MSc Physics and Astronomy
Track Science for Energy and Sustainability
Master Thesis
High Precision Assembly of Nanocubes on Silicon
by Floris Leon Taminiau
10661301
9 July 2019 60 ECTS
September 2018 - July 2019
Supervisor:
prof. dr. E.C. (Erik) Garnett
Second examiner:
Abstract
The fabrication of high-performance (opto)electronical devices at the micro- or nanoscale requires accurate control over the placement of the materials used to build such devices. Bottom-up approaches such as directive assembly mediated by capillary forces have been proposed as methods for accurate arrangement of nanostructures independent of material properties. High spatial organisation of single materials on structured templates has been achieved, but the feasibility of combining multiple differing materials within a 100 nm scale has not yet been explored. Here, we demonstrate high spatial accuracy for the placement of nanocubes through sequential template-assisted capillary assembly. Silicon substrates were patterned using EBL and made hydrophobic through fluoro-silanization, creating a high contact angle for polar solvents such as water and ethanol. We performed assembly with silver and palladium nanocubes (75 nm and 19 nm, respectively), although any material that is capable of forming a stable colloid can be used. We show the possibility of achieving a succession of silver-palladium-silver within a spacing of 30 nm through smart trap design. Even though assembly yield and reproducibility remain issues, the concept highlights opportunities for real-world device fabrication.
Acknowledgements
I would like to start by thanking prof. dr. Erik Garnett for providing me with the oppor-tunity to perform this research in his group at the AMOLF institute. He has supervised me throughout this entire year, given me valuable comments and steered me in the right direction on countless occasions. Next, I would like to thank Harshal Agrawal for showing me much of the practical skills involved in synthesis, fabrication and assembly, as well as providing me with abundant literature to increase my knowledge. Then, I would like to thank Dimitry Lamers, Bob Drent and Andries Lof from the AMOLF NanoLab for train-ing me on all of the equipment needed and helptrain-ing me immensely with precise sample fabrication. Furthermore, I would like to thank the entire Nanoscale Solar Cells group for the discussions, talks and fun we had. Lastly, I would like to thank my friends and family for supporting me through this year, listening to me ramble about cubes and also helping me take my mind off it.
Contents
1 Introduction 1 2 Nanocube synthesis 2 2.1 Theory . . . 2 2.2 Methods . . . 3 2.3 Results . . . 4 2.4 Conclusion . . . 5 3 Substrate fabrication 6 3.1 Methods . . . 6 3.2 Results . . . 9 3.3 Conclusion . . . 11 4 Capillary assembly 12 4.1 Theory . . . 12 4.2 Methods . . . 15 4.3 Results . . . 18 4.4 Conclusion . . . 235 Conclusions and outlook 24
References 25
1
Introduction
The ability to manufacture high-performance (opto)electronical devices is vital for progress in fields such as photonics, photovoltaics and biotechnology [1]. Over the last two decades, these devices have been made smaller and smaller, while increasing in performance. Producing devices at the micro- or nanoscale requires accurate control over the placement and properties of the materials used to build them. These devices are traditionally fabricated using top-down methods, such as electron-beam lithography, in which material is being etched away with high precision to obtain the desired dimensions [2]. However, top-down methods are intrinsically limited by the properties of the bulk material and often result in defects, grains and non-perfect geometries. The demand for progress in these advanced fields illustrates the need for novel fabrication methods of nanoscale devices, independent of material properties.
In the last five years, bottom-up approaches implementing directive assembly have been suggested for the manufacturing of nanoscale devices [3]. In directive assembly, the placement of a component is driven by either chemical bonding or by magnetic, electrostatic or capillary forces. Directive assembly mediated by capillary forces has recently been a topic of particular interest to many [4]. Capillary assembly, as it is called, assumes that the surface tension of a solvent exerts a force on colloidally dispersed particles in that solvent. The force exerted by surface tension depends solely on the geometry of the object that causes a surface deforma-tion. It therefore answers the need for an assembly technique which is independent of material properties. Capillary assembly has been used to produce arrays of a single material with high spatial accuracy [5]. To date, however, little attention has been devoted to the application of capillary assembly for placing two (or more) different materials in close proximity.
The current study explores the possibilities of directive assembly through capillary forces to create arrays of multiple differing materials with nanometre-scale precision. In particular, we concentrate on creating hydrophobic silicon substrates with shallow topographical traps, such as grooves and trenches, and using these traps to act as pinning sites for colloidally dispersed silver and palladium nanocubes. The cubes are dispersed in ethanol, which shows a high contact angle when deposited on hydrophobic substrates. This high contact angle is essential in establishing a contact line that drives the cubes both over the substrate and into the traps. The traps are designed to prefer a certain cube size: wide traps are made to fit the 75 nm silver cubes, while thinner traps can fit 19 nm palladium cubes, as can be seen in Figure 1. By patterning trenches into a silicon substrate with alternating wide and thin sections, a line of cubes alternating between silver and palladium is created.
This novel assembly method is not limited by any material properties other than the abil-ity to form a stable colloid. The entire assembly process itself is based on purely geometrical features of the traps and the cubes. It is not limited to silver or palladium cubes but can be ap-plied to any combination of metals, semiconductors or insulators required to fabricate a device. The extreme accuracy that can be achieved by template assisted capillary assembly makes it a promising approach for the fabrication of high-performance nanoscale (opto)electronical devices.
2
Nanocube synthesis
The capillary assembly of cubes is determined only by the geometry of the cubes and limited only by the ability to form a colloidal suspension. Any material that is shaped into a cube with a specific size can be assembled using the same technique and parameters. Therefore, obtaining cubes with well-defined features is of vital importance for this study. Here, 75 nm silver and 19 nm palladium cubes are used. The silver cubes are bought from nanoComposix, while the palladium cubes are synthesised in-house. The size disparity between the two types is large enough to show preferential trap sizes for each of them. Lastly, the size distribution of the cubes is narrow, which is essential for large-scale assembly. This section starts with theory on nanocube synthesis, then shows how the palladium cubes are synthesised. Lastly, the size distribution of the cubes will be quantified.
2.1
Theory
Nucleation A colloidal dispersion of nanoparticles is generally synthesised in a solution based reaction. An inorganic compound containing a metal is reduced to individual ions [6]. When the concentration of ions reaches a specific threshold, nucleation occurs, which is the first step of nanocube synthesis [7]. Nuclei bind together to further decrease the energy of the system, which causes the ion concentration in the solution to drop. When this concentration drops below the critical threshold, it is no longer possible to form new nuclei. Adding the remaining metal ions to the nuclei happens through diffusion, and some binding sites are preferential over others. This causes the nuclei to assemble into well-defined shapes called crystals [8].
A crystal is simply an ordered period arrangement of an atom, ion or molecule. Most metals and semiconductors crystallise in a specific fashion. The crystalline type is determined by the symmetries in the unit cell of the crystal, which is the simplest repeating unit in the crystal. Examples are: simple cubic, body-centred cubic (bcc) and face-centred cubic (fcc). To form the strongest metallic bonds, most metals crystallise in the fcc arrangement, which allows for the maximum packing density of spheres [9].
Growth In the fcc arrangement (Figure 2a), the crystal can grow along specific crystallo-graphic planes. Growth along these planes determines the macro geometry of the crystal. To minimise free energy, the crystal tries to minimise the surface to volume ratio by resembling a sphere. To do so, it grows along the {001} and {111} facets, shown in Figure 2b and 2c, respectively. Growth along these planes forms an edge-truncated octahedron, which closely resembles a sphere and is shown in Figure 2d.
Figure 2: Schematic representation of a face-centred cubic crystal. a: Unit cell. b: The {001} facet shown in red. c: The {111} facet shown in green. d: An edge-truncated octahedron with the h001i and h111i directions shown in red and green, respectively.
