DISS ETH NO.: 18942
Structural Influences on Self-cleaning Surfaces A dissertation submitted to
ETH ZURICH for the degree of Doctor of Sciences
Doris Madeleine Spori Dipl. Werkstoff-Ing. ETH born on November 27, 1979 citizen of Boltigen (BE)
accepted on the recommendation of
Prof. Dr. Nicholas D. Spencer, examiner Prof. Dr. Anthony B. Brennan, co-examiner Dr. Tanja Drobek, co-examiner
For Martin and my parents
You can never know everything and part of what you know is always wrong. Perhaps even the most important part. A portion of wisdom lies in knowing that. A portion of courage lies in going on anyway. James Oliver Rigney, Jr.
Abstract How a liquid contacts a solid and how it spreads on the surface is of utmost importance to many aspects of everyday life. The effects range from rain drops on a window pane obscuring a clear view, or soap residues in the bath tub, to research on the drop retention of fungicides on plant leaves and applications in the coatings industry. The spreading of a droplet is influenced by the surface energies of the liquid and the solid and of the solid’s surface roughness. Usually, very smooth surfaces are easier to clean than rough surfaces, where the contaminants remain stuck within the surface topography. However it was found that in nature some very rough surfaces remain clean even in very dirty environments. The lotus plant is a prime example from nature that show this extraordinary cleanliness and it led to the term self-cleaning surface. A surface is self-cleaning when it can be cleaned by a gentle rinsing with water without the use of detergents or mechanical force (wiping). Such an effect is only observed on very rough, usually hydrophobic surfaces, where the drop sits on top of surface features like a fakir on a bed of nails. Due to the few contact points with the surface, the drop is very mobile, and by rolling over the surface it carries away contaminating particles. It is clear that technical surfaces displaying such an effect would find an application everywhere where the surface is occasionally exposed to a flow of water (e.g. outdoor applications in areas with occasional rain). This would reduce costs for detergents and labour hours. In order to make use of this effect and design technical surfaces, the parameters influencing the self-cleaning effect have to be characterised and understood. But also the effects of the case where the rough surface fails to suspend the drop on top of the features need to be understood more closely. Indeed, the knowledge of this case is relevant in an even wider field of applications, since most technical surfaces exhibit a certain degree of surface roughness. On rough surfaces in particular, the mechanisms governing wetting are not yet fully understood. Therefore, the aim of this thesis is to increase the understanding of wetting by systematic studies on very rough surfaces and over a wide range of surface energy. Four different, very rough surfaces with precisely controlled surface chemistries were investigated as to their capability to pin the contact line of a moving drop. The four surfaces were sand-blasted glass slides as well as replicas of acid-etched, sand-blasted titanium, lotus leaves and photolithographically manufactured golf-tee shaped micropillars (GTMs). It was found that on slightly hydrophilic samples, pinning on the roughv
ness causes the drop to adopt much larger contact angles than would be expected from purely energetic considerations. The pinning effect was found to be independent of the surface chemistry. A parameter was identified that can distinguish the different surfaces according to their pinning strength. All rough surfaces with a very hydrophilic coating were invaded by water—the so-called hemi-wicking effect. With a hydrophobic coating, the lotus leaf and the GTM replicas showed a wetting behaviour as expected for self-cleaning surfaces. Thanks to their re-entrant edges, the GTM replicas could maintain this effect even into the slightly hydrophilic regime. This finding is of significant technical interest, since hydrophobic surfaces may become oxidised or contaminated during use and the hydrophobic property deteriorates. The influence of the fraction of the area under the drop in contact with the solid (f1 ) was explored on density gradients of randomly placed holes and pillars. These gradients were prepared with four different surface chemistries: native PDMS (polydimethyl siloxane), perfluorosilanized PDMS, epoxy and CH3 -terminated thiols on gold. f1 was identified to be the key parameter influencing the static water contact angle on a composite surface consisting of air and substrate. Additionally, it was shown that dynamic contact angles (contact angles measured during increasing and decreasing of the drop volume) are sensitive to the type of surface feature, hole or pillar, that predominates. Roll-off angles have also been measured and found to be influenced not only by the drop weight, but also suction events at holes, edge pinning and f1 . The main experimental method used in this thesis, the replica technique, was thoroughly characterised regarding artefacts occurring during replication or during characterisation with the scanning electron microscope. It was also tried to identify the precision of the replica technique. It was found that the replica technique is actually so precise that it is hard to find methods that can quantify the difference between the master and the replica. In addition, an alternative replica technique was introduced, which is capable of transferring photolithographic structures onto a ceramic substrate. In addition to scanning electron microscopy, contact-angle measurements were the main characterising tool for the surfaces investigated. Influences of several parameters, such as drop size, flow speed and measuring method, and a few theoretical assumptions were tested on model surfaces (alkane thiols on gold). PDMS is a very versatile elastomeric polymer, which is used in many fields of research. It is shown that PDMS is a complex material, as far as contact-angle measurements are concerned. Besides the ability of the polymer’s backbone to adapt to the phase in contact, the z-component of the water surface tension is also large enough to pull up a rim along the drop’s perimeter, which artificially “roughens” flat PDMS. It is also presented that accurately measuring very high contact angles, as observed on self-cleaning surfaces, is complicated due to optical issues and failing fitting routines. Researchers are still trying to define all the parameters necessary to create mechanivi
cally and chemically stable self-cleaning surfaces. This work has shed more light on the parameters influencing wetting on rough surfaces. It has also revealed a further condition to be considered, when choosing when to use the two most frequently applied models of Wenzel and Cassie-Baxter. Additionally, a few promising paths for future research have been outlined.
Zusammenfassung Wie eine Flüssigkeit eine Oberfläche berührt und wie sie sich auf der Oberfläche ausbreitet, ist von größter Bedeutung für viele Aspekte des täglichen Lebens. Die Auswirkungen reichen von die Sicht trübenden Regentropfen auf einer Fensterscheibe, Seifenresten in der Badewanne, über die Haftung von Fungiziden auf Pflanzenblättern bis zu Anwendungen in der Beschichtungsindustrie. Das Ausmass der Ausbreitung eines Tropfens auf der Oberfläche wird durch die der Oberflächenenergien der Flüssigkeit und des Festkörpers und der Oberflächenrauhigkeit des Festkörpers beeinflusst. Gewöhnlich sind sehr glatte Oberflächen leichter zu reinigen als raue Oberflächen, auf denen Verschmutzungen in der Oberflächentopographie haften bleiben. Allerdings wurde festgestellt, dass es in der Natur auch sehr raue Oberflächen in sehr schmutziger Umgebungen gibt, die ungewöhnlich sauber bleiben. Die Lotuspflanze ist ein gutes Beispiel, das diese außergewöhnliche Reinheit zeigt. Untersuchungen an Lotusblättern führten zur Prägung des Begriffs „selbstreinigende Oberfläche“. Eine Oberfläche ist selbstreinigend, wenn sie durch sanftes Spülen mit Wasser ohne Einsatz von Seife oder mechanischer Mittel gereinigt werden kann. Eine solche Wirkung ist nur auf sehr rauen, in der Regel hydrophoben Oberflächen zu beobachten, wo der Tropfen wie ein Fakir auf dem Nagelbrett nur auf den Spitzen der Oberflächestruktur sitzt. Aufgrund der wenigen Kontaktpunkte mit der Oberfläche ist der Tropfen sehr beweglich, und kann durch eine Rollbewegung über die Oberfläche Schmutzpartikel aufsammeln und entfernen. Es ist klar, dass technische Oberflächen mit einer solchen Eigenschaft überall Anwendung fänden, wo die Oberfläche in regelmässingen Abständen einer Benetzung mit Wasser ausgesetzt ist (z. B. Fassadenfarben in Regionen mit gelegentlichem Regenfall). Da man sollche Oberflächen nicht reinigen müsste, würden die Unterhaltskosten drastisch gesenkt. Um technische Oberflächen mit einem solchen Effekt entwerfen zu können, muss man die Parameter, die den selbstreinigenden Effekt beeinflussen, charakterisieren und verstehen. Auch der Fall, dass die rauhe Oberfläche nicht fähig ist, den Tropfen nur auf den Rauhigkeitsspitzen zu stützen, ist in diesem Zusammenhang sehr wichtig. Die meisten realen Oberflächen zeigen eine gewisse Oberflächenrauhigkeit, sodass Erkenntnisse über das Verhalten von Tropfen im Kontakt mit nicht-idealen Oberflächen für ein viel breiteres Feld von Anwendungen relevant sind. Die Benetzungsmechanismen auf rauen Oberflächen sind noch nicht vollständig verstanden. Deshalb war es das Ziel dieser Arbeit, das Verständnis über die Benetzung ix
durch systematische Studien auf sehr rauhen Oberflächen und über einen weiten Bereich an Oberflächenenergien zu vertiefen. Vier verschiedene, sehr rauhe Oberflächen mit genau kontrollierter Oberflächenchemie wurden auf ihre Fähigkeit untersucht, die Kontaktlinie eines Tropfen an der Fortbewegung zu hindern (Pinning). Die vier Flächen waren sandgestrahlte Glasplättchen sowie Replikate von säure-geätztem, sandgestrahltem Titan, Lotusblättern und photolithographisch hergestellten Golftee-geformten Mikrosäulen (GTMs). Es wurde festgestellt, dass auf leicht hydrophilen Oberflächen das Pinning der Kontaktlinie, das durch die Rauhigkeit bedingt ist, zu wesentlich grösseren Kontaktwinkeln führt, als es aus rein energetischen Überlegungen zu erwarten wäre. Der Pinning-Effekt ist von der Oberflächenchemie unabhängig. In dieser Arbeit wurde ein Parameter identifiziert, welcher die verschiedenen Oberflächen anhand ihrer Pinningstärke unterscheiden kann. Wenn die vier rauhen Oberflächen mit einer sehr hydrophilen Beschichtung funktionalisiert wurden, zeigten alle das sogenannte Hemi-wicking, ein Effekt, bei dem Wasser aufgrund der Oberflächenenergie und Oberflächenrauhigkeit in die Struktur gesogen wird, und kein Kontaktwinkel mehr gemessen werden kann. Ein Tropfen auf hydrophob beschichteten Lotusblatt- und GTM-Replikaten verhielt sich gemäss den Erwartungen für selbstreinigende Oberflächen. GTM-Replikate können dank ihrer unterschnittenen Kanten auch im leicht hydrophilen Bereich den Tropfen auf den Säulenspitzen halten. Dieser Befund ist von großer technischer Bedeutung, da hydrophobe Oberflächen während der Benutzung verunreinigt oder oxidiert werden können, und sich ihre hydrophobe in eine leicht hydrophile Beschichtung umwandeln kann. An Dichtegradienten von zufällig platzierten Löchern und Säulen wurde der Einfluss des Flächenbruchteils (f1 ) auf den Kontaktwinkel untersucht. f1 ist der Flächenbruchteil unter dem Tropfen, der in Kontakt mit dem Festkörper steht. Die Gradienten wurden mit vier verschiedenen Oberflächenchemien hergestellt: unbehandeltes PDMS (Polydimethylsiloxan), perfluorosilanisiertes PDMS, Epoxy und CH3-terminierte Thiole auf Gold. Es wurde nachgewiesen, dass f1 der wichtigste Parameter ist, der die statischen Kontaktwinkel auf einer Oberfläche beeinflusst, auf der der Tropfen nur auf den Spitzen der Oberflächenstruktur sitzt. Darüber hinaus wurde gezeigt, dass dynamische Kontaktwinkel (Kontaktwinkel gemessen während der Zu-und Abnahme des Tropfenvolumens) empfindlich auf die Art der vorherrschenden Oberflächestruktur (Löcher oder Säulen) ist. Auch Abrollwinkel wurden gemessen, und es wurde festgestellt, dass der Abrollwinkel nicht nur durch die Tropfengrösse, sondern auch von Saugeffekten an Löchern, Pinning an Kanten und dem Faktor f1 beeinflusst wird. Die wichtigste experimentellen Methode dieser Arbeit war die Replika-Technik. Diese wurde gründlich charakterisiert bezüglich Artefakte, die während der Replikation oder während der Charakterisierung mit dem Rasterelektronenmikroskop entstehen können. Ausserdem wurde auch versucht, die Genauigkeit der Replika-Technik zu identix
fizieren. Allerdings musste festgestellt werden, dass die Replika-Technik tatsächlich so präzise ist, dass es schwer ist, Methoden zu finden, die den Unterschied zwischen dem Original und dem Replikat bemessen können. Darüber hinaus wird eine alternative Replika-Technik vorgestellt, die die Übertragung von photolithographischen Strukturen auf ein keramisches Substrat erlaubt. Kontaktwinkelmessungen waren neben der Rasterelektronenmikroskopie das wichtigste Instrument zur Charakterisierung der untersuchten Oberflächen. Der Einfluss verschiedener Parameter wie Tropfengröße, Fließgeschwindigkeit und Messverfahren, und ein paar theoretische Modelle wurden an Modelloberflächen (Alkanthiole auf Gold) getestet. PDMS ist ein sehr vielseitiges elastomeres Polymer, das in vielen Bereichen der Forschung eingesetzt wird. Es wird gezeigt, dass PDMS kein einfaches Material für Kontaktwinkelmessungen ist. Neben der Fähigkeit der Polymerketten, ihre Konformation je nach der Phase, mit der sie in Kontakt stehen, zu ändern, ist die zum Substrat senkrechte Komponente der Oberflächenspannung des Wassers auch groß genug, um das Polymer entlang des Tropfenumfangs zu verformen. Dies führt zu einer lokalen Rauhigkeit, die die Kontaktwinkelmessung beeinflusst. Es wird auch gezeigt, dass eine genaue Messung sehr hoher Kontaktwinkel wie sie auf selbstreinigende Oberflächen beobachtet werden, durch optische Probleme und fehlschlagenden Fitting-Routinen erschwert wird. Es sind noch nicht Parameter gefunden und definiert worden, die es erlauben würden, mechanisch und chemisch stabile selbstreinigenden Oberflächen herzustellen. In dieser Arbeit sind einige Parameter, die die Benetzung rauher Oberflächen beeinflussen, genauer untersucht und beschrieben worden. Es wurde auch gezeigt, dass es zu den bereits bekannten noch eine weitere Bedingung zu berücksichtigen gilt, die bestimmen, wo die beiden am häufigsten angewandten Modelle von Wenzel und Cassie-Baxter verwendet werden dürfen.
Contents 1 Introduction
1.1 The Lotus Effect – Phenomenological Approach to the Term Self-Cleaning
1.1.1 Surface Morophology of Self-cleaning Plants, Exemplified on the Example of the Lotus Leaf . . . . . . . . . . . . . . . . . . . . . . . .
1.1.2 Self-Cleaning – Particle Removal . . . . . . . . . . . . . . . . . . . .
1.2 Definition of States and Expressions . . . . . . . . . . . . . . . . . . . . . .
1.3 Scope and Outline of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . .
2 Theoretical Background - Liquid in Contact with Solid
2.1 Surface Tension and Surface Energy . . . . . . . . . . . . . . . . . . . . . .
2.2 Interactions During Wetting . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Laplace Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4 Young’s Equation – the Ideal Case . . . . . . . . . . . . . . . . . . . . . . .
2.5 Drops on Chemically Heterogeneous Surfaces . . . . . . . . . . . . . . . . .
2.6 Wetting States on Rough Surfaces . . . . . . . . . . . . . . . . . . . . . . .
2.7 Wenzel and Cassie-Baxter Model, Current Debate . . . . . . . . . . . . . .
2.8 Roughness, Edge Effects, Contact Line Pinning and Contact Angle Hysteresis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.9 Model Substrates to Determine Parameters Influencing Cassie-state Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.10 Super-Oleophobic Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3 Materials and Methods 3.1 Self-Assembled Monolayers (SAMs)
33 . . . . . . . . . . . . . . . . . . . . . .
3.2 Contact-Angle and Roll-off Angle Measurements . . . . . . . . . . . . . . .
3.2.1 Static Contact-Angle Measurements . . . . . . . . . . . . . . . . . .
3.2.2 Dynamic Contact-Angle Measurements . . . . . . . . . . . . . . . .
3.2.3 Roll-off Angle Measurements . . . . . . . . . . . . . . . . . . . . . .
3.3 Scanning Electron Microscopy (SEM) . . . . . . . . . . . . . . . . . . . . . .
3.4 Light Microscopy (LM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5 White-Light Profilometry (WLP) . . . . . . . . . . . . . . . . . . . . . . . .
3.6 Differential Scanning Calorimetry (DSC) . . . . . . . . . . . . . . . . . . .
Contents 4 Parameters Influencing Contact Angle Measurements
4.1 The Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.4 Conclusions and Summary . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Surface Functionalisation with Trichlorosilanes . . . . . . . . . . . . . . .
4.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3 PDMS (Polydimethylsiloxane) . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3.4 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . .
4.4 Fit Routines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4.4 Conclusion and Summary . . . . . . . . . . . . . . . . . . . . . . . .
5 Replica Technique
5.1 Lotus Leaf Replica (LLR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.1 Replication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Limits of Provil Novo When Replicating Holey Structures . . . . . . . . . .
5.3 Vacuum Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4 Precision of the Replica Technique . . . . . . . . . . . . . . . . . . . . . . .
5.4.1 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.2 Results and Discussions . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.3 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . 100 5.4.4 Excursus: Curing Properties of Thermosets . . . . . . . . . . . . . . 101 5.5 Ceramic Replicas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 5.5.1 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 5.5.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 106 5.5.3 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . 107 xiv
Contents 6 Roughness Influences on Wetting over a Wide Surface-Energy Range 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.5.1 A Side Remark: Calculating with the Equation by Cassie & Baxter 6.5.2 Evaluation of Dynamic Contact Angle Measurements . . . . . . . .
109 109 112 114 120 120 120 121
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient123 7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 7.2 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 7.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 7.4 Conclusions: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 7.5 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 7.5.1 Mask Generation and Characterisation . . . . . . . . . . . . . . . . 140 7.5.2 Influence of Gravitation and Discretisation . . . . . . . . . . . . . . 142 7.5.3 Light Microscopy of Air Enclosure . . . . . . . . . . . . . . . . . . . 142 8 Conclusions
9 Outlook and Future Directions 9.1 Self-cleaning Effect: Single Versus Double Structures . 9.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . 9.1.2 Experimental . . . . . . . . . . . . . . . . . . . . 9.1.3 Results and Discussion . . . . . . . . . . . . . . . 9.1.4 Conclusions and Outlook . . . . . . . . . . . . . . 9.2 Visualisation of the Drop Base in the Cassie-state . . . 9.3 Smooth Gradient of Adhesion . . . . . . . . . . . . . . . 9.4 Gradient Design by Means of Mathematical Algorithms 9.5 Cassie-state in Oil and Under Pressure . . . . . . . . . 9.6 Raspberry-like Particles – A Move Into Applications? .
149 150 150 151 152 155 156 158 159 160 161
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
1 Introduction First, a short, rather phenomenological introduction on the term self-cleaning on the example of the lotus leaf is given. Then different states of drops on “superhydrophobic” surfaces are presented and the most important definitions are introduced. A short overview explains how such surfaces can be achieved. At the end of the chapter, the scope of this thesis is formulated and the topics of the single chapters are outlined.
1.1 The Lotus Effect – Phenomenological Approach to the Term Self-Cleaning In south-east Asia, the lotus plant (lat. Nelumbo Nucifera) grows out of muddy water, but its blossoms and leaves are always free of dirt. Thus, centuries ago in Buddhist cultures, it became a symbol for purity and divine birth. It is said, the spirit of the best men is spotless like the blossom of a lotus plant.1 What is the secret of this magnificent plant? How does it keep its surfaces clean? Contamination occurs on every surface, therefore some sort of self-cleaning mechanism must be involved. Even here in Europe, many plants such as tulips, nasturtiums or leeks show the same degree of cleanliness, as long as their leaves remain untouched. Figure 1.1 shows a large drop of water (diameter is about 12 mm) on a lotus leaf. Two things are striking when looking at the drop: first, the drop shape. The drop looks very similar to a drop of water in a hot pan. It is very round and seems to minimise its contact with the surface, very different to a drop sitting on a glass plate, for example. Also, the drop is extremely mobile on the lotus leaf, again comparable to the drop in a hot pan. Second: the blue-greyish colour at the bottom side of the drop is a reflection of the ceiling. In order for this to happen, it needs another interface at the bottom of the drop where total reflection of the light can occur: another air-water interface. Thus, this drop of water actually sits on an air layer and touches the leaf only at very few spots. This is what makes the drop on the lotus leaf so similar to the drop in the hot pan1 ; both 1
A drop sitting on his own vapour is called Leidenfrost drop, after the German physician who first studied the effect in 1756. The Leidenfrost effect occurs when water is placed on a very hot solid, at a temperature typically above 200 °C. The drop is very mobile, riding on a cushion of steam, and never touches the solid.)
1 Introduction are sitting on a gaseous layer: the drop in the pan on its own vapour, vaporised upon contact with the pan, the drop on the leaf on ambient air.
Figure 1.1: Water drop on lotus leaf. The bottom of the drop (blue-greyish colours) reflects the ceiling of the room, an indication of the additional air layer underneath the drop, causing total reflection of the light at the air-water interface. Thanks to this air layer, the drop is highly mobile on the surface. So, the drop on a lotus leaf takes up a shape that is nearly spherical, is highly mobile and has a layer of air enclosed underneath. These are the basic necessities of the selfcleaning effect, as was shown by Barthlott and Neinhuis.2 They were investigating plant-epidermis morphology and found that there were certain plant species that hardly ever required any cleaning before analysis. They found a connection between the leafsurface morphology and its cleanliness, and proposed a mechanism for this self-cleaning effect.
1.1.1 Surface Morophology of Self-cleaning Plants, Exemplified on the Example of the Lotus Leaf First, the surface morphology of the lotus leaf is discussed (see Figure 1.2). During evolution, at the point when plants left the aquatic environment, they had to develop a means to exist in the new dry and sunny conditions. Their solution was a continuous extracellular membrane, the cuticle (see Figure 1.2a) whose primary function is to be a transpiration barrier. Additionally it helps to stabilize the plant tissue. By reflecting visible light and inducing turbulent air flow it is a protection against harmful irradiation and is the means to control the leaf temperature. If the surface chemistry is rather hydrophilic, as often found on flower petals, it offers a good grip to pollinators. Hydrophobic leaf surfaces have a reduced particle and pathogen adhesion.3 2
1.1 The Lotus Effect – Phenomenological Approach to the Term Self-Cleaning a.)
b.) epicuticular wax
cuticle cutin + wax pectine layer cell wall plasma membrane 50 µm
Figure 1.2: a.)Stratification of the outer part of epidermis cells as shown in Koch et al.3 The cuticle consists of the cutin and the flat wax layer. On top of the cutin, some plants show three-dimensional epicuticular waxes. The pectin layer is a connective layer to the cell wall. b.) – d.) SEM images of the lotus leaf and its replica. b.) a stereo-SEM reconstruction of the lotus-leaf-replica surface from SEM images acquired at the angles 0°, 15° and 30°. c. and d.) are cryo-SEM images of a fresh leaf. c.) morphology on the micrometer scale: conical protrusions (papillae) immerge from the surface. Between the papillae, stomata (pores used for gas exchange) are visible (arrow). d.) The papillae are covered by a layer of crystalline wax tubules which builds up the nanostructure. The cuticle is a composite material consisting of the cutin network (a polyester-like polymer) and hydrophobic waxes. On all leaves there is a thin two-dimensional wax layer, and on many leaves there are three-dimensional wax structures in addition. These structures are crystalline4 and most probably formed through self-assembly. The cutin layer may serve as a crystallization template. The pectin layer connects the cuticle to the cell wall of the outermost cells3 (Figure 1.2a). There is a large diversity of plant surface structures. These topographies are built up either by single cells or by multiple cells. Cell shapes range from concave to tabular to convex. Convex cells are classified according to their aspect ratio (width / height) into papillae or hairs. Papilla cells, in particular, often show a hierarchical structure; a micro3
1 Introduction structure formed by the cell itself and a nanostructure formed by the three-dimensional wax crystals. Such plants are commonly self-cleaning. A prominent representative of this species is the Lotus plant (Nelumbo Nucifera).3 Figure 1.2c and d show the two scales of roughness found on the lotus leaf on images acquired by cryo-SEM (Scanning Electron Microscopy) on a fresh leaf. The micro-scale roughness is formed by the papillae (Figure 1.2c). The distance between neighbouring papillae is about 21 ± 7 µm. Between the papillae, stomata are visible (arrow in Figure 1.2c). Stomata are pores for gas exchange on the leaf (O2 , CO2 and H2 O). The whole surface of the leaf is covered by the second component of roughness; hydrophobic, threedimensional crystalline wax tubules consisting of nonacosanol (C29 H60 O)3 (Figure 1.2d). Figure 1.2b is presented to give a more spatial impression of the lotus leaf surface. It shows a reconstructed 3D view of a lotus leaf replica (LLR, more in chapter 5.1 and 5.4) using stereo-SEM. The papillae with their steep sides and the nearly regular distribution on the leaf are a challenging substrate for stereo-SEM, thus this technique was not further used in this thesis.
1.1.2 Self-Cleaning – Particle Removal Due to its high surface tension and the hydrophobicity of the waxy surface, a drop of water on the lotus leaf will only touch the top of the papillae covered with wax tubes, and bridge the space in between, thus enclosing air underneath. This is how the air layer discussed above is formed. The few contact points allow the drop to almost retain its spherical shape which it naturally adopts when it is only contacting air. Additionally, adhesion of the drop to the surface is greatly reduced by this low number of contact points, and therefore, only a very small amount of energy is necessary to bring the drop into motion. Thus, the nearly spherical drop forms a very high contact angle of about 160° with the surface (if it was a full sphere, it would be 180°), and the leaf surface has to be tilted by only a few degrees (roll-off angle < 5°) to set the whole drop into motion. The movement of the drop is more of a rolling motion, than the sliding observed on a flat window pane, for example. The findings of Barthlott and Neinhuis2 on the mechanisms behind the self-cleaning effect on the lotus leaf are summarised in Figure 1.3. Contaminating particles larger than the distance between the papillae will only sit on top of the papillae. Smaller particles will fall in between and lie only on the hydrophobic tubular wax crystals. Due to hydrophobic nature of the leaf the interaction between the contaminating particles and the surface occurs only over weak Van-der-Waals interactions. Additionally, due to the very small contact area between the particles and the rough surface (small number of contact points), adhesion of the particles to the surface is minimal. If the highly mobile, rolling water drop encounters a particle lying on the leaf, the area of contact of 4
1.1 The Lotus Effect – Phenomenological Approach to the Term Self-Cleaning the drop to the particle will be larger than that of the particle with the leaf. Thus the particle will adhere to the water drop’s surface and be removed with the drop. Barthlott and Neinhuis investigated a number of different contaminant particles on leaves of different plant species, ranging from silicon carbide dust (different sizes), to toner particles to conidiospores of a grey mould. They exposed the dirt-covered leaves to either artificial fog or rain. They found that with the fog treatment, very small particles remained on the surface. Upon rinsing lotus leaves with artificial rain, all particles, independent of their nature and size, were washed away. Rain has a certain momentum upon contact with the surface and drops will deform elastically between the papillae and therefore also reach smaller particles between them.2 The momentary water pressure upon impact can reach values of 105 Pa in heavy rain.5 Analysing the performance of other plant species with flat leaf surfaces Barthlott and Neinhuis found that the wettability of the leaf plays an important role. A drop on a flat hydrophobic leaf moves only slowly and mainly redistributes the contaminants on the surface. However, hydrophilic surfaces were also quite efficient in being cleaned by water. The drops spread quickly and ran off the surface with considerable speed. Particles are thus carried along with the moving liquid front.2 There are also some plant species with tabular (flat) cells, such as the tulip, which display superhydrophobic behaviour, solely due to their dense three-dimensional wax layer. These species show this effect mainly on young leaves, in which the biosynthesis of the wax is still active. On mature leaves, damage and erosion reduce the superhydrophobic effect. It is crucial for the plant that the first few leaves stay healthy and provide the plant with the necessary energy for growing.3 What are the benefits of a self-cleaning surface for the plant? The low adherence of the water and the consequential rolling-off leads to leaf surfaces that are clean and nearly always dry. Thus, temperature control is guaranteed and particles cannot occlude stomata. Pathogens such funghi and bacteria do not find the water necessary for germination. In conclusion, the lotus effect, or self-cleaning effect, can be observed on surfaces where a drop of water balls up to a spherical shape enclosing air underneath, is highly mobile and drags contaminating particles off the surface. The mechanism is caused by the rough topography of the surface and the hydrophobic nature of its wax layer. It is emphasised that the “self-cleaning” still needs water to carry away the dirt, and thus is of little use in arid climates. 5
Figure 1.3: Diagram summarizing the connection between surface roughness and selfcleaning. a.) On the smooth surface, drop mobility is rather low and particles are mainly redistributed by water. On a sufficiently rough surface, water droplets and particles only adhere to the tips of the surface roughness. Adhesion between water and particles is thus larger than to the surface and particles are removed from the leaf when the drop rolls off. With kind permission from Springer Science+Business Media: Planta, Purity of the sacred Lotus, or escape from contamination in biological surfaces, 202, 1997, pg. 1-8,W. Barthlott, C. Neinhuis, figure 8
1.2 Definition of States and Expressions A self-cleaning surface commonly consists of a rough, hydrophobic surface that is able to enclose air underneath the drop, and thus reduce the contact points with the drop. However, there are many different states that can occur on a rough, hydrophobic surface (see Figure 1.4). All of these states can lead to very high contact angles, but they strongly influence the behaviour of the drop on the surface. In general, two states can be distinguished: the Wenzel- and the Cassie-state. If the surface is rough, but not able to build up an air layer between surface and drop, the drop is in the Wenzel-state6 (Figure 1.4a). This state is characterised by a rather high or no roll-off angle and high contact angles, which get rarely above 150°. Thus, it is a “sticky” surface. If roughness is within the critical range and surface energy of the substrate low, an airsubstrate composite is formed beneath the drop. The interface is heterogeneous, since it consists of solid-liquid and air-liquid contacts. Thus, any drop which bridges from surface feature to surface feature and by doing so enclosing air underneath, is in the Cassie state7 (Figure 1.4b). Researchers from many different fields have published in this field. Thus, a number of terms have arisen to describe the effects observed, and no general agreement has yet been found. Surfaces exhibiting large contact angles are often referred to as “superhydrophobic” or “ultrahydrophobic”. If they additionally display very low roll-off angles, “superhydrophobic” and “self-cleaning” are used interchangeably. There have 6
1.2 Definition of States and Expressions been suggestions to unify the terms,8 e.g. using the term “ultrahydrophobicity” to describe surfaces that show contact angles between 120° and 150° and “superhydrohobicity” for surfaces with contact angles above 150° and roll-off angles below 10°. As shown in Figure 1.4, there are many different states possible on a surface and simply distinguishing according to a number does not seem to match the complexity of the observed phenomena. In this thesis, the wetting state is usually described by calling it Wenzel- or Cassie-state. The latter is more closely specified according to the subspecies of Cassie-state that was observed. A drop in the Cassie-state shows distinct differences in mobility, depending on how the air layer is comprised. It can be a free flowing air layer with a connection to the atmosphere (Figure 1.4c), as it is found e.g. when a drop is sitting on pillars, spikes or papillae. On the right of the drop, a top view of such a surface is illustrated. The black squares represent pillar tops that are in contact with the drop. In this state, the drop shows a considerable-to-high mobility. Contact-area reduction can also be achieved with quite the opposite surface structure: holes or cavities are formed, instead of pillars, and the enclosed air cushions have no direct contact to the atmosphere (Figure 1.4d). These drops show a reduced-to-no mobility on the surface, but are nevertheless in the Cassie-state. Drops on such surfaces reach high contact angles, but rarely exceed 150°.9, 10 A special case occurs when free-flowing air and sealed air cushions are combined in the surface design. Such surfaces exhibit higher contact angles than those only having sealed air cushions, but still have a low drop mobility.11, 12 An application of such sticky superhydrophobic surfaces might be found in open microfluidic devices. Hong et al.13 reported that no volume loss (generally a problem in these devices) was observed when they repeatedly transported superparamagnetic microliter-sized liquid droplets in alternating magnetic fields from the bottom substrate to the top substrate. The two substrates were installed parallel to each other at a distance slightly larger than the drop height. There exists a vast number of ways to produce surfaces that suspend drops in the Cassiestate.8, 14–17 Drops in the Cassie-state with a free-flowing air film can be sorted into those exhibiting a single structure and those combining different scales of structures. Single structures on the micrometer scale often consist of photolithographically produced pillars, which are used as model surface to study wetting parameters. Such surfaces offer a precise design of feature size, spacing, shape and order, see for example in publications.9, 10, 18–22 7
1 Introduction a)
sealed air cushions
free flowing air
i.) double: “nano” on micro
fractal-like / multiple
Figure 1.4: Illustration of different wetting states on rough, hydrophobic surfaces. a.) Wenzel-state: the liquid contacts the whole area underneath the drop; b.) Cassie-state: the drop encloses air underneath the drop. The Cassie state can be subdivided judging on how the air can flow: c.) the air can freely flow if the drop sits on pillars or spikes and has a direct contact to the ambient air, d.) the air is sealed in depressions underneath the drop and has no contact to the ambient air. e.) The adhesion to the drop can be tailored, if the states of c.) and d.) are combined e.g. by tubular pillars. Surfaces with a potential to show the self-cleaning effect can be built up by a single layer of structures, consisting of features in f.) the micro-range or in g.) the nano-range. Many surfaces show h.) a double structure consisting of two types of features or i.) are prepared by combining multiple methods of structuring or have a fractal-like appearance. Single-layer roughness with feature sizes below 1 µm are mostly prepared by bottomup processes. Phase separation leads to 3D porous networks when the second phase is removed. Differential etching, such as plasma etching on polymers (e.g. PMMA) or electrolyte or wet chemical etching on metals, leads to sharp tips, which (mostly after hydrophobisation) can stabilise an air layer underneath the drop.8 Gas-phase coating of poly8
1.2 Definition of States and Expressions methylsilsesquioxane nano-filaments also leads to a stable self-cleaning surface.8, 23, 24
Double-structures usually combine a nano-structured layer with a micro-structure (Figure 1.4h), as illustrated on the lotus leaf with the tubular wax crystals on top of the papillae.2 A versatile approach is also given by the assembly or aggregation of spherical particles. Spherical particles forming a dense hexagonal layer can show high contact angles, independent of their size,25 but the combination of smaller and larger particles strongly increases the self-cleaning effect. Such superhydrophobic surfaces can be achieved by bringing raspberry-like particles26 (silica nano-particles bound to silica microparticles) to the surface, by layer-by-layer assembly27 or by a templating method, in which a mould of a larger structure is filled by nano-silica beads,28 thus also forming a double-structure.8 If fibres and textiles are coated with particles or carbon nanotubes, these are also classical double-structured surfaces. Etching photolithographically produced pillars,29 or coating them with carbon nanotubes,30 also yields double-structures. Many structured surfaces need to be chemically hydrophobised after creation.
Surfaces involving crystal growth, electro-deposition or gas-phase deposition often form structures that are fractal-like, resembling cauliflower florets (Figure 1.4i). Multiplescale roughness can also be achieved by combining different structuring methods. Multiple-scale roughness enhances the ease of roll-off, allows impacting drops to more easily remove particles and helps to prevent the conversion from the Cassie-state to the unfavourable Wenzel-state. There are also indications that multiple roughnesses can convert condensed Wenzel-state water drops into Cassie-state drops, which then roll-off easily.8
Technical surfaces usually consist of single, nanostructured or double-structured surfaces. If transparency is a prerequisite, then only structures of around 100 nm feature size can be utilised in order not to scatter the light. Such structures often are mechanically weak and break upon touching. Additionally, most of the surfaces presented above can become contaminated or oxidised during use and lose part of their chemical hydrophobicity, and thus fail in suspending drops in the Cassie-state.8
While Figure 1.4 mainly illustrates in what different states a drop can sit on a rough surface, many more effects can be observed or introduced on such superhydrophobic surfaces. By adding asymmetry into the surface structures, it is possible to direct droplets; either by vibration,31 or by bouncing.32 Other surfaces can switch between the Wenzeland the Cassie-state, or even between being superhydropobic and superhydrophilic14 upon external stimuli (change of temperature, exposure to light or application of an external electrical field). Superhydrophobic surfaces have also been used as antifouling surfaces.15, 33 9
1.3 Scope and Outline of the Thesis Despite the innumerable ways that were found to produce superhydrophobic surfaces and the exciting number of phenomena that can be observed, many associated mechanisms are not yet fully understood. The scope of this thesis is to shed more light on how water interacts with rough surfaces, investigated by contact angle measurements. Hydrophobic coatings may be contaminated or degenerated e.g. upon UV irradiation, and become more hydrophilic. How this affects the wetting behaviour of rough surfaces is a key focus of this thesis. A few comments and insights on the most frequently used and debated predictive models by Wenzel and Cassie-Baxter6, 7 will be presented. As pointed out above, most technically relevant surfaces consist of stochastically distributed surface features. But most investigations of the parameters determining the Cassie-state have previously been made on periodic photolithographic patterns. In order to move such investigations closer to the application, a model surface consisting of stochastically placed pillars that increase in density from one side of the substrate to the other was designed and models from the literature tested on them. Chapter 2 is intended to provide the reader with a theoretical background to integrate this work into the science of wetting and to understand the principles that were investigated. During the study of the literature, it was recognised that there are many different ways to determine contact angles. People who are introduced to the principles of contactangle measurements are often amazed at the simplicity of the technique; it does indeed simply involve placing a drop on a surface. However, frustration often sets in when the results have to be evaluated. Chapter 4 is intended to provide a deeper introduction into contact-angle measurements and the parameters influencing their outcome. Contact angles were measured on a system of alkyl-thiols on gold and flat polydimethylsiloxane (PDMS) substrates. The same surface chemistries were then used to prepare substrates for the analysis of the influence of surface roughness on contact-angle measurements in Chapters 6 and 7. A critical discussion of methods for contact angle evaluation at high contact angles is given. In order to produce many substrates of equal roughness, the replica technique was used to copy the topography of a master. To this end, a mould was prepared by casting a silicone species onto the master. After curing, this mould was used to either cast epoxy or PDMS. Epoxy substrates have the advantage that they can be coated with a layer of gold / chromium and subsequently be functionalised by thiol chemistry to adjust their wetting properties. PDMS already has a hydrophobic nature which can be further functionalised by silane chemistry. In Chapter 5, it was tried to quantify the precision of the replica technique and to characterise artefacts that occur during analysis of a polymeric replica, mainly in scanning electron microscopy (SEM). Additionally, a lost-form 10
1.3 Scope and Outline of the Thesis approach of the replica technique is presented to create ceramic replicas of photolithographic structures. Chapter 6 investigates roughness influences on wetting over a wide surface-energy range.22 Here, four different, heavily structured surfaces were multiplied by the replica technique and their surface energy adjusted from hydrophobic to hydrophilic by mixed self-assembled monolayers (SAMs) of mercapto-undecanol and dodecylthiol. The surfaces were those of a sand-blasted glass slide (SBG), a lotus-leaf replica (LLR), sandblasted (large grit) acid-etched titanium (SLA) and photolithographically produced golftee-shaped micropillars (GTM). It was found that pinning events play a major role on arbitrarily rough surfaces such as SBG, LLR and SLA. GTMs showed a very stable Cassie-state wetting, even in the hydrophilic regime. Chapter 7 examines Cassie-state wetting by means of a hole-to-pillar-density gradient.9 A mask for photolithography was prepared, having a stochastical distribution of a low number of black dots which increases along the gradient, the dots slowly merging together until the mask is fully black. This was used for photlithography with SU-8 (epoxy species, negative photoresist), spincoated onto a silicon wafer, and formed the master for the replica technique. Four different surface energies were investigated: native epoxy, native PDMS, SAM of dodecylthiol on gold / chromium / epoxy and perfluorinated PDMS. Several methods were tested on the acquired data to reveal the accuracy of their predictive power. A special focus was placed upon the contact angle hysteresis (a measure for adhesion) found on theses structures. In Chapter 8 general conclusions and in Chapter 9 an outlook are presented.
