized. The experiment is done with buffer type 1 to prevent aspecific binding during functionalization. Only in the last washing step, after the DNA is added, buffer type 2 is added to increase specific binding be-tween complementary DNA strands. As can be seen in figure 3.6b, specific dimers, trimers
3. Results and discussion
(a)Specific clusters (b)Aspecific clusters
Figure 3.5: Cluster formation over time. Inert DNA is added for stability. Specific and aspecific
clusters are analyzed
and also tetramers are forming in time. For each time point between 310 and 460 particles were analyzed. The percentage of single particles implies the overall aggregation, specifically and aspecifically. We can see that the number of single particles decreases a bit in time, but considering the increase of specific clusters, we conclude that the amount of aspecific clusters has not increased significantly.
(a)Phase diagram for the amount of clusters
(b)Clustering in time
Figure 3.6: The effect of salt on clustering. 3.6a illustrates the amount of salt needed to obtain
aggregation. Figure adopted from [23]. 3.6b shows aggregation in time. After preparation of the sample, buffer type 1 is replaced with buffer type 2.
The time needed to form specific tetramers is still very long, therefore we tried to hybridyze new DNA. DNA, even when stored at 4°C can be digested by enzymes called deoxyribonu-clease (or shortly DNase). This enzyme cleaves the structural linkages in the DNA backbone
3.1.3. Cluster properties
[24]. Hybridizing freshly DNA strands might increase the fraction of available DNA linkers for binding. With more DNA able to bind, we move up in the phase diagram in figure 3.6a and more aggregation should be happening. When empoying freshly obtained DNA linkers, figure 3.7 indeed shows a rapid increase in specific dimers and larger specific structures, that we have not seen before, have formed not long after sample preparation.
Figure 3.7: Clustering in time. New hybridized DNA effects in a much faster specific clusterforming.
A type 2 buffer was used for the high salt concentration and inert DNA provides stability against aspecific aggregation.
3.1.3 Cluster properties
Once we have enough specific tetramers, we can analyze their dynamic behavior. One of the properties that we have looked at, is the joint mobility, which we define as the change in the opening angle θ, which is enclosed by the lines connecting the center of a middle particle with the centers of the two neighboring particles, over time [12].
The opening angles are depicted in figure 3.8a. The tetramer has two opening angles and fig-ure 3.8a shows the mean squared angular displacement as a function of lagtime for a mobile tetramer. The joint flexibility we found for θ1was J = 178±24 deg2/s and for θ2was J = 309
±42 deg2/s. When we averaged over all experiments we find for both opening angles θ 1and
θ2, J = 154±3 deg2/s. For trimers this behavior was explored by Rinaldin and Verweij et al. and they found a joint flexibility of 184±101 deg2/s for the opening angle θ [18].
Because of the two opening angles, the tetramer has more configurations to explore than the trimer. The configurations given by the angles are depicted in 3.8b. Combinations of θ1+θ2<
180° and θ1+θ2>540° are not possible, because two particles cannot overlap.
The free energy as a function of both opening angles θ1 and θ2that was found from 49
3. Results and discussion
(a)Mean squared angular displacement of θ1and
θ2
(b)Some possible configuration
Figure 3.8: Mobility and configurations of self-assembled tetramers. By fitting the MSAD
shown in 3.8a we find a joint mobility of 178± 24 deg2/s for θ1 and 309 ±42 deg2/s for θ2 for this
particular tetramer. Figure 3.8b shows an overview of a of the possible configurations a tetramer can form due to the flexibility in the joints.
derived by calculating the rotational entropy in 2 dimensions (figure 3.9b). Symmetries are taken into account and the free energy is averaged over the line at θ1+θ2 =360° and θ1−θ2=
240°. Therefore, the graphs only shows the free energy for θ1 < 180°. The free energy that
we find by experiments agrees well with the free energy landscape that we find by calculation the rotational entropy. Similarities are the maxima in the free energy for opening angles θ1+
θ2 =180°. Besides these closed configurations, the free energy also shows a maximum at θ2=
60° and θ2 = 300°, which is a closed configuration as well. On the contrary, θ1 = 180°and
θ2 = 180° is preferred on the order of magnitude about 0.4 kBT for the theoretical calculated
free energy. For the experimentally derived free energy, the order of magnitude between open and closed configuration is higher, which may be the effect of other contributions to the free energy that we did not encounter for. For the trimers, Rinaldin and Verweij et al. found no preference for the opening angle within the experimental error [18]. This means the particles in a trimer are freely jointed and if there is a contribution in the rotational entropy, it cannot be measured outside the error. In our work on the tetramers, we experimentally find a larger difference between the minimum and maximum of the free energy and therefore, we can more easily distinguish preferred and not preferred configurations for a tetramer.