Growth control Obtaining well-defined cubic structures relies on proper capping of un-wanted growth facets. Poly(vinylpyrrolodine) or PVP is a long chain water soluble polymer and is used in many technical applications as a stabilising agent, adhesive, emulsifier or sur-factant [10]. The PVP molecule contains both a strong hydrophilic pyrrolidone moiety and a moderately hydrophobic alkyl group. When dispersed in a polyol, PVP strongly binds to the {001} facet, allowing growth only in the h111i direction. As can be seen in Figure 2d, growth along the green planes primarily forms a cube [11].
2.2
Methods
The palladium nanocubes were synthesised according to a previously reported procedure, which should yield cubes with a diameter of 15.2 nm [12]. First, 8 ml ultrapure Milli-Q water (Mil-lipore) was added to a new 20 ml vial. Then, 60 mg ascorbic acid (Sigma Aldrich), 600 mg KBr (99%, Sigma Aldrich) and 105 mg PVP (55000 MW, Sigma Aldrich) were added, resulting in a clear solution. The vial was suspended in a silicone oil bath at 80 °C under magnetic stirring (300 RPM). Then, in a 10 ml vial, 57 mg Na2PdCl4 (Sigma-Aldrich) was added to
3 ml Milli-Q water, which yields a dark brown colour. Using a 5 ml Eppendorf pipette, the Na2PdCl4 solution was added to the 20 ml vial, also turning it brown. The vial was quickly
capped with Parafilm and the reaction was let to proceed for 3 hours.
The product was collected by a series of centrifugation and dilution steps. First, 2 centrifuge tubes were filled with each 1 ml product and 4 ml ethanol, which were spun at 4000 rpm for 10 minutes. Then, 2 ml of the supernatant was taken from each and was distributed over 4 new centrifugation tubes. The tubes were filled to 5 ml with ethanol and spun at 8000 rpm for 30 minutes. Next, the supernatant was discarded and only the precipitate was left in the tubes. The tubes were filled up to 5 ml again and spun at 8000 rpm for 30 minutes. This process was repeated 6 times and is done to wash off residual PVP from the cubes. In this report, no sonication was performed in between each repetition, it is however advised to do so to ensure redispersion and avoid the trapping of excess PVP in the pellet. With each repetition the supernatant becomes more transparent. In the final step, the precipitate from all 4 tubes was taken and added into one, which was spun at 4000 rpm for 10 minutes. From this tube the supernatant (about 2 ml) was taken and put in a vial with a screw cap, which was labelled with ”Pd Cubes”. The colloidal dispersion has a semi-transparent dark brown colour with no visible chunks or precipitate, even after several months.
Figure 3: Histogram showing the distribution of diameters of the synthesised palladium nanocubes. The average size of the cubes is 18.7 ± 1.3 nm.
2.3
Results
Size distribution The size distribution of the palladium cubes was determined using an imaging and measurement software called ImageJ. First, 10µl colloidal dispersion was dropped on a flat untreated silicon substrate. Ethanol was left to evaporate for 5 minutes, after which the substrate was placed in a scanning electron microscope (SEM, FEI Verios). High resolution images taken with the SEM were cropped, increased in contrast and converted to binary black and white images (Figure 4). ImageJ was used to count the number of particles and calculate their surface area. The assumption is made that the particles are cubic on average so the square root of the surface area gives the diameter. Figure 3 shows the distribution of the diameter. The size of the cubes is 18.7 ± 1.3 nm.
Figure 4: Original SEM images of palladium cubes on silicon and corresponding binary black and white masks. The bottom row shows the overlay between the mask and the SEM image. Three cubes are left out from selection because they are not lying flat on the silicon, which would result in a miscalculation of the diameter. It can be seen that the size of the cubes is not being over- or underestimated.
The silver cubes from nanoComposix have a diameter of 74 ± 7 nm (determined using a JEOL 1010 Transmission Electron Microscope and ImageJ), are dispersed in ethanol and coated with PVP (55000 MW) [13].
2.4
Conclusion
In this Chapter the synthesis of palladium nanocubes is shown. The synthesis yields a colloidal dispersion of cubes with a diameter of 18.7 ± 1.3 nm. Furthermore, silver nanocubes are bought from nanoComposix which show a diameter of 74 ± 7 nm. The next Chapter will cover the other key component required for capillary assembly: the substrate.
3
Substrate fabrication
Figure 5: Schematic representation of the five main steps for the fabrication of a patterned silicon substrate.
This Chapter reports on the fabrication of nanoscale trap patterns on silicon substrates and the characterisation of the trap dimensions. Trap dimensions are an integral part of achieving suc-cessful assembly, and even slight deviations can cause a drastic reduction in yield [5]. There-fore, accurate and consistent fabrication at the nanoscale is of utmost importance for this re-search. The basic steps for making patterned silicon substrate are: resist spincoating, electron beam exposure, resist development, silicon etch-ing and lastly resist strippetch-ing (Figure 5). In this research, multiple large area (20 × 20 µm) fields of traps have been etched into silicon with indi-vidual trap sizes down to 20 nm. The traps show uniformity over the entire fields with little irreg-ularities or defects. The traps also show fairly vertical side walls (95° - 100°) which is of partic-ular interest when performing capillary assembly. In the next section, the methods and optimisation of the substrate fabrication are discussed.
3.1
Methods
Preparation All steps were performed in an
ISO-class 7 cleanroom (AMOLF NanoLab Ams-terdam) to minimise the contamination from
air-borne particles. First, nine 12 × 12 mm p-type silicon (Siegert, <100> orientation, 525 ± 20 µm thick) samples were taken. These samples were cut from a 4 inch wafer using a diamond saw, resulting in chunks of silicon on the surface. To remove these, the samples were placed in a teflon basket, submerged in water in a 100 ml vial and placed in a Bransonic Ultra sonicator for 15 minutes, after which pieces of silicon are observed at the bottom of the vial. Next, the basket was taken out and sprayed with water to remove more residual silicon chunks. Then, a base-piranha solution was prepared by mixing 10 ml of NH4OH (30% in water, Sigma-Aldrich)
and 10 ml H2O2 (30% in water, Sigma-Aldrich) with 50 ml type 1 H2O (18.6 kOhm resistivity)
heated to 75 - 80 °C. After placing the basket in the solution it begins to bubble vigorously, otherwise an additional 5 ml of H2O2 was added. The samples were left to react in solution
for 15 minutes, after which they were thoroughly rinsed in a large volume of water. The H2O2
in base-piranha decomposes to form water and oxygen, which removes most organic residues, while the NH4OH is a strong complexant for many metals, removing them from the silicon
surface. The samples were removed from the teflon basket one by one and blown dry with a N2 gun. The last cleaning step involved a O2 descum recipe (30 s, 25 sccm O2 flow, 20 °C, 5
mTorr pressure, 50 W HF forward power) in the Oxford Plasmalab 80+. This step cleans off organics and activates the surface of the sample for further fabrication steps.
Resist spincoating A fresh surface of silicon-dioxide, which is produced in the plasma-cleaning step, is hydrophobic. However, water vapour in the air reacts with this surface to form silanol groups (Si—OH), making the surface hydrophilic [14, Chapter 11]. Most resist materials
do not adhere well to hydrophilic surfaces, therefore, an adhesion layer was first coated on the surface. Hexamethyldisilazane (Sigma Aldrich), also called HMDS, converts hydrophilic silanol groups into hydrophobic siloxanes (Si—O—Si). A small drop of HMDS was dropped on the sample which was spun with a Delta 80 Suss microtec spincoater for 35 seconds at 4000 RPM. After spinning, the sample was baked at 150 °C for 1 minute. Then, the photoresist can be applied. A photoresist is a light-sensitive material that acts as a coating layer on the sample. Exposing specific regions of the photoresist to radiation degrades or strengthens the material, making those regions either soluble or insoluble in a developer. The first is called a positive resist, where the exposed regions become soluble, while the latter is called a negative resist, where the exposed regions become insoluble [15, Chapter 1]. CSAR 62.09 (also known as SX AR-P 6200, Allresist) is a positive electron beam lithography resist and has a 10 nm resolution [16, 17]. CSAR contains a photoacid generator that helps break down the polymer backbone during exposure, making it soluble. Since CSAR is stored in a fridge the bottle has to warm up to room temperature before it is opened, otherwise water vapour in the air condenses and contaminates the resist. After roughly 30 minutes the bottle was opened and 200µl was taken using an Eppendorf pipette. The resist was dropped on the sample until it was coated, while the residual resist and the pipette tip are discarded. The sample was spun at 4000 RPM for 45 seconds, after which it was baked at 150 °C for 3 minutes. The thickness of the resist is roughly 107 nm, determined with a KLA-Tencor P7 profilometer.