2 Theoretical Background - Liquid in Contact with Solid This chapter is intended to provide the reader with the tools to understand the content of this thesis. It will start with an introduction into the relevant physical background of wetting (2.1-2.4), leading to the models used to explain wetting on heterogeneous surfaces (2.5-2.6). Then the current controversy in the field is outlined (2.7) and the influence of roughness on contact-angle measurements presented (2.8). A short summary of the current knowledge of the mechanism behind superhydrophobic surfaces is given (2.9-2.10). The interplay of liquids with solids has been a research topic for centuries. In the early 19th century Thomas Young published the qualitative theory on surface tension.34 The study of the contact between three phases is called wetting and describes how a liquid deposited on a solid spreads out.35
2.1 Surface Tension and Surface Energy While molecules in a liquid attract each other via intermolecular attraction forces (cohesion), there are nearly no interactions among molecules in the gas phase. Thus the attractive forces of a molecule in a liquid at the interface to the gas phase are unbalanced; in fact a molecule at the surface loses about half of its cohesive interactions. Therefore the loss of cohesive interactions at the interface cannot be compensated by interactions with the molecules in the gas phase. Thus, work has to be done to bring a molecule from the bulk to the surface. Consequently molecules at the surface are in an unfavourable, higher energy state, causing a drop to always adopt the shape with the least surface area. The work δW to bring one molecule to the surface can be written as:
δW = γLA · δA
where γ is the surface energy [mJ/m2 ] (LA denotes the liquid-air interface) and δA the area occupied by the molecule at the surface [m2 ]. The surface free energy γ can be 13
2 Theoretical Background - Liquid in Contact with Solid calculated by deriving the free energy F by the change of surface area A and is therefore a thermodynamic property of the surface.35
δF γ= δA
The surface tension γ [mN/m], also called capillary force, is a force which works tangentially to every point on the surface and will oppose any distortion. It is in magnitude identical to the surface energy γ. The work δW , e.g. along the length l of a distortion, can be written as:
δW = F · dx = γLA · l · dx
where the directions of l and dx are perpendicular to each other.35 Interfacial tensions between the liquid and air or vacuum, or between two liquid phases can be measured by pendant-drop measurements36 (see also Chapter 2.3). Typical values for interfacial tensions are: for water 72.8 mN/m, for ethanol 22.7 mN/m and for n-hexane 18.4 mN/m.37 The attractive interactions between interfaces can be interpreted as energy per area or as force per line. These two forms of expression are often not distinguished in the literature and also intermingled (see Equation 2.9). De Gennes et al.35 solved this ambiguity by simply referring to the quantity as surface tension γ. If they want to specify the nature of the surface tension, they call it surface energy or capillary force. The same notation will be used in this thesis.
2.2 Interactions During Wetting When a liquid droplet comes into contact with a solid substrate (or another, immiscible liquid) then again, cohesive interactions are lost, but also attractive intermolecular interactions with the substrate are gained, thus forming the surface energy γSL (SL denotes the solid-liquid interface). According to Dupré the work of adhesion per area WSL can then be written as38, 39
WSL = γSA + γLA − γSL
where γSA and γLA denote the solid-air and liquid-air surface tensions. The chemical nature of the liquid and the solid strongly influence the surface energy γSL . In principle, the attraction occurs due to molecular interactions at the interface and could be determined (in principle!) by exact solutions of the Schrödinger equation.39 All non-covalent 14
2.3 Laplace Pressure bonds between the molecules contribute to the attraction: dipole – dipole interactions for molecules with a permanent electrical dipole (Keesom forces), dipole – induced dipole interactions (Debye forces), induced dipole – induced dipole interactions (London or dispersion forces) induced by fluctuations in the electron clouds of the molecules, hydrogen bonds, charge-charge interactions (anion-cation interaction). Dispersion forces are found between all molecules, e.g. alkanes only interact via dispersion forces. All other electrostatic forces depend on the (polar) groups present in the phases. A precise description of intermolecular forces can be found in the book by Israelachvili.39 But, although the origin of the surface tension can be explained on a molecular level, the surface tension is an inherently macroscopic parameter defined at a macroscopic scale.35 A comment on the measurement of γSA and γSL is made in Chapter 4.1.1. Finally, the interfacial tension is a measure for the dislike of two phases X and Y. A large γXY indicates a large difference in the type of interaction between the phases X and Y, thus only little attractive interactions can occur between them. Conversely, a small γXY shows that many of the lost cohesive interactions from within one phase can be compensated by interacting with the other phase.
2.3 Laplace Pressure The surface tension of a drop causes the pressure pi inside a drop to be slightly larger than the atmospheric pressure po . The difference in pressure is called Laplace pressure ∆p. It describes the increase in the hydrostatic pressure observed when passing across a curved surface or interface. It is equal to the product of the surface tension γ and the curvature C of the surface. The curvature C of a surface is described by the sum of the reciprocal values of the radii of curvature R and R’. In the case where the centre of the corresponding circle lies within the liquid phase, R or R’ is positive, otherwise it is negative. The larger the curvature C, the larger the pressure inside the drop.35 Or, in other words, the smaller the radii of curvature R and R’, the larger the pressure inside the drop. ∆p = pi − po = γ · C = γ ·
1 1 + R R’
The Laplace pressure can also be used to explain capillary adhesion between two plates, between hairs or fibres, or between particles. The liquid forms a layer between two surfaces (capillary bridge). At least one of the two radii of curvature is small and negative (centre of the corresponding circle lies outside of the liquid phase) and thus gives rise to a negative Laplace pressure. This leads to a strong attractive force between the solids with which the liquid is in contact.35 15
2 Theoretical Background - Liquid in Contact with Solid The curvature-dependent Laplace pressure is also responsible for effects such as Ostwald ripening, where in an emulsion, for example, smaller droplets with large curvature empty themselves into larger ones with smaller curvature.35 The Laplace equation can be used as basis to calculate surface tensions. The capillary force holds a drop in shape, whereas gravity distorts its profile. Thus, in the hydromechanical equilibrium, capillary and gravitational forces must be balanced at each location on the drop profile.
where ρ is the density of the fluid and g the acceleration due to gravity. If the drop is axisymmetric to the coordinate z, R and R’ can be described for each point r(z) on the drop profile and the height h is a function of z.35, 36, 40 This leaves γ as the only adjustable parameter. Thus, γ is iteratively varied until the calculated profile fits to the experimentally measured profile. The value for the best fit is the surface tension which was to be determined. This approach is used for pendant and sessile drops.
2.4 Young’s Equation – the Ideal Case Wetting is the term used to describe the study of how a drop of liquid spreads out on a solid (or other liquid). Wetting phenomena are influenced by surface chemistry and surface roughness. In this and subsequent Chapters (2.4 and 2.5), surface-energy influences are discussed. Surface roughness influences are introduced in Chapters 2.6 to 2.9. Wetting is divided in two different regimes (see Figure 2.1): total wetting and partial wetting. In order to distinguish between the two, de Gennes41 defined the spreading parameter S, being the surface-energy difference between the dry and the wetted state of the substrate:
S = ESubstrate−dry + ESubstrate−wet = γSA − (γSL + γLA )
If S > 0, the energy of the dry surface is higher than of the wet surface, and the liquid completely spreads to reduce the energy of the system. This is called total wetting. In the opposite case where S < 0, only partial wetting occurs, meaning the liquid forms a drop on the substrate. Small drops with a radius r smaller than the capillary length, take up the shape of a truncated sphere, touching the substrate at a specific angle θY (definition of capillary length in Chapter 4.1.1). This specific angle can then be used to split the partial wetting regime in two more types of wetting: “mostly wetting” and 16
2.4 Young’s Equation – the Ideal Case “mostly non-wetting”. If the liquid is water (oil) then these two types are called “hydrophilic” (oleophilic) and “hydrophobic” (oleophobic). The change from one type to the other occurs, nominally, at θY = 90°. “Mostly wetting” combinations will spontaneously invade in a porous medium such as a sponge or a wick, or form capillary bridges between two solid surfaces (see Figure 2.1). The technical term for liquid invasion into porous media is “imbibition”.
total wetting S>0
wetting partial wetting S 90° θ < 90°
Figure 2.1: Flow chart summarising the wetting states found when a liquid comes into contact with a solid. Two major influences govern wetting phenomena: surface chemistry and surface roughness. “Wetting” is the term used to describe the study of how a drop spreads out on a surface. If a drop fully wets the surface it is called “total wetting”, otherwise “partial wetting”. Partial wetting is the state where a distinct drop is formed on the surface and can be subdivided into “mostly wetting” and “mostly non-wetting”. Surface roughness can cause a drop on a mostly wetting substrate to fully wet the surface (“hemiwicking”). On a mostly non-wetting surface a drop can span from surface feature to surface feature and thus be in the “Cassie-state”. The “Wenzelstate” occurs when the drop follows the surface topography and can occur on partial wetting, rough substrates. More on the influence of surface roughness in Chapter 2.6 to 2.9. The contact angle θY corresponds to the ideal case where only the surface tensions are involved (see Figure 2.2). It is calculated by summing the surface tensions acting along the contact line L of the drop. The system is at equilibrium, and therefore the forces must be equal to zero. Only forces acting in the plane of the substrate need to be considered, since the vertical portion of γLA is balanced by the rigidity of the substrate. This leads to the famous Young’s equation:34 γSA − γSL L · γSA − γSL − cos θY · γLA = 0 → cos θY = γLA
The Young’s equation consists of four parameters, of which two are directly measurable 17
2 Theoretical Background - Liquid in Contact with Solid
Figure 2.2: Sessile drop on a rigid, flat and smooth surface. The Young’s equation describes how the surface tensions (γLA , γSL , γSA ) involved cause the liquid drop to adopt one specific contact angle at the three-phase contact point. (θ and γLA ). A comment on the determination of γSA and γSL is made in Chapter 4.1.1. The case of contact angle measurements on a substrate that is not rigid enough to compensate for γLA−Y is described in Chapter 4.3.1. The Young’s equation predicts one single contact angle for a specific surface – liquid – air (third phase) combination. Experimentally, even on nearly “perfect” substrates this cannot be observed, since on every surface there exists a certain degree of hysteresis. The hysteresis is the difference between a maximum contact angle and a minimum contact angle. Hysteresis on flat and smooth surfaces is discussed in Chapter 4.1. Hysteresis on rough surfaces is discussed in Chapter 2.8. The difference between the surface-air and surface-liquid surface tensions is sometimes also referred to as the adhesion tension:6 γadh = γSA − γSL = γLA · cos θ. By combining Equation 2.8 with Equation 2.4, the work of adhesion can be determined as
WSL = γSA + γLA − γSL = γSL · (1 + cos θ)
This equation is the Young-Dupré equation of adhesion.
2.5 Drops on Chemically Heterogeneous Surfaces Most surfaces are chemically inhomogeneous. They often comprise different chemical species, be it due to contamination or prepared intentionally by using micro-contact printing42 or UV-photopatterning,43 for example. If a drop is placed on such a surface, the contact angle will be influenced by the species present on the surface. Cassie44 presented a model capable of predicting the contact angle on such surfaces, if the contact angles on the pure substances are known. Preconditions for the applicability of the equation are that the surface consists of well-separated and distinct patches of different species; e.g. of hydrophilic and hydrophobic nature.45 Additionally, the patches need to be small compared to the base diameter of the drop and homogeneously distributed over 18
2.6 Wetting States on Rough Surfaces the surface.46 Then, if f1 denotes the area fraction of phase 1, f2 the area fraction of phase 2 and θ1 , θ2 , the corresponding contact angles on the pure homogeneous surfaces, the average over the work of adhesion leads to the contact angle θC observed on the heterogeneous surface.44, 45 cos θC = f1 · cos θ1 + f2 · cos θ2 ,
f1 + f2 = 1
If the two components are mixed at the atomic or molecular scale then a relation developed by Israelachvili and Gee considering the polarisability of the involved molecules should be employed.42, 45
2.6 Wetting States on Rough Surfaces Figure 2.3 shows three wetting states, which only occur due to surface roughness. As indicated in Figure 2.1, each state is a combination of surface energy and surface roughness.
Figure 2.3: Different wetting states: a.) Wenzel state, the liquid follows the topography of the substrate, b.) Cassie-Baxter state, the drops rests on top of the topographical features, c.) hemi-wicking, the liquid penetrates in between the topography of the substrate. The Cassie-state (Figure 2.3b) is specifically only possible for surfaces having a rather low surface energy, whereas hemi-wicking (Figure 2.3c) can only occur on rough highenergy surfaces. However the Wenzel-state, where the drop base follows the entire topography (Figure 2.3a) can occur over the whole range of surface energies. Wenzel-state: Wenzel6 precisely described this state and proposed a model to explain how apparent contact angles on rough surfaces are modified by roughness to differ from the Young’s contact angle θY . The term “apparent contact angle” is often used to specify that the contact angle observed is not the Young’s contact angle θY , but a contact angle altered by surface roughness. If this macroscopically observed contact angle on the rough surface belongs to the Wenzel-state, it is often denoted θW . 19
2 Theoretical Background - Liquid in Contact with Solid A drop sitting on a rough surface encounters more contact area per unit area than if it were sitting on a flat surface. In this sense, “a greater intensity of surface energy” is found for the solid-air and liquid-solid interface6 . Since the surface tensions of the phases involved are characteristic material properties unchanged by roughness, Wenzel introduced a roughness factor r to take into account the “physical condition” of the surface underneath the drop.
real surf ace area projected area
This roughness factor r magnifies γSA and γSL and thus causes the drop to assume more extreme values than would be expected for a flat surface (Young’s equation). cos θW =
γSA − γSL r · γSA − r · γSL =r· = r · cos θY γLA γLA
The effect of this model is that roughness enhances the initial wettability; with roughness, a hydrophilic surface should exhibit a more hydrophilic contact angle and a hydrophobic surface a more hydrophobic contact angle. While the description of how the drop sits on the surface is very accurate, doubt has been cast on the validity of the model itself (see Chapter 2.7). Since it is based on the Young’s equation, it does not predict any contact-angle hysteresis. In Chapter 6 it is shown that it does not fit experimental data acquired on very rough surfaces. Cassie-state: If the surface is hydrophobic, water can span the distance between two asperities and enclose air underneath (Figure 2.3b). If the features have flat tops (no roughness on top) then Equation 2.11 from Cassie for chemically heterogeneous surfaces can also be employed for a heterogeneous surface where one phase (phase 2) consists of air. The contact angle of water with air is 180°, the contact angle θ1 of phase 1 is then simply the Young’s contact angle θY of the solid. This leads to the Cassie-Baxter7 equation:
cos θCB = f1 · cos θY + f2 · cos θair = f1 · cos θY − f2 ,
f1 + f2 = 1 f or f lat tops
However, f1 in the original equation proposed by Cassie & Baxter was derived from a study of a droplet wetting fibres. In Figure 2.4, the original sketch of Cassie and Baxter is shown. They explain the derivation of f1 and f2 as follows: “The plane geometrical area may be taken as OA, when f1 is given by the arc DC divided by OA, and f2 is given by BC/OA.”7 (see Figure 2.4). cos θCB = f1 · cos θY − f2 , 20
f1 + f2 > 1, arbitrary surf ace
2.6 Wetting States on Rough Surfaces Thus, f1 in the original paper corresponds to the normalized area in contact with the drop, leading to f1 + f2 ≥ 1.
θa r C 0
fibre cross-section 0’
(d+r) Figure 2.4: Original sketch from Cassie and Baxter in 1944.7 The two circles represent cross-sections through two fibres, the dashed area is the drop only wetting the top of the fibres. The arc length DC divided by the length of OA is equal to f1 , the wetted area fraction and the distance between C and B divided by OA represents f2 , the area fraction in contact with air. In small drops, the Laplace pressure is assumed to be constant and the influence of gravity can be neglected. For such drops, the liquid-air interface f2 is understood to be flat.47 Some researchers48, 49 combine Equation 2.12 and Equation 2.13 in a way that the parameters f1 , f2 describe the projected surface-liquid and liquid-air area fractions, and the roughness factor rwet the roughness of the wetted area fraction f1 . cos θCB = rwet · f1 · cos θY − f2 ,
f1 + f2 = 1
Hemi-wicking: If the surface energy of the substrate is rather high, hemi-wicking,50–52 also called surface wicking53, 54 can occur. Hemi-wicking (see Figure 2.3c) is used to describe the phenomenon when a liquid that would show partial wetting on the smooth surface, fully spreads on the rough surface with the same surface chemistry. The rough surface can be described as a 2D-porous surface. The driving force for liquid invasion into the roughness is on the one hand the Laplace pressure and on the other hand the fact that in contrast to a drop on a flat and smooth surface the liquid-air area does not have to be increased by the same extent as on the flat substrate. Therefore spreading within the topography costs less energy than spreading on the flat surface, and is thus favoured. If all depressions in the surface are filled with liquid, the excess volume builds a drop, sitting on a composite interface with liquid and substrate. Therefore, the contact angle can be predicted by Equation 2.10, where the contact angle θ2 for the second phase is 0°, corresponding to the contact angle the liquid adopts when contacting itself. 21
2 Theoretical Background - Liquid in Contact with Solid To induce hemi-wicking a certain minimum of surface energy is necessary. The critical Young’s contact angle θCY for this to happen can be estimated for flat-top pillar surfaces with the parameters of Wenzel and Cassie-Baxter,47 by considering that the area fraction of the pillar tops, f1 , remains dry during wetting: cos θCY =
1 − f1 r − f1
2.7 Wenzel and Cassie-Baxter Model, Current Debate The current debate was stimulated by Lichao Gao and Thomas J. McCarthy with their provocative 2007 publication “How Wenzel and Cassie Were Wrong”.55 Figure 2.5 shows a sketch of the experimental demonstration Gao and McCarthy used to illustrate their criticism. They prepared a flat, hydrophobic substrate having a circular patch consisting of pillars. The area covered with the pillars allowed the drop to sit in the Cassie-state (contact angle ≈ 168°). If the drop volume was then increased, such that the contact line of the drop crossed the boundary of the patch, the contact angle of the drop dropped to the value of a contact angle measured on the homogeneously flat part of the substrate (contact angle ≈ 117°).55 The same effect was observed on chemically hetergeneous surfaces, where instead of pillars the patch was either a hydrophilic spot in a hydrophobic matrix, or vice versa.56 +∆V
Figure 2.5: Sketch of the demonstration performed by Gao and McCarthy. They prepared a flat hydrophobic substrate which had a circular patch consisting of pillars. A drop formed on these pillars was in the Cassie-state and showed a superhydrophobic contact angle. If the drop was then enlarged until the contact line crossed the boundary of the patch, the contact angle would change at this to the contact angle expected for an evenly flat surface. The main statement of Gao-McCarthy is that all experimental data indicate “that contactangle behavior (advancing, receding, and hysteresis) is determined by interactions of the liquid and the solid at the three-phase contact line alone and that the interfacial area within the contact perimeter is irrelevant”.55 Therefore they claim that the models of Wenzel and Cassie-Baxter are “fundamentally flawed”57 because they are based on areafraction parameters. In their argumentation they follow the line of thought which was 22
2.8 Roughness, Edge Effects, Contact Line Pinning and Contact Angle Hysteresis first expressed by Gray.58 Surface-energy (area dependent) and capillary-force (line dependent) approaches should not be intermingled. Gray and Gao-McCarthy believe that surface energy and capillary forces are discrete, distinct and different quantities, which are only mathematically equivalent at equilibrium.57–59 Equilibrium at the liquid-air interface can be achieved due to the high mobility of the molecules, whereas equilibrium on a solid is an uncertain concept since an approach to it may be infinitely slow.58 The area-based parameters in the models of Wenzel and Cassie-Baxter lead to the intuition that interfacial free energies dictate wetting and a change in the contact area has an effect on the contact angle.57 Gao-McCarthy recommend using these equations only on homogeneous surfaces, where “fortuitously” the conditions of the contact area reflect the ground-state energies of contact lines.55, 57, 59, 60 The replies to the article mentioned above did not question the statement, that it is only the surface roughness faced by the contact line, which determines the contact angle, but they tried to specify under which conditions the models by Wenzel and Cassie-Baxter were accurate.46, 61–63 Nosonovsky63 emphasises that superhydrophobicity is a multiscale phenomenon, involving effects on the molecular, micrometer and millimeter scales. The macroscale equations of Wenzel and Cassie-Baxter are valid for uniformly rough surfaces, where surface features are small compared to the drop. The two models do not consider the absolute value of the drop free energy, but the net energy change dE in the system free energy during the advancing event of the liquid drop.46, 63 Therefore only the roughness parameters found at the triple line should be entered in the models of Wenzel and Cassie-Baxter.46, 62, 63 For non-uniformly rough surface, Nosonovsky49, 63 proposes generalised equations where r, f1 , f2 are functions of their position f (x, y). Marmur61 states that local conditions at the contact line determine the actual contact angles. The Wenzel and Cassie-Baxter equations are global considerations regarding the solid-liquid interfacial area and provide the most stable apparent contact angle on rough surfaces, if the contact angle is measured with sufficiently large drops.
2.8 Roughness, Edge Effects, Contact Line Pinning and Contact Angle Hysteresis One of the core criticisms of the models presented above to describe or predict contact angles (Young, Wenzel, Cassie-Baxter) is that they neglect the existence of contact-angle hysteresis. By increasing the drop volume of a sessile drop, the contact angle increases without any movement of the contact line, until the contact angle adopts a maximum value—the advancing contact angle. Having reached this contact angle, the contact line starts to move while the advancing contact angle remains constant. The same is true if the drop evaporates or if its volume is actively reduced. The contact angle will be redu23
2 Theoretical Background - Liquid in Contact with Solid ced to a minimum contact angle, the receding contact angle, which then stays constant as soon as the contact line starts to move. The difference between the advancing and the receding contact angle is commonly referred to as hysteresis (Equation 2.17). Contactangle hysteresis is usually ascribed to chemical heterogeneity and surface roughness.41 If both are present, surface roughness has the stronger influence on the hysteresis. Hysteresis on flat surfaces and its influence on static, advancing and receding contact angle measurements are discussed in Chapter 4.
∆θ = θa − θr
The hysteresis is a measure for the energy that is lost between placing to removing the drop. In considerations where drop retention plays a role, the quantity
∆ cos θ = cos θr − cos θa
is sometimes also called contact-angle hysteresis. Examples can be found on sliding drops (see Chapter 4.1, Furmidge Equation 4.5),64 contact-angle hysteresis on diluted defects (see below)65, 66 and investigations on the asymmetric wetting hysteresis on microstructures (see below).10 It has its origin in the difference in the work of adhesion (Equation 2.9):
∆WSL = WSL−rec −WSL−adv = γLA ·(1+cos θr)−γLA ·(1+cos θa) = γLA ·(cos θr−cos θa) (2.19) Contact-angle hysteresis has remained a central wetting phenomenon without quantitative predictive models.10, 65
Contact-Line Pinning: Oliver et al. investigated the inhibiting effect of sharp edges on liquid spreading.67 Their aim was to confirm experimentally the Gibbs inequality condition (Equation 2.20) for a drop whose contact line is pinned at an edge. The Gibbs condition was derived from purely geometric considerations. The apparent contact angle θapp of a drop at an edge can adopt any value given by the inequality condition. θ0 is the equilibrium contact angle (in their work the advancing contact angle θa) measured on the flat and horizontal surface, and φ the angle of the edge (see Figure 2.6a).
θ0 = θa ≤ θapp ≤ (180° − φ) + θa 24
2.8 Roughness, Edge Effects, Contact Line Pinning and Contact Angle Hysteresis This means the contact line is at equilibrium between the large range of contact angles from θa to (180° –φ). The maximum possible apparent contact angle θc at the edge should be (see Figure 2.6a, bottom)
θC = (180° − φ) + θa
Circular discs with different edge angles φ were prepared, and a capillary was drilled in the middle of the discs. Through the capillary, the drop was enlarged. Different kinds of liquids were tested. During the approach of the contact line to the edge, the advancing contact angle was measured. When the contact line first reached the edge the drop volume V0 was read. Then the liquid volume was increased until the contact angle reached the maximum apparent contact angle θC , the contact angle before the contact line started to move again, or the drop fell off the disc. At this point, also the maximum drop volume VC was recorded. 120° θa θapp φ
θa θC φ
75° 105° 120°
Figure 2.6: Drops suspended on circular discs. a.) shows the experimental set-up used in the work of Oliver et al.67 with two exemplary discs. Circular discs with a capillary in the middle and having different edge angles φ were prepared. The advancing contact angle θa, the volume V0 at the edge, the maximum contact angle θC and the maximum volume VC before contact line movement were recorded. b.) The ability of an edge to effectively inhibit contact line movement seems to depend on the edge radius. While the tangent to the drop at the three phase contact can remain constant on a blunt edge (bottom), the contact angle is forced to change on a sharp edge (top).
They found that discs with undercut shapes could suspend much larger drop volumes than discs with large values of φ. The maximum contact angles θC coincided well with the maximum angles predicted by the Gibbs condition, independent of the chemical nature of the liquid. After reaching θC , drops on undercut shapes would suddenly jump 25
2 Theoretical Background - Liquid in Contact with Solid and spontaneously spread. Drops on discs with large values of φ were also observed to keep their maximum contact angle θc and the contact line would move down the slope. Since they used two different kinds of discs, hydrophobised sapphire discs with an edge radius of 0.05 µm and hardened aluminium discs with an edge radius of 1 to 5 µm, they found that the drops on the blunter edges showed a kind of creep (Figure 2.6b). The contact line could move on the edge, thus the tangent to the drop most probably remained constant when traversing the edge. There was no apparent sign of this behaviour for the sharper edges; there the contact line remained fixed and the contact angle changed. This investigation showed that very high contact angles and drop suspension are not only possible by air enclosure like described by the Cassie-Baxter equation, but also by a very strong pinning of the contact line. Extrand68 showed with quite a similar experimental set-up that by adding a second, smaller-scale roughness, the creep down the slope can be inhibited and thus larger drop volumes be suspended. By using circular holes or circular rims made of photoresist on silicon wafer, Kalinin et al.69 investigated the influence of slope steepness, slope sign (upward or downward), rim height h and rim width w on the maximum apparent contact angle θC (see Figure 2.7). A step upward (positive slope) did not alter the advancing contact angle of the drop, whereas a step downward led to strong contact-line pinning. For rim heights above 2 µm, θC was similar to the prediction by the Gibbs condition. For smaller heights, the measured θC decreased with decreasing height. The rim width did not have any effect on the observed contact angle.
contact line movement θa
Figure 2.7: Contact-line-pinning dependence on slope steepness φ and sign (upward, downward), and rim height h and width w. Kalinin et al.69 found that a step up (positive slope) has no pinning effect on the maximum contact angle (point A), whereas a step down (negative slope) has a strong influence on contact line pinning and the maximum apparent contact angle θc. The steeper the negative slope the higher θc. The rim width and height above 2 µm do not have any influence on θc. For smaller rim heights, θc decreases with decreasing rim height and the Gibbs criterion is no longer met.
2.8 Roughness, Edge Effects, Contact Line Pinning and Contact Angle Hysteresis Contact-line movement: a.)
Figure 2.8: Pinning of the contact line on crests and pillars. a.) shows the two positions on a system of grooves where the contact line (normal to the sheet) of the drop is at equilibrium. The position at the top of the crest is metastable, the position at the bottom unstable.54 b.) The lines represent schematically the top of the crest (C) and the bottom of the trough (T). The dark line is the receding contact line. The arrows indicate the direction of movement of the two jogs after depairing.41 c.) unlikely dewetting mechanism of the receding contact line on a set of periodic pillars d.) probable mechanism: successive dewetting from one post at a time.48
On a surface consisting of crests and troughs as depicted in Figure 2.8a, the contact line finds two positions of equilibrium, one metastable and one unstable.54 The stable position lies near the top of the ridge and the unstable position near the bottom of the trough. Thus, a drop moving perpendicularly over a parallel set of grooves will always want to jump from one crest to the next (Figure 2.8b). If the whole line were to move at once this would encounter a huge energy barrier. This energy barrier is reduced by the nucleation and depairing of two adjacent jogs;41 similar to the dislocation movement found in metals in the special case of kink pairs.70 Thus the contact line will always choose the energetically most favourable route to move forward. Dorrer and Rühe20 applied the same argumentation on the receding line of a drop in Cassie-state wetting, sitting on a array of square-top pillars (see Figure 2.8c and d). The contact line will 27
2 Theoretical Background - Liquid in Contact with Solid successively dewet from one post at a time rather than jumping from one row to the next over the whole length at once. This pinning of the contact line on top of the crests leads to a strong anisotropy in contact-angle measurements. Advancing contact angles measured at the side where the contact line moves perpendicularly to the grooves are significantly higher than those measured at the sides going along the parallel grooves. The advancing contact angles measured along the parallel grooves are still higher than those measured on a flat substrate. Receding contact angles show the opposite behaviour; due to contact-line pinning on the crests, the receding contact angle is greatly reduced compared to the contact angles measured parallel to the grooves or on a flat surface.33, 71, 72
Contact-line pinning at rectangular pillar and hole structures: Priest et al.10 closely investigated the movement of the contact line over hydrophobised, square flat-top pillars at different densities and their reverse structures, square holes (see Figure 2.9). They found that the depinning behaviour of the contact line is different on pillars than on holes. The contact line moving over pillars was pinned at the same edge of the pillar, whereas the receding contact line moving over the hole was pinned at the opposite edge to the advancing movement.
b.) Adv Adv
Figure 2.9: Priest et al.,10 in a light microscopy study, observed the locations of pinning sites of the contact line of an advancing or receding drop over a substrate consisting of pillars or holes. a.) pillared surface: the contact line is pinned on the same side of the pillar; b.) holey surface: the contact line is pinned at opposite sides of the holes. Additionally they found that the contact-angle hysteresis is sensitive to the kind of surface feature the drop sits on. The difference was mainly observed in the asymmetry of the hysteresis; on pillared surfaces the advancing contact angle θa remained insensitive to f1 whereas the receding angle qualitatively followed the Cassie-Baxter equation (Equation 2.13). On holey surfaces it was quite the opposite, and the receding contact angle remained constant over f1 , whereas the advancing contact angle followed Equation 2.13 quite accurately. Only drops in the Cassie-state were analysed. Air is one of the most “hydrophobic” materials (contact angle of 180°), much more hydrophobic than the employed hydrophobised 28
2.9 Model Substrates to Determine Parameters Influencing Cassie-state Stability substrates. Therefore, the pillared surfaces can be looked at as a very “hydrophobic” surface (air cushions) with “hydrophilic” defects (pillar tops). The holey surface on the other hand represents a surface with a “hydrophilic” matrix (bars between the holes) with “hydrophobic” defects (air enclosures). By plotting the contact angle hysteresis (Equation 2.18) versus the defect-area fraction fD , Priest et al found that the hysteresis increases linearly with fD (Equation 2.22). Thus, fD corresponds to f1 in the case of the pillar surface and to f2 in the case of the holey surface.
4 cos θ = 4 cos θf lat + fD ·
Epin , → fD = f1−pillars , fD = f2−holes γLA
The hysteresis measured in Chapter 7 is analysed by this approach.
2.9 Model Substrates to Determine Parameters Influencing Cassie-state Stability Model substrates are surfaces with a controlled design. Wenzel’s roughness factor r and Cassie-Baxter’s area fractions f1 , f2 can be calculated from geometrical considerations. Johnson and Dettre53, 54 tested the models available at their time with a computational analysis on an idealised sinusoidal roughness and qualitatively compared the predictions achieved by the free-energy-minimisation approach to contact angles measured on waxy substrates with different degrees of roughness (Figure 2.10). De Gennes called “some aspects of these calculations very artificial (the energy barriers are proportional to the total length of line involved), but some aspects are instructive”.41 Energy barriers between adjacent metastable configurations of the drop on a rough surface are responsible for contact-angle hysteresis. If roughness is increased, the energy barriers between the metastable configurations become larger and thus, the hysteresis of a drop in the Wenzel state also strongly increases. As soon as the roughness of a substrate exceeds a certain value, the Cassie-state becomes energetically more favourable. The receding contact angle drastically increases since free-energy barriers for the composite configurations are very low. By spraying diluted fluorocarbon waxes on glass slides, Dettre prepared substrates exhibiting a very profound roughness after evaporation of the solvent. This surface showed water contact angles around 160° with a very high mobility. By heating cycles the roughness was reduced and contact angles decreased. Roughness was not quantitatively determined. By monitoring the advancing and receding contact angle, a change from Cassie-state to the Wenzel-state was observed. The behaviour of the advancing and receding contact angle was qualitatively predicted by their computational experiment. 29
2 Theoretical Background - Liquid in Contact with Solid
Figure 2.10: a.) Famous experiment where Dettre and Johnson53 prepared rough wax structures and measured advancing and receding contact angles on them. By repeated heating cycles the surface roughness was reduced, which caused a change from the Cassie-state into the Wenzel-state after the 7th heat treatment. A vast number of publications has appeared in the last decade trying to elucidate the parameters favouring and stabilising Cassie-state wetting. Most investigations were made on model surfaces consisting of pillar structures, prepared by lithography. A large number of reviews47–49, 57, 73–75 tried to summarise the advances in the theoretical knowledge on superhydrophobic surfaces. The Wenzel (Equation 2.12) and the Cassie-Baxter equation (Equation 2.13) define the contact angle of the lowest energy state.61, 76 Therefore by combining the two equations the contact angle corresponding to the critical surface energy of the transition for a given surface topography can be calculated48, 72 cos θCW = cos θCCB → r · cos θCY = f1 · cos θCY − (1 − f1 ) ⇒ cos θCY =
f1 − 1 r − f1
However, it is often observed that drops are in the Cassie-state, even though the Wenzelstate was the energetically more favourable. This is due to pinning at the edge of the pillars74 and due to local energy barriers between the numerous metastable states in 30
2.9 Model Substrates to Determine Parameters Influencing Cassie-state Stability the energy landscape on a superhydrophobic surface.76 A transition from the Cassie-state into the Wenzel-state is induced by an increase in the Laplace pressure inside the drop. This can be achieved by exerting a pressure, vibrating the sample, applying an external electrical voltage or letting the drop evaporate.48 The Laplace pressure must overcome the cost in surface energy for wetting down the surface structure.49 Two mechanisms can then be identified to accomplish the transition (see Figure 2.11). Either, the base of the drop touches the ground between the pillars (Figure 2.11a), or the contact angle of the sagging drop reaches the value of the inherent advancing contact angle θa0 and the drop slides down the structure48 (Figure 2.11b).