3.2 Silica coating on anisotropic colloids
(a)Free energy derived from measurements (b)Free energy calculated from rotational entropy in 2 dimensions
Figure 3.9: Free energy landscape of self-assembled tetramers.
3.2
Silica coating on anisotropic colloids
Anisotropic particles are interesting in several ways and will be discussed later in this chapter. Here we use polystyrene (PS) dumbbell shaped colloids as anisotropic particles. In order to use them in our experiments, we have to coat them with silica, because the lipid bilayer is not mobile on polystyrene [18]. Secondly, polystyrene is a much lighter material than silica and dumbbell shaped particles made of polystyrene go out of focus very easy when looking at them under the microscope. Because we are looking at the properties of clusters, confined in the 2D surface, a silica coating on the light PS dumbbells could confine them to the 2D surface. 3.2.1 Silica coating of PS particles
Silica can be coated on PS by a sol-gel method [25]. The reaction consists of a hydrolysis and condensation process. The hydrolysis process is given by reaction 3.1 and shows that a alkoxy silanes such as TEOS reacts with water to a silanol, a silica group bonded to an hydroxyl group. The hydrolysis can be increased by increasing the concentration of alkoxy silanes, or water [26]. The silanol groups are necessary for the condensation reaction, which is given by reaction 3.2 and 3.3. Condensation is the formation of Si-O-Si groups by either alcohol or
3. Results and discussion water elimination. Si-O-Si, also siloxane, is the final result, where silica atoms are bonded by an oxygen atom.
Si OR+H2O Si OH+ROH (3.1)
Si OH+ Si OR Si O Si +ROH (3.2)
Si OH+ Si OH Si O Si +H2O (3.3)
We varied the amount of TEOS, ammonia and water to find what gives the thickest and smoothest silica coating without aggregation of the dumbbells. We tested four different ex-periments, described in table 3.1. In the first experiment 50 and 48 mL ethanol was used for PS and PS+TPM sample respectively and we added 5 mL of 3%wt. of dumbbell dispersion, 1.5 mL ammonia and 0.8 mL TEOS. In the second experiment 10 mL of ethanol was used and 1 mL of the dumbbell dispersion, 0.3 mL ammonia and 1.8 mL TEOS was added. The third and fourth experiment, the amount of ammonia compared to ethanol was varied and the rest is left the same. Therefore I refer to these experiments as 3a and 3b. In these experiments 0.36 mL TEOS and 1 mL of dumbbell dispersion where the medium was taken out, was used. In experiment 3a 10 mL of 4.4% v/v ammonia in ethanol was used and in experiment 3b 8.4% v/v ammonia in ethanol was used.
The resulting concentrations and the silica shell thickness are summarized in table 3.1. In some of the experiments, the dumbbells were clustered as a result of the silica coating. Ultrasonica-tion during the coating helped in keeping the particles dispersed. However, in experiment 3b the silica shell broke at the points where the dumbbells were connected, leaving a hole with-out silica. To reduce the amount of secondary nucleation found in experiment 1, the sample was washed at a lower rotational frequency to only sediment the dumbbells and wash out the small homogeneous solid silica spheres. This worked for experiment 2 and 3a, but experiment 3b still contained a lot of secondary nucleation. Van Blaaderen, van Geest en Vrij propose that the rougness the colloids with a silica layer are the result of the ”building units” composed of siloxane structures. Interpreting their proposal means that a high ammonia concentration leads to larger building units.
The thickness of the silicashell was analyzed for the four different experiments and is shown in figure 3.10. Here the silica shell on the dumbbells in experiment 1, 2 and 3b was success-ful, leading to a silica shell thickness between 40 and 80 nm. For experiment 3a there was no silicashell we could measure the thickness of, therefore the thickness is set to 0 nm. Including properties such as aggregation, show that experiment 1 and 2 were preferred. However, the dumbbells are still not heavy enough, therefore we proceeded with more silica coatings.
3.2.1. Silica coating of PS particles
exp seed particles description SEM image shell thickness aggregation 1a PS PS + TPM [H2O ] = 5.75 M [NH3 ] = 0.38 M [TEOS ] = 0.063 M [H2O ] = 5.95 M [NH3 ] = 0.39 M [TEOS ] = 0.065 M 78±17 nm 45±13 nm Almost none Idem 2b PS PS + TPM [H2O ] = 5.03 M [NH3 ] = 0.33 M [TEOS ] = 0.62 M Idem 53±21 nm 51±11 nm Almost none Idem 3ab PS PS + TPM [H2O ] = 1.62 M [NH3 ] = 0.58 M [TEOS ] = 0.16 M Idem 0 nm 0 nm None Idem 3bb PS PS + TPM [H2O ] = 3.24 M [NH3 ] = 1.17 M [TEOS ] = 0.16 M Idem 66±21 nm 76±28 nm A lot Idem
aThe reaction was terminated after 3 hours by washing at 3000 rpm bThe reaction was terminated after 3 hours by washing at 1500 rpm
3. Results and discussion
Figure 3.10: Thickness silica shell. For the different experiments described above, we measured the
thickness of the silicashell.