Electron beam exposure After the sample has been coated with the photoresist, it is ready for radiation exposure. The desired trap dimensions are below a size of 100 nm, meaning that using light as a radiation source provides an insufficient resolution. Therefore, it is necessary to use electron beam lithography (EBL). The Raith VOYAGER is an EBL system with a beamwidth of 1 nm and is capable of writing lines down to 7 nm width [18]. Samples were placed in the VOYAGER and the accompanied Raith software was used to align and focus them. A column mode of LC40 was used and a beam current of 0.2200 nA was measured. Two types of patterns were written, trenches and squares (Figure 6). Three 20 × 20 µm fields of trenches and five fields of squares were written, each with varying trap dimensions. The designs for these patterns were produced by a Python script written by Andries Lof. A crucial parameter for exposure is the dose that the sample is subjected to. The dose is specified in coulombs per square centimetre, and with the beam current it is a measure of the amount of power the electron beam puts through. A higher dose lowers the exposure time needed to write a patterned area. Writing time, which is determined by the total charge density deposited in the sample, follows the equation: D · A = T · I, where D is the dose, A the exposed area, T the writing time and I the beam current. However, a high dose also hinders the ability to write sharp corners, especially for small designs. After optimising the writing time versus writing accuracy, the dose selected was 170 µC cm-2 for trenches and 150 µC cm-2 for square designs, resulting in an exposure time of roughly 30 minutes per sample.
Resist development During exposure, the polymer backbone of the CSAR is broken down by chain scission, making those regions soluble in a developer. Developing the resist consisted of dipping the sample in four different solutions. First, the sample was dipped in pentyl-acetate (99%, Sigma Aldrich) for 60 seconds, which dissolves most of the exposed CSAR. However, some unwanted residue is left on the sample which was dissolved by dipped the sample in o-xylene (>99%, VWR) for 6 seconds. Next, the pentyl-acetate was rinsed off by dipping the sample in a mix of methyl-iso-butyl ketone (99.8%, VWR) and 2-propanol in a 9:1 ratio for 15 seconds. This mix is commonly abbreviated to MIBK/IPA 9:1. Lastly, the sample was thoroughly rinsed in 2-propanol to remove any residual organic compounds and stopping all development. The sample was put back in the samplebox after being blown dry with N2.
Figure 6: SEM (FEI Verios) images of the silicon substrates with the trench and square trap designs. a: Trenches top-down. b: Squares top-down. c: Trenches 45° tilt. d: Squares 45° tilt.
Silicon etching When the resist is developed, the exposed regions of silicon are still coated by a native oxide layer (SiO2) of at least 1 nm thick [19]. When dry-etching, the etch rate
for SiO2 is different than the etch rate for pure silicon. The time it takes to etch the SiO2
layer is called the breakthrough time, and is important to consider when a very specific trap depth is required. Different gasses have different etching speeds, some etch isotropically while others etch straight down. In this research, traps with straight side walls are essential, so a specialised etching recipe is created. Optimisation of this etching recipe can be found in the Results section of this Chapter. First, the plasma etcher (ICP/RIE Cobra) was pre-heated to 60 °C. The sample was stuck on a 4 inch carrier wafer using Fomblin oil and placed in the Cobra when the temperature reached 60°C. Etching the sample using only HBr (30 s, 50 sccm HBr, 20 sccm He backing, 60°C, 118.0 V DC Bias, 7 mTorr pressure, 19 W HF forward power) resulted in a trap depth of roughly 50 nm, which was confirmed by tilted SEM images and by making cross-cuts using a focused ion beam (FEI Helios dualbeam).
Resist stripping and fluoro-silanisation After etching, the stripping of residual resist was done by putting the samples in a base-piranha solution (as described in the Preparation subsection) for 15 minutes. The fabrication itself is now complete, but for successful capil-lary assembly the sample needs to be treated such that it shows a high contact angle with the assembly solvent. In this research the solvent used is ethanol, which shows a low con-tact angle on hydrophilic surfaces. Therefore, the sample was coated with a monolayer of fluorosilane (trichloro[1H,1H,2H,2H-perfluorooctyl]-silane, Sigma Aldrich) in a process called fluoro-silanisation. First, the sample was plasma cleaned in the Oxford Plasmalab 80+ using the same O2 descum recipe as in the Preparations subsection. Meanwhile, a vacuum oven
(SalvisLab Vacucenter VC20) was pre-heated to 50 °C and the fluorosilane was taken from the refrigerator. Once the plasma cleaning was done, the sample was immediately placed in the vacuum oven together with 20µl of fluorosilane, which was deposited on a small teflon hourglass using an Eppendorf pipette. The oven was quickly closed and pumped down to vacuum. At 50 °C the fluorosilane evaporates and fills the oven with vapour. The silanol (Si—OH) groups at the surface of the sample hydrolyse and form siloxane linkages (Si—O—Si) with the fluorosi-lane, much like HMDS does. The sample was left in the oven for at least 45 minutes. More was generally not a problem as the surface of the sample eventually becomes saturated with a monolayer of fluorosilane. This process increases the contact angle of the sample with ethanol significantly, as is shown in Figure 12a. The sample was rinsed one last time in 2-propanol, after which it was blown dry with N2 and was finally ready for assembly.
3.2
Results
Dose test During electron beam exposure the dose is an important parameter to consider. A higher dose results in a faster write time, but it can influence the accuracy of the writing. A dose that is too low might not even break through the full resist layer. This subsection shows images of the trenches and square trap patterns written at different doses (Figure 7). From these images a selection is made for the optimal dose based on qualitative features: sharpness of the corners, smoothness of the edges and width of the barriers.
Figure 7a shows the square design written at different doses. At 130 µC cm-2 large regions of the resist are not fully penetrated and details in the design are not transferred. However, doses of 170µC cm-2and up cause the barriers between traps to disappear, creating a line traps
instead of individual squares. For squares, the optimal dose was chosen to be 150 µC cm-2. This dose results in nice squares with defined corners, while leaving the barriers intact.
Figure 7b shows the trench designs written at different doses. It shows that doses up to 140 µC cm-2are too little to write the thin parts of the trench and result in disconnected line-pieces.
doses of 180 µC cm-2 and over cause deformation of the side walls, with many imperfections
appearing. For trenches, both 160 and 170 µC cm-2 show beautiful traps, but 170 µC cm-2 was
chosen as the optimal dose since the writing time is lower.
Figure 7: Dose test for etching silicon substrates. Scale bar: 100 nm. a: Square trap designs written with doses varying from 130 µC cm-2 to 190 µC cm-2. b: Trenches trap designs written with doses varying from 120 µC cm-2 to 190µC cm-2.
Figure 8: a: Top-down SEM image of a patterned silicon substrate with a platinum block evaporated on top. The block is about 1µm in height. b: SEM image of a silicon sample tilted at 52°. The block of platinum, as well as the underlying silicon, have been etched away using a FIB, so the profile and depth can be observed. The inset shows a zoom-in of the trench.
Etch test Another crucial optimisation is the etching recipe used to etch down silicon. In this fabrication dry-etching is employed, where the substrate is subjected to a stream of glow discharge from a specific gas mixture. The gas mixture and the etch time determine the profile and the depth of the etch. For applications in capillary assembly, specifically a sharp etch profile with vertical side walls is important. This subsection covers various gas mixtures that have been tested for their etch profile and chooses an optimal etching recipe for assembly.