Figure 2.11: In order to induce a transition from a Cassie- to a Wenzel-state either the drop base has a.) to touch the ground between the asperities or b.) the contact angle at the side of the asperities has to reach the inherent advancing contact angle (advancing contact angle which would be observed on the flat surface). Thus, especially in the case where the Cassie-state is only metastable, the state in which a drop sits on a surface is mainly determined on how it was placed.47, 49 Transitions from the Wenzel-state into the Cassie-state have been much less investigated, but there are indications that it is possible when the Cassie-state is energetically preferred.48 In clouds, the water droplets are quite polydisperse, with a typical radius of 10 µm,47 thus much smaller than the droplets used for contact-angle measurements. It has been observed that water condenses mostly in the Wenzel-state on superhydrophobic surfaces which only display a microscale roughness. Upon fusing, the droplets turn into a mixed state, where Wenzel-type and Cassie-type wetting occur within the same drop. It has been observed that drops condensed on the double-structured lotus leaf end up being in the Cassie-state. There is an indication that nano-structures may exhibit condensationresistant properties. In this case, the critical feature size is probably given by the smallest stable drops formed upon condensation. If the features are smaller, the Cassie-state is the preferential state.47, 48 The feature height is an important parameter for the thermodynamic calculation of the energy barrier between the Wenzel- and the Cassie-state. For a hydrophobic surface the larger the height, the more favourable is the Cassie-state.18, 48 31
2 Theoretical Background - Liquid in Contact with Solid
2.10 Super-Oleophobic Surfaces A drop can be held in the Cassie-state, even if this is no more the lowest energy state. This can be achieved either by adding a roughness to the sides of the surface features to pin the contact line of the drop on top of the feature or by preparing surface features with undercut (overhanging) topographies. Tuteja et al.77 have shown that with a structure of hydrophobised “micro-hoodoos” (Figure 2.12) even alkanes (much lower surface energies than water) could be supported in the Cassie-state, and Ahuja et al.78 showed that a topography consisting of “nano-nails” could support even ethanol. Such a behaviour was also predicted around the same time by Liu et al.79
Figure 2.12: Schematic drawing of a micro-hoodoo (grey) as used by Tuteja et al.77 In Chapter 6 it is shown that such undercut structures can also be made slightly hydrophilic and still support a drop in the Cassie-state. In Chapter 7 undercut side-walls were employed to ensure that the water only wetted the flat pillar tops (precise definition of f1 ).
3 Materials and Methods In this chapter, it is introduced how the surface chemistry of substrates can be modified by self-assembled monolayers. Then, principles and conditions of the main characterisation techniques used in this work are presented. More details and additional information on the experimental setup will be given in the individual chapters.
3.1 Self-Assembled Monolayers (SAMs) A SAM is a monomolecular adlayer of surfactants in which all molecules bind with their head group to the substrate and expose their functional end group, which is separated with a spacer chain (normally alkane or fluoroalkane) from the head group, to the phase in contact with the substrate (see Figure 3.1). These surfactant molecules can be adsorbed from solution or the gas phase and organise (sometimes epitaxially) spontaneously into crystalline (or semi-crystalline) structures. Maximisation of van der Waals forces between neighbouring spacer groups is very often the main driving force to achieve dense and highly oriented structures. Depending on the substrate and the constitution of the spacer, a 2D crystalline organisation in the spacer layer is only possible above a certain chain length. For chain lengths that are too short for the van der Waals forces to be significant, the SAM remains in an unordered state. Thus, in such a case, not only the end group is facing e.g. air, but also groups from the spacer.80, 81 SAMs are excellent systems to study fundamental phenomena such as ordering and growth of surfactant molecules, wetting, adhesion, lubrication and corrosion.81 In this thesis, two different kinds of SAM-forming molecules were used: alkylthiols and perfluorinated silanes. Thiols were always adsorbed from ethanolic solutions and their chain length was sufficiently long to form an ordered structure on the (111) phases of the polycrystalline gold.80 The silanes were adsorbed from the vapour phase and remained in an unordered state due to their rather short chain length and weaker interaction between perfluorinated chains.82, 83 Sulfur compounds such as thiols have a strong affinity to transition-metal surfaces such as gold, silver, platinum and copper. Thiols on gold are the most thoroughly investigated SAMs. This is mainly due to the fact that gold is easily available and can be deposited 33
3 Materials and Methods
CF3 F2C CF2 F2C
CF2 F2C CF2 F2C
head group substrate
Figure 3.1: a.) Illustration of self-assembled monolayers (SAM), ordered – unordered SAM; b.) surface-active species used in this thesis. The thiol species (1: dodecyl-thiol, 2: mercapto-undecanol) are long enough to form ordered structures on gold, the shorter silane species (3: 1H, 1H, 2H, 2HPerfluorooctyltrichlorosilane) forms a disordered SAM on silica surfaces. by a simple physical vapour deposition. Additionally, it is an inert material and does not oxidise at room temperature.80 The adsorption of thiols can be divided in two main steps: in a first step, a layer of 80 – 90 % completeness forms within the first few seconds to minutes, depending on the molecule and the solution concentration used. Then the reaction rate considerably slows down in the second step where holes are filled and the order in the SAM is maximised. This reorganisation can take place since thiols are mobile on gold. This second step can last up to a week.80, 81 When working with Au / Cr-coated epoxy substrates (Chapters 6 and 7), there was a trade-off between adsorption time for perfect SAMs and the swelling of the epoxy due to uptake of ethanol. Therefore, the adsorption time for the 0.1 mM ethanolic thiol solution was reduced to 20 minutes, thus leading to nearly full SAMs, not yet perfectly ordered. This preparation time led to a reproducible SAM quality. The sulfur-metal bond and the spatial arrangement on the underlying gold lattice are still controversial. The Au-S bond is strong and estimated to be ∼50 kcal/mol.80 Despite the strength of the Au-S bond, the bond can be oxidised in ambient air by ozone attack. The attack starts at defects in the SAM, such as grain boundaries. This makes incomplete SAMs more prone to oxidation. Therefore, SAMs on gold should always be measured directly after preparation. If they need to be stored, then air-tight containers are needed.84 Mixed SAMs can be formed by co-adsorption of two different species. This co-adsorption is a way to tailor the surface energy of a substrate, if one compound has a hydrophilic and the other a hydrophobic end group, e.g. mercapto-undecanol and dodecylthiol (see Figure 3.1b.). The ratio within the SAM often is not equal to the ratio in solution. 34
3.2 Contact-Angle and Roll-off Angle Measurements The difference occurs due to different adsorption kinetics of the two species which is influenced by their solubility for a given solvent. If dodecylthiol (CH3 -terminated thiol) and mercaptoundecanol (OH-terminated thiol) are adsorbed from ethanol, ethanol is the slightly worse solvent for the CH3 -terminated thiol than for the OH-terminated thiol and will therefore adsorb preferentially on the surface. So, from a solution with equal molar quantities of both species, the amount of adsorbed CH3 - terminated thiols will be slightly larger on the surface. There can occur some degree of phase separation, but with co-adsorption this is minimal.80, 85 Silanes require hydroxylated surfaces for SAM formation, and the formation is sensitive to humidity. If there is no humidity, no full monolayer is formed. If there is too much humidity the silanes polymerise before reaching the surface, leading to the adsorption of polysiloxanes. They covalently bind to many different oxides, including silica and alumina surfaces (see also Chapter 4.2). Trichlorosilanes with their three reactive groups not only bind to the substrate but also polymerise with neighbouring silane groups. Thus the mobility of the individual molecule is strongly restricted and healing of defects after adsorption prohibited. Therefore well-ordered SAMs are harder to achieve with silane species than with thiols on gold. Nevertheless, due to their stability, alkyl silane SAMs are widely used in industry as adhesion promoters and boundary lubricants.81
3.2 Contact-Angle and Roll-off Angle Measurements Most surfaces were characterised by sessile-drop measurements. For a sessile-drop measurement, a drop of the test liquid is placed on the surface and the contact angle θ at the edge of the drop is measured. The setup to measure contact angles is fairly simple (see Figure 3.2). The observer’s eye, assisted by some magnifying optics and a camera has to be placed in front of the drop on the same axis as the substrate surface with a light source behind the drop. The drop, though transparent, due to its spherical shape disperses the light out of the line of sight and the drop appears therefore black in front of a bright background. The contact angle is found by drawing a tangent to the liquid phase at the three-phase-contact point between liquid, air and substrate and measuring the angle θ within the drop. The test liquid used in this thesis was always water. This section is subdivided into three parts, describing static and dynamic contact-angle measurements and the roll-off angle measurements. Roll-off angle measurements determine the angle to which a substrate has to be tilted until the drop starts to move. The influence of different parameters, such as drop size, analytical methods to measure the contact angle etc. on the result of the analysis will presented more precisely in Chapter 4. Most work was done on a DSA 100 equipped with a tilting frame (Krüss GmbH, Ger35
3 Materials and Methods
eye or camera
Figure 3.2: Sketch of the contact-angle-measurement principle. A drop of liquid is set on a substrate and illuminated by a light source. An optical setup magnifies the drop image and the image is either recorded by a camera or directly analysed by eye and a goniometer built in the optical setup. many), which is supplied with DSA 3 analysis software. Additionally, the results of the dynamic measurement shown in Chapter 6 were performed on an older system, the G2/G40 2.05-D, also from Krüss GmbH. In this chapter, static contact angles were measured on a Ramé-Hart contact-angle goniometer where analysis is done by eye, right at the machine. Some roll-off angle measurements were also made on a home-built tilting table. The tilting table was attached to a goniometer and could be slowly lifted by turning a crank handle.
3.2.1 Static Contact-Angle Measurements For a static contact-angle measurement (see Figure 3.3), a drop of about 6 µl is produced and then gently set on the surface. Then, both the left and the right contact angles are recorded. In order to obtain a meaningful value, at least three drops have to be measured on different spots of the substrates. The static contact angle is then the average over the six values. In the case of a smooth and homogenous surface, a standard deviation of about 1° can be expected . Despite hysteresis, static contact-angle measurements are highly reproducible, if they are always performed in the same way. The volume of 6 µl works for most surfaces. If a surface is very hydrophilic, a 6 µl drop can already spread over a large area, and therefore the substrate dimensions have to be large enough to enable three measurements, or a smaller volume has to be chosen. On superhydrophobic surfaces (contact angle above 150°) 6 µl drops do not detach from a syringe with diameter 0.5 mm. For these measurements a larger volume has to be chosen. With the 0.5 mm diameter syringe, 9 µl drops can usually be detached onto the superhydrophobic surface.
3.2.2 Dynamic Contact-Angle Measurements Dynamic contact angle measurements (Figure 3.4) describe the measurement of the two maximum contact angles possible with a certain liquid-substrate combination. In order 36
3.2 Contact-Angle and Roll-off Angle Measurements
Figure 3.3: Static sessile-drop measurement. θs denotes the two angles recorder during the measurement. to measure the advancing (receding) contact angle, the syringe remains within the drop and the drop volume is enlarged (reduced) at a constant flow speed. In this work it is typically 15 µl/min. During volume change, a movie is recorded and subsequently analysed by means of the tangent method 2 routine of the Krüss Drop-Shape Analysis software. This routine fits the slope in the vicinity of the three-phase-contact point with a polynomial function of the type (y = a + bx + cx0.5 + d/ ln x + e/x2 ). It is a very stable evaluation routine, because each side is evaluated independently and influences from the syringe are omitted by limiting the range of analysis of the silhouette for fitting. a.)
Figure 3.4: Sketch of dynamic contact-angle measurement: a.) advancing contact angle; b.) receding contact angle. Naturally, as long as both sides move, both sides are also used for the angle evaluation. In Figure 3.5, a typical data set from a dynamic contact-angle measurement is shown. The start of the experiment is indicated with a circle. At this position, the drop roughly adopts the static contact angle. Then, once the drop growth has started (drop base diameter increases) the contact angle scatters around one specific value, the advancing contact angle (here 118° ± 1°). From the moment where the liquid is sucked back into the syringe the contact angle also reduces to one specific contact angle, the receding contact angle (here 98° ±4°). Also here, only those contact-angle data points are taken into the evaluation, for which the drop-base diameter was decreasing. Towards the end where the drop volume starts to be small compared to the size of the syringe, an effect of the syringe on the contact angle can be detected. In a case such as that shown in Figure 3.5, it would be correct to discard the last bit of the contact-angle data and only construct the average from the first part of the receding experiment. This would lead to a receding contact angle of 100° ± 1°. 37
3 Materials and Methods On strongly pinning surfaces it is frequently observed that the receding contact line only moves in the very last moment of the experiment (very small residual drop volumes) and thus shows a small influence of the syringe. This receding contact angle then might be slightly underestimated, but also the standard deviation would increase, therefore nevertheless leading to a meaningful value. Contact-angle measurements on rough surfaces: Advancing: Rough surfaces in the Wenzel state are characterized by numerous and strong pinning events. This leads in the advancing experiment, for example, to socalled “stick-jump” behaviour, and thus to higher standard deviations. In some cases the contact line is strongly pinned on one side of the drop and only free to move on the other side, in which case only the contact angles measured on the moving side were evaluated. Drops in the Cassie state, especially those with contact angles above 150° (superhydrophobic), squeeze out preferentially on one side during the measurement. In this case, it is not pinning events that prevent one side from moving, but the high mobility of the drop on superhydrophobic substrates, which allows the drop to escape the confinement between the syringe and the substrate. Also in these cases only the contact angle on the moving side was evaluated. Receding: Clearly, rough surfaces in the Wenzel state strongly pin the receding contact line. This pinning strength can be so strong that no receding contact angle apart from 0° can be recorded (i.e. the area of the surface that was in contact with the drop remains wet).
3.2.3 Roll-off Angle Measurements The roll-off angle of a surface is the angle to which the substrate has to be tilted before the drop starts to move. There are two terms for this very same angle: roll-off angle and sliding angle. The value they describe is the same, but the mechanism of motion is not. On flat surfaces a drop tends to slide, while on superhydrophobic, self-cleaning surfaces a drop rolls off the surface. Since in this work the mechanism of movement was not investigated, all angles evaluated with a tilting experiment are denoted as roll-off angles (see Figure 3.6). This type of measurement is not as frequently used as the two methods described above. During tilting sometimes it is rather the front line that moves first, sometimes the back line. Therefore, the roll-off angle is defined as the angle at which the whole drop starts to move, i.e. both sides of the drop are in movement. 38
3.2 Contact-Angle and Roll-off Angle Measurements
Water Contact Angle [°]
start: ~ static
influence of syringe 100
1.5 2.0 2.5 drop base diameter [mm]
Figure 3.5: Measurement of the dynamic contact angle on a smooth PDMS sample. The contact angles on both sides of the drop are plotted against the drop-base diameter. The start of the experiment is indicated with a circle. The advancing contact angle is detected as soon as the drop base diameter increases. When drop enlargement stops, the drop relaxes to its equilibrium angle. As the drop reduction starts, the contact angle decreases. The receding contact angle is averaged from the contact angle data where the angle stays constant and the drop base diameter starts to decrease. In this data set an influence of the syringe can be detected.
3 Materials and Methods
α Figure 3.6: Sketch of a drop rolling off a surface. The angle α denotes the angle to which the surface has to be tilted before the drop starts to move.
3.3 Scanning Electron Microscopy (SEM) Since most surfaces in this work showed at least a micrometer-scale roughness that was too rough for visualisation by atomic force microscopy (AFM), but also features smaller than the resolution of light microscopy, the scanning electron microscope (SEM) was best suited for surface topography analysis. SEM is a well-known and frequently used imaging technique. It offers high magnification and resolution combined with a large depth of focus. Depending on the microscope, a resolution in the range of 10 nm can be achieved and structural features over a large length scale up to hundreds of micrometers can be analysed. Since electrons are used as illumination source, the samples are imaged under high vacuum (10-5 to 10-6 mbar). The probing beam consists of highenergy electrons (short wavelength) produced by an electron gun (see Figure 3.7a) and focusing is achieved by means of electromagnetic lenses. Scan coils move the focused electron beam across the specimen where it interacts with the specimen’s surface. The image is obtained by collecting and counting the emitted electrons. The more electrons emerging from one spot, the brighter the grey tone. Preconditions for analysing a substrate in SEM are: the substrate must be stable in high vacuum, it must withstand the high-energy electron flux and it needs to be conductive, in order to prevent charging. The high-energy electron beam (primary electrons) interacts with the surface and produces different types of signals. Information can be gained from the elastically backscattered primary electrons (BSE), the low-energy secondary electrons (SE, < 50 eV) produced by inelastic scattering of the primary electrons, the characteristic X-rays emitted from electrons of higher shells filling the holes in lower energy electron shells and cathodoluminescence. Often, only selected information is used from a specimen. Generally the BSE emerge from deeper areas of the sample and carry information about the material itself, whereas the SE mostly provide topographical information, emerging from the first few nanometers of the surface. Since interest in this work is focused on the topographical information, only SE images were collected. Different types of SE electrons can be distinguished. SE1 are generated at the spot where the primary beam hits the surface. They provide a high resolution signal, since these electrons are only emitted 40
3.3 Scanning Electron Microscopy (SEM) from within the beam diameter. The result is an image with high clarity (low noise) showing surface characteristics (topography, structures) very precisely. SE2 are generated by BSE that have returned to the surface after several inelastic scattering events in the bulk sample. They emerge from an area larger than the beam diameter and have therefore a slightly lower resolution to SE1. But since the detector is placed at an angle to the surface, the image appears more like an actual photograph with a light source and shadows. Images from SE2 are also less prone to charging effects. A SEM image consists of pixelated greyscale information. As the focused beam is scanned over the surface, each pixel corresponds to one measurement of the signals produced by the SEM beam on one spot of the sample surface. The brighter the pixel, the more electrons that emerged from this spot. Increasing the dwell time on one spot increases the signal-to-noise ratio but can also induce damage in the substrate. Magnification in SEM is achieved by changing the size of the area scanned by the beam. Since electromagnetic lenses are always convex, the lenses are only there for focusing and moving the beam and not a means to enlarge the image. The size of the chosen area and the fixed-pixel-array of the image then lead to the specific magnification.
a.) electron gun
emitted SE display specimen
scan coils objective lens detectors specimen
SE not leaving the sample
amplifier e-beam interaction volume
Figure 3.7: Working principle of a SEM. a.) E-beam path: High-energy electrons are emitted at the electron gun and focused over a lens system on the specimen. The detectors acquire the emitted electrons and are directed to the signal amplifier. The image evolves when scanning the beam (scan coils) over the specimen area. b.) grey tones in an SEM image represent how many electrons emerged from one spot. On steep slopes, more electrons can escape the surface, thus giving topographical information. 41
3 Materials and Methods If the beam hits the surface perpendicularly, a certain number of electrons is emitted. In case the surface shows a slope, then more area of the surface is within the mean free path of the electrons to escape the substrate. Therefore more electrons are emitted from steep sides of the substrate (see Figure 3.7b). When acquiring a SEM image, the most information is obtained if the full dynamic range from white to black is utilised. The system used in this work was a Zeiss Gemini 1530 (Carl Zeiss SMT, Germany) equipped with a field-emission gun and a BSE and SE detector. On this machine the acceleration voltage (EHT) ranged from 0.3 to 30 kV, but most images were acquired with acceleration voltages of 3 -5 kV. Since all analysed samples were basically insulators, they were always sputter coated with a 5 nm thick layer of platinum. If the substrate was already gold coated from the experiment, then naturally no additional platinum coating was necessary to ensure conductivity.
3.4 Light Microscopy (LM) Light microscopy is one of the oldest techniques used to analyse small objects. Visible light, being either reflected from or transmitted through the specimen, is gathered and magnified by a system of lenses. A precondition for transmission is that the specimen is either very thin or transparent. All samples investigated were of micrometer thickness and often not transparent. Therefore, only bright-field reflected light microscopy was used in this thesis. In principle the light emerging from a halogen lamp source is focused on the spot of interest on the specimen. The reflected light passes through an objective lens, which inverts and greatly enlarges the image of the specimen owing to its short focal length. This image of the specimen is then gathered by a second lens, the ocular or eyepiece. The ocular lens inverts the image once more (back to normal) and also enlarges the image once more (virtual image). The visual magnification of a microscope is calculated by multiplying the magnification values of the objective and the ocular. The strongest objective used here magnified the specimen 64 fold.
3.5 White-Light Profilometry (WLP)
3.5 White-Light Profilometry (WLP)
Profilometry is used to characterise the topography and roughness of a substrate. Light profilometers are state-of-the-art, non-contact, non-destructive analysis tools. They can be used for line profiles as well as for topographical maps, when scanned over an area. The data achieved consists of a z-value for each x-y-position, very similar to the height maps obtained with atomic force microscopy (AFM). In contrast to AFM it does not consist of a tip mapping the topography, but a beam of white light. This leads to a comparable resolution to AFM in the z direction, but to significantly poorer resolution in the x and y-directions. With the machine used in this work (FRT MicroGlider, Fries Research & Technology GmbH, Germany, equipped with a CWL 300 µm detector), the x-y-resolution was 1 µm. The advantage over the AFM is that large areas (such as 10 x 10 mm2 ) with profound topographical features can be measured accurately and reasonably fast. Naturally, only topographies directly accessible by the beam of light can be mapped by the profilometer. Undercut structures can not be measured by the instrument. Depending on the sensor, light profilometers use three different principles to detect the height. The z information can be gained by finding the focus on the surface (confocal), by converting a phase map obtained from fringe-interference patterns into a height profile (interferometric) or by making use of chromatic aberration (chromatic). The machine in this work was equipped with a chromatic white-light sensor (CWL) (see Figure 3.8). First, white light is generated by a halogen lamp. The light is transferred into the passive sensor. It is called passive because focusing does not require any movement in the z-direction, from the objective or from the sample. The optics in the sensor disperse the white light into its constituent wavelengths λ. Each wavelength has its focal point at a wavelength-specific distance. Blue, for example, has a very short focal distance, whereas red has a longer focal distance. The reflected light is gathered and analysed in a spectrometer. Since only one wavelength (e.g. green in Figure 3.8) is in focus on the sample, most light will come from that wavelength. The peak λ corresponds then to a height. The z range (∆z) measurable with the sensor used is 300 µm. By measuring several ∆z ranges after each other, topographies of up to 3 mm can be mapped with a z-resolution of below 3 nm. The chromatic-aberration technique (CWL detector) has the advantage of working under high-angular-slope conditions. For strongly reflecting surfaces, slopes down to 30° to the horizontal are possible and on samples with low reflectivity (much dispersion due to roughness) down to nearly 90°. Neither sample colour nor transparency strongly affect the measurement; black materials or glass slides can also be analysed. 43
3 Materials and Methods white light source
spectrometer sensor: lens system
Figure 3.8: Sketch of the principle of white-light profilometry. The white light is focused via fibres and the sensor on the surface. Due to the strong chromatic aberration of the last lens the white light is split up in its constituent wavelengths, leading to different focal distances of the wavelengths. This way, only one specific wavelength is in focus (here: green). The reflected spectrum is analysed in the spectrometer and then the peak wavelength is transformed into a z-value.
3.6 Differential Scanning Calorimetry (DSC) Differential Scanning Calorimetry (DSC) is a thermoanalytical technique and mainly used to monitor phase transitions and define heat capacities. It can also be used as a quality control for sample purity or to study polymer curing. The measurement works as follows: A reference and the sample are placed in two different heating chambers and then heated up such that the temperature of both increases linearly and equally. The reference needs to be a substance that shows no phase transition in the temperature range of interest and has a known heat capacity, such as sapphire. It is also possible to calibrate the machine, e.g. with measurements of indium. Thereafter the reference can also consist of a can filled only by air. By heating, the sample experiences endothermic (such as melting, transition over the glass temperature, desorption) or exothermic (such as crystallisation, curing, decomposition) transitions. To undergo these transitions, the sample either needs more (endothermic reaction) or less (exothermic reaction) energy to increase in temperature at the same rate as the reference. This difference in amount of energy required can be measured directly and allows the determination of 44
heat flow [mW]
3.6 Differential Scanning Calorimetry (DSC)
decomposition Temperature [°C]
Figure 3.9: Typcial DSC curve heat flow [mW] versus temperature [°C]. Heat capacities (C) can be extracted from constant ranges in this curve. In addition, a transition over the glass temperature (TG), two exothermic (here: curing, decomposition) and one endothermic peaks (desorption) are visible. the heat capacity of the material, transition temperatures (glass temperature T g, melting temperature T m, crystallisation temperature T c, decomposition temperature T d) and the corresponding enthalpies. This information can then be used to produce phase diagrams for various chemical systems. The outcome of a measurement with DSC is a curve of heat flow [mW] versus temperature [°C] or time as shown in Figure 3.9. In the range where the curve stays constant (C), the heat capacity of the material in this state can be extracted. A step-like transition (T g), indicating a change in the heat capacity but not in enthalpy describes a transition over the glass temperature T g, at which point, in polymers, the amorphous glassy components transform into a rubbery and soft state. The T g is the temperature in the middle between the onset and the endpoint of the transition. Integrating the area of the endothermic and exothermic peaks allows calculating the enthalpies of the transition. For polymeric substances, DSC measurements (and especially T g) strongly depend on the thermal and processing history of the material. Therefore, if one wants to compare different materials care should be taken that they always had the same thermal history. The history can be deleted if the heating cycle is run twice, thus the second cycle will be revealing rather material properties than thermal history. In this work the aim with the DSC was to characterise different epoxy thermosetting polymers used in the replica technique to find out in what state they are after different curing protocols. Therefore, the values obtained in the first heating cycle were of interest, not those from the second cycle. DSC measurements were performed on a Mettler Toledo DSC 822e (Switzerland) instrument, calibrated with Indium. Samples had a weight of about 17 mg and were tested 45
3 Materials and Methods under a stream of nitrogen. The heating/cooling rate was 10 °C/min, as is recommended for T g measurements. Two heating/cooling cycles from -50 °C to 250 °C were run. For transition-temperature evaluation, the temperature at the peak maximum was taken.
4 Parameters Influencing Contact Angle Measurements The aim of this chapter is to summarise measurements and data that offer a basic understanding of the materials and methods used in the two larger (published) chapters (6 and 7). Using standard, model surfaces (thiols on gold/silicon wafer), the influence of several parameters (except profound surface roughness, covered in chapters 6 and 7) on contact angle measurements is investigated, different measurement techniques are compared and a few theoretical assumptions tested. Perfluorinated silanes were used throughout the thesis either as a demoulding or a hydrophobising agent. A quick contact-angle survey shows which materials relevant to this thesis can be functionalised by silanes, adsorbed from the vapour phase. PDMS (polydimethylsiloxane) is a very versatile elastomeric polymer and is used in many fields of research (such as micro-contact printing,80 micro-fluidics,86 cell studies87 and anti-fouling surfaces88, 89 ). It is known that a native PDMS surface90 experiences a change in conformation of the surface layer upon contact with water, influencing contact angle measurements. A few tests were performed to better understand the effect of different treatments on the contact angle. In the last section some issues that arise when measuring very high contact angles are presented and discussed.
4.1 The Measurement 4.1.1 Introduction In the literature, several different ways of measuring contact angles have been described. In order to find out how these measuring methods can be compared, different methods were tested on standard substrates (thiols on gold/silicon wafer) with a standard liquid (water). 47
4 Parameters Influencing Contact Angle Measurements A standard substrate for contact angle measurements needs to be clean, stiff and rigid, flat and smooth, chemically homogeneous and inert towards the test liquid. For lab purposes, mica sheets (layer silicate), polished silicon wafers and glass microscope slides (float glass) are often used for contact angle measurements. They all have the advantage that they are reasonably smooth, can be modified by surface chemistry or can be coated with any metal or polymer of choice. They are resistant to ultra-high vacuum and to many acids, bases and solvents. They are readily available and affordable. The choice of the thiol/gold/silicon wafer system was made because self-assembled monolayers of thiols on gold (Chapter 3.1) are well-known and well-characterised80 and the surface energy can easily be tailored by using either an OH-terminated (hydrophilic) or CH3 terminated (hydrophobic) thiol species or mixtures of them (see Chapter 3.1). An ideal liquid for contact angle measurements has a low vapour pressure, is pure, does not take up any substances from the atmosphere (no self-contamination), has a low viscosity and is preferentially non-toxic.35 Silicone oils best meet these conditions; additionally they can be tailored in viscosity by varying the chain length of the silicone species and are therefore often used to study wetting phenomena. But they have one big disadvantage: silicones are known to contaminate any surface with low-molecular weight species.91 Due to their low surface tension (fully spreading on nearly every surface) and non-volatility, they form a precursor film ahead of the drop via a surface-diffusion process, which is one monolayer to 10 nm in thickness, and will propagate over any surface until the drop is entirely depleted, covering the surface with a layer of silicone.41, 92 This silicone contamination is to be avoided in every surface science lab that is involved in any activity besides working with silicones. Therefore, water is often chosen as test fluid. It is of great technical interest and readily available. Due to its high surface tension it shows a finite contact angle on many surfaces. When measuring with water one must take care to always use clean and fresh water (minimise self contamination) and not to prolong the measurement longer than necessary (to minimise the effect of evaporation). If these conditions are met, contact angles will show values close to that predicted by the Young’s equation.35 Other parameters that may influence the contact-angle measurements are temperature, atmospheric pressure and humidity (relative vapour pressure). As already shown in Chapter 3, contact angles can be measured either statically or dynamically. Figure 4.4 shows a compilation of different ways in which contact angles can be measured. This compilation of methods is far from complete; it only shows the methods that were tested in the present work and are reported in this chapter. More approaches can be found in the books by de Gennes et. al (chapter 2)35 and Adamson and Gast (mainly chapter 10).38 Generally, the static measurement is performed by producing a specific drop volume that is still hanging from a syringe needle, which is then gently brought into contact 48
4.1 The Measurement with the solid. Upon contact with the solid, the drop will detach from the syringe and spread on the surface, decreasing the contact angle from initially 180° (prior to contact) to the static (“equilibrium”) contact angle θs. For small drops, the driving forces are only capillary forces (surface tension). During the measurement the syringe tip does not remain within the drop. If one uses drop volumes larger than the maximum volume that stays attached to the syringe, then the drop needs to be formed between surface and syringe. This method can lead to a slightly higher contact angle, but it reduces the influence of impact of falling droplets. Static measurements are usually quicker than dynamic measurements, but the single value yields less information about the surface than the dynamic measurement. It is mainly used as a rapid characterisation tool. If, by a chemical reaction or other type of surface coating, a significant change in surface energy can be expected, then θs provides rapid information on whether, and within certain limits, to what extent the experiment was successful. The spatial resolution of a sessile-drop measurement is defined by the drop-base diameter. The resulting contact angle is an average over the whole length of the contact line,9, 55 (whole area of contact44 ), if the surface features (chemical or topographical) are small compared to the drop. Usually micro-litre-sized droplets are used for measurements. The spatial resolution can be increased by the use of pico-litre-sized droplets. Pico-litre droplets have to be measured immediately after deposition to minimise the influence of evaporation (reduction of contact angle), but are indeed comparable to data measured with micro-litre drops.93 If surfaces are smooth and homogeneous, static measurements provide data for calculations in thermodynamic equilibrium, and the surface tensions γSL and γSA can be calculated by applying the appropriate model (see comment further below). On surfaces where chemical heterogeneities or surface roughness are homogeneously distributed along the surface and are small compared to the drop’s footprint, repeated static contact angle measurements can lead to values with small standard deviations, if the drop is always set on the surface by the same method and with care. Such contact angles most probably are far from the thermodynamic equilibrium but nevertheless allow the comparison of different surfaces. θs measurements suffer from the disadvantage that heterogeneities and surface roughness cannot be distinguished from intrinsic surface energy. In the course of the measurement, swelling or dissolution may also occur and may not be detected due to the limited time scale of the measurement. Therefore, static contact angle measurements must always be treated with caution.94 Influence of drop size V on contact angle θs, drop base diameter r and roll-off angle α: Surprisingly, the literature does not provide a conclusive answer to the question as to the influence of the droplet volume V on contact angle θs, drop base diameter r and roll-off angle α. The shape of the drop is determined by the surface tensions involved and by gravity. For every liquid there exists a particular length, the capillary length 49
4 Parameters Influencing Contact Angle Measurements κ−1 , beyond which gravity becomes important.35 When capillary forces and hydrostatic forces are in equilibrium, the capillary length can be calculated by comparing the two forces; with g being the acceleration due to gravity and ρ the density of the liquid: −1
While on the one hand the capillary length defines the characteristic length within which a perturbation of a surface decays, on the other hand it splits the observation of sessile drops into two regimes. If the radius r of the drop is smaller than the capillary length, then gravitational effects on the shape of the drop can be neglected. In the opposite case when r > κ−1 , gravity flattens the drop to a puddle with an equilibrium height e. e itself is a function of κ−1 and the equilibrium contact angle θY . In contact-angle measurements it defines the maximum radius that a drop can adopt before gravity starts to distort its profile from the ideal truncated sphere (see Figure 4.1).
Figure 4.1: For small droplets it can be assumed that gravity has no effect on their profile. The limit for the maximum radius is given by the capillary length κ−1 . a.) small droplet showing no sign of gravitational effects, b.) large puddle with a diameter larger than twice the capillary length, which is flattened by gravity. In fluid mechanics, the Bond number is a numerical expression of the ratio of gravitational (volume) to surface forces. For a sessile drop measured in air the Bond number Bo is defined as follows:38
ρ · g · r2 γLA
If the Bond number is equal to one, the expression is equal to the capillary length if solved for the drop-base radius r. Numbers well below unity describe static systems that are only governed by surface forces, and in these cases, gravitational forces can be neglected; for very high numbers it is the opposite. Numbers around unity describe a system where the two forces are in a non-trivial balance. As mentioned above, if a drop is unperturbed by gravity, it assumes the form of a truncated sphere. By knowing the dispensed volume V0 and the measured contact angle, the drop-base radius r can be calculated (see Equations 4.3 and 4.4). For a given (or 50
4.1 The Measurement expected) wettability (θ) of a surface the drop-base diameter for a certain volume can be predicted, which provides useful information for the selection of specimen size or what can be the maximum spatial resolution for contact-angle measurements. By actually measuring the radius r on a standard substrate, conclusions can be drawn as to how accurate the predictions are and whether an influence of gravity can be detected and whether it is significant. In Figure 4.2, the different geometrical parameters are outlined and the diameter (2r) of the drop’s footprint is calculated for the hydrophilic and hydrophobic case as follows:
6 · V0 · sin 3θ π (1 − cos θ) 3 · sin2 θ + (1 − cos θ)2
−3 · V0 · sin3 β π (1 − cos β)2 (2 + cos β) − 4
!1/3 , where β = 180° − θ
r β hydrophobic
Figure 4.2: Schematic illustration of which parameters were used to calculate the radius of the drop’s footprint from the dispensed volume. In 1962 Furmidge64 was looking into the conditions necessary for drop retention on inclined surfaces and, by equating all the forces involved, defined a correlation between the inclination α of the surface with the drop weight m, drop diameter perpendicular to the sliding direction w, advancing and receding contact angles (θa, θr) and the surface tension of the liquid.
m · g · sin α = w · γLA · (cos θr − cos θa)
Furmidge (see Figure 4.3) assumed the footprint of the drop to be rectangular during sliding, which may be a source of error, but he found the equation to be sufficiently accurate to be able to account for sliding drops on a flat surface. The Young’s equation (Chapter 2.4) correlates the involved surface tensions γ by using one specific contact angle θY . Therefore, on an ideal surface, no hysteresis would occur.95 51
4 Parameters Influencing Contact Angle Measurements
γLS θa γLS
Figure 4.3: Sketch of a sliding drop in the side and top view with indication of all the parameters involved in the Furmidge equation: drop weight m, acceleration due to gravity g, inclination angle α, drop width w, liquid-air surface tension γLA and dynamic contact angles θa and θr However, measuring on real surfaces reveals the fact that on every surface the contact angle θ can adopt any contact angle between a minimum (receding, θr) and a maximum (advancing θa) contact angle. The difference (θa − θr) between those two quantities is known as the hysteresis. But even if a surface were perfectly smooth, chemically homogeneous and perfectly elastic, hysteresis could only be prevented if the experiment was carried out infinitely slowly, to ensure thermodynamic reversibility. Since this is not possible, energy-dissipating processes occur. Upon contact, chemical effects, such as interdiffusion, interdigitation, molecular reorientations and exchange processes occur. Due to the finite time of measurement, the work of adhesion W (Young-Dupré: W = γLA (1 + cos θ), Equation 2.9) of the separation (receding) is larger than that of the approach (advancing).96 More significant causes of hysteresis are attributed to chemical heterogeneities and surface roughness.41 Hysteresis occurs due to contact-line pinning, e.g. on strong and sparse defects.35, 41, 43, 65 More information on the hysteresis induced by roughness can be found in Chapters 2, 6 and 7. In general, hysteresis is still a subject that is not fully understood. Many issues remain to be solved on what the nature and location of energy dissipation is, and how solid-liquid interactions, surface heterogeneity and roughness, substrate viscoelasticity and stiffness influence wetting dynamics.97 Even if no comprehensive model for hysteresis has been found, its experimental value can be determined by dynamic contact-angle measurements. Dynamic measurement: Influence of choice of method During a dynamic measurement the contact line is forced into movement. The contact angle facing fresh surface (larger contact angle) is known as the advancing contact angle θa while the contact angle moving over surface which had already been wet by the drop (smaller contact angle) is denoted as the receding contact angle θr. In principle, during wetting the contact line of the drop measured statically encounters the same underlying mechanisms as the advancing contact angle, and therefore usually θs and θa generally show similar effects.98 Some have even replaced θs with θa in their calculations95 —a 52
4.1 The Measurement
Figure 4.4: Measuring contact angles: Contact angle measurements can basically be divided into static and dynamic measurements. a.) depicts a static measurement. The syringe is out of the drop and the drop is at rest. b.-e.) show dynamic measurements where the contact line of the drop is forced to move: b.) by changing the drop volume incrementally; c.) by changing the drop volume at a constant speed; d.) by dragging the drop along the surface and e.) by tilting the substrate until the drop starts to move. procedure that others do not endorse.97
Figure 4.5: Schematic drawing of contact-angle-hysteresis evolution with increasing contact line speed U.98 The main parameters that quantify the dynamics of wetting are the contact-line speed U and the dynamic contact angles θa and θr. θa and θr are macroscopic quantities, since they are usually measured by relatively simple optical setups. θa and θr can get pinned, and sometimes displays a stick-slip behaviour that renders its interpretation difficult and the achievement of steady velocities experimentally challenging. Advancing (receding) contact angles increase (decrease) with increasing contact line speed U (see Figure 4.5). Two main approaches (models) were developed to explain dynamic wetting. They differ from each other both in terms of conceptual framework (mechanical versus chemical model35 ) and the channel of energy dissipation.98 The first is known as hydrodynamic theory and the energy dissipation occurs due to viscous flow within the wedge of the liquid near the moving contact line. The second model is known as molecular-kinetic theory. Liquid transport is described as a statistical process of thermally activated local displacements at the contact line. Energy is mainly dissipated by contact-line friction.92, 97, 98 53
4 Parameters Influencing Contact Angle Measurements There are indications e.g. in the process of dewetting (receding), that the hydrodynamic theory fits better for higher contact line speeds (speeds occurring right after placement of a drop),whereas the molecular-kinetic theory describes the slower range shortly before complete relaxation. Additionally, models that involve pinning have been proposed.97 Researchers have developed several different approaches to moving the contact line. Figure 4.4b shows the case where the drop volume is enlarged incrementally until the contact line moves and then the contact angle is measured. It is employed in cases when no automated machine is available.42, 90, 99 The automated version of this type of measurement is shown in Figure 4.4c where the drop volume is changed by enlarging and reducing the drop at a constant dosing speed. This is the type that has been used most frequently in our group,22, 100, 101 and will be referred to as “automated” in the subsequent discussion. Figure 4.4d forces the contact line to move by moving the substrate underneath the drop10, 19 and will be referred to as “drawing”. If the substrate is tilted to the point where the drop starts to move, the front angle can be considered as the advancing contact angle and the back angle as the receding contact angle.33, 64 This type of measurement will be abbreviated as “tilting”. A comment on the possibilities of measuring γSA and γSL : The Young’s equation consists of four parameters, but only θ and γLA can be measured independently, which leaves γSA and γSL undefined. From the static contact angle data of different, carefully chosen test liquids measured on the surface of interest, the surface tensions γSA and γSL can be calculated?, 102, 103 by employing one of the models developed by Zisman,104 Fowkes,37 Owens and Wendt105 and vanOss et al.106 These models, starting with Fowkes, made the assumption that the kinds of interactions occurring at the liquid-solid interface play an important role. They distinguish the interactions as either polar or dispersive,37, 105 or, within dispersive, acid and base interactions.106 Thus, only interactions of the same nature will combine, the resultant being calculated from the geometric mean of the fraction of the total surface tensions γLA and γSA . For example, the surface tension γLA of water (72.8 mN/m) consists only of 30 % dispersive interactions,37 meaning that only these 30 % of the total surface tension γLA will attractively interact with a dispersive surface such as polyethylene (purely dispersive, γSA ≈ 35 mN/m37 ), leading to a low adhesion and large contact angles. The group of Neumann has cast serious doubt on the accuracy of these models95 and proposes an alternative, equation-of-state approach. They claim that other authors developed their approaches because they did not have reliable contact angle data, and further state that the models “contradict physical reality”. Their main criticism was that these models were developed after measuring mainly static contact angles. By measuring static contact angles, effects like stick-jump behaviour, dissolution of the surface, adsorption of molecules onto the surface cannot be recognised as they can by measuring slow advancing contact angles. Thus, to explain the scatter in their data, Fowkes and 54
4.1 The Measurement others came up with the idea that it is not only the surface tension between the phases which defines the contact angle, but also the type of intermolecular forces at work. This assumption is what the Neumann group strongly doubts, since they can show that by choosing the liquids that provide (in their eyes) reliable data, polar and apolar liquids fall smoothly on the same curve in the γLV cos θ vs. γLV plot, which would not be expected, if the specific intermolecular interactions indeed played a role.