3.2.2 More silica layers
For multiple silica coatings, we first looked into the literature, and here results of more silica layers are given by [27–29]. Caruso et al. showed in 1999 a linear increase in silica multi-layer film thickness with deposition multi-layer number, see figure 3.11a. In their experiments they showed they could achieve a large increase in silica shell thickness (36±10 nm) by increasing the stability. Between every silica coating, the colloids are coated with polydiallyldimethy-lammonium chloride (PDADMAC) polymers. The particles have a potential of 45 mV when PDADMAC forms the outermost layer and between -20 and -30 mV with silica as the outer layer. The increase of magnitude in zeta potential indeed indicates a higher stability. Van Blaaderen also proposed stability is an important fact in silica growth [30].
(a)Figure adopted from [29] (b) Thickness of silica layers after one and two coatings on PS and PS+TPM seed particles
Figure 3.11: Thickness of silica layers. Thickness as function of layer number 3.11b and compared
3.2.2. More silica layers
Because the silica coated dumbbell particles we found in experiment 1 formed very few aggre-gates, we assume that the stability of our particles is high enough and in our work, we coat the silica coated particles of experiment 1 with another layer by the same protocol. We found a silica layer growth of 44± 24 nm for PS and 42 ±19 nm for PS+TPM, see figure 3.11 and indeed we can coat our particles with more silica layers to produce a uniform thick layer.
(a)Growth mechanism explained by Han et al. [27]. Pathway I is the nucleation and growth of small silica particles and path-way II is the in situ seeded growth.
(b)5 growth steps by addition of TEOS and H2O, including SEM images
of the dumbbell particles and silica coatings.
Figure 3.12: In situ seeded growth. The growth mechamism of in situ seeded growth proceeds via
the classic Stober method for ammonia concentration larger than 0.95 [27]. However we did not find an increase in shell thickness as shown in figure 3.12b
Han et al. showed a slightly different method for a silica growth process with the in situ seeded growth model [27]. They state that the reaction up to a certain amount of time Ti is dictated by
3. Results and discussion the hydrolysis to form monomers and is responsible for nucleation and growth of small silica particles, shown by pathway I in figure 3.12a.
Afterwards, the reaction is dictated by condensation of newly formed silanol monomers onto the earlier formed silica particles. This process is responsible for the enlargement in size of silica particles. The process is illustrated by pathway II in figure 3.12a. Also, since TEOS hy-drolysis is strongly affected by the ammonia concentration, the truly in situ seeded growth of silica particles takes place only at [NH3] ≥0.95M. Below this concentration the induction
period of the reaction time Ti prolongs and pathway I interferes with pathway II, resulting in
secondary nucleation due to pathway I during the seeded growth process described in figure 3.12a by pathway II.
We imitate the process described by Han et al. [27] and the result is shown in figure 3.12b. The figure shows that the silica coated particles do not grow in time after the first silica coating. Besides, we see a lot of secondary nucleation, despite using an ammonia concentration above the threshold of 0.95 M. We are not sure if the secondary nucleation has formed in the first stage of the silica growth or during the seeded growth. A possibility is that the ammonia evaporated during or consumed by the reaction. Ammonia is a base and in a next experiment we measured the pH to determine if the pH decreases when ammonia evaporates. We added extra ammonia during the seeded growth process as a control.
Figure 3.13: Thickness of the silicashell. Measured for two silica coatings produces by in situ
seeded growth and by adding extra ammonia to increase pH
Directly after addition of all reagents the pH was 10. The reaction was left for 8+ hours and afterwards the pH was measured again and we found a value of 8.5, which is lower than the initial pH, indicating indeed evaporation or consumption of the ammonia. After a seeded growth where extra ammonia was added the pH increased to its initial value. In the sample where only TEOS and water were added, the pH stayed at 8.5. The reaction of the second coating was left for 8 hours and afterwards the pH was determined at 8.5 and 9.5 for the sample without and with an extra addition of ammonia. Figure 3.13 shows the increase in shell thickness for the in situ seeded growth with and without extra addition of ammonia. It seems that the silica shell did grow for both samples and for the normal seeded growth procedure the
3.3 Properties of anisotropic colloids
shell thickness increased by 70±47 nm. For the sample where we added extra ammonia the silica shell increased by 91±35 nm, hinting at that extra addition of ammonia indeed effects the silica growth.