The most commonly used silicon etching recipe at the NanoLab cleanroom is Cl2 + HBr/O2.
In this recipe the first step is to use Cl2 gas to remove the native oxide layer on the silicon.
Then, the Cl2 is pumped out and a mixture of HBr and O2 is used. The HBr selectively etches
Si2 over SiO2 while the O2 binds to the side walls, passivating them from further HBr etching.
It is observed, however, that this etching recipe often leads to slanted side walls, which is detrimental to capillary assembly. A second etching recipe that is tested included only Cl2 gas.
While Cl2 is normally preferred as a native oxide etcher, it can also etch silicon. The last etching
recipe that is tested only uses HBr. It is observed that HBr etching can lead to very straight side walls, but the etch depth depends greatly on the amount of contact the samples have had with air. Since HBr shows a high selectivity of etching Si over SiO2, the initial breakthrough
of the native oxide layer takes a relatively long time. If the sample contains inconsistencies in the thickness of the native oxide layer, some regions might be etched substantially deeper than others. All three etching recipes are optimised for an etch depth of 50 nanometres by etching multiple samples and fitting an etch speed.
The profile and depth of the etch are observed by making a crosscut through the sample using a focused ion beam (FIB, FEI Helios dualbeam) with a gallium ion source. In order to
Figure 9: Cross-section of a trench made with FIB. Dark-grey is silicon, the lighter grey with small white grains is the deposited platinum. a: Recipe: Cl2, then HBr/O2. b: Recipe: Cl2. c: Recipe:
make a clean cross-cut first a block of platinum is deposited on the substrate using electron beam induced deposition (Figure 8a). Next, the substrate is tilted to 52° to align it with the ion beam column and a large hole is etched (Figure 8b). The trench is filled with the deposited platinum which results in a clean crosscut that is observed using the SEM. An added benefit of depositing platinum is that it contains many small grains, which can be used to focus on. A zoomed in image is shown in the inset of Figure 8b.
Figure 9 shows the cross-section of a trench for the three different etching recipes. The angles have been calculated using imageJ. It can be seen that the recipe using only HBr shows the most vertical side walls. Optimising the depth of the etch involves simply etching for a longer or shorter time. For HBr, 30 seconds of etching yielded roughly 50 nm.
3.3
Conclusion
In this Chapter the fabrication of patterned silicon substrates is shown. Two types of patterns are written: trenches and squares. A single substrate contains three 20 × 20 µm fields of trenches and five 20 × 20 µm fields of trenches. The patterns are etched down to a depth of 50 nm and show fairly vertical side walls (95°- 100°). All substrates are coated with a monolayer of fluorosilane, making them very hydrophobic. Now that the nanocubes are synthesised and the substrates are fabricated, assembly can begin.
4
Capillary assembly
This Chapter reports on the implementation of capillary assembly as a tool to create nanoscale patterns of silver and palladium cubes and explores the parameters impacting assembly accuracy and yield. In capillary assembly, a solvent containing colloidally dispersed nanoparticles is carefully dragged over a patterned substrate using a glass slide. A schematical representation is shown in Figure 10. By accurately controlling the evaporation of the solvent the particles are brought to the meniscus and down to the substrate, where they are pinned by the geometry of the pattern. Capillary assembly has been used to assemble metallic nanorods and spheres [5, 20, 21], semiconducting nanowires [22], polystyrene beads [23, 4] and live cells [24, 25]. The approaches taken in all types of assemblies are similar but have to be optimised for their respective particles and patterns, so no singular method can be followed blindly. In this research sequential assembly of 75 nm silver cubes and 19 nm palladium cubes dispersed in ethanol on a patterned silicon substrate was studied. The theory behind capillary assembly and the methods for optimising deposition are described next.
4.1
Theory
Principle workings In essence, capillary assembly is a controlled dewetting process of a solvent containing nanoparticles. It relies on two driving forces: a long-range driving force and a short-range driving force [3], depicted by the red and green arrows in Figure 10, respectively. Properly managing these two driving forces is essential for a successful assembly.
The long-range driving force is a convective flow responsible for bringing colloidally dis-persed particles to the surface of the solvent. Evaporation of molecules at the surface causes a deformation of the meniscus. It is energetically favourable to have a smooth air-liquid interface due to surface tension, so new molecules flow to the surface. This flow of molecules carries the dispersed nanoparticles towards the meniscus, building an accumulation zone of particles. However, this effect is counteracted by two other forces [21]. Firstly, a concentration gradient is created, with many particles near the surface and few particles in the bulk solvent. This
Figure 10: Side-view perspective of a schematic representation of capillary assembly. Shown are the silicon substrate (dark blue), the nanoparticles (violet) and the glass slide (grey). The long- and short-range driving forces are depicted by the red and greens arrows, respectively. The figure is not to scale: the cubes are 10 - 100 nm in size while the distance between the glass slide and the substrate is 50 - 500µm.
Figure 11: a: Top-down view of an assembly. The black background is the silicon substrate and the yellow lines are stick-slip events of silver nanocubes. In this assembly, either the evaporation rate was too high or the assembly speed was too slow. b: Side-view of an ethanol droplet on a silicon substrate. The contact angle is defined as the angle between the two yellow lines. c: Coffee stain on a desk. The thick coffee deposit around the edges is clearly visible.
concentration gradient causes a diffusive particle flux opposite to the convective flow. Sec-ondly, a recirculation flow is induced by the meniscus movement, which is also opposite to the convective flow. Only when the convective flow is greater than the other two forces combined, an accumulation zone is formed. The well-studied coffee-ring effect relies on the same mech-anism [26]. Solid coffee particles that are dispersed in the solution flow to the edges during evaporation, causing a spilled drop of coffee to leave behind a thick ring-like deposit along the perimeter (Figure 11c). Establishing a stable accumulation zone of particles near the surface is the first key factor in achieving a successful assembly.
Next, the cubes have to be placed and pinned inside the traps, which happens in several steps [1, 4]. First, the cubes fill up all available space in the accumulation zone, most importantly the traps themselves, due to minimisation of free energy. The cubes are placed in the correct positions long before the meniscus moves over the traps, which stresses the importance of a stable accumulation zone. Once the meniscus does move over the traps, the short-range driving force takes over. Just like the long-range driving force, the short-range driving force depends on the desire of the solvent to retain its surface profile. When the meniscus moves over a trap, it is pinned on the trailing edge, as shown in the inset of Figure 10. If the trap is filled by a particle, the meniscus is then locally disturbed. This deformation of the contact line causes a force, called the capillary force, to be exerted on the particle, pinning it into the trap. When the meniscus moves further, it eventually breaks and is depinned from the trailing edge. A small amount of solvent resides in the trap together with the cube and slowly evaporates.
Once the cubes are placed inside the traps they are kept there through Van der Waals forces. Even though these forces are comparatively weak, the cubes are so small that gravitational forces are not strong enough to eject them.
Most importantly, this entire assembly method is independent of material properties. Only the colloidal stability and the geometry of the cube impact the process. As long as the assembly material can be made into cubes and can be stably dispersed in a solvent any material can be assembled in an arbitrary array. It is for this reason that capillary assembly is an attractive method for building nanoscale (opto)electronic devices.
Assembly control In this subsection the parameters that control an assembly are discussed, and general values for these parameters found in literature are given.
The first step in achieving a successful assembly is controlling the long-range driving force through the temperature. This convective flow is caused by the evaporation of the solvent at the surface. A higher evaporation rate means a stronger convective flow, which is needed
to build up a stable accumulation zone. However, when the evaporation of the solvent is too high, the accumulation zone gets overfilled with particles which will cause the meniscus to periodically break, causing the so-called stick-slip phenomenon (Figure 11a) [27]. To achieve the desired temperature, a heating element is generally installed underneath the substrate. Reported processing temperatures for nanoparticles are 30 - 50 K above the dew point [3, 28, 29]. However, most reported assemblies have been performed using water as their solvent while this research uses ethanol, which has a substantially lower boiling and dew point. The consequences of this choice are discussed in the Results section of this Chapter.