4.1.2 Experimental Substrates: Single-side-polished silicon wafers (Si-Mat Silicon Materials, Germany) of dimensions 10 x 40 mm2 were cleaned by ultrasonication, first for 10 min in toluene (for HPLC, Acros Organics, Belgium), then twice 10 min in ethanol (analytical grade, Scharlau, Spain) to remove glue residues from the cutting and then blown dry in a stream of nitrogen. Additionally, in order to be able to measure large drops on a substrate with a hydrophilic coating, one microscope slide (40 x 40 mm2 , custome size, Menzel-Gläser, Germany) was cleaned with Piranha solution (30 v% H2O2 in 70 v% fuming sulphuric acid 95 – 97 %, both pro analysi, Merck, Germany). Subsequently the silicon wafers were cleaned for 2 min in air plasma (Harrick Plasma, USA) and then, together with the glass slide, coated by resistance evaporation (MED 020 coating system, Baltec, Liechtenstein) with 10 nm of Cr and 80-100 nm of Au (purity >99.99 %, Umicore, Liechtenstein). Surface modification: Directly after coating, the samples were immersed in a 1.2 mM ethanolic solution of 11-mercaptoundecanol or dodecanethiol (Aldrich Chemicals, USA). Self-assembly took place overnight. Directly before the measurement the samples were removed from the solution, rinsed with ethanol and dried under a stream of nitrogen. Quality control: Before starting the measurement, the quality of the water was checked by a pendentdrop experiment (DSA100, Krüss GmbH, Germany) and found to be very pure with a surface tension of above 72.4 mN/m.107 Room temperature was 23 °C and the relative humidity 51 % while measuring the dodecanethiol SAMs and 26 % for the 11mercaptoundecanol SAMs. Additionally, the crystallinity and the organization of the alkyl chains of the SAM were controlled by polarization-modulation infrared reflection absorption spectroscopy (PMIRRAS). The position of the peak maxima and the broadness of the absorption bands indicate that the SAMs were of high quality.108 Contact-angle measurements: All contact-angle and roll-off angle measurements were performed on the DSA100 (Krüss GmbH, Germany) and evaluated by the tangent-method-2 of the DSA3 software (Krüss 55
4 Parameters Influencing Contact Angle Measurements
Table 4.1: Volumes used to measure static and roll-off measurements on the two different substrates. SAM / Volume [µl]
GmbH, Germany). Contact angles were then exported into Excel (Microsoft, USA) for further evaluation (average, standard deviation, deletion of artefacts occurring from faulty profile detection). The static measurements were carried out with drops of between 1 and 30 µl volume (Table 4.1), which were gently placed onto the substrates. An image was taken of each drop and subsequently the contact angles were evaluated. Drop-base diameters were extracted with the straight-line-selection tool in the ImageJ program (rsbweb.nih.gov/ij). Pixels were calibrated by the width of the syringe (0.516 mm), which was also visible in the image. During the automated measurement a movie is recorded while enlarging or reducing the drop volume at a constant flow speed. The contact angles are then subsequently evaluated from the movie. The measurements were performed at different flow speeds (5, 10, 15, 20, 25, 30 µl/min). First, a drop of 5 µl was set onto the substrate, with the syringe still within the drop. Then, the drop was enlarged in two steps of 4 µl to 13 µl at the corresponding speed. The reason to divide the advancing measurement into two movies was to provide the possibility of re-centreing the drop to the syringe, in case it had initially grown asymmetrically. The experiment with the hydrophilic substrates was repeated with additional measurements of the advancing contact angle by increasing the drop in one step by 8 µl at the corresponding speed. The recording settings were chosen such that at least 60 images of the drop could be evaluated (up to 170). The receding measurement consisted of only one movie. Only those images were evaluated in which the side of the drop was actually seen to be moving. For the drawing measurement a drop of 8 µl was placed on a substrate coated with the hydrophobic SAM, with the syringe close to the surface. Then, the stage was moved underneath the drop at three different speeds offered by the machine: fast (11.1 mm/sec), normal (6.3 mm/sec) and slow (2.9 mm/sec). During the movement, a movie was recorded. The tilting measurement was performed directly after the static measurement. Therefore the same drop volumes as mentioned above were investigated. All samples were tilted with a speed of 1.2 °/sec (machine parameter “slow”) and during tilting a movie was recorded. As soon as the drop started to move, tilting was stopped at the correspon56
4.1 The Measurement ding angle and advancing and receding contact angles were evaluated from the moment where the drop started to move on both sides. Images for evaluation varied between 3 to 10 images per drop, depending on the distance available for sliding and the sliding speed of the drop.
4.1.3 Results and Discussion Static contact angles: Figure 4.6 shows the results from the static-contact-angle measurements versus droplet volume. The same contact-angle range is shown for the hydrophilic and hydrophobic SAMs (Figure 4.6a and b), to allow comparison of the standard deviations. For both substrate types, no distinct dependence of the contact angles on the investigated drop volumes can be observed. All data points scatter around one specific value (20° ± 2° for the hydrophilic, 108.7° ± 0.6° for the hydrophobic substrate). The same behaviour was observed in another study93 where contact angles were also investigated by the tangent method. Others109, 110 have reported a slight decrease with drop size, a tendency which might be hidden within the (small) standard deviation of the data presented here. On the hydrophilic substrate, the contact angle of the 30 µl drop seems to be slightly higher than expected. Since the 30 µl drop had a diameter of nearly 10 mm, it could only be measured on the large sample—the OH-thiol/gold-coated glass slide. The different substrate is indicated by an open instead of a full symbol in Figure 4.6a. Other drop sizes (data not shown) were also measured on the very same sample and they all showed the same trend: they scattered around 23° instead of 20°. Since the standard deviations from the data measured on the silicon wafer and the glass slide slightly overlap, no definite conclusion can be drawn that this influence is due the choice of the substrate. But all other parameters (gold coating, solvent preparation, glass cleaning) were held constant. The standard deviation on the hydrophilic substrate is significantly larger than that observed on the hydrophobic substrate. This may have several origins: hydrophilic surfaces with their low contact angles are more difficult to measure.36 Also, the apex of the drop reflects the light, which can confuse the software. Additionally, hydrophilic surfaces are more prone to airborne contamination than hydrophobic surfaces, even though all the surfaces used here were exposed to air as briefly as possible. They were measured immediately after removal from the thiol solution. On each substrate coated with the hydrophobic SAM, one additional drop of 6 µl was measured. It was intended as a control to be sure that all substrates had indeed the same quality. But because space was limited this 6 µl drop was always placed at the end where the sample was held with the tweezers. When evaluating the data it was 57
4 Parameters Influencing Contact Angle Measurements
110 θs [°]
clear that the contact angles measured on this spot was again higher than on the untouched areas, namely 110.1° ±0.4° rather than 108.7° ± 0.6°. In order to achieve very reproducible results it is necessary to always measure on untouched surfaces.
10 15 20 25 drop volume [µl]
10 15 20 25 drop volume [µl]
Figure 4.6: Influence of drop volume on static contact angle measured on a) OH- and b) CH3 -terminated thiols on gold/silicon. For the observed drop volume range no significant influence of drop size on contact angle was found. The data points in Figure 4.7a were obtained by extracting the drop-base diameters from the images of the static measurements. The dashed line represents the limit given by the capillary length (Equation 4.1) and the solid lines show the evolution of the drop-base diameter when calculated from the contact-angle average from the static measurement (θphilic = 20°, θphobic = 109°) and the dispensed drop volume as shown in Equations 4.3 and 4.4. For both substrates and for drop sizes up to 15 µl, the calculated curves lie within the standard deviation of the measured drop-base diameters. Distinct underestimation occurs for 30-µl-drops, and this was slightly more pronounced on the hydrophilic than on the hydrophobic substrate. For the hydrophilic case, the diameter already intercepts the capillary length criterion at a volume of ∼6 µl and indeed, the curve starts to deviate, even though, as mentioned, the deviation only starts to become obvious for drops larger than 15 µl. The intercept for the calculated hydrophobic case with the capillary length would only be at a drop volume of 73 µl. For the hydrophobic SAMs, the Bond numbers (Equation 4.2) vary from 0.06 to 0.63, and for the hydrophilic SAMs from 0.26 to 3.27. None of these values is far away from unity (as is the case for pico-litre droplets, for example93 ), so all measurements are probably influenced by surface and volume forces. The two SAM species (OH- and CH3 terminated) were chosen such that contact angles on the respective surface are close to the lower and higher limit of finite contact-angle measurements observable on smooth 58
4.1 The Measurement
drop base diameter [mm]
8 6 4 2 0 0
10 15 20 25 drop volume [µl]
Figure 4.7: Drop-base diameter as a function of drop volume. a.) The data points represent diameters actually measured on the sessile drop, the full lines show the diameter calculated with the truncated sphere assumption and the dashed line corresponds to the capillary length. b.) Examples of drop images, 1 and 30 µl drops on hydrophilic (top row) and hydrophobic (bottom row) substrates. The syringe outside diameter is 0.516 mm and can be used as a scale bar. surfaces. Contact-angle measurements are usually performed with micro-litre droplets. Therefore the Bond number may not be the correct criterion to choose a suitable size for sessile-drop measurements. In Figure 4.7b, drop images of 1 and 30 µl drops on the hydrophilic (top row) and the hydrophobic SAM (bottom row) are depicted. The outside diameter of the syringe is 0.516 mm, and serves as a scale bar. By eye, no gravitational influences can be detected for the hydrophilic case, but in the hydrophobic case the large 30 µl drop no longer resembles a spherical drop. The deformation is probably even more pronounced for a drop of 73 µl, as suggested by the capillary length criterion to be the maximum drop size without gravitational influence. To conclude, the influence of the drop volume on the static contact angle is very small. For drop volumes below 15 µl, the drop diameter does not strongly deviate from the calculation with the truncated sphere assumption. Nevertheless, it should be mentioned that Taylor et al.93 investigated the accuracy of contact-angle determination of microlitre-sized drops by evaluating the drop profile with the truncated sphere assumption (DSA 3: circle method) on a surface with medium wettability (θ ≈ 69°). They found that with the circle method, contact-angle underestimation increases with increasing drop volume. Thus, even though the radius does not strongly deviate when calculated with the sphere assumption, the contact angle cannot be fitted adequately by the circle method, if used for medium or low wettability. Roll-off angle α: The substrates for all drops used in the static measurement were tilted to determine the roll-off angle as a function of drop volume (Figure 4.8). The smallest 59
4 Parameters Influencing Contact Angle Measurements drops did not move at all, the weight being too small to overcome the pinning of the drop. All other drops moved and the roll-off angle was recorded. All parameters used in the Furmidge64 equation (Equation 4.5) were measured independently; drop width w as in Figure 4.7, dynamic contact angles (θr, θa) defined by the automated method as in Figure 4.11, and the drop weight calculated from the dispensed droplet volumes. The Furmidge equation fits the data reasonably well. By solving the Furmidge equation for the weight of the drop and defining the drop width as a function of the volume of a spherical drop, one can calculate which would be the smallest volume to still be able to move on the surface: for the hydrophobic surface it would be 2.0 µl and on the hydrophilic 2.7 µl.
0.8 0.6 0.4 0.2 0.0 0
10 15 20 25 drop volume [µl]
Figure 4.8: Roll-off angles versus droplet size. The equation of Furmidge64 (lines) correlates well with the measured roll-off angles. Dynamic measurements: Since the automated approach is the most commonly used method for measuring dynamic contact angles in our group, the technique was scrutinised a bit more closely. Initially the advancing contact angle was measured in two steps, since sometimes the drop did not grow axio-symmetrically to the syringe but tended to grow to one side. A pause between two movies (successive measurements) allows the syringe to be readjusted to the middle of the drop and thus leads to a better measurement while recording the second movie. Examining the contact angles derived from two such consecutive movies on the hydrophobic substrate did not reveal any influence of the break on the contact angle behaviour, but there was clearly an influence seen on the hydrophilic substrate. Therefore it was decided to repeat the measurement on the hydrophilic SAM, measuring drop growth in one continuous movie. In Figure 4.9a, the contact-angle data of the 1-movie and the 2-movie measurements at a dosing speed of 15 µl/min are plotted versus time. By looking at the contact angles of the 2-movie dataset, the break can easily be seen at 15 seconds. There is a possibility that there was an air bubble trapped unnoticed in the dosing system, which would change the dosing speed due to its compressibility, but also in this case one longer movie would be more 60
4.1 The Measurement favourable, because there would be a longer time to build up a steady flow, despite the trapped air bubble. In the end, the number of movies really did not have an influence on the averaged contact angle, but the standard deviation with one movie appears slightly lower. When measuring receding angles, the values of the hydrophilic substrate did not show any dependence on the decreasing volume. On the hydrophobic SAM, another contact angle behaviour was observed, see Figure 4.9b. With decreasing volume the data seems to show a downward tendency. Since it is right from the beginning and the drop was still pretty large, the syringe argument does probably not hold here. The scatter and the tendency do not seem to be velocity dependent therefore it is assumed that indeed the volume has an influence on the receding contact angle on this hydrophobic substrate.
1 movie 2 movies
5 µl/min 15 µl/min 30 µl/min
15 20 time [°]
8 10 drop volume [µl]
Figure 4.9: Datasets from the automated measurement. a.) comparison of the data measured at a speed of 15 µl/min on the hydrophilic surface. The 2-movie-data set was measured in two steps, which is clearly visible in the contact angles. This artefact does not occur with the long single measurement (1 movie). b.) With decreasing volume, the receding contact angle is not constant, but also increases on the hydrophobic surface. Both hydrophilic and hydrophobic surfaces were measured with the same series of dosing speeds. By extracting the drop-base diameter from the movies and plotting the radius versus time, the contact-line speed could be extracted from the slope of the linear regression (Figure 4.10). Contact line speeds from 0.002 to 0.025 mm/sec are considered as low-rate dynamic contact angle measurements.95 Such velocities lead to data extremely close to the y-axis of Figure 4.5. In other words, with these velocities the inherent, speed-independent hysteresis can be characterised. The upper limit (0.025 mm/sec) is indicated with a dashed line in Figure 4.10. When measuring advancing angles with speeds in this range on relatively smooth surfaces, they are thought to be “essentially identical to the static contact angles”94 and could therefore be used for surface-energy 61
4 Parameters Influencing Contact Angle Measurements
contact line speed [mm/sec]
calculations.95 More generally it means that contact angles measured with speeds within the range given above will not show influences of the dosing speed. Interestingly, the contact-line speed on a hydrophilic substrate is significantly higher for the same dosing speed than on a hydrophobic substrate, and crosses in the advancing case the limit of 0.025 mm/sec at a dosing speed of slightly larger than 15 µl/min. On both surfaces the advancing contact-line speed increases linearly with the dosing speed. The same is true for the receding contact line on the hydrophobic surface. The receding contact-line speed for the hydrophilic sample was harder to define, since it was only the very last portion of the drop volume that moved over the surface, and the influence of the syringe was readily detectable by a change in the decreasing rate of the radius r. Due to the very small residual volume that is actually moving, starting from a large contact diameter, the contact line speed is naturally a lot higher than on the hydrophobic surface. Additionally, contact angles as low as 5° (see Figure 4.11a) are very hard to extract reliably with the sessile-drop measurement. The placement of the horizontal line can also have a strong influence on the extraction of r. As a conclusion, 15 µl/min is clearly the upper limit for dynamic contact-angle measurements with the automated method. On strongly hydrophilic samples that have a θa of about 30 – 35°, even lower dosing speeds are advisable.
0.05 0.00 -0.05 -0.10 -0.15 CH3 adv CH3 rec OH adv OH rec
-0.20 -0.25 -0.30 0
10 15 20 25 dosing speed [µl/min]
Figure 4.10: Contact-line speed versus dosing speed (automated measurement). The contact-line speed increases linearly with the dosing speed, and much more steeply on the hydrophilic (OH) than on the hydrophobic (CH3 ) substrate. The receding contact line speeds on the hydrophilic sample are significantly higher than on the hydrophobic sample since movement only starts for the very last portion of residual volume. Comparison of three different techniques to measure dynamic contact angles: The three different techniques, automated, tilting and drawing (see Figure 4.4) are compared with each other in Figure 4.11. Full symbols correspond to advancing, open symbols to receding contact angles. The first row was measured on the hydrophobic SAM, the second 62
4.1 The Measurement on the hydrophilic SAM. The drawing method was only tested on the hydrophobic substrate. The data points (Figure 4.11) clearly show a strong dependence on stage speed, which is essentially equal to the contact-line speed. The contact-line speed is therefore about 100 to 1000 times higher than with the automated method (Figure 4.10). Since the receding contact angle deviates strongly from the values measured with the other methods, the stage speeds offered by the machine are clearly too fast to obtain representative results. The automated method leads to contact angles (θa and θr) that scatter around one specific angle on both substrates, quite independently of the dosing speeds investigated. The tilting method also leads to consistency of the contact angle as a function of the drop volume. However, on the hydrophilic sample in particular, the scatter in the θa values is larger than that obtained with the automated method. The automated method does not lead to advancing contact angles that are equal to static contact angles, even though many of the data points fulfil the requirement for slow dynamic measurements. The same is true for the tilting measurement on the hydrophobic surface. On the hydrophilic substrate however, there are indeed two measurements that are close to 20° and thus similar to the results obtained from the static measurement. However, since only 2 out of 6 measurements fit, no general conclusion can be drawn.
100 90 2 4 6 8 10 12 stage speed [mm/sec]
30 20 10
0 0 10 20 30 0 10 20 30 dosing speed [µl/min] drop volume [µl]
Figure 4.11: Dynamic contact angles measured with three different methods: automated, tilting and drawing. Full symbols denote θa, open symbols represent θr. top row: hydrophobic SAM; bottom row: hydrophilic SAM. Subtracting the value of θr from θa, the hysteresis can be calculated (Table 4.2). The 63
4 Parameters Influencing Contact Angle Measurements
Table 4.2: Hysteresis calculated by subtracting the average of the θa minus the average of the θr measured by the automated and tilting methods. hydrophobic SAM
10° ± 2°
27° ± 2°
10° ± 2°
17° ± 4°
contact angle hysteresis on the hydrophobic substrate is very similar for the two types of measurement. On the hydrophilic substrate however, a considerable difference is visible. One reason for the smaller hysteresis achieved by the tilting method might be found in the fact that only two spots of the drop’s footprint are fully in the advancing and receding mode. All other spots are a mixture of both modes. And since all spots are connected especially via the liquid-air interface, which acts like a membrane, the hysteresis might be slightly reduced. On the hydrophilic substrate, the larger adhesion and substrate contact area may cause this effect to become visible. A few closing remarks: With the tilting method it is harder to control contact-line speed. As Furmidge64 has described, the drop usually increases in speed once it has left its original position. This effect is more pronounced the longer the drop was at rest. Therefore, Furmidge reduced the tilt angle after the drop started moving to achieve a speed of about 2 mm/sec, which is quite fast. Precise tilt-angle reduction while the drop is moving would be a challenge on the DSA 100, and was therefore not done. If the drop was initially slightly pinned, e.g. as seen on superhydrophobic surfaces (data not shown), the drop moves very fast once it has overcome the pinning energy barrier. θa and θr determination then become difficult and require a high-speed camera. A few effects were not discussed in this chapter, but which are also believed to have an influence on the contact angle: - Some authors claim to see an effect of the contact-line tension on the contact angle,109 while others do not,93 and the literature is not even conclusive on the sign and the magnitude of the contact-line tension. According to de Gennes35 it is in the range of 10−11 J/m and too small to influence the macroscopic contact angle as determined here. -Usually, contact angles are not measured in vacuum, and therefore the substrate is most probably not completely dry but has a certain amount of the adsorbed molecules from the test liquid. If such a film is present, an additional film pressure π 38 (also called surface pressure92 ), has to be considered, which has the same units as the surface tension. The influence of π is more pronounced if liquid and solid attract each other. In the case of water it is more likely to play a role on the hydrophilic than on the hydrophobic SAM. But again, this value is important for solid surface tension (γSA ) determination, and not so much for comparative contact-angle measurements. Control of π (e.g. humi64
4.2 Surface Functionalisation with Trichlorosilanes dity) increases the reproducibility of contact-angle measurements.
4.1.4 Conclusions and Summary Water contact angles were measured on standard substrates coated with two different SAMs: Mercaptoundecanol on gold/silicon renders the surface very hydrophilic and dodecanethiol very hydrophobic. Thus, investigations were performed on two substrates with a large difference in surface energy. Static measurements: The influence of drop volume on contact angles is minimal for the investigated volumes (1 to 30 µl). Up to drop volumes of 15 µl, drop-base radii can be approximated by assuming the dispensed volume takes up the shape of a truncated sphere. The capillary-length criterion is of the right magnitude to distinguish between drops that are or are not influenced by gravity. However, it seems, especially on the hydrophobic substrate, that it greatly underestimates the influence of gravity on the drop profile. The Furmidge equation leads to a reasonable fit of roll-off angles. Dynamic measurements: For the automated method, the contact-line speed was evaluated as a function of the dosing speed. The contact-line speed at the same dosing speed is considerably higher on hydrophilic substrates than on hydrophobic surfaces. In order to stay in the range of slow dynamic measurements, a dosing speed of 15 µl/min is the upper limit for hydrophilic surfaces. It might even be advisable to reduce the speed to 10 µl/min when measuring on hydrophilic surfaces. Comparison of three different methods to characterise dynamic contact angels: For each method, different parameters were varied to investigate the parameter influence and compare the methods among each other. The drawing method soon proved not to be applicable with the DSA 100, due to the excessive speed of the stage. The automated and tilting methods lead to consistent contact-angle values, which were not strongly influenced by the investigated parameters. The tilting method seems to give slightly lower hysteresis on hydrophilic substrates. Generally, with both methods, advancing contact angles were not found to be equal to static contact angles.
4.2 Surface Functionalisation with Trichlorosilanes 4.2.1 Introduction Alkyl-trichlorosilanes are known to react with a number of oxide surfaces, such as silicon, glass and metallic oxides.111 Also polymers, such as Polydimethylsiloxane (PDMS), which form a superficial silicate layer upon plasma-oxidation, are known to be good substrates for silane-based SAMs.83, 90 Other polymers such as polyethylene have been 65
4 Parameters Influencing Contact Angle Measurements successfully functionalised by creating a thin silicate layer (adsorption and hydrolysis of SiCl4 ) before exposure to the silanes.83 Silanisation is not a trivial reaction. Under ambient conditions, a clean oxide surface, such as that of silica, is hydrated (silanol groups) and a thin layer of adsorbed water is present. It is believed that a trichlorosilane head group arriving on the surface from the gas phase becomes hydrolysed once the molecule gets close enough to the oxide surface, where it forms hydrogen bonds to the surface and neighbouring silane groups. By means of water elimination, this unstable situation then leads to the formation of a network in which each chain head is covalently linked to the surface and neighbouring silane groups.112 In this chapter, the static water contact angles of a few materials of relevance to this thesis (silicon wafer, glass, polydivinylsiloxane (PROVILnovo) and epoxy) were measured before and after exposure to a perfluorinated trichlorosilane.
4.2.2 Experimental Static water contact angle measurements: All measurements were performed on a simple Ramé-Hart contact-angle goniometer (Ramé Hart model 100, Ramé Hart Inc., USA). Prior to air-plasma treatment and after exposure to the perfluorinated silanes, static contact angles were measured with 6 µl drops of high-purity water (18.3 MΩ, EASY®pure, Barnstead, USA) on two different spots per sample. Substrates: A single-side-polished silicon wafer piece (Si-Mat Silicon Materials, Germany) was cleaned with toluene and ethanol (see Chapter 4.1.2) to remove glue residues arising from the cutting process. A microscope slide (Menzel GmbH & Co KG, Germany) was measured directly from the box. Two negatives (imprints) made from a commercial polydivinyl siloxane species (PROVILnovo, Light C.D. 2, fast set; Heraeus Kulzer GmbH, Germany) were prepared from a freshly cleaved mica surface. One of the negatives was used for contact-angle measurements; the other was used as a mould to cast a replica from an epoxy blend (EPO-TEK 302-3, Epoxy Technology, USA). The replication technique is more closely described in Chapter 5. Surface functionalisation: All substrates were exposed to air plasma (RF Level high, 0.1 torr, PDC—32G, Harrick Plasma, USA) for 2 min and then put in a desiccator. The humidity of the air formed the necessary silanol groups for the reaction. A small glass vial containing 1H, 1H, 2H, 2HPerfluorooctyltrichlorosilane (ABCR GmbH, Germany) was placed in the desiccator. A 66
4.2 Surface Functionalisation with Trichlorosilanes gentle vacuum was applied and the SAM was allowed to form overnight. Contact angles were measured immediately after removal from the desiccator.
4.2.3 Results and Discussion The results of the static water-contact-angle measurements are summarised in Table 3. Since all surfaces are fully wettable after plasma treatment, “Before” denotes measurements on the native surfaces as received, while “After” refers to the samples after functionalisation. Only the two surfaces with a silicon oxide layer (glass and silicon wafer) show a distinct change in surface energy. Their contact angle increases to ∼110°. This value is commonly found on CF2 -rich surfaces;113 therefore it can be assumed that the monolayers formed by this experimental procedure did not lead to crystalline structures, exhibiting mostly their CF3 -end group, which would correspond to a contact angle of 120°.113 This conclusion is supported by the fact that alkylsilanes of short alkyl chainlengths (i.e. C10 or lower) are known to be in a liquid-like state on the surface.83 The main component of the commercial material PROVILnovo is a polydivinylsiloxane, a silicone species closely related to PDMS (whose contact angle data on the native and the perfluorinated surface is displayed in Chapter 4.3). Therefore, it was interesting to see whether it could be modified in a similar way to PDMS. But even by its colour, opaque green instead of transparent, it is clear that it is a complex mixture of many components. It is commonly used by dentists to get an imprint of the dentition. Due to its low viscosity, short curing times (minutes) and ability to be removed from any surface, it is a useful tool for replication of any surface of interest. When depositing a drop of water on the PROVILnovo substrate, the contact angle changed from hydrophobic to ∼60° within seconds, most probably revealing the reason why it does not stick to any surface: it contains a high concentration of surfactants. The surfactants can also be made evident by immersing a PROVILnovo substrate in water and shaking it: A foam is generated. Plasma treatment and exposure to the perfluorinated silanes did not change this behaviour upon contact with water. The case is slightly different with the epoxy sample. The epoxy surface does show a stable static contact angle at ∼73°. After “functionalisation” the contact angle starts, very similar to the PROVILnovo, at a hydrophobic value and then it decreases to slightly below its initial contact angle. This could be interpreted that, indeed, a layer of perfluorinated silanes was physisorbed on the epoxy, but reacted immediately upon contact with water. Therefore, it can be concluded, with the chosen experimental setup, epoxy surfaces cannot be functionalised with trichlorosilanes via the vapour phase. This finding increases the understanding of the situation when photolithographic structures made of SU-8 (epoxy polymer) are used as a template for PDMS replication. These templates consist of SU-8 patterns on a silicon wafer. If such a template were to be used for 67
4 Parameters Influencing Contact Angle Measurements
Table 4.3: Static water contact angle measurements on different substrates before and after fluorosilanisation. Substrate / water θs
64° ± 2°
111° ± 1°
35° ± 3°
107° ± 3°
61° ± 6° *
57° ± 5 *
73° ± 3°
68° ± 2° *
* all these contact angles started initially at a contact angle > 90°, but decreased within approximately 60 seconds to the values stated here . PDMS replication without further treatment, the PDMS would react with the substrate and could not be removed anymore. Therefore it is common to functionalise the template with perfluorinated silanes. However, as is seen here, these only stick covalently to the bottom of the template, where the silicon surface is exposed. The fact that PDMS removal still works would lead to the conclusion that the silicone either does not stick to the epoxy or the silane layer also works as a mould release agent when it is only weakly adsorbed.
4.2.4 Summary In this short study, four different smooth surfaces were activated by air plasma and then exposed to a vapour of perfluorinated silanes. It was found that with this treatment silicon wafers and glass slides can be hydrophobised. The contact angle is equal to a contact angle expected for a disordered, CF2 -rich surface. The commercial PROVILnovo material reacts with water and cannot be modified by trichlorosilanes. Contact-angle measurements on epoxy revealed that the silanes only adsorb weakly on the substrate.
4.3 PDMS (Polydimethylsiloxane) 4.3.1 Introduction Cured PDMS is a polymeric elastomer consisting of an inorganic backbone with alternating silicon and oxygen atoms. To each silicon atom two methyl groups (CH3 -groups) are bound (see Figure 4.12). Ambiguity of PDMS: This structure imparts a certain ambiguity to PDMS; the silicone backbone is rather hydrophilic whereas the methyl-side chains are hydrophobic. When exposed to air, the 68
4.3 PDMS (Polydimethylsiloxane)
Figure 4.12: Simplified chemical structure of cured PDMS.
methyl groups of the native PDMS form a close-packed layer to shield the high-energy silicone backbone. Upon contact with water, the chains reorient such that the hydrophilic silicone chain is facing the polar liquid.90 Thus, the high flexibility of the loosely cross-linked backbone allows the polymer to adopt the conformation leading to the lowest energy state. If the PDMS faces air, which can be considered as a very hydrophobic environment, the CH3 -groups will face the atmosphere. If the PDMS is in contact with the highly polar water then the silicone backbone will favourably be at the interface. The methyl groups bend away from the surface normal.114 This process is fast and occurs already during the course of a dynamic contact angle measurement,90 leading to a large hysteresis. However, this is not the only ambiguity occurring with PDMS. One of the preconditions for common contact-angle measurements is that the solid has to be rigid. The rigidity of the surface has to balance the force in y-direction of the liquid-air surface tension, γLA−Y . Therefore if a drop is placed on another immiscible liquid, it assumes the shape of a floating lens,35 instead of a drop with a flat base. If placed on a soft solid such as a hydrogel or an elastomer with a small E-modulus E, γLA−Y can pull up a ridge around the edge of a drop (see Figure 4.13). The height h of this ridge is given by the formula developed by Shanahan and de Gennes:115–117
3 · γLA sin θ E
θ denotes the apparent contact angle. By measuring dynamic contact angles (tilting method), Extrand115 investigated the maximum E-modulus of elastomers for which such a ridge is formed. To this end he prepared a series of natural and polybutadiene rubbers with different cross-linking densities to modify the E-modulus. He found that the limit for contact-angle-influencing ridge formation lies at an E-modulus of about 5 MPa. The ridge is a local increase of roughness and therefore it increases the advancing and decreases the receding contact angle, thus increasing hysteresis. The PDMS used here has an E-modulus of 1.448 MPa,118 i.e. well below 5 MPa. The ridge height h would be 146 nm. It is generally expected that feature sizes above 100 nm have an influence on the contact angle.92 69
4 Parameters Influencing Contact Angle Measurements
γLA γLA-Y θ h
Figure 4.13: Sketch of a drop on a soft elastomer. The y-direction of the liquid-air surface tension γLA−Y pulls up a ridge of height h around the edge of the drop. Oxygen Plasma Treatment of PDMS: By exposing PDMS to an air or oxygen plasma, a superficial silica-like phase is formed.119 This brittle, glassy layer increases in thickness with plasma exposure time until it reaches its maximum thickness, which is around 100-150 nm. The exposed surface exhibits a high surface energy, leading to full spreading of water. Thus, upon contact with the humidity of the air, the high-energy surface of the plasma-treated PDMS will immediately form a wet silanol layer, as is necessary for the reaction with silane molecules.112 PDMS is known to recover its hydrophobicity over time by conformational changes, condensation of silanol groups90 and the diffusion of low molar mass PDMS species to the surface.119 The low-molar-mass species are residues from the polymerisation process and are formed during plasma treatment due to chain scission. This recovery is very slow as long as the silica-like layer remains closed. Upon introduction of cracks, recovery is much faster. Even though the recovery restores a low-energy surface, it is not equivalent to a native PDMS surface. The brittle, silica-like layer is still present, only now covered by lowmolecular-weight species. The chains have lost their mobility, and thus contact-angle hysteresis is smaller on a recovered PDMS surface.90 Extraction: The silicone elastomer kit that was mostly used in this work (Sylgard 184) consists of a base and a curing agent. Both components consist of numerous silicone species and solvents. A full description of the single compounds can be found in the thesis of Raphael Heeb.120 In this commercial elastomer kit, reinforcing fillers and uncrosslinked silicone species with high mobility are to be found. These are known to diffuse from the bulk to the surface. By using a strongly swelling solvent121 these species can be largely removed from the polymer. 70
4.3 PDMS (Polydimethylsiloxane)
4.3.2 Experimental Substrates: The two components of the Sylgard 184 elastomer kit (PDMS, SYLGARD® 184 silicone elastomer, base and curing agent, Dow Corning, USA) were mixed in an oil-to-hardener ratio of 10 : 1. The air bubbles introduced into the mixture due to vigorous stirring with a glass rod were removed by evacuation in a desiccator. Then the mixture was cast into two different “moulds”: a single-use Polystyrene Petri dish (TC Dish 92x17, 150350, NuncTM, Thermo Fisher Scientific, Denmark) and a perfluorinated (see below) singleside polished silicon wafer (Si-Mat Silicon Materials, Germany) of 2 by 2 cm2 . Curing took place at 70 °C overnight., The cured PDMS was cut into pieces of 1 x 3 cm2 with a razor blade. Most samples were measured on the surface facing the petri-dish or the perfluorinated silicon wafer during curing. For one set of measurements the side facing the air was also used. A set of samples was functionalised with a monolayer of perfluorinated silanes. Perfluorination: The silicon wafer was cleaned with toluene and ethanol before usage to remove glue residues from the cutting step. The PDMS samples were used directly after cutting, without further treatment. The conditions for perfluorination were the same for all substrates. They were first treated with an air plasma (RF Level high, 0.1 torr, PDC32G, Harrick Plasma, USA) for 2 min and then loaded into a desiccator. A small glass vial with a small amount of 1H, 1H, 2H, 2H-Perfluorooctyltrichlorosilane (ABCR GmbH, Germany) was also placed into the desiccator. Then, a low vacuum was applied. SAM formation via the vapour phase was allowed to take place for 1 h. PDMS extraction: Low-molecular weight species were extracted from four PDMS samples. The samples were placed in a piranha-cleaned glass vessel and covered by a large amount of hexane. Hexane strongly swells the samples and thus extracts non-bound species. The hexane was exchanged after 3.5 and 6 h. After 24 h the samples were rinsed with hexane and loaded into a desiccator. With a gentle vacuum the absorbed hexane was removed from the PDMS samples. Contact-angle measurements or fluorination were performed after another 24 h, to allow residual traces of hexane to evaporate under ambient conditions. Cleaning methods: Intensive rinsing with ultra-pure water (> 18.2 MΩ, TKA-GenPure, TKA, Germany) and high-purity ethanol (analytical grade, Scharlau, Spain) was undertaken, holding the samples with tweezers such that no liquid could flow back from the tweezers on the sample and possibly contaminate it. The samples were measured immediately after drying under a stream of nitrogen. A few samples were ultra-sonicated (Sonorex TK30, 71
4 Parameters Influencing Contact Angle Measurements Bandelin, Germany) in ethanol for 10 min in piranha-cleaned glassware. After removal, the samples were rinsed with ethanol and dried with nitrogen. An overview of the cleaning methods and treatments to the samples is given in Table 4.5. Contact-angle measurements: All measurements were performed on the DSA 100 and evaluated with the DSA3 software (Krüss GmbH, Germany). On each substrate, three 9 µl drops were used to measure static and roll-off angles. The stated static contact angles and roll-off angles are an average over three drops. Dynamic measurements were conducted with a dosing speed of 15 µl/min. For the advancing contact angle, two movies were recorded, for the receding only one. For the advancing measurement, the initial volume was 5 µl and was increased twice by 4 µl to a 13 µl drop. After recording the receding movement, 15 µl were deposited from the syringe to ensure a high water quality for the next measurement. Only contact angles where the contact line was actually moving were considered. Roll-off angle measurements were performed at the slowest tilting speed, which is 1.2 °/sec.