3.3
Properties of anisotropic colloids
In general, anisotropic colloids have different properties than isotropic colloids. First of all, anisotropic colloids can be made by selectively create different surface properties at two sides of a spherical colloid (called Janus particles). These particles have been used for their different surface properties in for example wettability [31], surface roughness [32] and magnetization [33]. Secondly, the shape of colloids can be adjusted to create anisotropic particles.
Here, we would like to compare the properties described in section 3.1.3 for clusters con-structed from spherical colloids with clusters of dumbbell shaped particles. The type 2 dumb-bells used for these experiments are described in section 2.1. Because these particles are larger compared to the dumbbells used in section 3.2, these dumbbells do not go out of focus. Dif-fusion is driven by equation 1.1 and thus the larger the radius of the particles, the smaller the diffusion coefficient. We experimentally found that this size was large enough to let the dumb-bells stay in the 2D plane, but small enough to experience diffusion on the substrate. On PAA coated cover glass type 1, we measured a diffusion coefficient of 0.1± 0.04 µm2/s, which is approximately the same as for as for the spheres on this type of substrate.
We coated the dumbbells with a silica layer described in section 2.3 and funcionalized the particles with a lipid bilayer, DNA linkers and inert DNA. We do not know the concentration of the dumbbells and therefore cannot tell what the final linker density on the dumbbells surface is. For the functionalization, we added the same amounts as for the spheres and we found a homogeneous bilayer (see figure 3.14a), that is coated on all particles (figure 3.14a and 3.14b). We also find specific aggregation.
(a)Confocal image (b)Brightfield image (c)Confocal image (d)Brightfield image
Figure 3.14: Self-assembled dimers of colloidal dumbbells with 20 nM (left) and 0.3 nM (right) added DNA linkers. In the confocal image (a and c), we see that the bilayer on the dumbbell
particles looks homogeneous. Comparing the confocal (a and c) with the bright field image (b and d), we also conclude that all particles have a lipid bilayer coating. Both added DNA concentrations result in specific binding.
3. Results and discussion However, the specific clusters that we found, were not mobile and the joint mobility averaged over 9 measurements was J = 8.2±0.7 deg2/s for θ1and J = 9.4±0.7 deg2/s for θ2, which is
much lower than the mobility that we found for clusters made of spheres (J = 154±3 deg2/s).
Chakraborthy et al. stated that colloids with a high linker density cause this effect of immobil-ity [12] and therefore, we decreased the linker concentration that we added to the dumbbells. Even at an added DNA linker concentration of 0.3 nM we see specific clusters one hour af-ter sample preparation (see figure 3.14c and 3.14d). One hour afaf-ter sample preparation, we analyzed the number of specific and aspecific dimers of dumbbells as function of the added DNA linker concentration, see figure 3.15. There seems to be an overall effect of decreasing the linker concentration. In figure 3.15b the percentage of specifically bound dimers decreases and adding even lower concentrations of DNA linkers (0.13-0.4nM), figure 3.15a shows that the number of specific dimers decreases even further. At this low linker concentrations we still have not found mobile clusters.
(a)Linker concentration ranging from 0.13-0.4 nM (b)Linker concentration ranging from 10-25 nM
Figure 3.15: Effect of linker concentration on cluster forming
Although bilayers are mobile on silica surface, a FRAP experiment could give insight whether the bilayer might be the problem why we did not see mobile clusters. A FRAP experiment was done on eight particles and averaged afterwards, see figure 3.16.
Fitting formula 2.10 to the measurement data gives us a half time of 10.4± 8.8 seconds and comparing this value with the recovery time of 3.2± 0.02 seconds that Rinaldin and Verweij et al. found for the lipid bilayer on the same silica spheres we use in chapter 3.1, we conclude that overall the recovery time is longer for dumbbells. The reason for this might be because the dumbbell particles are coated with a silica layer, which has a rougher surface than the spheres. Rinaldin and Verweij also reported that the surface roughness has a large influence on the mo-bility of the bilayer and for very rough surfaces, they even found that the bilayer is not mobile at all. Here we do see a recovery of the bilayer, however we do not see a full recovery. The fact that only 60% of the fluorescence intensity recovered, may be because part of the lipids in the bilayer is stuck to the surface and therefore immobile.
3.3 Properties of anisotropic colloids
Figure 3.16: FRAP experiment. Graph is the averaged FRAP over 8 measurements and the errors
are the standard deviation. The bilayer on dumbbells fully functionalized show a recovery after photo bleaching, which means that the bilayer is mobile.
The reason that we have not found mobile clusters, probably is a combination of the partly immobile lipids and not having found the right DNA linker density yet. Since we do not know what the concentration of the dumbbell dispersion is, it is hard to say what the total surface area of the dumbbells is, when we add the dispersion during functionalization.