Secondly, the assembly speed must be controlled. The speed and the temperature have to be balanced to sustain the accumulation zone. Slow evaporation with a high assembly speed results in depletion of the accumulation zone and no filling of the traps. Fast evaporation with a slow speed results in stick-slip on the substrate and random deposition of the particles. Moreover, the meniscus must move with a constant velocity to ensure reproducible results over a large area, which is why the glass slide is dragged over the substrate using a motorised stage. Reported assembly speeds range from 12 - 300 µm min-1, with more complex trap patterns
requiring slower speeds [3]. This low speed is required to ensure that the recirculation flow caused by the meniscus movement does not overcome the convective flow. For more complex trap patterns where traps are located very close to each other, pinning and depinning happens at a faster rate, stressing the importance of a low speed.
Figure 12: a: Difference between contact angle of ethanol on a fairly dirty silicon substrate (left) and a freshly fluoro-silanised silicon substrate (right). b: Dependence of the direction of the capillary force vector and the contact angle on the distance between the substrate and the glass slide.
Thirdly, the assembly is influenced by the contact angle of the solvent on the substrate. The contact angle is defined as the angle between the edge of a drop of solvent on a substrate and the substrate itself (Figure 11b). The contact angle can be controlled through solvent choice, substrate surface treatment or slide height. Contact angles reported for capillary assembly range from 30° to 75° [3, 29]. Generally, the solvent used is water, which has a higher contact angle on silicon than ethanol. To compensate the silicon surface is treated with a fluoro-silanisation step, making it more hydrophobic (Figure 12a). The contact angle also determines the direction of the capillary force vector. A low contact angle results in a large vertical component of the force vector, while a high contact angle results in a stronger horizontal component, as can be seen in Figure 12b. By increasing or decreasing the distance between the silicon substrate and the glass slide the contact angle can be fine-tuned such that the direction of the capillary force vector is optimal for assembly.
Trap geometry Two types of trap arrays have been designed and fabricated, both with a different approach to assemble two materials in close proximity. Trenches, as shown in Figure 13a are long etched lines with alternating wide and narrow segments. Note that Figure 13a1
Figure 13: Schematic representation of the traps etched in silicon. The lighter blue is flat silicon, the darker blue are the etched regions (roughly 50 nm deep). The top row is simply the etched silicon, the middle row is after the first assembly with silver cubes and the bottom row is after a second assembly with palladium cubes. a1, a2: Top-down and side-on (resp.) view of the trenches design. b1, b2: Top-down and side-on (resp.) view of the squares design.
is a top-down view and Figure 13a2 is a side-on view. The larger silver cubes do not fit in the narrow segments, and only assemble in the wide parts, shown in the middle row of Figure 13a. The bottom row shows the second assembly, where the palladium cubes are pinned in the narrow segments in between the silver cubes.
For the squares design the approach is slightly different (Figure 13b). First, the large silver cubes assemble in a tight array of square traps, which is shown in the middle row of Figure 13b. The silver cubes are larger (75 nm) than the traps are deep (50 nm), making them protrude from the traps. Therefore, they act as new pinning sites for the palladium cubes in the second assembly, as can be seen in the bottom row of Figure 13b.
4.2
Methods
This section describes the procedure of performing a capillary assembly and how the previously described parameters are controlled.
Preparation One 12 x 12 mm patterned silicon substrate was taken and checked under an optical microscope (Zeiss Axio Imager.A2m) for large defects or dust particles. The substrate was observed using the dark-field mode under a magnification of 20x. If large (>20µm) particles were found on or near the trap patterns the substrate was cleaned with opaque adhesive tape and sonicated (Bransonic 2510 E-MT) in a 50 mL vial of ethanol for 30 minutes. After drying the substrate with a pressurised N2 nozzle, the presence of particles was checked again. Most
of the time the substrates were clean enough after this procedure but if large chunks remain on the substrate it is plasma cleaned and fluoro-silanised again as described in the Chapter on Substrate Fabrication. The substrates are stored in a sample-box with corresponding labelling when not in use. Next, one 1 mm thick microscope slide (VWR) was taken which has been plasma cleaned and fluoro-silanised in the same way as the silicon substrates. The glass slide was labelled on one side to distinguish the side that has been made hydrophobic from the untreated side. The slide was stored in an aluminium foil when not in use. Further, a 50 mL capped vial of ethanol and a Eppendorf pipette capable of holding 10 - 20 µl with the corresponding pipette tip were gathered. Lastly, the vial containing the nanocube solution was sonicated for 30 seconds to break up possible large clumps of cubes.
Figure 14: Image of the home-built capil-lary assembly setup. a: Nikon SMZ800N stereo microscope attached with a Basler acA2440-35um Area Scan Camera, used for continuous top-down viewing of the assem-bly. b: Basler acA1920-155um Area Scan Camera, used for side-view height alignment of the substrate and the slide. c: Tubes attached to a Gast DOA-P704-AA vacuum pump and a Laird PR-59 temperature con-troller. d: Silicon substrate. e: Glass slide. f: Thorlabs MTS25-Z8 motorised transla-tion stage oriented in the Z-directransla-tion, used to move the glass slide up or down. g: Thor-labs stage oriented in the X-direction, used to drag the substrate out from underneath the glass slide. h: Thorlabs R-stage, used to align the substrate and the glass slide paral-lel to each other. i: Thorlabs stage oriented in the X-direction, used to align the sub-strate and the glass slide to be in focus for the back-microscope.
Assembly Figure 14 gives an overview of the home-built capillary assembly setup. After all preparation has been performed, the substrate was taken from the sample-box and placed on top of several small holes in the sample-holder (arrow d in Figure 14). The holes are connected through tubes to the vacuum pump (Gast DOA-P704-AA) which was turned on to hold the substrate in place. Next, the desired temperature was set through the Laird software provided with the Laird PR-59 temperature controller. Warm or cold water flows through the blue tubes (arrow c) heating or cooling the sample-holder. Temperatures used range from 14 °C - 22 °C.
Then, the glass slide was put in place by tightening the screws on the Z-stage (arrow f, Thor-labs MTS25-Z8 motorised translation stage). The glass slide and the substrate were checked to be level using a spirit level. The setup is built inside a box that can be closed off but is outfitted with two viewing windows for the top and back camera. At this point the box was closed and camera software (Pylon viewer) was used to view the image from the back camera (arrow b). LED lights installed around the perimeter of the top microscope allow for various lighting conditions focused on the substrate and glass slide. Using the software designed by Jorijn Kuster (Controlled Particle Placement) the X- and Y-stage (identical Thorlabs transla-tional stages to the Z-stage) were moved to bring the substrate and the glass slide in focus of the back camera (Figure 15a). When both were in focus the Z-stage was used to adjust the distance between them. Distances used range from 50 µm - 800 µm.
Next, the substrate was centred under the glass slide and moved such that the left-most patterns on the substrate were 2 mm to the right of the edge of the glass slide. This 2 mm distance is needed for the assembly to create a stable accumulation zone before the meniscus passes over any patterns. In most experiments the next step was to create an ethanol environ-ment in the box. This was done by putting an open vial of ethanol in the box and then quickly closing the box off.
Figure 15: a: Side perspective of the alignment between the glass slide (top right) and the substrate (bottom). b: Top-down perspective of an assembly in progress. The black background is the silicon substrate, the orange squares are the patterned fields of traps, and on the right the yellow nanocube solution is visible which is trapped between the glass slide and the substrate. In this image, the meniscus moves from left to right, and has already moved over all of the traps. It is also visible that some defects on the substrate surface have produced clusters of nanocubes (the yellow specks).