4.3.3 Results and Discussion The aim of this chapter was to learn what influences water-contact-angle measurements on PDMS. In Figure 4.14a, dynamic and static contact angles were measured seven times on the very same spot. Even though it is always better to measure on a fresh spot, the water does not seem to alter the surface such that it would lead to a significant change in contact angle. For Figure 4.14b, contact angles were measured on different spots over time after removal from the Petri dish. No change in contact angle can be observed over the time span of two days. The PDMS was also allowed to cure against different mould (negative) materials. Their dynamic contact angles are summarised in Table 4.4.
Table 4.4: Dynamic water contact angles evaluated on the three different mould materials. Material
76° ± 1°
13 ± 2°
114° ± 1°
98° ± 1°
Figure 4.15 shows the dynamic measurements from the native PDMS as a function of the mould material. The advancing contact angles are very similar for the three moulds, no difference can be seen here. The case looks very different when comparing 72
4.3 PDMS (Polydimethylsiloxane) b.) 120
100 static advancing receding
90 80 70
contact angle [°]
contact angle [°]
100 static advancing receding
90 80 70
10 10 time [log sec]
Figure 4.14: Repeatability of contact angle measurements on native PDMS. a.) 7 subsequent measurements on the same spot; b.) influence of time after exposing the sample to air; measured on the same sample but on different spots. the receding contact angles. The PDMS sample which was cured against the rather hydrophilic Petri dish shows a receding contact angle which is 23° lower than the sample that was cured against air. It is likely that the surface energy of the phase facing the PDMS during curing has an influence on the superficial microstructure of the PDMS. Low energy surfaces such as air or the perfluorinated silicon wafer will cause the CH3 groups to orient along the surface normal, shielding the hydrophilic silicone backbone; on a high-energy surface the opposite occurs. Even though reorientation of the chains upon contact with water is fast, the initial order of the polymer chains might play a role for contact-angle measurements. Nevertheless, the results are not unambiguous, since it is expected that the roughness of the moulds decreases from the Petri dish to the silicon wafer and then to air. Reduction in roughness invariably also leads to a reduction in hysteresis. During the course of several measurements it was also recognised that receding contact angles of samples cured against a Petri dish are not always identical. Usually they are around 73 – 75°, but also values as low as 65° were measured. Considering the influence of the E-modulus on ridge formation it is very well possible that slight variations in the E-modulus strongly influence the receding contact angle (as it can happen e.g. by using an older, thus less effective hardener, leading to a lower cross-linking density and Emodulus). In the following section it was tried to define how different treatments influence the contact angle. If this study would have been performed only by measuring static contact angles, no conclusions could have been drawn except that none of the treatments has any influence on the PDMS. All static contact angles θs scatter around ∼105° ± 2°, and 73
4 Parameters Influencing Contact Angle Measurements
110 100 90 80 70 Petri dish
Figure 4.15: Dynamic contact angle measurements as a function of the mould (negative) material during curing of the PDMS. Only the receding contact angle shows an influence of the mould on the contact angle. The finding is somewhat ambiguous, because both surface energy and mould roughness decrease from the Petri dish to air. no influence of the extraction or the perfluorination could be detected. This experiment is therefore a very good example to emphasise the fact (see Chapter 4.1) that static contact angles only carry a limited amount of information. In Table 4.5 all treatments carried out and their corresponding advancing and receding contact angles are presented. For the native PDMS (left column, without perfluorinated layer), the advancing contact angle scatters reliably around 120° and is therefore independent of the treatment. Solely the extracted PDMS shows a slightly higher θa, which could be interpreted in terms of slightly higher surface roughness after extraction. This assumption is supported by the receding contact angle, which is about 10° lower than on the unextracted PDMS samples. Surface roughness leads to an increase of the advancing and a decrease of the receding contact angle due to pinning events (see more in Chapter 6). Rinsing with water or ethanol does not change the wetting properties of PDMS. Perfluorinated PDMS samples usually show a lower θa and a higher θr compared to the native PDMS. At first, the lower θa for the perfluorinated species may appear surprising, since it is generally assumed that perfluorinated surfaces show higher contact angles than surfaces consisting only of hydrocarbons.113 However, this observation confirms the fact that the native and perfluorinated surfaces are very different in nature. On the native surface, the surface tension of water is able to pull up a rim at the edge of the drop, thus increasing the advancing contact angle. This effect is a lot more difficult to achieve on the plasma-treated and perfluorinated PDMS. The superficial, brittle silica-like layer is a lot stiffer (higher E-modulus) than the bulk PDMS, thus the surface tension of the water is not strong enough to pull up the ridge. Additionally, the chains of the native PDMS are free to conform to the phase with which they are in contact. This 74
4.3 PDMS (Polydimethylsiloxane)
Table 4.5: Dynamic contact angle data evaluated on PDMS and perfluorinated PDMS substrates after different treatments. PDMS
(1) no treatment
119.9 ± 0.5
75.7 ± 2.6
115.1 ± 1.0
98.2 ± 2.6
(2) rinsed with water
119.6 ± 0.6
115.3 ± 0.8
95.1 ± 2.0
(3) rinsed with ethanol
120.0 ± 0.8
74.8 ± 1.8
115.8 ± 0.5
102.8 ± 1
(4) 10 min US ethanol
117.2 ± 0.3
93.5 ± 0.8
120.9 ± 0.7
66.5 ± 1.5
(120.9 ± 0.7)
(81.5 ± 1.3)
115.0 ± 0.4
92.0 ± 1.4
(6) extracted, 10 min US ethanol
freedom is lost for the perfluorinated PDMS substrates, resulting in higher receding contact angles. Rinsing seems to slightly influence the receding contact angle. Water reduces, while ethanol increases the θr. The increase in contact angle upon ethanol rinsing is somewhat contradictory to the decrease in θr for the sample that was ultrasonicated in ethanol (no. 4). One could speculate that it may indicate that by rinsing with ethanol, physically adsorbed silane residues are washed away, thus leading to a smoother surface. However, upon ultrasonication roughness might be introduced by removal of parts of the brittle silica-like layer. Contact angles on the ultrasonicated perfluorinated substrate (no. 4) are similar to those on the sample that was extracted and then ultrasonicated in ethanol (no. 6), which could indicate a similar surface roughness. The data for the extracted and perfluorinated sample (no. 5) are in brackets because due to an experimental error this sample could only be measured four days after preparation. The high hysteresis (compared to the other perfluorinated samples) may indicate that a hydrophobic recovery of low molecular species through the cracks in the silica-like layer has already occurred. Thus, the contact angles measured resemble those measured on the native PDMS and can therefore not be reliably used for interpretation in this study. The numbers were still added into the table to show that pefluorinated PDMS is not as inert over time as might be expected from other perfluorinated surfaces. Also roll-off angles were measured on all PDMS samples. No roll-off angle could be detected on the native PDMS. Reasons for this may be once more the ridge which is pulled up by the surfaces tension of water, leading to high pinning and the ability of the 75
4 Parameters Influencing Contact Angle Measurements PDMS chains to conform to the polar nature of the water, thus increasing the adhesion of the drop to the surface. All perfluorinated samples showed a roll-off angle (see Figure 4.16a). Samples that were not extracted or ultrasonicated showed significantly lower roll-off angles. Rinsing seems to increase the scatter (standard deviation of samples 1 to 3) in the measured data. Plotting the roll-off angle versus the hysteresis shows that generally, a higher hysteresis also causes a higher roll-off angle, as expected from the Furmidge equation (see Equation 4.5 and Figure 4.16b).
b.) 80 roll-off angle [°]
roll-off angle [°]
80 60 40
2 3 4 (5) treatment number
4 6 3
20 30 hysteresis [°]
Figure 4.16: a.) Roll-off angles measured on the perfluorinated PDMS samples. The numbers on the X-axis denote the treatment applied to them and correspond to the numbers given in Table 5. b.) plotting roll-off angles versus hysteresis shows that generally the roll-off angle is higher with increasing hysteresis. The dashed line represents the linear regression through the data and is meant as a guide to the eye.
4.3.4 Summary and Conclusions PDMS is a very interesting, but ambiguous material for contact-angle measurements. The advancing contact angle is very constant between 119° and 121° and nearly unaffected by any treatment. If no parameter is changed, the receding contact angle is also very constant. When comparing different conditions, however, the receding contact angle shows a high variability. There is an indication that the receding contact angle is influenced by the mould material against which the PDMS was cured. The more hydrophobic the mould material, the higher the receding contact angle of the PDMS. However, in this study, the effect of roughness was not determined and might have an effect leading in the same direction as surface energy. 76
4.4 Fit Routines The high mobility of the hydrophilic silicone backbone and the hydrophobic methyl groups allow the material to adopt the lowest energy state by changing conformation. The low E-modulus of PDMS was identified as another cause to increase the hysteresis on PDMS. PDMS is not rigid enough to compensate the force of the water-air-surface tension perpendicular to the surface. Therefore, a ridge is pulled up around the edge of the water drop, locally increasing the surface roughness and thus hysteresis. Native PDMS can be rinsed with water or ethanol without changing the contact angle. Extraction of low molecular species reduces the receding contact angle slightly, possibly due to an increase in surface roughness. Plasma-treatment and subsequent perfluorination by adsorption of perfluorinated silanes significantly changes the properties of the surface. The formation of a silica-like layer during plasma treatment increases the stiffness of the topmost layer, thus inhibiting the formation of a ridge around the drop. This reduction in roughness causes the advancing contact angle to be slightly smaller after functionalisation than that of the native PDMS. Additionally the silica formation destroys the flexibility of the chains. Thus, reorganisation upon contact with the liquid is no longer possible. The functionalisation with a perfluorinated monolayer reduces the adhesion of water. These reasons lead to a greatly reduced hysteresis when compared to native PDMS. The effect of rinsing with water and ethanol is not very clear. It seems to slightly change the receding contact angle. Ultrasonication and extraction of low-molecular-weight species also seem to reduce the receding contact angle, thus increasing the hysteresis.
4.4 Fit Routines 4.4.1 Introduction While fitting the recorded contact angles in the DSA3 program from Krüss it was realised that the fit was not always optimal. Deviation was mainly observed for contact angles above 135°. Therefore an evaluation was performed, as to which of the analysis tools provided with the software would fit a superhydrophobic drop best. The circlefitting method fits the drop contour to a circle segment function. The tangent-method-1 attempts to fit the drop profile by a general conic-section equation, whereas the tangentmethod-2 fits the slope in the vicinity of the three-phase-contact line with a function of the type (y = a + bx + cx0.5 + d/ ln x + e/x2 ) and was generally used in this work. The YoungLaplace fitting procedure is the most complicated fitting model and takes interfacial and gravitational effects into account. Additionally, two methods for the program ImageJ were employed. One was a semiautomated method, the drop-snake analysis122 (snake tool) and the other the simple 77
4 Parameters Influencing Contact Angle Measurements “angle tool” provided in ImageJ. Criticism of contact-angle data stated in literature including many significant figures was already pointed out by J. Zimmermann123 and emphasised in the review of Dorrer and Rühe.48 J. Zimmermann evaluated the reproducibility of data extracted by the Young-Laplace method by small changes in the optical settings during the acquisition of the images. He found that lighting and contrast, setting of the horizontal line and focus had a great influence on the extracted contact angle value. The very same drop varied then between 166° and 176°. His conclusion was that the automated evaluation is good and reliable for contact angles between 20° and 140°, but has deficiencies for very high contact angles.
4.4.2 Experimental Images (see Figure 4.17) of four 9 µl drops were recorded on a perfluorinated PDMS substrate with an f1 (on a porous substrate, areal fraction in contact with the drop, see Chapter 2.6) of 5.6 %, thus clearly showing very high contact angles (around 160°). Four scientists were asked to evaluate the images with the ImageJ tools; drop snake analysis and angle tool. For the evaluation with the DSA3 software (Krüss GmbH, Germany), three different sensitivities in the optical settings were used for the analysis (gradient Threshold value 20, 40, and 60). The circle fit, tangent-1 and tangent-2 method and the Young-Laplace method were tested.
Figure 4.17: 4 images on a superhydrophobic perfluorinated PDMS substrate as used in Chapter 7 at an f1 value of 5.6 %. 78
4.4 Fit Routines
4.4.3 Results and Discussion The extracted contact angles differ a lot depending on the method that was used. The circle fitting was developed to measure small contact angles and clearly fails to measure such high contact angles. The two tangent methods give similar results and the YoungLaplace method extracts very high contact angles (see Figure 4.18a). The values extracted from the four scientists with the snake and the angle tool were clearly higher than those obtained from the tangent methods, but lower than that from the Young-Laplace fit. The difference between the use of the snake tool and the simple angle tool are negligible (Figure 4.18b), therefore for the time-intensive evaluations of the dynamic data (many drop pictures) it was decided to use the simpler and faster tool, the angle tool. Thus, if interest lies in the apparent contact angle that can actually be seen, drops with a contact angle above 135° should be evaluated by the use of ImageJ, because automated fits either under- (tangent methods) or overestimate (Young-Laplace) the apparent contact angle.
155 θs [°]
135 40/20 60/20 20/20 gradient threshold values tangent-1 circle tangent-2 Laplace
snake angle ImageJ tools scientist 3 scientist 1 scientist 4 scientist 2
Figure 4.18: Comparison of different evaluation techniques to extract the contact angle out of a drop picture. a.) automated methods delivered with the program DSA3; b.) semi-automated (snake tool) and manual tool (angle tool) provided with the program ImageJ.
4.4.4 Conclusion and Summary On superhydrophobic surfaces, reliable contact-angle measurements are difficult. Not only do the optical settings of a contact-angle measurement strongly influence the experimental values of the contact angles, but also the choice of method. The tangent 79
4 Parameters Influencing Contact Angle Measurements methods in DSA3 tend to underestimate, and the Young-Laplace method overestimates the apparent contact angle. The simple angle tool provided with ImageJ can be used instead to measure contact angles above 135°.
5 Replica Technique The replica technique has been used in many fields, such as cell studies,124, 125 wetting22, 126 and SEM studies, micro-contact printing,127 microfluidics86 and dentistry as an attractive low-cost technology to multiply or copy an expensive or labour-intensive topography In addition, it can reduce the variability of topography as an experimental parameter by always copying from the same master, to transform a brittle or sensitive substrate into a rigid and treatable substrate22 or to prepare a master in a material best suited for machining / preparation and then transform the topography into a material of interest. Depending on the field in which the replica technique is used, it is also called replica molding127–129 or templating.8 For Xia and Whitesides127 it is a component of “soft lithography”. The aim of this chapter is to summarise the efforts that have been performed to characterise the replica technique as used in this thesis and in LSST. Chapter 5.1 introduces the replica technique and shows, by the example of the lotus leaf, what kind of artefacts can be introduced when replicating a freshly cut leaf. Chapter 5.2 shows the limits of the Provil Novo mould material when replicating holey structures. In Chapter 5.3, artefacts that can occur when imaging epoxy replicas in SEM are shown and a solution is presented as to how they can be minimised. Chapter (5.4) concerns a toolbox that has been developed to characterise the precision of the replica technique. The idea was to find masters and methods which can be used to define shrinkage and edge smoothing and thus to enable the easy evaluation of new materials and compare them to those materials currently in use in a quantitative manner. Different epoxy and silicone species were tested as replicating materials. Additionally, the replica technique was taken one step further to achieve replicas made of ceramic instead of polymers (Chapter 5.5). It is shown that by employing this replica technique, photolithographic structures can be transferred very precisely into a ceramic substrate.
5.1 Lotus Leaf Replica (LLR) The replica technique with the EPO-TEK 302-3M / Provil novo light combination for replica / mould, as has been mostly used in LSST, was first employed by Wieland et al.125 He showed that this material is capable of faithfully replicating complex two81
5 Replica Technique scale structures such as an acid-etched, large-grit-sandblasted (SLA) titanium surface (replica surface of SLA shown in Chapter 6). It was therefore a logical next step to try the technique on the equally complex, two-scale structure of a living surface, such as the lotus leaf—the prime example of a self-cleaning surface.
5.1.1 Replication In Figure 5.2, the technique is exemplified by the replication of the lotus leaf (lat. Nelumbo Nucifera, Figure 5.1), as was used in Chapter 6. Lotus leaves are large, with diameters between 25 to 40 cm. Leaf veins emerge radially from the centre of the leaf, providing up to 25-mm-wide areas between neighbouring leaf veins that are reasonably flat and thus suitable for replication and contact-angle measurements.
Figure 5.1: Lotus leaf in the pond of the old botanical garden in Zurich. The body of the water strider on the leaf is about 10 mm long. Generally, lotus leaves were obtained from the botanical garden in Zurich (Figure 5.2a). A frame of size 42 by 15 mm2 was placed between the leaf veins and a low-viscosity and fast-curing silicone (PROVILnovo, Light C.D. 2, fast set; Heraeus Kulzer GmbH, Germany) was cast into the frame via a mixing gun (Figure 5.2b). The silicone form obtained in this way was then inserted into a PTFE frame, to provide side walls for epoxy casting (EPO-TEK 302-3, Epoxy Technology, Billerica MA). The epoxy was allowed to cure overnight at 60 °C (Figure 2c). After removal of the mould it was post-cured for 1 h at 150 °C, sawn into 5 x 15 mm2 pieces and cleaned with a 2 vol% aqueous solution of Hellmanex (Hellma, Germany). 82
5.1 Lotus Leaf Replica (LLR) In order to compare the initial surface of the lotus leaf with the replica used for the experiments, a SEM study was performed. The aim was threefold: firstly, to verify how well the initial surface was replicated; secondly, to check whether the replication process caused a change on the site of replication on the lotus leaf itself and thirdly, to investigate the influence of the metal layer on the topography of the replica. Therefore, the lotus leaf that was used as master was dried in ambient conditions. A small piece from the area of replication and a small piece near the area of replication were coated with a 5-nm-thick layer of platinum. One of the pieces of the replica taken at this spot was also coated with a 5-nm layer of platinum, another being coated with 6 nm of chromium and 50 nm of gold, as was always done for the thiol experiments in Chapter 6.
Figure 5.2: Replica technique used on the lotus leaf. a) Lotus leaf from the botanical garden b) application of fast-curing silicone with a mixing gun c) casting epoxy blend with subsequent curing and post-curing d) Epoxy replica coated with a metal layer.
5.1.2 Results and Discussion To investigate the replication fidelity between the original lotus leaf and the replica, SEM images were taken from both specimens (Figure 5.3). In Figure 5.3, the three images on the left were taken from samples of the dried lotus leaf (a, b, e) and the three images on the right were taken from pieces of the replica (c, d, f). Images in the same row all have the same magnification. The lotus-leaf replica (LLR, images on right) was made on the same site of the lotus leaf as shown in Figure 5.3b. Nevertheless, the images in Figure 5.3b, c and d do not correspond to exactly the same spot (no direct correlation of features). The lotus-leaf surface consists of a two-scale structure, which consists of papillae, forming the microstructure (Figure 5.3e), covered by a layer of wax tubes, comprising the nanostructure (Figure 5.3a).2 The nanoscale wax structure consists of wax tubes with a diameter of 60 – 80 nm and a length up to 1 µm. The microscale papillae have a diameter of 7 - 11 µm and are evenly distributed, with a nearest-neighbour distance of 15 - 28 µm. An area from which a replica was made is shown in Figure 5.3b. The wax tubes, having a low mechanical stability,5, 126 were compressed by the silicone during the mould-forming process. This compression results in a nanostructure on the LLR that displays significant structural differences from the original leaf (compare Figure 5.3a 83
5 Replica Technique anf c). For the thiol experiments in Chapter 6 the LLR had to be coated with chromium and gold. This additional metal layer of 60 to 65 nm slightly increased the flattening of the nanostructure (compare Figure 5.3c and d). Thanks to the high mechanical stability of the papillae, the microstructure of the lotus leaf was well reproduced (compare Figure 5.3e and f).
Figure 5.3: Different steps of the replica process. The scale bar in d.) is valid for the whole upper row, the scale bar in f.) for the bottom row. Nanostructure: a) dry lotus leaf, b) area on dry lotus leaf, on the spot where the negative was made, c) LLR with a Pt-coating of 6 nm for SEM, d) LLR with 15 nm Cr and 50 nm Au for thiol experiment. Microstructure: e) dry lotus leaf, f) lotus-leaf replica with 6 nm Pt. If the LLR is coated with gold and hydrophobised with dodecylthiol as described in the experimental part of Chapter 6, contact-angle data reveal that the exact replication of the lotus leaf’s nanostructure is not necessary, in order to achieve superhydrophobic and self-cleaning behaviour (see Table 5.1). Differences in contact angles between the lotus leaf and the LLR cannot be ascribed to different surface morphologies since all values are within the error bars. Contact-angle measurements on these naturally grown and wavy surfaces pose a special challenge. Since drops roll off at the slightest incline, which often is already provided by the natural waviness of the leaf, the values stated here most probably represent the lower end of possible contact angles on the lotus leaf and the LLR surface. 84
5.2 Limits of Provil Novo When Replicating Holey Structures
Table 5.1: Static water contact angle and roll-off angle on the lotus leaf and its hydrophobised replica. hydrophobic wax / dodecylthiol
157° ± 2°
3° ± 2°
160° ± 3°
3° ± 1°
flat (gold coated Si-wafer)
109° ± 1°
30° ± 1°
The replication of natural leaves exhibiting a soft wax structure is not as precise a technology as the replication of the rigid SLA surface, since the wax nanostructures are compressed during replication. Nevertheless, as long as the leaf also has a stable microstructure in the form of papillae, the replica can still show a self-cleaning effect following hydrophobisation. Thus it can be concluded that the nanostructure is not needed per se for the self-cleaning effect and that the difference in morphology is not revealed by simple contact-angle measurements.
5.2 Limits of Provil Novo When Replicating Holey Structures The replica technique, as described in 5.1.1, was also used to replicate the photolithographic structures described in Chapter 7. For these structures, the master has single holes that increase in density over the length of the substrate and slowly merge to larger holes. A single hole has a top square-side length of about 23 µm and a depth of 30 µm. If the mould material is Provil Novo, then the replica appears as shown in Figure 5.4. The air filling the holes has no path to escape when the Provil Novo is cast over the master. Thus, the rapidly curing silicone also replicates these air bubbles. This effect can be largely reduced by using Sylgard 184 PDMS (Dow Corning, USA) as a mould material. It has to be cast under low vacuum in a desiccator. The vacuum, combined with the lower viscosity, which is further reduced at the curing temperature (70 °C), allows the air bubbles to be removed from the holes, thus leading to high-quality moulds, as shown in Chapter 7. 85
5 Replica Technique
Figure 5.4: SEM images on four different positions of an epoxy replica from the photolithographic gradient described in Chapter 7 (holey side) made of EPO-TEK 302-3M from a Provil Novo mould. Dark areas are holes. When the holes are well separated from each other, the air cannot escape during mould formation. Thus, air bubbles (arrows) are replicated.
5.3 Vacuum Stability While analysing epoxy replicas in SEM, it was observed that wrinkles started to appear after some time. The effect is hard to see on complex surfaces, such as LLR or SLA replicas, but on the golf-tee-shaped micropillars (GTM, see Figure 5.5) the effect is quite visible, since they have quite large, flat and smooth areas. The preparation of the GTM replicas (EPO-TEK 302-3M / Provil Novo) shown in Figure 5.5 is described in Chapter 6. They have a top diameter of about 23 µm and a height of 30 µm. The replica of Figure 5.5a was simply cured at 60 °C without any subsequent post-curing. After 30 min exposure to the high vacuum of the SEM chamber (∼10-6 mbar), wrinkles were clearly visible that had not been apparent when the sample was introduced into the chamber. If the epoxy replica was post-cured at 150 °C for 1 h, as seen in Figure 5.5b, no wrinkles were visible after high-vacuum exposure for a similar time. One sample was under analysis by XPS (X-ray photoelectron spectroscopy) and was therefore exposed to the ultra-high vacuum of the XPS chamber (∼10-9 mbar, Figure 5.5c) for 24 h. Even though this sample was post-cured, it showed small, finely distributed wrinkles, which were probably caused by the long exposure to the vacuum. It is assumed that the wrinkles occur due to a slight shrinkage caused by degassing in the vacuum. Thus, to increase the stability of the epoxy replica, the post-curing step was added for the preparation of the samples used for gold-coating and SAM functionalisation. A side remark: Wrinkles on the epoxy substrate can also be induced by the electron beam (see Figure 5.6). Given a non-post-cured sample with features in the micron-range or smaller, the high energy density of the electron beam at a large magnification can cause the features 86
5.4 Precision of the Replica Technique
Figure 5.5: Vacuum stability of epoxy replica (EPO-TEK 302-3) shown for the example of gold-coated, golf-tee-shaped micro-pillars (preparation see Chapter 6). a.) cured at 60 °C, 30 min exposure to SEM vacuum; b.) cured at 60 °C, postcured at 150 °C for 1 h, 36 min exposure to SEM vacuum; c.) cured at 60 °C, post-cured at 150 °C for 1 h, 24 h exposure to XPS vacuum to shrink. Before the image in Figure 5.6 was taken (magnification 15,000x), a closeup of the area in the middle was recorded (magnification 100,000x). This caused the imaged squares to shrink and to wrinkle. The conditions for acquisition used here were an acceleration voltage of 5 kV, acquisition time about 40 sec, working distance of 4 mm, low current mode. The damage could be reduced by using a smaller acceleration voltage and even less current.
Figure 5.6: Wrinkles induced by the electron beam on an epoxy replica (EPO-TEK 301-2) from a AFM calibration grating (TGX01) as used in Chapter 5.4.
5.4 Precision of the Replica Technique If a 2D profile of a rough surface is acquired (z = f (x, y)), then by employing fast-Fourier transformation (FFT), roughness at specific wavelengths can be identified. Wieland et. al. used this property130 to render the statistical analysis by the root mean square (rms) parameter (also called Rq) more precise. The rms parameter calculates the standard deviation of the profile heights from a centre line and is sensitive to peaks and grooves.131 By comparing the plots of the rms as a function of the roughness wavelength of the original SLA titanium surfaces and its replicas, Wieland et al.125 found that the two curves were congruent in a range from 100 nm to 1 mm. 87
5 Replica Technique Schuler et al.132 characterised the replica process for cell studies. The masters were the SLA titanium surface and the bottom of a 24-well plate (smooth surface). As in the study by Wieland, polydivinylsiloxane (PROVIL novo light) for the mould and an epoxy for the replica (EPO-TEK 302-3M) were used. They found that replication of the smooth surface (Ra = 2.3 ± 0.2 nm) resulted in a surface less than 0.4 nm rougher than the original. Replication proved to be precise in the millimetre to micrometer range, but seemed to smoothen submicron structures to a certain extend. The smoothening effect was mainly attributed to the TiO2 -coating (60 nm), necessary for the cell experiments. Subsequent replication of the same mould did not lead to any observable effect in topography after at least eight generations. The interest of other research groups has been more in the replication fidelity of PDMS (Polydimethylsiloxane). Gordan129 identified the ratio of surface tension to the elastic (Young’s) modulus of the cured sample as being the responsible factor for the smoothing out of sharp corners. Zhang et al.128 found that the choice of the replica material according to its stiffness defines the maximum aspect ratios (diameter / height) at which replicas can be made. PDMS works up to an aspect ratio of 6; epoxy and polyurethane even higher than 12. Lee133 et al. prepared a master on a 4-inch-silicon wafer with scale marks, rulers and align keys, by photolithography. The cured PDMS negative (Sylgard 184) was then realigned with the master and shrinkage was measured on the ruler lines. They defined 2-D shrinkage to be 1.52 %, when cured at 80 °C. Despite these investigations, a quantitative understanding of the accuracy of the replica technique is still missing, especially in terms of shrinkage and edge smoothening. The investigations by Wieland et al.125 showed that roughness features with a wavelength greater than 100 nm are identical between the master and replica, as far as was detectable by stereo-SEM. In this technique, 3D information is reconstructed from several (2D) SEM images acquired at different tilt angles (stereo-photogrammetry). It is known134 that this reconstruction leads to higher uncertainties for small surface features than for large ones. Height values might be underestimated by 20 %, whereas lateral information is very accurate. Thus it was attempted to find masters and surface patterns to investigate replica fidelity, which can be analysed by macroscopic tools such as callipers or simply by SEM investigation. SEM investigations are fast and offer laterally good measures. First, candidates for master patterns were identified and then examined as to their suitability as systems for testing the limits of replica fidelity on the replica materials available. The characterisation tool-box that was hoped for, has not been found yet, but a few masters have showed promising results. Accurate measurements in the 10 nm to 1 µm range remain a challenge.135 Chrome steel cubes with a volume of 1 cm3 were used to measure shrinkage, and whether there was a directionality or temperature dependence. AFM (Atomic Force Microscope) calibration gratings were replicated to observe shrinkage at the nano- to mi88
5.4 Precision of the Replica Technique croscale, and as structures that can give a measure for smoothing / relaxation (surface tension effect) of sharp edges. This analysis proved to be difficult. A first attempt was made by carving a custom-designed structure into a silicon wafer with a focused ion beam. The idea was to achieve a structure that shows relaxation and surface tension effects in the XY plane. In this thesis, most epoxy replicas were coated with gold and self-assembled monolayers of alkanethiols after preparation. In order to ensure complete curing and to reduce deformation under the highly focused electron beam in SEM (as described in Chapter 5.3), these substrates were exposed to a post-curing step at 150 °C. It is shown that this step does not lead to significant additional shrinkage.
5.4.1 Experimental Masters: Cubes and cuboids: One pair of chrome steel cubes with dimensions 10 x 10 x 10 mm3 and one pair of cuboids of 5 x 10 x 20 mm3 were produced in the machine shop. For the cube, each face was marked for easy identification. Before the first use they were cleaned by ultrasonication first in petroleum ether (technical grade) then in acetone (technical grade). Calibration gratings: AFM calibration gratings (MikroMash, USA) TGG01 and TGX01 were ultrasonicated in toluene and 2-propanol before use. Specifications of the gratings are given in Table 5.2. Custom-made FIB structure: In order to have a tool to quantify the smoothening effect, a bitmap-pattern was designed in Adobe Illustrator CS3 (see Figure 5.7). This pattern was transferred into a silicon wafer by Focused-Ion-Beam milling (FIB; CrossBeam NVision 40, Carl Zeiss SMT, Germany) at the electron microscopy centre of ETHZ (EMEZ). One star consists of 10 rays emerging from the centre at an angle of 36°. The other star was designed with varying emerging angles between the rays (2, 10, 30, 45, 60, 90, 120°). The FIB mills out all black pixels, therefore a black and a white star was designed to have elevation and depression on the same sample. Additionally, a sketch of the comic figure Calvin from Calvin & Hobbes comic by Bill Watterson was added to see qualitatively how the replica technique reproduces fine and non-linear lines. FIB conditions: A Ga-ion beam of 40 pA at 30 kV was rastered over the target area. The binary-bitmap structure was employed as mask for the “feature mill” pattern generator. Horizontal lines and offsets (see Figure 5.11a) are artefacts due to the block wise milling routine of this tool. 89
5 Replica Technique
Table 5.2: Dimensions of AFM calibration gratings used as masters. Illustrations were taken from the specification sheet of the producer (MikroMash, USA) Grating
3.0 µm ± 5 nm
3.0 µm ± 8 nm
TGG01 (ridge structure) edge
TGX01 (chess-board structure) edge
Mould materials (negative): Mould materials (see Table 5.3) were either polydimethylsiloxanes (PDMS, SYLGARD® 184 silicone elastomer, base and curing agent, Dow Corning, USA and Gelest OE 41, 2-part Silicone RTV Encapsulant, ABCR, Germany, Prod no: AB129245) or polydivinylsiloxane (PROVIL novo light; Heraeus-Kulzer, Germany). Sylgard 184 is a PDMS frequently used in research, but has the disadvantage of being a contaminant due to many uncrosslinked and mobile low-molecular-weight silicone species. Gelest OE 41 is a PDMS without any silica filler. Provil Novo is a complex, commercially available silicone containing fillers and low-molecular-weight species, which cures within 3-5 min at room temperature. For simplification, the polydimethylsiloxanes will be referred to as Sylgard and Gelest respectively, while polydivinylsiloxane will be referred to as Provil Novo. All three silicones are two-component systems consisting of a base resin and a catalyst that have to be mixed together in order to solidify.
Table 5.3: Details on mould materials Silicone
Mixing ratio base : catalyst
Curing temp. [°C]
10 : 1
1:1 (Mixing gun)
5.4 Precision of the Replica Technique
Figure 5.7: FIB design at a 1000 x 1000 px resolution Replica material (positive): The materials used to cast the positives were either epoxies (EPO-TEK 301, EPO-TEK 301-2 and EPO-TEK 302-3M; Epoxy Technology, Polyscience AG, Switzerland; see Table 5.4) or the above mentioned PDMSs. The three epoxies are two-component systems consisting of a resin and a hardener, which have to be mixed in specific ratios, and are then cured at temperatures indicated by the manufacturer. They mainly differ in their viscosity at room temperature.
Replication: Mould (negative) preparation: Provil Novo was mixed in the mixing gun and directly cast onto the master. Sylgard and Gelest had to be mixed with a glass rod in a plastic beaker and then degassed in a desiccator at a low vacuum until all bubbles were gone. Rectangular Teflon frames for casting the moulds of the cube and cuboid masters 91
5 Replica Technique
Table 5.4: Details on epoxies used for replication. Epoxy type
Mixing ratio resin : hardener
Curing time: Requested minimum / effective curing time
Curing Glass temp. Temp. [°C] [°C]
Viscosity at 100 rpm / 23 °C [cPs]
20 : 5
> 1 h / Over night
100 : 35
> 3 h / Over night
10 : 4.5
> 3 h / Over night
were prepared by the machine shop. Mould casting of the AFM calibration gratings and the FIB structure was performed in a well of a 24-well plate (TPP, Switzerland). The moulds were cured according to the usual protocol. After curing, the master was removed. Moulds made of Provil Novo could be used right away, while moulds made out of PDMS had to be perfluorinated, to be replicated by another silicone. This was done by exposing the samples to 2 min air plasma with subsequent exposure to a lowpressure atmosphere of 1H-, 1H-, 2H-, 2H-Perfluoro-octyltrichlorosilane (ABCR GmbH, Germany) in a desiccator for 1 h. Replica (positive) preparation: If the replicas were made of PDMS, the preparation was the same as described above. The epoxies were prepared by mixing resin and hardener in the ratios described by the manufacturer, and were degassed in a desiccator for 30 min. After casting, epoxies and PDMS were cured in the oven overnight at the specified temperatures. Additionally, one set of cube epoxy replicas was cured at room temperature for three days. In order to define the influence of a possible post-curing effect, the cube epoxy replicas were post-cured at 150 °C for 1 h.
Metrology: Shrinkage of the cube and cuboid replicas: The side lengths of all epoxy cube and cuboid replicas were measured by the use of a calliper (precision ± 0.5 µm, 0 - 25 mm, Tesa, Switzerland). The data is an average over 2 to 4 casts, two cubes per cast and 2 sides evaluated per cube. The PDMS replicas deformed under the use of the calliper and were therefore measured with the optical system of the DSA 100 contact angle machine (Krüss GmbH, Germany). 92
5.4 Precision of the Replica Technique Images (bitmaps) of the masters and the PDMS replicas were recorded. For calibration, a 1.83 mm wide syringe was also placed in the image. The data is an average over 2 cubes and 2 casts. Post-curing at 150 °C: The mass of all epoxy cubes cured at room temperature according to the protocol were weighed before and after the post-curing. DSC measurements were made of samples cured at room temperature, according to the protocol and after postcuring. The samples were heated twice from -20 °C to 250 °C in 10°/min on a Mettler Toledo DSC 822e (Mettler Toledo, Switzerland). Samples with an average weight of 15 mg were tested under nitrogen flow. Investigations on the replicas of the AFM calibration gratings and the FIB structure: The masters and the epoxy replicas were analysed by SEM (Zeiss Gemini 1530, Carl Zeiss SMT, Germany). The masters remained uncoated, whereas the epoxies were sputter-coated with a Pt layer of 5 nm thickness (MED 010 coating system, Baltec, Liechtenstein). The SEM conditions were held constant: acceleration voltage 5 kV, working distance 4 mm, aperture size 30 µm. For the AFM-calibration-grating specimen, the magnification was 15,000x. The images of the FIB structures were acquired with magnifications of 20,000x for the overview, 80,000x for the stars and 100,000x for the Calvin figures. On the AFM-calibration-grating specimen, distances from ridge to ridge (TGG01) or chess-board-square side to chess-board-square side (TGX01) were measured with the line tool in ImageJ (rsbweb.nih.gov/ij). Replicas of TGG01 were cooled down in liquid nitrogen and freeze-fractured with a pair of pliers to get a side view of the ridge. On the SEM images of the FIB structure, the diameter of the centre of the stars with the homogeneous ray distribution was measured by the circle tool of ImageJ.