While waiting at least 5 minutes for enough ethanol to evaporate the stage control software was used to set up the assembly speed. Since the stages are moved by rotating screws with gears the speed at which they move is not linearly adjustable. They can move at for example: 200 µm min-1, 300 µm min-1, 400 µm min-1 and higher, each in increments of roughly 100 µm min-1. In the initial stages of this research the lowest possible speed (200µm min-1) was often
used, as literature suggested that lower speeds are better for assembly yields [3]. However, 200 µm min-1 still seemed too fast so some adjustments have been made to the setup to reach even lower speeds, which will be discussed in the Results Section of this Chapter. Lastly, the relative humidity was checked with a thermo-hygrometer (VWR Traceable Jumbo Digital Thermo-Hygrometer) that is installed in the box.
When all preparations are performed, 10 - 20 µl of the nanocube solution was taken using the Eppendorf pipette. The box was opened briefly to drop the solution on the substrate near the edge of the glass slide and was then closed again to maintain a high ethanol vapour pressure. Capillary forces moves the droplet in between the glass slide and the substrate as can be seen in Figure 15b. The assembly was then started which takes 30 - 60 minutes, depending on the assembly speed. In total, the glass slide moves 6 mm over the substrate, enough to move over all patterned fields and to evaporate all the solvent. The assembly direction is parallel to the lines of patterns.
During assembly the top-view microscope was used to check for any stick-slip events, large clusters forming or any other inconsistencies that could lead to a failed assembly. The temper-ature was routinely checked as the heating element tends to fluctuate. After the assembly was completed, the box was opened slightly, the vial of ethanol taken out and capped and an air vent was placed over the opening of the box to extract the ethanol vapour. After 5 minutes the vapour was mostly gone and the box was opened completely to gather the glass slide and the substrate. The glass slide was cleaned with ethanol, blown dry with pressurised N2 and
wrapped in an aluminium foil. The substrate was placed back in the sample-box for optical characterisation or was used directly for a subsequent palladium assembly.
4.3
Results
First of all it should be stated that directive assembly with size-dependent placement of cubes was not realised in this research. Silver and palladium nanocubes both show a preference to assemble within traps over flat silicon, but exact silver placement and subsequent palladium trapping as shown in Figure 13 requires more thorough investigation of multiple factors. These factors determining reproducibility and precision of assembly are discussed in the following Subsections, together with the achieved results. All images are obtained by SEM using either the FEI Verios or FEI Helios dualbeam at 5 kV and 100 pA.
Silver assembly I The assembly shown in Figure 16 was obtained by closely following the parameters used in a previous report [5]. Here, 50µl of silver nanocube solution was used. The temperature T was set at 17 °C and the assembly speed v was 102 µm min-1, the slide height
D was 500 µm and the relative humidity RH was measured to be 29%.
As can be seen in Figure 16a several silver cubes are placed in their respective traps consec-utively. However, as depicted by the black arrow, the traps are slightly larger than the cubes, leaving some space around them. The traps in this image are 140 nm wide. It is observed that square traps with a smaller diameter (<100 nm) showed no assembly of cubes at all, which is most likely due to the fact that the cubes topple into the traps over their diagonal. A quick cal-culation of the diagonal length of the cubes agrees with this observation (l =√752+ 752 ≈ 106
nm). The first thing to note is that these gaps are detrimental to subsequent palladium as-sembly. As shown in Figure 13b, the silver cubes should closely assemble and form tight traps for the smaller palladium cubes. Without this tight packing, the pinning and depinning of the meniscus is disturbed, leading to unpredictable assembly.
Figure 16b shows that the cubes prefer traps over flat silicon, with no cubes outside of patterns and all cubes inside of patterns. This behaviour was observed for the entire assembly area. This means that no stick-and-slip events occurred during assembly and the evaporation rate and assembly speed are properly balanced. However, Figure 16b also shows that the yield of cubes in traps is extremely low. The initial response is that the concentration of cubes in the colloidal suspension might be too low. It is shown in the Appendix that 10 µl of cubes contains 2.6 × 109 silver cubes, while eight fields of trap patterns can maximally contain about
4.3 × 105 cubes. Thus, the concentration of cubes in the solution is sufficient. Furthermore, the low yield can be ascribed to a capillary force vector that is too horizontal. As shown in Figure 12b, the direction of the force vector is determined by the contact angle of the solvent, which can be altered by adjusting the distance between the slide and the substrate. This concern is also addressed in the Appendix. It is shown that adjusting the slide height from 150 µm
Figure 16: Silver nanocubes (white) assembled on a patterned silicon substrate (grey). a: Close-up tilted image of five consecutive cubes. b: Zoom-out showing the preference of cubes to the traps. c: Close-up showing two cubes forming a gap of 30 nm.
to 900 µm changes the angle from 63° to 50°, which is both still well within reported values [3, 4, 5]. Moreover, this observation could mean that the accumulation zone is not properly established and filled. If the evaporation rate is too low, the long-range driving force is weaker than the recirculation flow and the flow induced by the concentration gradient. Increasing the evaporation rate by increasing the temperature means that the assembly speed has to be adjusted again to prevent stick-and-slip. However, increasing the temperature above 17°C was observed to lead to many large clusters of cubes congregating and randomly depositing on the substrate, which is also shown in the silver assembly II. Lastly, the low yield can be explained by the assembly speed being too high. Since the trap pattern is extremely dense, pinning and depinning of the meniscus occurs very rapidly. An assembly speed of 100 µm min-1 is roughly equal to 1700 nm second-1. This means that the meniscus is pinned and depinned over 20
times per second. Attempts were made to decrease the assembly speed but limitations to the setup only allowed speeds down to 100 µm minute-1 to be realised, which is discussed in the Subsection on Assembly speed.
Figure 16c shows two silver cubes closely assembled next to each other. This type of tight assembly is observed in some places on the substrate but is not predictable or reproducible. The gap indicated by the arrow is 30 nm wide, which is where the palladium cubes should assemble in the next run, as shown in Figure 13b.
Silver assembly II Another silver assembly shows drastically different results (Figure 17). The assembly parameters are: 10 µl silver nanocube solution, T = 20 °C, v = 204 µm min-1,
D = 400 µm, RH = 25%. First of all it should be noted that this assembly was done on a substrate with more shallow traps. The trenches and squares on this specific substrate are about 30 nm deep (determined from tilted SEM images). Compared to the previous assembly, the temperature was increased by 3 °C and the assembly speed was doubled. Also, much less nanocube solution was used since the amount of stock solution available is limited.
Figure 17: Silver nanocubes assembled on a slightly more shallow substrate. a: Large area overview of an assembled area. Many clusters of cubes are deposited randomly on the surface. b: Slight zoom-in of the same area. It is clear that the cubes show a preference for the traps over flat silicon. c: Close-up of a line of assembled cubes. Even these shallow traps of 30 nm provide enough capillary force to pin cubes. d: Top-down view of a 5µm long line of assembled cubes.
Figure 17a shows a large area image of the assembly. It can be seen that many large clusters of cubes are deposited on flat regions of the substrate. This is the consequence of improper control on the accumulation zone. In this assembly, the evaporation rate was too high or the assembly speed was too low. Also, since less nanocube solution was used, the assembly was started with the meniscus closer to patterned fields than in the previous assembly. While this could have impact on the stability of the accumulation zone, the assembly was monitored with the top microscope and the deposition rate was observed to be stable at the moment the meniscus moved over the patterns. Another consequence of the limited amount of solution used together with the increased evaporation rate is that the solution was observed to be depleted before it had moved over all eight fields, which results in less data to be observed.