5.4.2 Results and Discussions First, a few general comments on the experience when preparing the moulds with the silicones: Provil Novo can not be used as a mould material for silicone species, because during curing, the silicone (Sylgard and Gelest) reacts with the Provil Novo mould. Additionally, Gelest is not suitable to be used as a mould material at all. After curing, it is still in a gel-like state and breaks during removal from the mould. Replicas will be usually specified by first mentioning the material of the replica, then the material of the mould, e.g. EPO-TEK 302-1 / Sylgard. 93
5 Replica Technique Shrinkage in 3D, measured on the cubes and cuboids Epoxies: All epoxies were cast into moulds made of Provil Novo. Care was taken that the liquid surface was horizontal in the mould, so the volume would correspond to exactly 1 cm3 . While curing, a depression occurred at the face facing the atmosphere. Due to the curved shape of the depression, the length of this side was not taken into account during the evaluation of shrinkage. All three examined epoxies cast into cubes show shrinkages between 0.3 ± 0.2 % and 0.5 ± 0.2 %, irrespective of whether they were cured at room temperature for 3 days or at the temperature suggested by the protocol. According to the producer, EPO-TEK 3012 would be expected to show a linear shrinkage of 1.4 %. This number is considerably higher than the values measured here on the sides facing the mould. The depression formation on the side facing the air seems to reduce the shrinkage at all other sides. Surprisingly, when analysing the cuboids, it was found that the shrinkage depends on the actual side length of the cuboid. The side with nominal length 10 mm showed a similar or slightly smaller shrinkage as the sides of the cube, but the side with the nominal length of 20 mm showed a significantly reduced shrinkage, in some cases less than half of the shrinkage of the 10 mm side. This finding was unexpected since the volumes of the masters are identical and a homogeneous shrinkage should not depend on the shape of the substrate. The most obvious difference between the cubes and cuboids is the area where the depression during curing occurs (1 vs 2 cm2 ). On this side shrinkage is not hindered by a side-wall of the mould, therefore most of the shrinkage due can occur on this side. This leads to the conclusion that shrinkage can be minimised for substrates where length and width are considerably larger than height. Silicones: Shrinkage of the two PDMS species seems to be slightly larger than for the epoxies. The analysis of the cubes showed shrinkage of Sylgard / Sylgard-cube replicas of 1.3 ± 0.5 % and of Gelest / Sylgard-cube replicas of 1.0 ± 0.5 %. There was hardly any depression formed on the side facing the air. The shrinkage found for the Sylgard replicas is comparable to that found by Lee et al.133 (1.52 %). It was also found that the calibration of the images taken with the DSA100 with the width of the syringe is not an easy task. The apparent width of the cylindrical metal syringe is dependent on the illumination settings. High brightness or contrast will make the syringe to appear thinner than it is. Therefore all images have to be taken with the same illumination settings. 94
5.4 Precision of the Replica Technique Replicas of AFM calibration gratings Replicas of the AFM gratings were prepared. Mould materials were Sylgard as well as Provil Novo. Gelest did not yield moulds that transferred the pattern of the masters. On replicas of TGG01 (ridge structure) the distance from ridge to ridge was measured to identify shrinkage on the scale of the pattern. Figure 5.8 shows the master (a) and three epoxy replicas (b-d). b.) and d.) were fabricated with the same epoxy (EPO-TEK 3012), but different mould materials and c.) and d.) had the same mould material (Provil Novo) but are different epoxies. The comparison of images taken with the two detectors of the SEM (InLens (SE1) vs. SE2) shows that the InLens detector provides a sharper image, which is easier to evaluate with the line tool of ImageJ. But, as it turned out after analysis of the measured lengths, the replicas are so precise that sources of error were in the same range as possible differences between the master (AFM calibration grating) and its replicas. The ridge tops on the master already have a certain dimension of 4 to 6 pixels. On the SEM images investigated, one pixel corresponded to a size of 25 nm (this size could be reduced by increasing the resolution of the image during acquisition). If it is assumed that the shrinkage on the replicated structure is similar to that measured over the whole side length of the cubes (see above, about 0.5 %), then the difference in length from the replica to the master over 4 ridges (= 12 µm, 480 pixels) would be 60 nm. 60 nm correspond to nearly 2 pixels. Therefore placement of the line of analysis needs to be exact, down to the pixel, to show any reliable difference between master and replica. This placing precision was certainly not achievable by placing the lines by hand. Another issue was found when replicas of TGX01 (chess board structure) were analysed. The lengths extracted along the y-axis were consistently larger than lengths extracted along the x-axis. Since the samples were mounted randomly, this could not be a characteristic of the replica technique, but an artefact from the SEM image. There also seemed to be either a drift of the sample during the measurement (Figure 5.9); distorting the squares to rhombi, or the orthogonality of the SEM machine was not well calibrated. This might also be a source of error influencing the measurement of lengths. It is therefore concluded that if such analysis needs to be reliably performed, special care should be taken in the mounting of the specimens, so that they cannot show drift during the measurement. Additionally, the machine should be freshly calibrated in the x- and y-directions. It is also advisable to always measure the master and its replicas in the same session under the same conditions. The occurrence of drift could be verified by changing the scanning speed (shorter acquisition time reduces the effect of drift) or reducing the magnification (reduces the energy density applied by the beam). A test of the calibration state of the machine would be to 95
5 Replica Technique
Figure 5.8: SEM analysis of TGG01 (uncoated) and its epoxy replicas (coated with 5 nm of Pt). The images are acquired in a mode that splits the photograph into two sides, one side showing the substrate acquired with the InLens (SE1) detector, the other side the same spot with the SE2 detector. a.) master, b.) EPO-TEK 301-2 / Sylgard, c.) EPO-TEK 302-3M / Provil Novo; d.) EPO-TEK 301-2 / Provil Novo. take two images of the chess-board master rotated by 90°, leaving all other parameters constant. If there is a measurable difference, it is an artefact of the calibration settings. Smoothening: The replicas of TGG01 were also used to obtain a view onto the side of the ridges. This was expected to shed light upon the smoothening of sharp edges during the replication process. The energy of the propagating crack during freeze-fracturing of the sample strongly distorted the ridges, as shown in Figure 5.10. One option to gain a side-view of the ridges could be by making an FIB cross-section.
Replicas of the FIB structure The bitmap file from Adobe Illustrator was successfully rendered into a silicon wafer (see overview, Figure 5.11a). First replicas showed that qualitatively there exists a difference between the master and the different species of epoxy (e.g. compare the structures of Calvin Figure 5.11 e - h). The width of the lines of Calvin on the master is about 75 nm (Figure 5.11e). In contrast to the replica made of EPO-TEK 301-2 / Provil Novo (Figure 5.11h), the replica made of EPO-TEK 302-3M / Provil Novo (Figure 5.11f) is missing one 96
5.4 Precision of the Replica Technique
2µm Figure 5.9: Distortion of squares to rhombi, exemplified on a EPO-TEK 301 / Sylgardreplica of TGX01. One side of the chess-board structure was aligned to be parallel to the image section (1). If a line is drawn perpendicularly to it (2), it clearly does not follow the sides of the chess-board squares, as does line (3) which is tilted by 2.4° to line (2).
3 µm Figure 5.10: SEM image side view of TGG01 EPO-TEK 302-3 / Provil Novo after cooling down in liquid nitrogen and freeze fracturing with a pair of pliers. Note the deformation of the features due to the energy of the crack propagation.
eye and the nose-mouth area is distorted. Since they have the same mould material, this may be a difference in the replication fidelity of the two epoxies. Sylgard seemed to be the worst mould material for the replication of these fine structures. Figure 5.11g and h show replicas with the same epoxy (EPO-TEK 301-2), but with different mould materials. The replica from the Provil Novo mould is clearly better. The stars with the homogeneous ray distribution can be used to measure the central, coalesced circle within the star, before the rays emerge. The diameter of the circle of the master is distinctively smaller (∼1024 µm) than of the replicas (> 1092 µm). There exists a difference of +20 µm between the diameter of the EPO-TEK 302-3M / Provil Novo to the EPO-TEK 301-2 / Provil Novo, which leads to the conclusion that the latter can fill corners better than the former. For Figure 5.11d, where the mould material was Sylgard, the rays of the star are broadened and it is more difficult to decide where to set the circle to measure the diameter. 97
5 Replica Technique
Figure 5.11: a.) overview master, uncoated; e.) close-up of Calvin on the uncoated master, i.) close-up of the depression of the irregular star on the uncoated master. b. – d.) replicas of elevated star, f. - g.) replicas of Calvin and j. – l.) replicas of the depression of the irregular star. b., f. & j.)) Epo-Tek 302-3 / Provil Novo, the dashed circle in b.) shows how differences between the different samples had been evaluated; c., g. & k.) Epo-Tek 301-2 / Provil Novo; d., h. & l.) Epo-Tek 301-2 / Sylgard. Cracks in the Pt-coating show that the non-post-cured replicas were slightly affected by volume changes while under vacuum and caused by the high energy density of the electron beam.
5.4 Precision of the Replica Technique Figure 5.11i to l show the master and the replicas from the star with different angles. In the mould-preparation step, the mould material had to flow into the carved pattern. If, during curing, surface tension flattens the structure, smaller mould depth and larger edge radii should result. Figure 5.11k and l show the replicas of the same epoxy (EPOTEK 301-2) obtained from the different mould materials. Since it is the same epoxy, the replica only shows the difference in the mould material. The EPO-TEK 301-2 / Sylgardreplica seems to pronouncedly show this smoothening effect, an effect not seen as clearly for Provil Novo-replica. Provil Novo probably has more filler in its composition which work against the forces of surface tension. If one intends to measure ray dimensions, the same problems occur as discussed above with the replicas of the AFM calibration gratings.
Post-curing at 150 °C:
The mass of the cubes and cuboids was measured before and after the post-curing step. The mass loss was averaged over all cubes and cuboids, since no difference could be detected. The samples cured following the protocol showed a mass loss of around 0.25 – 0.3 %. If they were cured at room temperature, the mass loss was slightly higher, around 0.4 – 0.55 %. These values are within the range to be expected. The producer specified the mass loss of EPO-TEK 302-3M at 250 °C as being 0.77 %. The change in side length was mostly below a change of 0.1 %, sometimes even slightly positive. This leads to the conclusion that shrinkage due to post-curing is negligible. However, in some rare cases, some sides did change shape by this heating up above the glass temperature, rendering their sides immeasurable. A reason for this behaviour may be that the material was under internal stress and, upon heating, the modulus decreased and the stress could deform the substrate. DSC measurements of the samples cured at room temperature showed an exothermic peak superimposed to the Tg, followed by a reaction peak. The low Tg and the reaction peak starting at temperatures above 80 °C indicate that the epoxy polymer was not fully reacted after curing at room temperature for 3 days. Samples that were cured according to the protocol showed the highest Tg and a small endothermic peak at temperatures above 120 °C, which most probably is due to desorption. This would correspond to the mass loss measured on the epoxy-cube replicas that were post-cured. Post-cured samples only showed a Tg and no further peaks below 160 °C, indicating a fully cured sample. 99
5 Replica Technique
Table 5.5: Tg measured by DSC measurements after the different curing conditions and post-curing. Curing history
5.4.3 Summary and Conclusion
Measurements with a calliper on 1 cm3 epoxy-cube replicas have shown that for all three examined epoxies (EPO-TEK 301, 301-2, 302-3M) shrinkage is very small (0.3 – 0.5 %) when cured in Provil Novo moulds. Epoxy-cuboid replicas show anisotropy in shrinkage depending on the actual side length; the longer the side, the smaller the shrinkage. It was concluded that the larger the area facing air during curing (large width and length compared to height), the smaller shrinkage of width and length of epoxy samples. Then, most of the necessary shrinkage occurs on the side facing air, forming a depression. PDMS-cube replicas were not found to form depressions or any anisotropy in shrinkage. Shrinkage was determined to be 1.3 % for Sylgard 184 and 1 % for Gelest OE 41. Defining differences between master and replica by measuring lengths in SEM images on epoxy replicas of AFM-calibration gratings proved to be difficult, because the replica fidelity is very high. If replicas of AFM-calibration gratings are to be analysed, the samples need to be mounted very carefully to reduce drift to a minimum. To avoid cracks caused by the vacuum, the samples should be post-cured before analysis. The machine should be freshly calibrated and all samples need to be measured in the same session under the same conditions. The FIB approach proved to be the most promising method to evaluate structural differences between master and replica. The star with the homogeneous distribution of rays shows differences in the size of the central circle from which the rays emerge. This circle can be measured on SEM images and actually indicates differences. The fine carvings of the Calvin figure allow for a quick qualitative evaluation of the replica fidelity. The star with the inhomogeneous distribution of rays also showed differences between the different material combinations. It was the most suitable test structure for comparing different materials, especially the influence of mould materials. The higher filler content of the Provil Novo increases replica fidelity compared to the Sylgard. 100
5.4 Precision of the Replica Technique
5.4.4 Excursus: Curing Properties of Thermosets One material involved in the replica process, the epoxy resin, is a thermoset. Thermosets are network-forming polymers whose polymerisation and cross-linking reaction is activated by heat or light. The polymer is obtained by mixing a resin and a hardener, often together with a catalyst to accelerate cure. In most cases the reaction is thermally activated, giving rise to the term “thermoset”, but also light-activated network polymers such as SU-8 are called “thermosets”. The basis of the following summary can be found in the book “Thermal characterization of polymeric materials”, Academic Press, Second edition, Volume 2.136 As depicted in Figure 5.12, the reaction of the monomers starts (Figure 5.12a) by forming linear chains that soon begin to branch (Figure 5.12b) and finally cross-link(Figure 5.12c). At full conversion, all monomers are reacted into a dense network (Figure 5.12d). From the time where the reaction is started, the viscosity of the liquid increases up to a point where cross-linking is advanced so much that the liquid loses its ability to flow. This point is called gel point and indicates the incipient formation of a network (transition from Figure 5.12b to c). The polymer is not yet fully cured (conversion 0.5 0.8, depending on the monomers involved). The onset of gelation defines the time span during which it is possible to process the mixture.
Figure 5.12: Sketch of thermoset cure: a) monomers reacting with each other to linear chains, b) the growing chains start to branch. c) cross-linking, and later gelation occurs, until d) the polymer is fully cured / cross-linked. Illustration taken from “Thermal characterization of polymeric materials”, Academic Press, Second edition, Volume 2.136 The reaction rate does not significantly change after this point, but with ongoing reac101
5 Replica Technique tion, the glass temperature (Tg ) increases due to increased cross-link density, which strongly influences the ultimate physical properties. Apart from the gelation, at any time during curing, vitrification can occur. This transformation from a liquid or rubbery state to a glass is a direct consequence of the ongoing reaction, and occurs when Tg becomes equal to Tcure (as long as Tcure < Tg∞ , the Tg of the final, fully cured thermoset). At this point, the reaction rate decays by two to three orders of magnitude from a chemically to a diffusion controlled reaction. Due to the diffusion controlled curing Tg can reach 10 – 15 °C above Tcure. At temperatures above Tg∞ the thermoset stays in the rubbery state, unless other reactions occur, such as thermal degradation or oxidative cross-linking (char region). These phenomena can be illustrated by the so-called TTT (time-temperature-transformation) cure-diagrams (Figure 5.13).
Figure 5.13: Generalized time-temperature-transformation (TTT) cure diagram. Tg0 corresponds to the Tg of the unreacted material, gelTg to the glass temperature at gelation, Tg∞ to the Tg of the fully cured thermoset. The plot of isothermal cure time versus cure temperature shows four different states of matter: liquid, gelled rubber, gelled glass and ungelled glass. Illustration taken from “Thermal characterization of polymeric materials”, Academic Press, Second edition, Volume 2.136 Epoxy-resin handling is restricted to the time it takes until the reaction reaches the gel point. After it has reached this point it has lost its ability to flow. The glass temperature Tg , which indicates the transition from a glassy to a rubbery state, can be tuned by the choice of the curing temperature Tcure , as long as Tcure is below the maximal reachable glass temperature Tg∞ . Tg∞ is a material property and can be reached by the appropriate curing temperature. If a post-curing step is used to ensure complete cure, it should be taken care that the char region is not reached, to avoid thermal degradation. 102
5.5 Ceramic Replicas
5.5 Ceramic Replicas During this thesis, it has also always been a wish to find mechanically strong selfcleaning surfaces. Ceramics are very hard materials, stable over a wide temperature range, and they can withstand many solvents, acids or bases and show functional properties as piezoelectricity, pyroelectricity and catalytic activity. If it was possible to design an appropriate surface structure and easily transfer it onto any ceramic of choice, e.g. self-cleaning high-voltage insulators, bath room tiles or dishes would be conceivable. Therefore it was attempted to apply the replica technique using ceramic slurries. The choice fell on an alumina slurry. A recipe to achieve high solid contents slurries was readily available, since it has already been used to replicate sand-blasted and acid-polished alumina sheets (C. Zink, unpublished work). High solids contents are necessary to minimise shrinkage during drying of the ceramic green body. Another advantage of alumina is that it can be functionalised by SAMs of alkylphosphates.137 C. Cremmel138 found in an X-ray-photoelectron-spectroscopy (XPS) study, that the alumina powder used here has a small silica contamination on the surface. This contamination accumulates during sintering at the surface of the ceramic and covers a significant fraction of the surface. Phosphates do not bind to silica surfaces. Thus, no further experiments were performed with the alumina replicas. Nevertheless, the replica technique works impressively well at replicating photolithographic microstructures and might inspire other areas of research and find application in electronics and micro-electromechanical (MEMS)139 or micro-chemical systems.140 In MEMS, ceramics are used as thin films or are deposited to form stripes by writing methods, by downsizing mechanical processing methods or by self-assembly.139 Most of the structures achieved by these methods have a lateral feature resolution of > 50 µm. Ceramic microchemical systems have the potential to be used for reactions under harsh conditions, such as high operating temperatures and corrosive gases. Expensive fabrication methods and reproducibility issues to generate non-deformed and crack-free centimetre-scale ceramic structures have hindered development of such ceramic devices. Our method might work equally well as the gelcasting method described by Christan et al.140 The work here is not the first attempt to make ceramic replicas. Marzolin et al.141 used PDMS moulds to shape Sol-Gel precursors. The mould was removed after gelation and the substrate sintered. This method allowed feature edges of 50 nm radius to be reproduced. Schönholzer et al.142 also used PDMS moulds of a photoresist structure, but used high-solids-content aqueous alumina slurries for ceramic replication. The soft mould yields to the stresses built up during drying of the slurry to the green body. Therefore, fine patterns do not break. After removal of the mould, the green bodies were sintered. 200 nm α-alumina particles replicated holes of 2 µm diameter and 3 µm depth 103
5 Replica Technique satisfactorily. Lines of 3 µm width and depth were also replicated with tin oxide, cerium gadolinium oxide and zirconia.142 In most approaches140, 141, 143–146 the mould material was PDMS whose silicone nature can form silica and thus had to be removed before sintering to prevent contamination.
5.5.1 Experimental Masters: The masters are photolithographic structures made from SU-8 on silicon wafers. Their production is described precisely in the thesis of C. Cremmel.147 In short, they consist of large pillars of diameters between 20 and 500 µm, ∼30 µm height and arranged in a hexagonal pattern. On top of the large pillars, smaller pillars were formed with diameters between 2 and 15 µm, ∼10 µm height, also in a hexagonal pattern. This second layer of pillars was expected to be washed away at places where no large pillar was crosslinked underneath. It needed several generations of parameter adjusting until this goal was achieved satisfactorily. Nevertheless imperfect masters could be used to explore the possibility of transferring photolithographic structures into a ceramic replica. The structured silicon wafer was cut into pieces of 20 x 20 mm2 with a diamond cutter. Any splinters were blown away under a stream of nitrogen. Moulds (negatives): Affinity VP 8770 (Dow Chemical Company, USA) was used for mould preparation. Affinity is a polyolefin plastomer (POP), comprising of a copolymer of ethylene and 1-octene. Moulds were prepared as described in Csucs et al.148 In a hot press with a small load applied, the POP pellets were melted into blocks of 25 x 74 x 4 mm3 at 190 °C, using an appropriate metal template. To avoid sticking, Polyimide foils were placed between the hot plates and the polymer / metal template. After cooling down, the POP bars were removed from the metal template and cut into three pieces of equal size. These samples were then rinsed with ethanol and dried under a stream of nitrogen. Then, one POP piece was placed on top of the 20 x 20 mm2 master and another piece of silicon wafer as spacer. These three substrates were sandwiched between a set of aluminium sheets and polyimide foils. The whole compound was then transferred onto another hot press, preheated at 130 °C. First, it was pressed with a weight of 200 g for 5 min, and then with 700 g for 4 min (see Figure 5.14). After cooling down, the master could easily be removed from the POP mould. Ceramic Replicas (positives): Many ways to remove the ceramic green body from the polymeric mould (Provil Novo and PDMS) were tested by C. Cremmel.147 But none of the methods enabled to replicate pillars smaller than 80 µm in diameter, because the pillars were broken off the 104
5.5 Ceramic Replicas Aluminum sheets Polyimide foils Affinity Master Spacer Weight
Figure 5.14: Set-up during hot-pressing of the Affinity mould. brittle ceramic green body. Since Schönholzer could remove much smaller features from the mould142, 149 it is assumed that the stress applied on golf-tee-shaped micropillars (all structures used for replication were undercut) of the green body are too large to be removed intact from the mould. The only successful results were achieved when burning the mould during sintering. This lost mould process is more time intensive than methods where the mould can be reused, but it leads to very satisfactory results when replicating complex structures. In the following, this method is described in detail. Slurry preparation: The alumina ceramics were prepared from 200 nm diameter Al2 O3 particles (α-Al2 O3 , d50, ∼200 nm (grade Ceralox HPA-0.5), Sasol North America Inc., USA). The slurry was prepared by mixing 57 vol% of Al2 O3 particles with deionised water (house supply) containing 0.05 mol/l NH4Cl (99.5 %, Sigma-Aldrich, Switzerland) and was adjusted to pH 2 with HCl (1 mol/l in water, Merck, Switzerland). By adding alumina milling balls twice the weighed in amount of the alumina particles, the suspension was ball-milled for 18 to 24 h to de-agglomerate the suspension. Then the slurry was separated from the milling balls, a few drops of octanol were added to the slurry and it was stirred with a magnetic stirrer in an evacuated desiccator for 30 min in an iced bath. Casting and drying: The slurry was sucked into a 20-ml syringe (B. Braun Melsungen AG, Germany) and cast into the Affinity moulds. In order to achieve homogeneous drying rates of the green body, the mould was covered by a lid made of Sylgard 184. The slurry was allowed to dry for at least 2 days. Sintering: The green body still in the mould was placed in a high temperature oven (Carbolite RHF 105
5 Replica Technique 1600, Carbolite Limited, UK). First the organic material was burned away by heating at 1 °C/min and held constant at 400 °C for 1 h, then the heating rate was increased to 5 °C/min and held at 1500 °C for 4 h.
5.5.2 Results and Discussion Affinity is a polymer that is not quite as flexible as PDMS, but certainly has a lower modulus than epoxy. Thus, it is assumed that it was still flexible enough to yield to the stresses occurring during drying of the slurry, because the pillars appeared unharmed in the SEM analysis (Figure 5.15). Schönholzer150 also made an analysis where he directly used the patterned SU-8 / silicon wafer as a mould for alumina patterning and burned the SU-8 layer during the sintering step. He found that the pyrolysis of the polymer induced cracks in the ceramic pattern. This was not observed with the Affinity mould. The polyolefin burned away cleanly. During burning, a considerable amount of soot is formed. But, upon further heating, this is fully transformed into CO2 and H2 O. The XPS measurements of a sample straight out of the oven after cooling down indicated a carbon contamination of apparent normalised atomic concentration of 1.4 %138 which is very low when compared e.g. to the carbon contamination of a freshly cleaned gold surface.151 The gases formed during burning of the polymer seem to degenerate the heating elements (accelerated aging due to SiC conversion in SiO2 , and breakage of the heating elements). Therefore, it seems advisable to use an oven where the heating elements are not placed in the heating chamber, but shielded by an additional wall. Masters: Figure 5.15 shows SEM images of a master (left column, b. and c.) and a ceramic replica (right column). Only the optical microscope image (Figure 5.15a) is from the master used for the ceramic replica in the right-hand column. The other images are not from the master that was used for alumina replication, but since the SEM images are more demonstrative, they are presented here to show what kind of defects already could occur on the master. The small pillars tend to stick together, especially at the boarder of a large pillar (Figure 5.15a. and b.). Small pillars can even be detached from the master and be quenched between neighbouring pillars (Figure 5.15b. white arrow). Additionally, small pillars may lie in the space between the large pillars (Figure 5.15a.). What varied from master to master was the shape of the small-pillar tops. Sometimes they were round, as on the chromium mask, sometimes they were of a hexagonal form. So, the hexagonal shape of the small pillars on the alumina replica is not due to the ceramic material, but because the master already had hexagonal pillars (compare Figure 5.15a. and d.). Replicas: Looking now at the ceramic replicas on the right hand side of Figure 5.15, it is striking how precisely all the features of the master were replicated. Even free-standing pillars, as indicated with the white arrow in Figure 5.15d (close-up in Figure 5.15g), 106
5.5 Ceramic Replicas
Table 5.6: Dimensions on the chromium mask of the substrates illustrated in Figure 5.15 Substrate
Diameter large pillars [µm]
Pitch distance large pillars [µm]
Diameter small pillars [µm]
Pitch distance small pillars [µm]
SU-8 / Si-wafer b.) and c.)
are replicated. There is no chance that this small pillar was moved to this location after sintering, since the ceramic is far too hard to be easily broken by tweezers during handling of the sample. Also undercut shapes as shown at the rim of the large pillars in Figure 5.15e are replicated with a high precision. No systematic investigation was done to determine down to which sizes such feature precision is possible. The very lowest limit is clearly the diameter of the ceramic powder. Since the particle diameter of the powder was 200 nm, features can probably be reduced to 1 µm in size. Reducing the powder size might be an option to replicate smaller features, as long as a sufficiently high solids content can be achieved in the slurry. Also, other species of ceramic powder might be used for replication. Functional ceramics, such as tin oxide, cerium gadolinium oxide and zirconia have already been successfully replicated by slurry casting.149 The free-standing pillar inspires the idea that even three-dimensional structures such as bridges, or short tunnels could be transferred. Methods to achieve complex structures from SU-8 have been reported152 . The size combination of the pillars shown in Figure 5.15 is summarised in Table 5.6. Comparing optical microscopy images Cremmel et. al.147 found that the alumina replicas shrink about 10 to 19 % with respect to the master.
5.5.3 Summary and Conclusion It has been shown that by the use of a polyolefin (Affinity) as a mould material, ceramic replicas of high precision can be achieved when the mould is burned during the sintering step. It is expected that even 3D structures could be replicated by this technique.
5 Replica Technique
Figure 5.15: Light microscopy (a.) and SEM images (b. – f.) of photolithographic structures. a. – c.) SU-8 structures on Si-waver, d. – f.) alumina replicas. Master and replica correspond only for a.) and d. – f.). The images b.) and c.) are of a SU8 / Si-Wafer masters of another generation, where SEM images were available. The intention here is merely to show how precise the alumina replica technique can transfer fine, even undercut structures. The arrow in e.) indicates a free standing pillar, suggesting that maybe even bridge or tunnel structures would be possible to achieve. 108
6 Beyond the Lotus Effect: Roughness Influences on Wetting over a Wide Surface-Energy Range This work was published in 2008 in Langmuir.22 The supplementary material was either added into an appendix at the end of the chapter or the references were replaced by links to the specific chapters in the thesis.
Abstract In order to enhance our understanding of liquids in contact with rough surfaces, a systematic study has been carried out, in which water contact angle measurements were performed on a wide variety of rough surfaces with precisely controlled surface chemistry. Surface morphologies consisted of sand-blasted glass slides as well as replicas of acid-etched, sand-blasted titanium, lotus leaves and photolithographically manufactured golf-tee shaped micropillars (GTMs). The GTMs display an extraordinarily stable, Cassie-type hydrophobicity, even in the presence of hydrophilic surface chemistry. Due to pinning effects, contact angles on hydrophilic rough surfaces are shifted to more hydrophobic values, unless roughness or surface energy are such that capillary forces become significant, leading to complete wetting. The observed hydrophobicity is thus not consistent with the well-known Wenzel equation. We have shown that the pinning strength of a surface is independent of the surface chemistry, provided that neither capillary forces nor air enclosure are involved. Also, pinning strength can be described by the axis intercept of the cosine-cosine plot of contact angles for rough versus flat surfaces with the same surface chemistries.
6.1 Introduction For many practical applications, such as coating or fluid handling, the wettability of a surface plays a crucial role. The contact angle, θY , that a liquid drop makes with an ideally flat surface, corresponds to a minimum in the energy of the liquid-solid-ambient 109
6 Roughness Influences on Wetting over a Wide Surface-Energy Range gas system34 (see Figure 6.1). The prediction of contact angles for real surfaces presents a significant challenge, however, since it is well known that roughness exerts a significant influence over wetting phenomena.15, 41, 50, 53 Only on ideally flat, uniform surfaces does θY have a unique value. On real surfaces, depending on how the drop is deposited, the contact angle θ can vary between the so-called advancing and receding contact angles. This hysteresis can be ascribed to inhomogeneities in the distribution of adsorbates or the presence of contaminants, to surface roughness, or to time-dependent surface rearrangements.41 On rough surfaces the surface morphology strongly influences the value of θ. On rough, hydrophobic surfaces the liquid can either follow the surface topography and show strong pinning, or can bridge from asperity to asperity while enclosing air beneath and showing almost no hysteresis in contact angle. For the first case, Wenzel6 introduced a roughness factor r to describe the roughness influence on θ (equation 6.1 and Chapter 2.6).
cos θW = r · cos θY
r is calculated by dividing the actual, roughness-enhanced surface area by its projection. This behavior is often referred to as Wenzel-type wetting. If the cosines in equation 6.1 are plotted versus each other, the effect of roughness is evident in the deviation from a straight line with slope 1. For the second case, Cassie and Baxter7 modified Wenzel’s equation by introducing the fractions f1 and f2 , where f1 corresponds to the area in contact with the liquid divided by the projected area and f2 to the area in contact with the air trapped beneath the drop, also divided by the projected area (see chapter 2.6, discussion on the Cassie-Baxter equation):
cos θCB = f1 · cos θY − f2
Since the introduction of these equations, wetting on rough surfaces has been the subject of intensive research,41, 53 which was significantly boosted by the discovery of the ‘self-cleaning’ properties of the superhydrophobic lotus leaves (“lotus effect”).2 While the many publications on this topic have increased our knowledge of superhydrophobic behaviour,15, 18, 153–156 showing, for example, that the most stable topographies to achieve superhydrophobicity are “undercut”,16, 77, 78 many aspects of this field still remain controversial.46, 55, 60, 62 Structured surfaces that exhibit superhydrophobicity can also show an effect known as hemi-wicking50 or superwetting if they are surface-chemically functionalized to be 110
6.1 Introduction hydrophilic. Hemi-wicking is complete wetting due to the presence of capillary forces in two dimensions.51, 52
γSL - γSA
Figure 6.1: Scheme of wetting phenomena. a) definition of contact angle on an ideally flat surface; b) Wenzel-type wetting; c) Cassie-type wetting; d) pinning: a growing drop pinned by one obstacle; e) hemi-wicking. We have chosen to examine wetting-property changes upon significant variation in the surface chemistry of samples with four different surface topographies. This is of both fundamental and practical relevance, since many surface coatings change their surface chemistry over time due to contamination or oxidation. Surfaces have been analyzed by scanning electron microscopy and roughness factors evaluated by means of white-light profilometry. Static (θs) and dynamic (θa, θr) contact angles have been measured. In order to vary surface chemistry over a wide range of surface energies, the surfaces were coated with gold and subsequently functionalized by means of mixed, self-assembled monolayers of methyl- and hydroxyl-terminated alkane thiols. In this way, contact angles between 20° and 105° can be obtained on a flat surface. Surfaces examined were sandblasted glass microscope slides (SBG), as well as replicas of sand-blasted (large-grit), acid-etched titanium (SLA), lotus leaves (LLR) and golf-tee-shaped micro-pillars of photoresist (GTM) on a silicon wafer. The GTM pillars show an extraordinarily stable, Cassie-type hydrophobicity. All four surface topographies are uniformly rough, such that it does not make a difference where a drop is placed. This precondition, as emphasized by McHale,62 has to be met in order to be able to compare contact angle data with equations 6.1 and 6.2. By applying the Wenzel equation over this wide range of surface chemistries, the roughness factor fails to predict the data. If the θ data are pre111
6 Roughness Influences on Wetting over a Wide Surface-Energy Range sented in a cosine-cosine plot of rough vs. smooth θ for the same surface chemistries, the three major classes of behaviour - “hemi-wicking”, “pinned”, “Wenzel- or Cassie-type” wetting (see Figure 6.1) - can be readily distinguished. For the surface-energy range where pinning has the most profound influence, it is suggested that the axis intercept is surface-morphology sensitive.
6.2 Experimental Substrates. Glass microscope slides were sandblasted with an air pressure of 8 bars. The sand jet was passed over the microscope slide twice for about 15 sec in perpendicular directions, in order to achieve a homogeneously roughened surface. Subsequently these sandblasted glass (SBG) substrates were ultrasonicated in ethanol for 10 min, rinsed with ethanol and dried under a stream of nitrogen. GTM masters were produced by standard photolithography on 4’’ silicon wafers, upon which, after cleaning, SU-8 2025 negative photoresist (MicroChem, USA) was spincoated for 60 sec at 2000 rpm to a thickness of 30 µm and soft-baked on a hotplate for 2 min at 65 °C and 3 min at 95 °C. Subsequently, the wafer was exposed to UV light (MA6, Karl Süss, Germany) with constant intensity (total energy 180 mJ/cm2) through a chromium mask having circular patterns of 20 µm diameter and 70 µm pitch distance (centre to centre). After postbaking for 2 min at 65 °C and 3 min at 95 °C, the wafers were developed for 5 min in SU-8 developer. Finally the wafer was hard baked at 190 °C for 10 min. The prepared wafer was used as master to make replicas. Replicas of acid-etched, large grit sandblasted (SLA) titanium (Straumann, Switzerland), the SU-8 pillar-structured silicon wafer and lotus leaves (Nelumbo Nucifera) obtained from the Botanical Garden in Zurich were prepared according to Wieland et al.125 In short, a low-viscosity and fast-curing silicone (PROVIL®novo, Light C.D. 2, fast set; Heraeus Kulzer GmbH, Germany) was cast in a mould affixed to the surface to be replicated. This silicone sample, exhibiting the original’s negative structure was then used in turn as a mould to cast positive replicas with an epoxy blend (EPO-TEK 302-3; Epoxy Technology, USA). The epoxy was then cured for 5 hours at 60 °C, removed from the mould and subsequently post-cured for 1 hour at 150 °C. The lotus-leaf replicas (LLR) were sawn into 5 x 15 mm2 pieces. The SLA replicas were cast in a circular mould of 12 mm diameter. All replicas were cleaned in a 2 v% solution of Hellmanex (Hellma, Germany) in ultrapure water (resistance 18.2 MW, EASY®pure by Barnstead, USA) and subsequently rinsed 5 times with ultrapure water. Single-side-polished silicon wafers (Si-Mat Silicon Materials, Germany) were cut into 10 x 10 mm2 pieces. To remove glue residues from the cutting step they were sonicated for 10 min in toluene and 10 min in ethanol. 112
6.2 Experimental Gold Coating. The rough surfaces and single side polished silicon wafers were cleaned for 2 min in air plasma and then coated by resistance evaporation (MED 020 coating system, BALTEC, Liechtenstein) with 10 to 15 nm Cr and 50 nm Au (purity >99.99 %, Unaxis, Liechtenstein). The rough surfaces were rotated during evaporation, and the pillars also tilted by 25° to achieve a homogeneous coating. The gold coated silicon wafers were used in every experiment as flat reference. Surface Modification. In order to achieve a wide surface-energy range, self-assembled monolayers (SAMs) of 11-mercaptoundecanol and dodecanethiol (Aldrich Chemicals, USA) were chemisorbed in both pure and mixed form on freshly gold-coated samples. The samples were immersed in 0.1 mM ethanolic thiol solutions for 20 min. For the solution preparation the total thiol concentration of 0.1 mM was held constant and the composition of the two compounds was varied in terms of dodecylthiol molar ratio from 100, 70, 50, 45, 30, 15, 10 (GTM only) to 0 %. After assembly, the samples were rinsed with ethanol (purity > 99.8 %, Merck, Germany) and dried under a stream of nitrogen. Contact Angle Measurements. Static water contact angle (θs) measurements were performed on a Ramé-Hart contact-angle goniometer on freshly prepared surfaces. A drop of 6 – 8 µl was produced and then gently placed on the surface. For superhydrophobic surfaces the drop had to be enlarged up to 12 µl in order for the drop to be able to detach from the syringe. The Contact angle is then defined as depicted in Figure 6.1. Dynamic water contact angle (advancing (θa) and receding (θr)) measurements were performed on a Krüss contact-angle-measuring system (G2/G40 2.05-D, Krüss GmbH, Germany) with a speed of 15 µl/min. Two movies with 40 images were recorded for the advancing contact angle, and only one for the receding. Analysis was carried out by means of the tangent-method-2 routine of the Krüss Drop-Shape Analysis program (DSA version 188.8.131.52 for Windows 9x/NT4/2000, 1997-2002 KRUESS). The movies were evaluated as follows: If the drop exhibited a stick-jump behaviour, but moved over the whole recorded time, all contact angles were evaluated. In this way the stick-jump led to a high standard deviation. If one side of the drop did not move at all during the recording, only the other, moving side was evaluated. In the superhydrophobic case, since the drop is confined between the syringe and the surface, and in the mean time very strongly repelled, it squeezes-out at one side. In this case only the mobile, squeezedout side was evaluated. The tangent method 2 routine, a 4th order polynomial function, has difficulty in fitting drops in the Cassie-type wetting regime, leading to a systematic underestimation of a few degrees. For many experiments, the receding contact angle could not be defined, since the drop was so strongly pinned to the surface. A value of 0° was assumed for these cases. Several exemplary images taken from the movies illustrate the dynamic contact-angle evaluation in the Appendix 6.5.2. Scanning Electron Microcopy. Replicas and SBGs were analyzed in a Zeiss Gemini 1530 FEG SEM at 3 to 5 kV, at room temperature, gold coated as described above. 113
6 Roughness Influences on Wetting over a Wide Surface-Energy Range White - Light Profilometry. For the evaluation of the roughness factor an optical profilometer was used (FRT MicroGlider, Fries Research & Technology GmbH, Germany). X, Y resolution is 1 µm, Z resolution is better than 10 nm. Areas of 1 x 1 mm2 were measured with a resolution of 1000 by 1000 pixels. The data points were assembled into adjacent, non-overlapping triangles and their areas summed up to achieve the actual surface area. The roughness factor was then obtained by dividing this actual surface area with the analyzed area (1 mm2 ).
6.3 Results and Discussion Surface-Morphology Characterization: SEM analysis of the investigated surfaces reveals considerable differences in surface morphology (Figure 6.2). The SBG shows a very profound micro-roughness with very rough patches, nano-sized features, and very flat conchoidal fracture areas. The SLA replica surface also contains two major roughness scales. The micro-scale roughness originates from the sandblasting step, leading to troughs. The superimposed nanoscale roughness was created by the acid-etching process.101, 124, 157 The LLR shows the wavy structure of the lotus-leaf cells topped with papillae. The papillae exhibit a diameter of 9 ± 2 µm, a spacing of 21 ± 7 µm and a height of around 20 µm. The nano-scale roughness is an approximate replica of the leaf’s wax structure, which was compressed during the replication process (see Chapter 5.1). The GTM pillars measure 24 µm in diameter on top and thin out to a diameter of about 15 µm towards the bottom. Their height slightly excess 30 µm, and the pitch distance (centre to centre) is 70 µm. The pillar tops constitute about 9 % of the projected geometric surface area. The pillar surface is highly ordered, whereas the other three samples show non-periodic structures. For the sake of clarity the roughness factors obtained with white-light profilometry are presented after the contact angle section. Water Contact Angle Data: Figure 6.3a is an explanatory graph. It illustrates the effects of roughness on a surface chemistry defined by the contact angle found on the flat reference. The maximum θ that can be ordinarily achieved on a flat surface is 120°. Lower surface energies are thus excluded and no data can be obtained beyond 120° on the flat surface (area A). With a hydrophobic coating (90° to 120° on the flat surface) the drop either shows Wenzel- or Cassie-type wetting (area B). Below 90° on a flat surface, even though the surface chemistry is intrinsically hydrophilic, the drop on rough surfaces shows a higher θ, and even hydrophobic values are possible (area C). In some cases, surface topography can be such that Cassie-type wetting persists in the hydrophilic regime79 (area F). In the high surface energy range (low θ), hemi-wicking can occur (area D). Surface topography determines the surface energy at which capillary forces come into play. 114
6.3 Results and Discussion
Figure 6.2: SEM images at two different magnifications of the examined surfaces. The scale bar on the left side corresponds to 100 µm and on the one on the right side to 2 µm. a&b) sand-blasted glass (SBG) (rough and flat regions enlarged), c&d) replica of acid-etched, sand-blasted titanium (SLA), e&f) replica of lotus leaf (LLR), g&h) replica of golf-tee shaped micro-pillars (GTM). Presenting the data in a cosine-cosine plot facilitates comparison with the Wenzel predictions (area E) - a linear function of surface energy (cosine of Young contact angle) with the roughness factor as slope and axis intercept of zero. Water contact angles were measured on both rough and flat surfaces that had been exposed to the same thiol mixtures. In Figure 6.3, the θ data are plotted, such that the cosines of the static and dynamic contact angles on the rough surfaces are plotted versus the cosine of the static contact angles measured on flat surfaces with the corresponding surface chemistry. Thus each data point corresponds to one comparative experiment. With these diagrams, a direct comparison with the Wenzel theory6 (Equation (6.1)) is possible. 115
6 Roughness Influences on Wetting over a Wide Surface-Energy Range
A 120° B
resultant contact angle for rough surface
E F hydrophobic
surface energy (determined by adsorbate chemistry)
0° 0° 0°
contact angle θ 120° 90° 60°
120° 90° 60° contact angle θ
1 cos θrough 0 0.5 -0.5 -1
0 0.5 cos θYstatic
120° 90° 60° contact angle θ
e) GTM 180°
0 0.5 cos θYstatic
0 0.5 cos θYstatic
cos θrough 0 0.5
120° 90° 60° contact angle θ -0.5
cos θrough 0 0.5 -0.5 -1 -1
contact angle θ 120° 90° 60°
120° 90° 60° contact angle θ contact angle θ 60° 120° 90°
0 0.5 cos θYstatic
d) LLR 180°
cos θrough 0 0.5 -0.5 -1 -1
c) SLA 180°
contact angle θ 120° 90° 60°
b) SBG 180°
Figure 6.3: a) Model graph clarifying the meaning of the axes in b – e) and the effects that can be observed in the plot. A) surface energies leading to contact angles above 120° on a flat surface are physically not possible, B) hydrophobic region, Wenzel-or Cassie type wetting occurs, C) pinning effects cause the drop to adopt (considerably) higher contact angles compared to the flat surface, D) capillary forces occur, E) series of Wenzel predictions, F) contact angles above 150°, Cassie-type wetting. b – e) Water contact angle data for four different surfaces; static (black symbol), advancing (grey symbol) and receding (white symbol) θ. The x-axis corresponds to the cosine of the static contact angles of the flat reference; the y-values correspond to the cosine of the advancing, static and receding contact angles on the corresponding rough surface. The dashed line corresponds to the prediction by Wenzel when calculated with the roughness factors derived from white light profilometry (displayed in Table 6.1): b) SBG (sand-blasted glass), c) SLA (replica of sand-blasted, acidetched titanium), d) LLR (lotus leaf replica), e) GTM (replica of golf-tee shaped micropillars). 116
6.3 Results and Discussion Figure 6.3b-e show the data obtained on SBG, SLA, LLR and the GTM surfaces. The static and the advancing θ values for SBG (Figure 6.3b) display two linear trends, changing from slope 1 to slope 2.3 (corresponding to Wenzel roughness factor, Table 6.1) at 90°. The surface cannot be made superhydrophobic, but becomes superwetting when functionalized with a pure OH-terminated SAM. As long as the surface chemistry remains hydrophilic, the receding contact angle is zero, because the contact line is pinned to the surface. The static and advancing θ values on SLA replica (Figure 6.3c) show an extremely strong shift towards hydrophobicity compared to the θ predicted by the Wenzel equation. All data points in the lower right-hand quadrant correspond to hydrophilic surface chemistry, while the apparent contact angle on the SLA surface is hydrophobic. This phenomenon can be explained by pinning of the contact line. If the intrinsic surface energy is hydrophobic, the surface is capable of achieving contact angles near 150°, but remains essentially in a Wenzel-type wetting regime. The pinning strength is still sufficiently strong that the drop exhibits no roll-off, even though the receding θ is high. Similarly to SBG, SLA surfaces can also be rendered hemi-wicking when coated with OH-terminated thiols. The LLR (Figure 6.3d) shows a shift to hydrophobic contact angles due to pinning effects, which is greater than that of the SBG, but less than the case of the SLA. In contrast to SBG and SLA, the LLR functionalized with a pure CH3 -terminated SAM shows extremely high dynamic and static contact angles – an indication of Cassie-type wetting. The papillae are able to support the drop, even in the absence of the tubular wax of the original leaf. As soon as there is a hydrophilic contribution present in the SAM, Wenzel-type wetting occurs. Note the high standard deviations, which arise from the fact that in some measurements the Cassie-state still persisted while in others the drop penetrated into the structures. Both types of behavior were sometimes even observed on the same sample, which is an indication of the presence of metastable states. An explanation for this metastable Cassie-regime might be found in the naturally grown leaf topography, which is inhomogeneous in pillar density, top perimeter and height. On some areas of the leaf the papillae can support the drop, whereas on others they cannot. Again, the LLR shows superwetting when coated with a pure OH-terminated SAM. The GTM surface (Figure 6.3e) behaves somewhat differently from the others. Thanks to the golf-tee, undercut shape, Cassie-type wetting is moderately stable, even at higher surface energies, as predicted by Liu et al.79 Since the cross section of the pillar becomes smaller towards the bottom, the drop would need to form a larger liquid-air interface to follow the topography, which would be energetically unfavorable. Therefore the surface energy of the solid must be quite high to overcome this energy barrier. Tilting the sample also helps to overcome this energy barrier. The hysteresis of dynamic measurements on the GTM is quite high, compared to that 117
6 Roughness Influences on Wetting over a Wide Surface-Energy Range of the LLR with the pure hydrophobic coating, for example. This is commonly observed for micrometer-sized pillars72 in the absence of a second, smaller-scale structure. If the GTM surface is coated with a pure OH-terminated SAM, water undergoes spreading until it completely fills the volume between the pillars. At this point it forms a drop coexisting with a liquid film within the structure. Although the water film fully spreads over the whole patterned area, and is visible by eye, a contact angle of around 6° can still be measured (see Appendix 6.5.1). If static θf lat vs static θrough and advancing θf lat vs advancing θrough are presented in the same cosine-cosine plot, as shown in Figure 6.4a, differences between advancing and static measurements are cancelled out. The difference between the advancing and the static contact angles on the rough surface is equal to that on the flat surface. This means that it does not matter whether θs or θa are measured, since the effects observed are governed by the environment at the contact line and describe the same tendency. In the intermediate surface-energy region, additional effects such as air enclosure or capillary forces can be excluded. Thus the slope for all three non-periodical surfaces tested (SBG, SLA, LLR) is close to unity (see kCA , Table 6.1) and no significant effect on the slope, originating from surface roughness, can be distinguished. Roughness mainly influences the axis intercept, which is an indication of the magnitude of the energy barrier pinning the contact line. We propose the following empirical description to characterize non-periodical surface topographies in the intermediate hydrophilic regime (θY ranging approximately form 40° to 90°). θS is the contact angle measured on the rough surface, k ≈ 1 is the slope and dS the axis intercept found in the linear regression through the data (Equation 6.3). cos θS = k · cos θY − dS
The explanation of why SBG, the surface with the highest roughness factor, has the lowest hydrophobic shift, can be ascribed to the presence of conchoidal fracture areas on the surface, which present the energetically most favorable route for the contact line to move forward.41 As soon as a step forward is made, the energy barrier is overcome and the rest of the contact line will follow. This behavior resembles that found in dislocation movement in metals, in the special case of kink pairs (Seeger’s dislocation mechanism).70 Pinning of the contact line is therefore not very effective in this case. The LLR has no flat patches but a pillar-like structure, which presents fewer pinning sites than the dimpled SLA replica surface. It therefore seems that, for an understanding of roughness effects, the energy barrier is a more fruitful avenue to pursue than the effect of increased surface area. The receding contact angle remains close to zero, as long as the corresponding angle on the chemically equivalent flat surface remains below 90°. The detachment of the contact line is facilitated by a hydrophobic coating. The energy barrier governing this detachment is significantly lowered by the presence of 118
6.3 Results and Discussion
Figure 6.4: Water Contact angle data and the derivation of an empirical description. a) water contact angle data of the three non-periodical topographies. Here, the x-axis consists not only of the static, but also the dynamic values on the flat surface. White symbols correspond to receding, black to static and grey to advancing contact angles. SBG: M, LLR: , SLA: ; the circles with a cross are advancing contact angle data found by Tosatti et al.101 by applying mixed SAMs of OH- and CH3 -terminated alkylphosphates on SLA titanium surfaces. The dashed line has slope 1, b) empirical description to describe the mechanisms found in the intermediate hydrophilic regime where pinning effects are predominant.
air on Cassie-type wetted surfaces, leading to extremely low hysteresis. The advancing contact angle data of Tosatti et al.,101 measured on SLA titanium with SAMs of mixed OH- and CH3 -terminated alkylphosphate monolayers, coincide extremely well with the mixed thiol data taken on SLA replica in the present study. This demonstrates that the approach is valid, independent of the specific adsorbate-substrate system. White-light profilometry: The white-light profilometry data was used to calculate the actual surface area, in order to be able to define the roughness factor r. The roughness factor for the GTM surface was calculated by measuring the relevant lengths of a detached pillar in a SEM image (see Table 6.1) Additionally, “roughness factors” kCA from Figure 6.4a were extracted by applying a linear regression through the data found in the surface energy region corresponding to the contact angles from 40° to 90° on the flat surface. Contrary to Wenzel’s predictions, all tested surfaces show at least a small axis intercept. Additionally, slopes for SBG, SLA and LLR in the region corresponding to 40° to 90° on the flat surface lie between 1 and 1.2, i.e. far from the roughness values extracted from white-light profilometry data in Table 6.1. Due to the pillar surface’s strong and persistent Cassie-type wetting, the data show a transition from composite behavior directly to hemi-wicking, leaving little room for roughness-dependent pinning. 119
6 Roughness Influences on Wetting over a Wide Surface-Energy Range
Table 6.1: Roughness factors r for SBG (sand-blasted glass), SLA (replica of sandblasted, acid-etched titanium), LLR (lotus leaf replica), GTM (replica of golftee shaped micropillars) extracted from white-light profilometry (WLP) data and SEM image. kCA is the corresponding slope from the linear positions of Figure 6.4a. SBG
6.4 Conclusion Four different, heavily structured surfaces have been analyzed over a wide range of surface energies via water contact angle measurements. The data show three wetting regimes: If the surface energy is high, wettability is indeed enhanced by the surface roughness, causing hemi-wicking in many cases due to capillary forces. At lower surface energies, pinning of the contact line results in a shift to more hydrophobic θ values. It was found that the surface topography defines the pinning strength and with it the energy barrier counteracting the wetting behavior of the drop. An indication of this barrier is the axis intercept seen in the cosine-cosine plot (see Figure 6.4). This plot is surface-topography sensitive, and the behavior of all tested surfaces can be readily distinguished. The topographical influence on θ cannot simply be predicted via a roughness factor. With the exception of the hydrophobic data in the case of the SBG surface, none of the measured contact angles could have been predicted by the Wenzel equation. With regard to superhydrophobic surfaces, the golf-tee-shaped (GTM) pillars show stable superhydrophobicity over a wide range of surface energies. This topography seems to be a very effective design for micro-structured, superhydrophobic surfaces.
6.5 Appendix 6.5.1 A Side Remark: Calculating with the Equation by Cassie & Baxter The GTM Pillar surface offers a useful opportunity to test the equation developed by Cassie & Baxter. Hydrophobic: The strong energy barrier due to the undercut shape leads to an interesting consequence. Usually, f1 and f2 from the Cassie-Baxter equation are a function of surface energy, f1 increasing with increasing surface energy, because a greater proportion of the area of the surface features becomes wetted with increasing surface energy. 120
6.5 Appendix This is not the case on the GTM surface, where f1 corresponds to the pillar-top area. Therefore, the slope of a linear regression through the contact angle data can be extracted, which corresponds to f1 . This yields a value of 12 %., slightly higher than the 9 % evaluated from the SEM images. Hydrophilic: On the hydrophilic side, the case when the drop coexists with the film, the more generalised equation (Equation 6.4) from Cassie and Baxter can be used.44 cos θCB = f1 · cos θY 1 + f2 · cos θY 2
θY 1 is as before the contact angle of the liquid with the solid and θY 2 the contact angle of the liquid with itself, namely 0°. This yields the following Equation 6.5: cos θCB = f1 · cos θY 1 + f2
Entering the known data (f1 = 0.09, f2 = 1 - f2 , θY 1 = 20°) leads to a predicted contact angle of 6°, exactly what we measured. This shows that the Cassie-equation is also valid in the hydrophilic case, as was also shown by Bico et al.158
6.5.2 Evaluation of Dynamic Contact Angle Measurements Figure 6.5 shows still images taken from the exemplary movies that were published in the supporting information. They indicate situations which occurred while measuring contact angles on the rough surfaces and which are not so commonly observed on smooth surfaces. They accompany the comments made in the experimental section on contactangle evaluation.
6 Roughness Influences on Wetting over a Wide Surface-Energy Range
Figure 6.5: a.) and b.) LLR (lotus leaf replica) with a SAM containing 100 % dodecanethiol. a.) LLR_100_A4: advancing contact angle squeezed between syringe and substrate; b.) LLR_100_R2: receding contact angle, also squeezed between syringe and substrate; c.) and d.) SBG (sand-blasted glass slide) with a SAM containing 30 % dodecylthiol and 70 % mercaptoundecanol c.) SBG_30_A4: advancing contact angle showing stick-jump; d.) SBG_30_R2: receding contact angle so strongly pinned that it is equal to zero; e.) SBG with a SAM containing 70 % dodecylthiol and 30 % mercaptoundecanol. SBG_70_A3: advancing contact angle.
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient This work has been submitted to Langmuir and is currently under review. The supplementary material was either referenced to the corresponding chapters in the thesis or added into the appendix. Abstract: The superhydrophobicity of rough surfaces owes its existence to heterogeneous wetting. To investigate this phenomenon, density gradients of randomly placed holes and pillars have been fabricated by means of photolithography. On such surfaces, drops can be observed in the Cassie state over the full range of f1 (fraction of the area under the drop in contact with the solid). The gradient was produced with four different surface chemistries: native PDMS (polydimethyl siloxane), perfluorosilanized PDMS, epoxy and CH3 -terminated thiols on gold. It was found that f1 is the key parameter influencing the static water contact angle. Advancing and receding contact angles at any given position on the gradient are sensitive to the type of surface feature, hole or pillar, that is prevalent. In addition, roll-off angles have been measured and found to be influenced not only by the drop weight, but also suction events, edge pinning and f1 .
7.1 Introduction Since the superhydrophobic, self-cleaning properties of the lotus leaf2 were attributed to its rough surface structure, many different approaches have been developed to produce similar superhydrophobic surfaces artificially (see e.g. the review by Roach et al.8 ). While both natural and industrially produced superhydrophobic surfaces generally have a stochastic distribution of surface features, most investigations that have sought to determine which parameters lead to stable self-cleaning properties have been performed on micrometer-sized, photolithographically fabricated, periodically arranged pillars.18, 19, 47, 72 This approach does not take into account the possibility that the inherently strong anisotropy of periodic structures, or even the periodicity itself, may 123
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient have an effect on the wettability of a surface. It is therefore of interest to examine controlled, aperiodic surfaces, which are also better analogues for technically relevant systems. Gradients are very helpful tools159 for the investigation of a system’s sensitivity to specific parameters over a given range, since they constitute an inherently high-throughput approach. On surface gradients, one parameter is changed along the length of the sample, and therefore all other conditions, such as temperature, pressure or surface energy can be maintained constant while multiple measurements are made simultaneously over the parameter range selected. The parameter investigated in the present study was stochastically distributed pillar/hole density. This substrate allows models from the literature to be tested as to their validity on surfaces that resemble technically relevant systems. Generally, on a rough, somewhat hydrophobic substrate two different wetting states are possible: Either the drop completely conforms to the surface topography or it bridges from asperity to asperity, spanning the air gaps beneath. The first case can be denoted as the Wenzel-state,6 the second case the Cassie-state.7 Both states can be associated with very high contact angles, but only in the Cassie-state are very low roll-off angles in evidence, along with the concomitant self-cleaning effect. In this work, most drops described are in the Cassie-state, but only few have low roll-off angles (≤ 0°). On rough surfaces, as a consequence of contact-line pinning, there is not a single contact angle that describes wetting. In fact the contact angle can adopt any value between a minimum (receding contact angle, θr) and a maximum (advancing contact angle, θa) value. The difference between θa and θr is known as the hysteresis. Findings on regular pillar surfaces: Among other parameters, Barbieri et al.18 have investigated the influence of pitch distance (in hexagonal, square and honeycomb periodic patterns) and pillar-top perimeter on the static contact angle of water drops atop perfluorinated silicon pillars. They generally found that the drop resides in the Cassie-state much longer than would be predicted by calculating the thermodynamic transition to the Wenzel-state. Pattern parameters, such as symmetry, pitch distance and the pillar perimeter only influenced the stability of the drop near the transition point from Cassie- to Wenzel-state, hexagonal and long perimeters favouring the Cassie-state. Öner21 and Dorrer20 have shown that the advancing contact angle of a drop suspended on pillars in the Cassie-state for a given f1 (fraction of the area under the drop in contact with the solid) is independent of the shape or spacing of the pillars. Dorrer and Rühe20 proposed that the receding motion of the contact line is governed by a jumping motion from one post to another, starting at the outermost post that inhibits the drop in assuming its preferred spherical shape. By holding f1 constant, they showed that increasing the pillar-top dimensions reduces the receding contact angle (thus increasing hysteresis), whereas increasing of the pitch distance leads to an increase in the receding 124
7.1 Introduction contact angle and reduced hysteresis. Öner21 suggests that the contact-angle hysteresis on a surface of random roughness should be smaller than that on a regular surface with the same f1 due to the higher distortion of the contact line. Callies19 et. al. measured constant advancing contact angles on perfluorinated silicon pillars for f1 values from 1 % up to 25 % and additionally showed how, for low pillar densities, it is possible to have drops in both the Wenzel- and Cassie-state, depending on the way in which the drop was put on the surface. Several frequently used models for wetting have been tested with the acquired data. The models investigated were the Cassie-Baxter approximation,7 the Furmidge equation for roll-off angles64 and a proposition from Patankar156 for receding contact angles. Models tested on the pillar-density gradients: On an ideal surface, surface tensions γXY (where the subscripts X and Y refer to the phases at the corresponding X-Y interface: Solid, Liquid, or Air) in balance lead to one specific contact angle, the Young’s34 contact angle θY : cos θY = (γSA − γSL ) /γLA . However, most surfaces are not ideal and have a certain degree of roughness. Roughness generally increases the contact angle,22 unless surface energy is high enough to induce hemi-wicking.50 Wenzel6 introduced an area-roughness factor r to modify the Young’s equation, which takes into account the increased surface area and its influence on contact angle (θW ): cosθW = r · cos θY . Concerns about this equation have been expressed, both from a theoretical58 and experimental22, 55 standpoint, but Wenzel’s description of the conformal wetting state is a useful one. On certain surfaces the roughness is so profound that the surface tension of the liquid is sufficiently high to bridge from asperity to asperity and enclose air beneath the drop. Cassie and Baxter7 modified Wenzel’s equation to take the behaviour of this composite surface into account (see also Chapter 2.6): cos θCB = f1 · cos θY − f2
f1 is the roughness factor corresponding to the area wetted by the drop, f2 the fraction of the area below the drop that is in contact with air. Since both parameters are normalised by the analysed projected area, f1 + f2 ≥ 1. For small drops where gravity can be neglected and the Laplace pressure in the liquid is assumed to be constant,47 sagging of the drop can be neglected and the area fraction constituting f2 is assumed to be flat. When working with flat-top pillars, the area fraction of f1 can also be assumed to be flat and then f1 + f2 = 1 . There remains a controversy in the field55, 59 that can be summarised by the phrase “area vs. line”. The models of Wenzel and Cassie-Baxter are based on area considerations. If they are tested by free-energy calculations and free-energy barriers, they correctly predict the equilibrium contact angles for the noncomposite (Wenzel) and the composite (Cassie & Baxter) state.160 Nevertheless, the 125
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient shape of the drop is also determined by the interaction at the three-phase-contact line; sharp edges influence the contact angle via pinning phenomena that are not considered in the thermodynamic approach.55, 161 Surprisingly the Cassie-Baxter approximation has not prompted as many criticisms regarding its validity as the Wenzel equation, even though they are based on similar principles. In 1962 Furmidge64 was looking into the conditions necessary for drop retention on inclined surfaces and, by equating all the forces involved, defined a correlation between the inclination α of the surface with the drop weight m, drop diameter perpendicular to the sliding direction w, advancing and receding contact angles (θa, θr ) and the surface tension γLA of the liquid. m · g · sin α = w · γLA · (cos θr − cos θa)
Furmidge assumed the footprint of the drop to be rectangular during sliding, which may be a source of error, but the equation is sufficiently accurate to be able to account for sliding drops on a flat surface. The Cassie-Baxter model was designed to describe drops in thermodynamic equilibrium. Advancing and receding contact angles deviate from the equilibrium contact angles. Therefore Patankar156 deduced, via energy considerations, a condition that would describe the case of the receding drop if it would not leave a dry, but a wet surface behind. In this case, the θr should obey the following rule cos θrP = 2 · f1 − 1
In this work, a novel, rapid method has been developed to explore the influence of quasirandomly placed pillars and holes on the wetting of surfaces over a controlled range of f1 values. In order to investigate a large parameter range, hole-to-pillar density gradients have been prepared that span the range from isolated holes to isolated pillars, both of 2030 µm diameter. A systematic study of the wetting mechanisms and the effect of the real contact area on static and dynamic contact angles and roll-off angles has been carried out on hole-to-pillar-density gradients prepared with four different surface chemistries. In contrast to earlier systematic studies,10, 18, 19 this gradient approach covers the full range of f1 (0 – 100 %) and is, by the choice of the pillar distribution, a useful model for technically relevant surfaces.
7.2 Materials and Methods A variety of morphological gradients with identical topography and different surface chemistry was prepared, following the procedure summarized in Figure 7.1. 126
7.2 Materials and Methods
Photoshop standard photolithography SU-8 Si-wafer master greyscale gradient
2-step replica technique PDMS
water contact angle θ
mould/negative master PDMS or Epoxy replica/positive
surface coating Figure 7.1: Scheme of experimental setup: a) A pixel image of a greyscale gradient going from black to white was prepared and then converted into a black and white bitmap. Printing this bitmap on a transparency foil leads to a “pixel” mask for standard photolithography. A topographical gradient master was prepared by using standard photolithography of SU-8 on silicon. In a twostep replication process, copies of the master were prepared out of epoxy or PDMS. These substrates were functionalised with hydrophobic surface coatings. The wetting properties were analysed by measuring dynamic and static water contact angles and roll-off angles. Masks: The design of the mask is based on a 900 dpi bitmap with a random distribution of black and white pixels, which was printed on a foil mask. Photoshop CS (Version 8.0 for Macintosh) was used to create an 8-bit greyscale image. The image of the black-to-white gradient was modified with the linear gradient tool, which was set to span from rgb 0 to 255 for the whole range of pillar density (or 0 to 127 and 128 to 255 respectively). Afterwards the image was transformed to a binary image of black and white pixels by using the option “diffusion dither” in the image mode menu. The resulting image was used for the printing process. For technical reasons the whole range of pillar densities was split onto two samples, thus giving a higher accuracy for positioning the water drops. Each substrate was 42 mm long and 12 mm wide. The gradient spanned over 35 mm of the substrate, leaving 7 mm of flat area for handling of the substrate and measuring the reference contact angles (f1 = 100 %). Due to the presets of gamma in the colour workspace, the distribution of black and white 127
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient pixels along the gradient shows a distinct nonlinearity. A detailed analysis of the pixel distribution on the mask is shown in Chapter 7.5.1. The mask manufacturing (Fotosatz Salinger AG, Zurich) was carried out with a drum exposure system (3810 dpi resolution) on a polyester foil of 0.1 mm thickness. Masters: The photolithographic masks were used to prepare the masters for the replica technique. Standard photolithography was performed on 4” silicon wafers. After wafer cleaning, SU-8 2025 negative photoresist (MicroChem, USA) was spin-coated for 60 sec at 2000 rpm to a thickness of 30 µm and soft-baked on a hotplate for 2 min at 65 °C and 3 min at 95 °C. Subsequently, the wafer was exposed to UV light (MA6, Süss MicroTec, Germany) with constant intensity (total energy 180 mJ/cm2) through the photolithographic mask described above. After postbaking for 2 min at 65 °C and 3 min at 95 °C, the wafers were developed for 5 min in SU-8 developer (MicroChem, USA). Finally the wafer was hard baked at 190 °C for 10 min. Replicas: In order to facilitate the replica process, the masters were exposed to air plasma for 2 min and then functionalized with fluorosilanes (1H, 1H, 2H, 2H-Perfluorooctyltrichlorosilane, ABCR GmbH, Germany) by vapour phase deposition in a desiccator (rough vacuum) for one hour. Then, a mixture of polydimethylsiloxane (PDMS, 1 : 10 curing agent to base, SYLGARD® 184 silicone elastomer, base and curing agent, Dow Corning, USA) was cast over the master and allowed to cure overnight at 70 °C. The cured PDMS replica represented the negative of the master. After that, the PDMS negative replica was exposed to 2 min air plasma (RF Level high, 0.1 torr, PDC-32G, Harrick Plasma, USA) and functionalized with a layer of fluorosilanes. Again a 10 : 1 mixture of PDMS was cast and cured over night at 70 °C, to yield a positive replica of the master. Contact angles were measured on the bare PDMS positive and on its surface following subsequent functionalization with fluorosilanes. Epoxy positive replicas were produced as substrates for surface functionalization by selfassembled monolayers (SAMs) with thiol headgroups, similar to the procedure described in our previous publication.22 The PDMS negatives were used as moulds to cast an epoxy blend (EPO-TEK 302-3; Epoxy Technology, USA). The epoxy was mixed according to the protocol provided by the producer (4.5 g Part A with 10 g Part B), cast from the negative under vacuum, allowed to cure at 60 °C overnight and post-cured at 150 °C for 1 h. Later, these epoxy replicas were cleaned by ultrasonication for 10 min in a 2 v% solution of Hellmanex (Hellma, Germany) and subsequently rinsed five times with ultrapure water (resistance 18.2 MW, TKA-GenPure, Huber & Co. AG, Switzerland). Contact angle measurements were performed after this step. By resistance evaporation (MED 020 coating system, BALTEC, Liechtenstein) the samples were coated with a layer of 128
7.2 Materials and Methods 10 nm Cr and 50 nm Au (purity > 99.99 %, Umicore, Liechtenstein). During coating, the stage was rotated and tilted by 25°. Directly after coating, the samples were immersed in a 0.1 mM solution of dodecylthiol (Aldrich Chemicals, USA) in ethanol for 20 min, rinsed with ethanol and blown dry under a stream of nitrogen. Contact angles were measured on the freshly prepared samples. Contact Angle and Roll-off Angle Measurements: Static and dynamic contact angle measurements were performed on a Krüss DSA 100 (Krüss, Germany). Static contact angles (θs) were usually measured with drop volumes of 6 µl, or 9 µl on surfaces with contact angles above 140°. The drop was produced, still hanging on the syringe, and then the stage with the substrate was slowly lifted until the substrate touched the drop. Upon lowering the stage again, the drop detached and after it came to rest, an image was taken. Thus, the history of the contact line was a purely advancing motion. The drop volume had to be increased to 9 µl for contact angles above 140°, since otherwise the drop would not detach from the syringe. For dynamic contact angles (advancing (θa) and receding (θr)) measurements the drop volume was increased and decreased with a speed of 15 µl per minute. For the advancing drop, 2 movies with 200 frames and for the receding one movie with 250 frames were recorded. In cases where the drop was pinned on one side, only the moving side of the drop was taken into account for the evaluation.22 All drops were fitted with the tangent-method-2 routine, a 4th -order-polynomial function, of the DSA3 software (Krüss, Germany). This routine has difficulties in fitting drops in the Cassie regime, leading to a systematic underestimation of a few degrees. Therefore all drops with contact angles higher than 135° were fitted in ImageJ (http://rsbweb.nih.gov/ij/). Static contact angles were evaluated by the usage of the plug-in drop analysis122 and dynamic contact angles with the simple angle tool delivered with ImageJ. The reasons are more closely described in Chapter 4.4. Roll-off angles were measured on a home-built device. It consists of a stage that is fixed on one side to a spindle and a goniometer that indicates the angle of the stage. The substrate is placed on the stage and a drop of 6 – 9 µl is placed onto the sample surface. The stage is tilted until the drop starts to move and the roll-off angle is recorded. Scanning Electron Microscopy: The epoxy substrates were analyzed in a Zeiss Gemini 1530 FEG SEM at 3 to 5 kV, at room temperature, gold coated as described above. Extraction of f1, f2: The grey-scale SEM images were imported into the ImageJ program and then transformed into a black-and-white-bitmap. These bitmaps can be analysed with the “analyze particles” tool, yielding pillar top area and perimeter for each pillar on the image. Summing this pillar-top area, then dividing the sum by the total analysed (projected) area 129
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient results in a value of f1 . For f2 it is assumed that f1 + f2 = 1, since the pillar tops are flat, exhibiting no significant roughness. The analysed areas were 1.88 x 1.27 mm2 , except for the positions at 51, 56, 61 and 66 mm (corresponding to f2 values of 80.2, 85.9, 88.2 and 94.4 %), where the pillar density fell below 1000 pillars per analysed area. Below this pillar density the contrast in grey values was too low for the program to distinguish between pillar and background. Therefore the contrast had to be enhanced by hand by colouring the pillar tops white and the analysed area was reduced to 0.39 x 0.26 mm2 .
7.3 Results and Discussion By the use of standard photolithography, the original bitmap from the Adobe Photoshop program was transferred into a morphological gradient by adding the third dimension (height) to the 2D pattern (see Figure 7.2). By using a negative photoresist, such as SU-8, all white pixels are cross-linked while the black pixels are etched away. This led to a morphology containing hole features on the “white” side of the gradient, gradually merging to larger holes, until islands were isolated, ending up as single pillars on the “black” side of the gradient. The same parameters for the photolithographic step were used as in our former publication,22 leading to all features being flat on top and slightly undercut, similar to golf-tees. The undercut shape helps to support the drop in the Cassie state77 and ensures that the drop only wets the tops of the pillars. By choosing different materials (PDMS and epoxy) and surface coatings (perfluorinated silanes and CH3 -terminated thiols on gold/epoxy) the surface chemistry of the substrates was varied. In Figure 7.3a the distribution of the black pixels (black dots) along the gradient extracted from the bitmap is compared to the f2 measured on SEM images on the positive (epoxy) replica (grey dots). The spots chosen for the SEM analysis were the same spots as used for the contact-angle measurements. The dotted line is a guide to the eye to show a gradient of black and white pixels having an entirely linear increase of black pixels along the distance. As already mentioned in the experimental section, the black pixels do not increase linearly along the gradient due to presets of the colour workspace. The slight deviations between the bitmap and the topography occur during the printing and photolithography steps of the master fabrication. In the middle range (20 – 45 mm), where f2 is higher than the bitmap data, colour bleeding in the printing process led to larger black areas on the mask. At positions 0 to 20 mm, f2 is actually lower than expected because of the cross-linking of the SU-8 epoxy. On an ideal substrate the corners of diagonally neighboured pixels would only touch, but on our master, the SU-8 actually forms narrow bridges across the corners. At the other end of the gradient (positions 130
7.3 Results and Discussion 45 to 70 mm) the larger printing of the black pixels compensates the bridges formed between the single, neighbouring white pixels (f1 ).
Figure 7.2: SEM image composite of the gold-coated epoxy replica, containing 16 images that have been sampled every 5 mm along the gradient. The original whiteto-black gradient stretched from the top and white end to the bottom and black end. The top end of the gradient contains isolated holes, whereas the bottom part is dominated by isolated pillars. In the middle part of the gradient, the pixels grow together to yield more complex structures. 131
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient Figure 7.3b is a correlation graphic. It shows f2 extracted from the SEM analysis plotted versus f2 calculated from the contact-angle data, solving the Cassie-Baxter approximation (Equation 7.1) for f2 : f2 =
cos θY − cos θCB cos θY + 1
This calculation was performed for all available surface chemistries and the obtained data entered into the graph. Since, in principle, the same parameter (f2 ) is plotted on both axes, the data should exactly follow the bisector. The data shows some scattering around the bisector but does not display a significant deviation that would imply another (or no) correlation. At f2 values exceeding 70 %, the scattering pattern changes slightly, which we suspect to be due to gravitational influences in the drop (see Chapter 7.5.2). Influence of f1 / f2 on contact angle: When plotting static contact angle data versus distance on the replicate substrates (Figure 7.3c), the data clearly shows the same trend as depicted in the statistical analysis of the distribution of the initial black pixels (and f2 ) versus distance (Figure 7.3a). The shoulder in the black and white distribution occurs at the same position in the contactangle data. This is an indication that on our quasi-random substrates, f1 and f2 are indeed the significant parameters influencing the contact angle of drops in the Cassiestate. Since the static measurement is closest to thermodynamic equilibrium, it is justifiable to compare the data with the predictions calculated according to the Cassie-Baxter approximation (Equation 7.1). In the first part of the gradient in Figure 7.3c, with f2 going from 0 to 70 % (equal to the distance of 0 to 41 mm) θs shows a good correlation to the Cassie-Baxter approximation. At the black end, where f2 approaches 100 %, θs drops down to the angle measured on a flat sample with the same surface chemistry. Here, the distance between the pillars gets so large that the drop cannot span from one pillar top to the next and wets in the Wenzel-state. Just before this transition at f2 values between 70 and ∼90 % (positions between 41 and ∼61 mm), a deviation from the Cassie-Baxter prediction occurs, which may be due to gravitational and discretisation effects (see 7.5.2).