Figure 17b shows a slight zoom-in, where the clusters can be distinguished. It also shows a line of traps that is completely filled with cubes, which shows that there is still a definitive preference for these traps. It was observed for this assembly and many before it that either improper accumulation zone control or imperfections on the substrate cause large clusters to agglomerate on the surface, and that around these clusters the nanocube yield is significantly higher than in regions free of clusters. Specifically, areas ’behind’ large clusters were observed to be more filled with cubes. If the meniscus moves from left to right (which is true for all images shown here) ’behind’ means to the right of a cluster. This is explained by the fact that imperfections on the substrate form large pinning sites for the meniscus, causing the volume to fill up with cubes. Once the meniscus does eventually break, it causes a large spilling event of cubes. Most still show a preference for the traps over flat silicon, resulting in regions with high assembly yield. The same breaking events occur when the evaporation rate it too high and too many cubes flow towards the meniscus; which is essentially what happens during stick-and-slip. Figures 17c and d show close-ups of a line of assembled cubes. While the assembly is not completely controlled it can lead to long lines of assembled cubes, making the approach promising for further investigation.
Palladium assembly I Next, the 19 nm palladium cubes can be assembled on the same
substrate used in Silver assembly I. This second assembly was performed three days after the
Figure 18: The silicon substrate from Figure 16 after palladium assembly. a: Overview of the assembly. b, c, d, e: Close-ups of various assembly sites and points of interest (scale bars = 100 nm).
initial one, during which the substrate was imaged with the SEM and stored in a samplebox in air. The assembly parameters for the palladium assembly are: 15 µl palladium nanocube solution, T = 14 °C, v = 192 µm min-1, D = 500 µm, RH = 34%. An optical image of this
assembly is shown in Figure 19c.
In Figure 18a an overview can be seen of a representative area of the assembly. It is observed that the palladium assembly behaves similarly to the silver assemblies. Barely any cubes are left on the flat silicon, and all cubes that assemble do so in the trenches. However, proper assembly of the palladium cubes in between silver cubes does not happen consistently. It is mainly due to the low silver assembly yield that there are few spots to observe the behaviour of the palladium cubes. Moreover, the palladium cubes assemble in a random orientation respective to the substrate. They do not show to have any preference to assemble face-to-face, which is believed to be caused by the lack of tight traps.
The black arrow in Figure 18b shows palladium cubes that are pinned by the protruding silver cube, which is exactly the desired behaviour. In this instance, the silver cube is assembled on the left side in the trap, which happens occasionally but is not a controlled or reproducible result. Figure 18c shows the same location as Figure 16c. While palladium cubes assembled in the surrounding traps, none were pinned between the two silver cubes. These pairs of closely assembled silver cubes are found more often but assembly of palladium cubes in between them is not observed. Figure 18d shows the full packing of palladium cubes in empty parts of the trenches. The arrow in Figure 18e shows an optimal pinning site for palladium cubes in the trenches design. However, none are assembled.
Temperature dependence The temperature greatly impacts the formation of a stable ac-cumulation zone, which is essential for assembly. Throughout this research, the solvent used for the colloidal suspension is ethanol, because the bought silver nanocubes were already dispersed in ethanol. While the cubes could have been filtered and re-dispersed in water, this approach was not explored. Ethanol has a boiling point of 78°C, which is significantly lower than that of water. In most literature, water is used as the solvent and the assembly temperature is listed as degrees above the dew point, generally between 30 and 50 K. In initial assembly runs, the box was filled with air, resulting in a low dew point for ethanol. It was observed that, even at the lowest operating temperature achievable in the setup (13 °C), the evaporation rate was much too high and many breaking or stick-and-slip events occurred, as is shown in Figure 19a. To solve this, the assembly box was first filled with ethanol vapour, increasing the dew point. Leaving an open vial of ethanol in the closed box for five minutes resulted in optimal assembly
Figure 19: Optical images from assemblies using the top-down microscope. a: Assembly in an air-filled box at 13°C, causing the evaporation rate to be too high, resulting in random deposition of cubes and drops of solvent slipping underneath the glass slide. b: Assembly in an ethanol-filled box at 20°C (left) and a to 16 °C (right). The difference in random assembly is clearly visible. c: Palladium assembly with clean sweeping.
Figure 20: Original measured speed of the stage versus the speed input in the CPP software. In reality, the stages did not move at speeds lower than 210µm min-1or at any 50µm min-1increments.
temperatures of 15 to 18°C. The effects of a temperature change from 20 to 16 °C in an ethanol environment is shown in Figure 19b. Moving forward, extracting the cubes from ethanol and redispersing them in water provides a basis of experimentation that could be better compared to results in established literature.
Assembly speed The assembly speed is in close relation with the temperature, since they must be balanced to form a stable accumulation zone. The assembly speed is also thought to be the main detriment to the low observed silver nanocube yield. Figure 20 shows the measured speed of the x-stage versus the speed input in the CPP software during initial testing by software engineer Jorijn Kuster. The stage speed is not linear and increases at set intervals. This is due to the stage being moved by a turning screw, which is attached to gears of different sizes. While the speed at a specific plateau is constant, the jumps between plateaus are large. More importantly, it was observed that during actual assembly the stage did not move at all when the set speed was lower than 210µm min-1. This is thought to be caused by the additional weight on the stage from the vacuum-pump and temperature controller, as well as the stages not being in pristine condition. This meant that in initial assemblies, the lowest possible speed was 200µm min-1, which has never led to accurate assembly without clusters on the substrate. Attempts to decrease the assembly speed initially involved purchasing new stages, but these were too expensive and had long delivery times. Then, the idea of using a differential speed was proposed. Rotating the y-stage by 90° and aligning it with the x-stage allowed them to move in opposite directions, making the effective assembly speed the difference between them. Using this technique, the minimum assembly speed was thought to be lowered to 50 µm min-1,
since that is the difference between plateaus. However, during experiments the stages only showed to reliably move at their 100 µm min-1 increments. Nevertheless, the actual minimum
assembly speed was lowered to 100µm min-1, which is a factor two decrease. Assembly at this
speed showed the most promising results as depicted in Figure 16. For further investigation of capillary assembly in complex and tight trap arrays we suggest acquiring stages capable of moving at constant speeds down to 10 µm min-1. This lower speed also means that the
4.4
Conclusion
In the Chapter the placement of silver and palladium nanocubes on patterned silicon substrates using sequential capillary assembly is shown. Two types of assembly are performed. First, 75 nm silver cubes are assembled in traps designed for their size. While the cubes show a distinctive preference to assemble in the traps over flat silicon, the assembly yield is very low. Secondly, smaller 19 nm palladium cubes are assembled on the substrates. Once again, the cubes show a preference for traps over flat silicon, and the yield seems higher. However, due to the low initial silver yield few spots are available for the palladium cubes to be pinned in between them, and this behaviour was not reproducibly observed.
5
Conclusions and outlook
In conclusion, we investigated the feasibility of implementing sequential template-assisted cap-illary assembly for the fabrication of (opto)electronic devices. Assembly dependant on cube size is achieved by designing smart trap patterns on a substrate. This technique is used to arrange multiple different materials with a high spatial accuracy in an arbitrary design. The 75 nm silver and 19 nm palladium cubes were assembled on fluoro-silanised silicon substrates with etched patterns. While assembly yield was low, the concept of size-dependant capillary assembly is demonstrated.
Optimisations of the process involve increasing the assembly yield by further lowering the assembly speed. Addressing the speed should lead to a better silver assembly and a more reproducible area to study the subsequent palladium assembly. In order to effectively compare achieved results to established literature the choice can be made to changing the solvent from ethanol to water. Also, by lowering the speed, the temperature relative to the dew point needs to be decreased, which is an added benefit of switching to a higher boiling point solvent such as water. Furthermore, using the silicon substrate as a master to create PDMS molds and performing the assembly on those can aid in the throughput of experiments.
The results presented in this work show the possibility of applying capillary assembly for the accurate placement of nanocubes on silicon. This bottom-up approach is independent of material properties and is therefore not limited to silver or palladium cubes. Development of this novel technique opens up the fabrication of high-performance (opto)electronic devices at
References
[1] Shafigh Mehraeen, Mohamed Asbahi, Wang Fuke, Joel KW Yang, Jianshu Cao, and Mei Chee Tan. Directed self-assembly of sub-10 nm particles: role of driving forces and template geometry in packing and ordering. Langmuir, 31(31):8548–8557, 2015.
[2] Chris A Mack. The future of semiconductor lithography? look to flash. Journal of Micro/Nanolithography, MEMS, and MOEMS, 12(3):030101, 2013.
[3] Songbo Ni, Lucio Isa, and Heiko Wolf. Capillary assembly as a tool for the heterogeneous integration of micro-and nanoscale objects. Soft matter, 14(16):2978–2995, 2018.
[4] Songbo Ni, Jessica Leemann, Ivo Buttinoni, Lucio Isa, and Heiko Wolf. Programmable colloidal molecules from sequential capillarity-assisted particle assembly. Science advances, 2(4):e1501779, 2016.
[5] Valentin Flauraud, Massimo Mastrangeli, Gabriel D Bernasconi, Jeremy Butet, Duncan TL Alexander, Elmira Shahrabi, Olivier JF Martin, and Juergen Brugger. Nanoscale topo-graphical control of capillary assembly of nanoparticles. Nature nanotechnology, 12(1):73, 2017.
[6] J¨org Polte. Fundamental growth principles of colloidal metal nanoparticles–a new perspec-tive. CrystEngComm, 17(36):6809–6830, 2015.
[7] Victor K LaMer and Robert H Dinegar. Theory, production and mechanism of formation of monodispersed hydrosols. Journal of the American Chemical Society, 72(11):4847–4854, 1950.
[8] Siva Rama Krishna Perala and Sanjeev Kumar. On the mechanism of metal nanoparticle synthesis in the brust–schiffrin method. Langmuir, 29(31):9863–9873, 2013.
[9] Thomas C Hales. An overview of the kepler conjecture. arXiv preprint math/9811071, 1998.
[10] F Haaf, A Sanner, and F Straub. Polymers of n-vinylpyrrolidone: synthesis, characteriza-tion and uses. Polymer Journal, 17(1):143, 1985.
[11] Kallum M Koczkur, Stefanos Mourdikoudis, Lakshminarayana Polavarapu, and Sara E Skrabalak. Polyvinylpyrrolidone (pvp) in nanoparticle synthesis. Dalton Transactions, 44(41):17883–17905, 2015.
[12] Tao Chen, Yuwei Zhang, and Weilin Xu. Size-dependent catalytic kinetics and dynam-ics of pd nanocubes: A single-particle study. Physical Chemistry Chemical Physics, 18(32):22494–22502, 2016.
[13] Nanoxact silver nanocubes.
https://nanocomposix.com/collections/all/products/ nanoxact-silver-nanocubes-pvp?variant=15906870132825. Accessed: 2019-06-13.
[14] Uzodinma Okoroanyanwu. Chemistry and lithography. SPIE, Bellingham WA, 2010. [15] Marc J Madou. Fundamentals of microfabrication: the science of miniaturization. CRC
[16] Csar 62, alternative to zep.
http://www.allresist.com/ebeamresist-positiv-csar62-alternative-zep/. Accessed: 2019-05-21.
[17] Success story of csar 62.
http://www.allresist.com/research-development/success-story-of-csar-62/. Accessed: 2019-05-21.
[18] Raith voyager product brochure.
https://www.raith.com/files/PDF%20Downloads/brochure-voyager.pdf. Accessed: 2019-05-22.
[19] M Morita, T Ohmi, E Hasegawa, M Kawakami, and M Ohwada. Growth of native oxide on a silicon surface. Journal of Applied Physics, 68(3):1272–1281, 1990.
[20] Tianhong Chen and Bj¨orn M Reinhard. Assembling color on the nanoscale: multichromatic switchable pixels from plasmonic atoms and molecules. Advanced Materials, 28(18):3522– 3527, 2016.
[21] Tobias Kraus, Laurent Malaquin, Heinz Schmid, Walter Riess, Nicholas D Spencer, and Heiko Wolf. Nanoparticle printing with single-particle resolution. Nature nanotechnology, 2(9):570, 2007.
[22] Ma´eva Collet, Sven Salomon, Naiara Yohanna Klein, Florent Seichepine, Christophe Vieu, Liviu Nicu, and Guilhem Larrieu. Large-scale assembly of single nanowires through capillary-assisted dielectrophoresis. Advanced Materials, 27(7):1268–1273, 2015.
[23] Yadong Yin, Yu Lu, Byron Gates, and Younan Xia. Template-assisted self-assembly: a practical route to complex aggregates of monodispersed colloids with well-defined sizes, shapes, and structures. Journal of the American Chemical Society, 123(36):8718–8729, 2001.
[24] C´ecile Formosa, Flavien Pillet, Marion Schiavone, Rapha¨el E Duval, Laurence Ressier, and Etienne Dague. Generation of living cell arrays for atomic force microscopy studies. nature protocols, 10(1):199, 2015.
[25] L Chopinet, C Formosa, MP Rols, RE Duval, and Etienne Dague. Imaging living cells surface and quantifying its properties at high resolution using afm in qi mode. Micron, 48:26–33, 2013.
[26] Robert D Deegan, Olgica Bakajin, Todd F Dupont, Greb Huber, Sidney R Nagel, and Thomas A Witten. Capillary flow as the cause of ring stains from dried liquid drops. Nature, 389(6653):827, 1997.
[27] Jun Xu, Jianfeng Xia, Suck Won Hong, Zhiqun Lin, Feng Qiu, and Yuliang Yang. Self-assembly of gradient concentric rings via solvent evaporation from a capillary bridge. Phys-ical review letters, 96(6):066104, 2006.
[28] Songbo Ni, Jessica Leemann, Heiko Wolf, and Lucio Isa. Insights into mechanisms of capillary assembly. Faraday discussions, 181:225–242, 2015.
[29] Laurent Malaquin, Tobias Kraus, Heinz Schmid, Emmanuel Delamarche, and Heiko Wolf. Controlled particle placement through convective and capillary assembly. Langmuir, 23(23):11513–11521, 2007.
Appendix
Nanocube concentration
In this section a quick calculation is given for the number of cubes in an assembly. The concentration of the silver nanocubes is 2.6 × 1011 particles/ml[13]. In a typical assembly we
use 10 µl of nanocube solution.
1 ml = 2.6 × 1011 cubes
10 µl = 2.6 × 109 cubes (1)
The patterned fields consist of five fields of squares and three fields of trenches. A single field consists of 200 lines, each 20 µm long. For simplicity, we assume that the entire line can be filled with face-to-face 75 nm cubes.
one line = 20 µm = 20000 nm one line = 267 cubes one field = 53333 cubes eight fields = 4.3 × 105 cubes
(2)
Therefore, there are four orders of magnitude more cubes in the solution than could possibly be assembled on the substrate, indicating that the nanocube concentration is sufficient.
Contact angle measurements
In this section the influence of slide height on the contact angle of the solvent is discussed. The distance between the glass slide and the substrate is increased from 150 µm to 900 µm. The solvent used is ethanol, and the measurements are performed on freshly silanised flat silicon sample. Also, the angle of the glass slide respective to the substrate is investigated. The contact angle is measured for a glass slide mounted at a 5° and 0° angle, which is shown in Figure A1. The contact angles are measured using imageJ and are shown in Figure A2.
As is expected, the contact angle decreases as the distance between the glass slide and the substrate increases. However, the impact is not very large, with only a 15° difference between the smallest and largest distances. Furthermore it should be noted that all these angles are still
Figure A1: Images made with the Basler acA1920-155um Area ScanCamera. a: Glass slide with a 5° incline at a distance of 500 µm. b: Glass slide with a 0° incline at a distance of 500 µm.
well within reported values for suitable assembly. Lastly, it should be noted that the 5° incline stage was used throughout this report, because it better pinned the nanocube solvent near the edge of the glass slide. Using the 0° incline slide, the solvent was observed to creep underneath the whole slide.
Figure A2: Contact angle measurements for a 5° and 0° inclined glass slide at distances from 150 µm to 900 µm.