7.3 Results and Discussion
Figure 7.3: Statistical analysis and static contact angles. a.) f2 distribution along the gradient. The black circles show the data of a black pixel analysis on the initial gradient bitmap, the grey circles show the extracted f2 from the SEM analysis on the epoxy replica and the dotted line represents a linear increase of f2 along the substrate. b.) Correlation between f2 measured on the epoxy replica and f2 calculated from the static contact angle data by solving the Cassie-Baxter equation. c.) The static water contact angle measurements on all four substrates and the corresponding Cassie-Baxter approximation (dashed lines): perfluorinated PDMS, gold and dodecylthiol coated epoxy, native PDMS, native epoxy. “line versus area” The question arises as to how the parameters f1 / f2 , which are area fractions, can be the main parameters influencing the contact angle in the Cassie-state after it was shown55, 56 that the contact angle of a drop is only defined by what is in the vicinity of the contact line and not by what is to be found underneath the drop. The explanation is quite simple: Since the topographical features are small compared to the base diameter of the drop, the value of f1 determined as an area fraction is equivalent to the fraction of the contact line in contact with the solid. In this case, f1 is actually a line parameter, not an area parameter. The question as to where the contact line truly 133
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient contacts the pillars is irrelevant for this consideration, since the pillar tops are small compared to the footprint of the drop and deviations from the ideal circular footprint are negligible. Dynamic-contact-angle data:
b.) 1.0 advancing receding CB adv / rec
cosine (Water Contact Angle)
cosine (Water Contact Angle)
-1.0 0 20 holes
40 60 f2 [%]
0 20 holes
40 60 f2 [%]
advancing receding CB adv / rec Patankar
1.0 cosine (Water Contact Angle)
c.) cosine (Water Contact Angle)
advancing receding CB adv / rec
-1.0 0 20 holes
40 60 f2 [%]
advancing receding CB adv / rec
0 20 holes
40 60 f2 [%]
Figure 7.4: Comparison of dynamic-contact-angle data on different surface chemistries with the prediction by Cassie & Baxter (two lines: advancing slope always greater than receding slope) and Patankar. a) perfluorinated PDMS surface, b) gold and CH3. -terminated SAM on epoxy substrate, c.) native PDMS surface, d.) native epoxy surface. Symbols: open symbols = advancing; full symbols = receding; dashed grey lines = corresponding Cassie predictions; dotted grey line = Patankar prediction (only for the graph of the native PDMS since its shape is the same for all other chemistries). The alternating dashed black lines are linear fits to the data. 134
7.3 Results and Discussion Figure 7.4 shows the cosine of the dynamic-contact-angle data measured on all four substrates versus f2 (fluorinated PDMS, PDMS; CH3 -terminated thiol and the pure epoxy surface). Additionally, the Cassie-Baxter predictions for the dynamic case were calculated, using the f2 extracted from the SEM images. The condition for receding contact angles as proposed by Patankar (Equation 7.3) was also computed. Since Equation 7.3 is only dependent on f1 and not on surface chemistry, it looks the same for all data sets. In order to keep Figure 7.4 readable, it is only added to the graph of the native PDMS, because there the strongest correlation between Patankar’s prediction and the receding contact angle can be seen. Similarly to the static contact angles, the advancing and receding contact angles show a distinct dependence on f2 . A detailed analysis shows that the gradients can be split into three different regimes (alternating dashed black lines in Figure 7.4): In the middle part of the gradients, advancing as well as receding contact angles rise with increasing amount of air enclosure. Receding angles show a tendency to level off for low values of f2 , whereas for advancing angles, a plateau is visible for large f2 . Looking at the advancing contact angles, they rise with increasing hole density. The rise can be approximated by a linear trend line, and in the case of the PDMS (Figure 7.4c) and the fluorinated PDMS (Figure 7.4a) even follows the Cassie-Baxter prediction exactly. As the features increasingly resemble single pillars, the θa levels off at about 160° (150° for the more hydrophilic epoxy substrate) and no change is detectable with decreasing pillar density. The phenomenon of levelling off at high advancing contact angles (around 160°) has also been reported by others,10, 19 working with periodically distributed micro-meter sized pillars (4x4 symmetry). In this work, stagnation starts between 60 and 80 % of air, and not at the same location for all substrates, even though they are replicates of the same master. Clearly, surface energy plays a role in the onset of stagnation. Additionally, since the sequence (CH2 -term. thiols, fluorinated PDMS, PDMS, Epoxy) does not strictly follow surface energy, other properties of the material can apparently play a role. For example, PDMS, an elastomer, has a much lower modulus than epoxy, which is a thermoset. The low modulus of PDMS is also the reason why the advancing contact angle on the flat part of PDMS is larger than on the perfluorinated PDMS: the y-component of the liquid-air surface tension can not be fully compensated and thus forms a rim around the base perimeter of the drop. This rim acts like surface roughness and slightly increases the contact angle115, 117 (see Chapter 4.3). The receding contact angles mirrored this behaviour to some extent. On the holey side they first stagnate or even decrease, again until the holes become larger and the features resemble pillars, at which point the θr begins to increase. The rise starts at around 10 % of air for the consistently hydrophobic substrates (CH3 -term. Thiols and perfluorinated PDMS, Figure 7.4a and b) and at 50 % of air for the more hydrophilic substrates (PDMS and epoxy, Figure 7.4c and d). PDMS is known to change the orientation of its me135
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient thyl groups upon contact with water, making it more hydrophilic in this situation.90, 114 This change of orientation and the aforementioned rim formation are the reason for the high hysteresis even on flat PDMS; besides the roughness of the rim, the advancing contact line encounters a more hydrophobic surface than the receding contact line. The Cassie-Baxter equation generally greatly overestimates θr and the Patankar derivation strongly underestimates the measured θr of the hydrophobic substrates and clearly overestimates the θr of the epoxy surface. Only in the case of PDMS does the θr rise follow the Patankar prediction from 45 to 80 % of air. This phenomenon is not yet a proof that a water film is actually left behind (the Patankar assumption), since other factors such as reorientation could be playing a role.
Priest et al.10 showed in their investigations on substrates consisting only of holes or of pillars that the contact angle hysteresis actually indicates on what type of feature the drop sits. Stagnation of the advancing and linear increasing of the receding contact angle with an increasing amount of air enclosure is a characteristic of the pillar surface. A linear increase of the advancing (a rather close agreement with the Cassie-Baxter approximation) and stagnation or even decrease of the receding contact angle is found on holey surfaces. On our gradient surface where holes slowly merge together and pillars are formed, this behaviour is also observed. However, although all four substrates exhibit the same topography, the transition point, at which the behaviour changes from hole-contact to pillar contact cannot be found at the same f2 value (see Figure 7.4, intercepts of the linear regressions (alternating dashed lines)). Most probably, between air percentages of 20 to 70 % the effects of topography intermingle and surface energy starts to play a more important role.
In Figure 7.5, the cosine of the contact-angle hysteresis (∆ cos θ = cos θr − cos θa) versus f2 is shown. With increasing air content the hysteresis increases for all substrates up to about ∼50 % air. Beyond this point, the hysteresis decreases rapidly until it reaches its minimum, before the collapse into the Wenzel-state (indicated by a sharp increase in hysteresis at large f2 values). For the two consistently hydrophobic substrates (CH3 term. thiols and the fluorinated PDMS) a plateau in hysteresis can be distinguished between 20 and 50 % of air. Nevertheless all curves show a minimum at around 45 % of air. Dividing the data sets into two parts (0 – 45 % and 45 – ∼100 % for all drops in Cassie-state); the first part is predominantly governed by the mechanism originating from the hole-structure, the second part is mostly governed by the presence of pillars. Each part can be fitted by a linear curve fit. According to Priest et al.10 the slope in this linear fit is a measure for the pinning energy Epin and equal to Epin /γLA (see Equation 7.4): 136
7.3 Results and Discussion
Table 7.1: Slopes of the linear fit in Figure 7.5 (dashed lines) are compared to the values found by Priest et al. on homogeneous substrates. Epin /γLA
Priest et al.10
CH3 -term. thiols
∆ cos θ = ∆ cos θ0 + fD ·
fD corresponds to the area fraction of the “defect”. A surface consisting of holes can be considered as a matrix of substrate with very low-energy defect patches (air enclosures). Therefore, fD corresponds to f2 on the holey side of the substrate. Similarly, a pillar surface consists of a matrix out of air with high-energy defects (pillars; even a hydrophobic surface has a higher interaction with the liquid than air). Thus, on the pillar side, fD corresponds to f1 .
The slopes found on the substrates used in this work clearly show the same tendency: pinning energy of the pillars is significantly higher than the pinning energy of the holes (Table 7.1). On the holey surface the contact line is mostly in contact with the solid. In this way the energy barriers to adopting the contact angle corresponding to the lowest free energy state are lowered, because of increased flexibility in the positioning of the contact line on the substrate. By being able to meander between the defects, the contact line does not have to follow the shape of the defects exactly, as it does on the pillar side, but can actually average over the defects on the surface, as was assumed in the CassieBaxter approximation. This averaging might also be the reason why the Cassie-Baxter approximation fits the data of the θa better on the holey side than on the pillar side. But, as mentioned above, the difference between the values of Epin /γLA (Table 7.1, holes vs. pillars) is not quite as profound as in the work of Priest et al., most probably due to the fact that both features are present in the middle of the gradient. The differences in Epin /γLA between the present work (stochastic) and that of Priest et al (periodic) for similar features and surface chemistries could also be an effect of the greater contactline distortion in the stochastic case, as was suggested by Öner.21 137
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient
Figure 7.5: Contact-angle hysteresis (cosθr-cosθa) as a function of f2 ( f2 = 1 − f1 ). The curves have been separated for clarity. The slopes of the linear regressions are a measure for the pinning energy.10 (circles = perfluorinated PDMS; triangles = thiols on gold/epoxy; squares = native PDMS; pyramids = epoxy).
Despite the strong hysteresis, all drops evaluated were still in the Cassie-state. In Chapter 7.5.3, examples of light microscopy images taken through a water drop are presented and clearly show the air enclosure over the full gradient. Despite its lower hydrophobicity, the epoxy surface is clearly in the Cassie-state, since it shows qualitatively the same behaviour as observed on the more hydrophobic substrates, which would not be the case if it were in the Wenzel-state.97 Figure 7.6 shows the roll-off angles measured on the four different substrates together with the approximation calculated by the Furmidge equation (see Equation 7.2, dashed lines). All parameters occurring in the Furmidge equation were measured independently from the roll-off angle (drop weight m, drop diameter w, advancing and receding contact angles θa, θr). Positions where no roll-off angles occurred are not displayed in that graph. For the pure epoxy surface no roll-off angle at any position could be observed for the given drop volumes. The native PDMS does not show any roll-off on the flat area due to the high hysteresis induced by the reorientation of the polymer side chains upon contact with water. For the two consistently hydrophobic substrates (perfluorinated PDMS and CH3 -terminated thiols on gold), on the holey side of the gradient the roll-off angle increases with increasing hole density, until no movement is observable even up to 90°. Movement is only possible again when the topography changes from holes to pillars. There, the sine of the roll-off decreases linearly with f2 (R2 of the linear regression > 97 %). 138
7.3 Results and Discussion
Figure 7.6: Sine of the roll-off angles of all analysed surfaces. Only data points are shown where actually a roll-off angle could be measured. All positions where the drop did not move up until a tilt angle of 90° are not displayed. Drops on the pure epoxy surface did not roll off at any angle. The dotted line shows the prediction by the Furmidge equation for each surface.
As observed in Figure 7.5, the holey side shows a lower pinning energy than the pillar side. But in Figure 7.6 the drop rolls off far more easily on the pillar side. One reason for this is that the contact line has far more contact (low f2 ) on the holey side than on the pillar side. Therefore the inherent hysteresis of the material plays a bigger role on the adhesion of the drop. At first this adhesion is increased by the introduction of pinning defects such as holes. Additionally, there is a range on the holey side of the gradient where the drop could not roll off at all. Two reasons may explain this phenomenon: 1. the receding contact line was strongly pinned, possibly aided by the finite curvature of the edges, at the front end of the holes;10 2. due to the suction induced by the sealed air cushions162, 163 the drop was held on the surface. We therefore conclude that there are at least three different influences beyond f2 that determine whether a drop can roll off a surface consisting of holes (sealed air cushions): One, the size and with it the weight of the drop defines the driving energy for downward motion (as seen by Reyssat65 et al.), two, pinning at the front edge of the drop and three, the magnitude of the suction effect. On the pillar side the drop started moving at an air content of 56.2 % (CH3 -term. thiols and fluorinated PDMS) and for the PDMS at 71.4 %. The onset of drop movement for the consistently hydrophobic substrates coincides with the position chosen to divide the data for the linear regressions in Figure 7.5, probably indicating a change in the prevalent pinning mechanism. The sine of the roll-off angle of all three substrates increases linearly with f1 (R2 ≥ 97.2 %). The Furmidge equation is able to predict the general behaviour of the roll-off angles. It is an equation consisting only of measured values. Since no distinct deviation of the measured values can be observed, it can be concluded that 139
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient all necessary parameters to describe the behaviour were identified. The suction or pinning events mentioned above also influence the receding contact angle and are therefore taken into account in the equation.
7.4 Conclusions: A novel, rapid method for creating photolithographic masks has proven invaluable for the fabrication of gradients of 3D structures. Pseudo-random hole–to-pillar gradients have been shown to be a useful tool for investigating the influence of structural effects on a variety of different wetting phenomena. By the use of a gradient surface consisting of nearly randomly placed pillars slowly agglomerating into holes, different wetting mechanisms of drops in the Cassie-state could be identified. Static contact angles increase more or less linearly with f1 , clearly indicating the importance of f1 for the wetting behaviour of a drop in the Cassie-state. Since the topographical structures here are small compared to the drop diameter, the originally area parameter f1 can also be considered as a line parameter influencing the contact line. When measuring advancing and receding contact angles, analysing the hysteresis can distinguish whether holes or pillars are the predominant topographical feature. Dynamic measurements have shown that pinning energy is higher on pillar than on holey surfaces. Drops of 6 µl can be pinned so strongly that they do not roll off at all. Several causes were identified: pinning at the front edge of the drop and suction events of sealed air cushions.
7.5 Appendix 7.5.1 Mask Generation and Characterisation There is a distinct nonlinearity in the Photoshop-based gradient in the correlation between the rgb values and the resulting number of white pixels. This is caused by the presets of gamma in the colour workspace. A homogeneous patch of rgb 128, which is transformed into a black-and-white image in the same way as the gradient, shows 36.3 % white pixels. The black and white bitmap shows distinct texture effects, which are most prominent at intermediate f2 values. For a detailed analysis, the bitmap was imported to Matlab (Version 7.6), and analysed with a homebuilt routine counting black and white pixels in particular orientations. In Figure 7.7, the probability of a black pixel having neighbours of the same colour is shown. The probability is similar for the pixel beside and the pixel 140
7.5 Appendix down and to the right, whereas the pixel up and to the right is overrepresented, and the pixel below is underrepresented. This feature of the screening routine in Photoshop helps to avoid clustering of black pixels, which would produce an unpleasant visual effect in images.
30 40 50 position [mm]
p (next neighbour)
The printing process used for the mask manufacturing is less precise than the production of chromium masks, which are used in state-of-the-art photolithography. The size distribution of the pixels shows three maxima, caused by a mismatch of the 900 dpi bitmap and the 3810 dpi resolution of the printer. The vast majority of printed 900 dpi patches contains 3x4, 4x4 and 4x5 printing pixels. The edges of the printed areas are rough, and the corners are rounded. Sometimes an overlap of corners occurs. These effects, although complicating a quantitative analysis of the patterns, are not expected to significantly influence on a systematic parameter study of superhydrophobic surfaces. Barbieri et. al18 have shown that the pillar perimeter only plays a role when f1 really is at the boundary to the transition to the Wenzel-state. It is well known31, 33, 72 that anisotropically patterned surfaces influence dynamic and roll-off angles. But the small anisotropy found on our surfaces is not homogeneous enough to be seen in the macroscopic contact angle and roll-off angle measurements.
Figure 7.7: Statistical analysis of black pixels in the gradient bitmap used for mask printing. The black line shows the normalized number of black pixels occurring within a 44x66 pixel patch along the gradient, corresponding to an averaged f2 underneath a drop. The grey curves show the probability for a black pixel to be neighboured by another black pixel in a particular direction (after considerable smoothing, binomial, 50, Igor pro 6.0). 141
7 Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient
7.5.2 Influence of Gravitation and Discretisation Generally it is assumed35 that as long as the base diameter of a drop is smaller than twice the capillary length κ−1 , gravitational effects can be neglected. The drop shape then adopts the shape of a spherical cap (truncated sphere). The volumes used in this publication are far below this limit (κ−1 = 2.7 mm). But even by eye it is clear that the drop silhouette is not spherical, but distorted on the sides belonging to the lower part of the drop. If it was justifiable to assume that the silhouette resembles a sphere, then the drop outline should follow the dashed circle in Figure 7.8. Indeed this raises the question as to whether the criterion of the capillary length is valid for superhydrophobic surfaces. On structured surfaces that show very high contact angles, the contact line is pinned at the outermost circle of pillars it can still reach. What if the weight of the drop in the torus outside of the drop (see vertical, dashed lines in Figure 7.8) is significant and actually causes the drop to wet more surface area than it would be expected from Laplace pressure considerations? To date we are not aware of a systematic investigation of drop size on the superhydrophobic contact angle. At extremely high contact angles, the effect of discretisation, namely the fact that the drop is in contact with only a small number of pillars, also needs to be taken into account, and can lead to some deviations from the values predicted by the Cassie-Baxter approximation. The precise number of wetted pillars may also be influenced by the mode of deposition of the drop, which necessarily becomes increasingly forceful as drops are placed onto extremely hydrophobic surfaces from relatively hydrophilic syringe tips.
Figure 7.8: Image of a 9 µl drop on the perfluorinated PDMS replica at the position with f2 = 94.4 %. Drop base radius is 0.62 mm (corresponding approximately to 113 pillars), well below the limit given by the capillary length κ−1 of 2.7 mm. The dashed circle shows the deviation of the drop silhouette from an ideal sphere.
7.5.3 Light Microscopy of Air Enclosure In Figure 7.9a, a light microscopy image of a dry perfluorinated PDMS sample is shown. The images were taken on the same spots as the contact angle measurements were 142
7.5 Appendix made. Figure 7.9b shows images from the very same gradient, but with a drop of water between the substrate and the lens. Dark areas correspond to wetted areas (f1 ), bright areas to air cushions (f2 ). This demonstrates the Cassie-state was present along the entire gradient.
Figure 7.9: Light microscopy images of the perfluorinated PDMS substrate, on all positions showing drops in the Cassie-state; going from the left and holey side to the right and pillar side. a.) in air; b.) through a drop of water. Bright grey corresponds to air, dark grey to wetted area.
8 Conclusions Wetting has been a research topic for at least two centuries, started in 1805 by Thomas Young’s essay on the cohesion of fluids.34 Many theories explaining wetting phenomena have been found; however, little is known about how surface roughness causes contact angles to depart from equilibrium theories, such as the Young’s equation. This work increased the level of knowledge in this area by means systematic studies on very rough surfaces and over a wide range of surface chemistries. The outcome is relevant to any investigations made with contact-angle measurements on rough surfaces, but especially also to the pursuit of mechanically and chemically stable self-cleaning surfaces. Since self-cleaning surfaces are per definition very rough surfaces, Chapter 6 shows that if for example the chemistry of a self-cleaning surface is altered by oxidation or contamination and transforms to Wenzel-state wetting, it turns into a very sticky surface with large hysteresis and thus poor cleaning abilities. Chapter 7 emphasises the fact that for a drop to roll-off to remove dirt, it is necessary to have free flowing air underneath the drop, otherwise the drop remains stuck to the surface. Two main investigations were performed throughout this thesis. In one study (Chapter 6), mainly Wenzel-type wetting was observed over a wide range of surface energies on three very rough substrates. A special case of Cassie-type wetting was found on golf-teeshaped micro-pillars. In the other study (Chapter 7), Cassie-type wetting was analysed over a wide range of f1 and with a quasi-random placing of the pillars. The Wenzel equation (Equation 2.12) was shown to fail to predict contact angles, especially in the hydrophilic regime, before hemi-wicking can occur. The Cassie-Baxter equation (Equation 2.13) however was found to be quite accurate, the area fractions f1 and f2 being parameters strongly influencing the contact angle. At first, these two findings might be surprising, since the two equations are based on the same theoretical principles; the Cassie-Baxter equation being an adaptation of the Wenzel equation to composite wetting. They are both purely based on energetic considerations and do not consider contact-line pinning and hysteresis. This lack of accounting for the pinning is most probably the reason for the difference in the predictive power for the two equations. While in the Wenzel-state the contact line can be pinned all around the perimeter, this is not the case for the Cassie-state. Everywhere where the drop is touching air (f2 ), pinning is reduced to zero, and thus the pinning strength of the surface decreases with decreasing f1 . Pinning on the flat top features in Chapter 7 is small compared to the 145
8 Conclusions surface roughness of the surfaces used in Chapter 6. Thus, applied to data acquired on flat top features, the Wenzel equation is much more affected by the neglect of pinning events than the Cassie-Baxter equation. The Cassie-Baxter equation may show a stronger misfit if the feature tops on the gradient were covered with a profound roughness, touching the liquid in the Wenzel-state. There, pinning events would cause the contact angles to deviate from the prediction. This fact leads to the conclusion that another condition for the validity of the Wenzel and the Cassie-Baxter equation has to be added to the known conditions that the roughness features need to be small compared to the drop and homogeneously distributed over the whole surface. The surface features also need to be such that contact-line pinning is weak, especially for the Wenzel-state wetting. One of the main results in Chapter 6 was that the contact angles in the slightly hydrophilic regime of the cosine-cosine plot (Figure 6.4) followed the unit slope, but were shifted by an axis intercept. Thus, a new equation was suggested to be valid in this regime (Equation 6.3): cos θS = k · cos θY − dS The axis intercept dS was assigned to be an indication of the magnitude of the energy barrier pinning the contact line. This parameter dS probably includes a roughness factor comprising the Gibbs criterion and edge radii, which probably has to be defined in a scale-dependent way, and take all components into account for multiscale surfaces. The other main result was that undercut shapes (GTM pillars) could also support drops in the Cassie-state, even if the surface energy was hydrophilic. This surface structure has proven to be of great value for studying drops in the Cassie-state (Chapter 7, Tuteja et al.,77, 164 Choi et al.76 ). Most probably, efficient technical self-cleaning surfaces, consist of two or more scales of re-entrant surface features, to prevent external pressure from inducing a transition into the Wenzel-state. In Chapter 7, it was shown that the parameter f1 from the Cassie-Baxter equation is a key parameter to define wetting in the Cassie-state. Additionally, also the advancing and receding contact angles measured on the gradient in f1 were analysed. It was shown that both the advancing and receding contact angles are sensitive to what kind of surface feature the drop rests on. By means of the schematic drawing in Figure 8.1, the results and conclusions are summarised. Figure 8.1 shows the advancing and receding contact angles found on a gradient surface with increasing f2 (air portion). The plot is equivalent to the plots shown in Chapter 7, Figure 7.Y. The surface features change from holes to pillars along the x-axis (same substrate as in Chapter 7). 146
The advancing movement of a drop on a composite surface is more like rolling than sliding.48 Thus, the advancing front is not affected by pinning, since it is always a “fresh” liquid-air area touching the surface. Therefore, the contact line senses the increasing amount of air (f2 ) and follows the Cassie-Baxter prediction on the holey side of the gradient (region A in Figure 8.1). There might be a threshold value in f1 , above which the amount of defects in the air layer (number of pillars) has no influence anymore and the contact angle stays constant as if it were on a constant surface chemistry (region B in Figure 8.1). It could also be assumed that this is the region where gravity starts to play a role and deforms the drop. By definition a drop sitting on its own vapour, a so-called Leidenfrost drop (see footnote Chapter 1), has a contact angle of 180° with its own vapour. In the side view of such a drop, one measures a contact angle of ~162° (see Figure 3 in the publication of Quéré and Reyssat73 ). Thus, in addition to the optical difficulties in determining the contact angle of such a drop, gravity can also mask the “maximum” contact angle. The receding contact angle remains constant (region C in Figure 8.1), as long as the drop sits on a holey structure. There, the contact line is continuous and finds its way around the pinning sites (holes). Thus the pinning capability of the hole is actually very low, the receding contact angle being close to the receding contact angle found on the flat surface. If it deviates nonetheless it is most probably due to a slight penetration of water into the holes due to blunt edges. Once the contact line is broken (pillar-side, region D) the contact line has to conform to the pillars and gets pinned at their edges. Pinning is reduced by decreasing the number of pillars present (increasing f2 ), thus the receding contact angle is less pinned and increases until the drop achieves the high mobility known on superhydrophobic surfaces. In conclusion, the findings in this thesis have helped to shed more light on partial wetting on rough surfaces. The results specify more clearly under which conditions the energy-based Wenzel and Cassie-Baxter equations are valid. The surface-feature sensitivity of advancing and receding contact angles has been emphasised and probable influences identified.
rec D A adv B f2 Figure 8.1: Schematic drawing of the advancing and receding contact angle on a hole-topillar-substrate with increasing f2 . Region A – “sensing f1 ”: the advancing contact angle is mainly influenced by the increasing amount of air (f2 ); region B – “constant chemistry - air”: on the pillar surface the contact line can avoid the pinning sites (pillars) and behaves as if it was on a constant surface chemistry, namely air; region C “constant chemistry - substrate”: the receding contact line finds its way around the defects (holes) and thus shows more or less the same value like on the homogeneous part; region D – “sensing f1 and pinning”: the receding contact line has to conform to the pinning cites (pillars) and increases with decreasing number of those.
9 Outlook and Future Directions For technical self-cleaning surfaces, the ultimate goal would be to find a universal coating that is easily applicable to numerous surfaces. It should be transparent, so it can be used on window panes or shiny surfaces such as car coatings, and mechanically stable so it is insensitive to abrasion. Chemical stability under exposure to UV-irradiation (sunlight) should be as high as possible. The production, use and the disposal of the coating should be non-toxic. Each of the conditions mentioned for this “perfect” coating triggers research. The coating would need a carrier material which easily spreads on many different surfaces without dewetting. Transparency requires the surface features to be small; but small features are usually much more easily damaged than large features. Thus, there is a need to find small surface features (below 200 nm) which would survive e.g. the force of an impacting bug on the windscreen of a car. For outdoor applications the stability of the hydrophobic surface chemistry is of utmost importance. If this prerequisite can not be achieved by chemistry alone, systems that contain a certain “self-healing” mechanism need to be conceived. The solution could be a polymer that also comprises hydrophobic, low-molecular weight species that diffuse to the surface over time, constantly renewing the hydrophobic surface chemistry. The most hydrophobic flat surfaces involve perfluorinated surface chemistry. This surface chemistry is also used for self-cleaning surfaces. Perfluorinated surfaces are normally safe to use and are not toxic. However, wherever highly toxic fluorine-containing gases or fluoride ions are involved, as can happen in the production or disposal of perfluorinated compounds, many precautions have to be taken. Thus, fluorine-free selfcleaning surfaces should definitely be preferred over those containing fluorine. Academic research can additionally provide industry with good models that facilitate the design of effective self-cleaning surfaces. Models are based on systematic studies, such as those performed in this thesis. Chapter 9.1 is a feasibility study for a new kind of model surface. These surfaces consist of two or more size scales of pillars with well-known dimensions and can serve to determine the role of two or more scales on self-cleaning properties of surfaces. Parameters 149
9 Outlook and Future Directions such as pillar diameter, pillar-top perimeter, pillar height and pitch distance can be tested with contact angle measurements and the data used to test Cassie-state stability criteria and results from data modeling. Owing to the precise knowledge of surfacefeature dimensions, equations with predictive power for contact-angle hysteresis might also be found, as was done for single structures by Forsberg et al.165 Chapter 9.2 presents two methods that could complement the current methods to visualise the base of a drop in Cassie-state. On surfaces not consisting of flat-top pillars the quantities of f1 and f2 are unknown, and thus methods to visualise and measure these area fractions under different conditions will increase our understanding of wetting of droplets in the Cassie-state. Chapter 9.3 introduces an approach to achieving a smooth gradient of adhesion. Such a gradient might be used as a template to generate a gradient of small objects, which adhere differently at different points along the gradient. Chapter 9.4 emphasises the potential of the novel mask design approach used in Chapter 7. Printing on polymeric foil allows for a quick and inexpensive testing of structures. By designing patterns with custom-made mathematical algorithms, full control over the surface features is gained and complex surface structures can be made. Chapter 9.5 takes the investigations on Cassie-state wetting to the next level by exchanging the air with another solvent and adding temperature and pressure to the system. Such investigations are of interest in the oil recovery industry, for example. Chapter 9.6 introduces raspberry-like silica particles and outlines which issues need to be overcome to bring them into an industrial application.
9.1 The Influence of Single Versus Double Structures on the Stability of the Self-cleaning Effect In the master’s project of C. Cremmel138 it was investigated how two-scale structures influence the stability of the self-cleaning effect towards the hydrophilic regime. In this sense it was a continuation of the work presented in Chapter 6.
9.1.1 Introduction Nature’s most stable and efficient self-cleaning surfaces exhibit a double-structured topography. Different factors were identified as responsible for destabilising the Cassiestate: capillary waves, nanodroplet condensation, hydrophilic spots due to chemical surface inhomogeneities and liquid pressure.166 These factors have different characteristic length scales, thus also hierarchical roughness is needed to resist them. However, 150
9.1 Self-cleaning Effect: Single Versus Double Structures in most approaches, the second scale is applied by a bottom-up technique, such as etching29, 167 or coating with a fibre-like polymer.168 Thus they do not deliver a known value for f1 . Therefore it was attempted to find a substrate which has two controlled levels of roughness. Undercut shapes such as the GTM pillars used in Chapter 6 exhibit a strong and stable Cassie-state wetting towards the mostly wetting regime, even as single structures.22, 77, 78 Exemplarily, two structures will be compared in the subsequent text to show that with this approach interesting phenomena can be observed. It opens the door to a new, large parameter space in which to investigate rough surfaces. The two structures are a singlelevel and a double-level GTM pillar surface having nearly the same f1 (7 %). Contact angles (static and dynamic) were measured on these two substrates and a flat reference. The samples were prepared such that a wide range of surface energy could be investigated. The data are presented in the cosine-cosine plots introduced in Chapter 6 and the results compared to the GTM surface already shown in Chapter 6. The different stages (hemi-wicking, Wenzel-state, pinning, Cassie-state, Figure 6.3) will be analysed.
9.1.2 Experimental The experimental parameters are precisely described in the reports of C. Cremmel.138, 147 Here, only a summary will be given to introduce the set-up. The approach was to create masters by standard photolithogaphy which exhibit two levels of golf-tee-shaped micro-pillars (GTM pillars) (see Table 9.1, first row). Masks: Two photolithographic masks were designed, one with patterns for large pillars and one with patterns for small pillars. The pillar diameters for the large pillars varied between 20 and 500 µm, and for the small pillars from 2 to 15 µm. All patterns were hexagonally dense-packed in order to have only one parameter for the distance between the pillars (pitch distance). Fields with patterns of different pitch distances p and top diameters d were designed to achieve different area fractions of f1 . The parameters p and d of the two substrates used here are shown in Table 9.1. Photolithography: First, a thick layer of SU-8 was spin-coated, then illuminated through the mask for the large pillars and then soft-baked. After that, a second, thinner layer of SU-8 was spin-coated, illuminated through the second mask and soft-baked. The conditions were chosen such that the second layer of pillars could not cross-link down to the silicon, but only to the level of the large pillars, so that with development of the structure all excess small pillars were washed away. Top view images of the two structures on the master (SU-8 pillars on silicon wafer) are shown in the bottom row of Table 9.1. The precise preparation conditions are given in Cremmel’s report.147 PDMS Replicas: Replicas made of PDMS were made as described in Chapter 7.2. 151
9 Outlook and Future Directions
Table 9.1: Illustration of the substrates used for the experiments. Single GTMs
Double GTMs D
10 µm p
P = 70, D = 20
P = 150, D = 100
Small Pillars [µm]
p = 23, d = 10
Top View (LM) Surface energy variation: In order to cover a wider range of surface energy, the hydrophobic recovery property of PDMS90 was used. The two structured substrates and a replica of a flat silicon wafer (reference) were plasma treated to render the surfaces hydrophilic. Then, the substrates were repeatedly measured over time. The flat reference reached contact angles above 90° after 850 h. Contact angle measurements were performed as described in Chapter 7.
9.1.3 Results and Discussion Figure 9.1 shows the contact angle data measured on the single and double GTM surfaces during the hydrophobic recovery (Figure 9.1a and b, f1 = 7 %) and, for comparison, the data measured on the mixed thiols on gold/epoxy system from Chapter 6 (Figure 9.1c f1 = 9 %). The plots correspond to the plots shown in Figure 6.3 in Chapter 6. The x-axis is the cosine of static contact angle on the flat surface and indicates the change in surface energy. The y-axis shows the cosine of the static (black symbol), advancing (grey symbol) and receding (open symbol) contact angle measured on the rough surface. Since the hydrophobic recovery of the PDMS is slow, many measurements can be made over time and thus the change in contact angle due to the change in surface energy can be closely observed. However, it is not quite clear whether the recovery speed is the same on the flat and on the rough surface. At the point of the arrow in Figure 9.1a, θY is nearly constant for four measurements, but the contact angle data on the GTM structured surfaces increases strongly. Thus it is not very conclusive whether here a 152
9.1 Self-cleaning Effect: Single Versus Double Structures very small change in surface energy caused a large change in contact angles on the rough surface, or if the hydrophobicity on the flat and the rough surface recovered at a different rate. Comparison between the PDMS and the mixed-thiols-on-gold/epoxy system (see Figure 9.1): The trends observed on all pillar surfaces are the same; at high surface energy (top right-hand side of the plots), hemi-wicking occurs. Upon decreasing surface energy a transition begins, where first static and advancing contact angles increase, but the receding remains zero. Then, on the PDMS substrates (Figure 9.1a and b) a region can be observed where the slope of the increasing advancing and receding contact angles changes (dashed circles). This region is not visible on the mixed-thiols-on-gold/epoxy system, but this might be due to the fact that on the thiol-coated substrates this particular range of surface energies was not tested. (region of x-values between 0.2 and 0.5, corresponding to contact angles on the flat substrate of 60° to 80°).
contact angle θ 60° 0° 180° 120° 90°
contact angle θ 0° 180° 120° 90° 60°
contact angle θ 180° 120° 90° 60°
cos θrough 0 0.5 -0.5
120° 90° 60° contact angle θ
single chapt. 6
0 0.5 cos θYstatic
0 0.5 cos θYstatic
0 0.5 cos θYstatic
Figure 9.1: The cosine of the static (black symbols), advancing (grey symbols) and receding (open symbols) contact angle of the rough surface is plotted against the cosine of the static contact angle on the flat surface. a.) single GTMs, hexagonal, PDMS; b.) double GTMs, hexagonal, PDMS, c.) single GTMs, square pattern, mixed thiols on gold/epoxy, figure taken from Figure 6.3e, Chapter 6.
Then, quite an abrupt jump into the Cassie-state occurs, corresponding to high contact angles above 150°. On the PDMS samples it seems that the hemi-wicking state is stable for lower surface energies than on the mixed thiols on gold/epoxy system. At this point, it is not clear whether this is an effect of a different f1 (2 % higher for the mixed-thiolson-gold/epoxy system), a difference between the patterns (hexagonal versus square) or another effect, such as a topography-dependant recovery rate. Surprisingly the receding contact angle on the mixed-thiols-on-gold/epoxy system rather resembles the receding 153
9 Outlook and Future Directions angles measured on the PDMS double GTMs than to the single GTMs. At the current state of investigation it cannot be excluded that the data points on the mixed-thiols-ongold/epoxy system in the range of x-values between 0.2 and 0.5 would show a different trend, comparable to the PDMS single GTMs. Comparison between the single and the double GTMs. The onset of leaving the hemiwicking state, indicated by a measurable static and advancing contact angle, is at slightly higher surface-energy values (higher value of cos θY ) for the single than for the double structure. Thus, hemi-wicking state persists longer on double GTMs than on the single GTMs. Hemi-wicking on the double GTMs occurs up to a contact angle corresponding to 50° on the flat surface. Considering the transition condition from Equation 2.15, Chapter 2.6 (cos θCY = (1 − f1 ) / (r − f1 )), such a behaviour is expected, since the roughness factor r is larger for the double GTMs than for the single GTMs. However, θCY was not yet determined. Once the surface energy of the sample has left the range where it can induce hemiwicking , the contact angles on the double GTMs increase more steeply than on the single GTMs, and cross the x-axis at a much higher surface energy than the single GTM. Then, both surfaces show a surface-energy region in which the contact angles only moderately increase (indicated by circles in Figure 9.1a and b). This region is more closely examined in Figure 9.2a. Both GTM surfaces reach the Cassie-state at around the same surface energy. Thus, probably surfaces with the same f1 and the same edge conditions are similarly stable towards the hydrophilic regime. However, whether the edge conditions (edge curvatures of the single and double GTMs) indeed are the same and whether it really is only f1 defining the surface energy at which the transition from the Wenzel- to the Cassie-state occurs has to be examined more closely. Table 9.2: Slope k and axis intercept dS from the linear regressions in Figure 9.2a (dashed lines). Single GTM
1.0 ± 0.1
1.1 ± 0.2
−0.15 ± 0.02
−0.36 ± 0.04
The region within the dashed circles is particularly interesting. In Figure 9.2a the cosine of the static and the advancing contact angle of the rough surfaces are plotted against the cosine of the static and advancing contact angle of the flat surface. This plot is equivalent to the plot in Figure 6.4 in Chapter 6, where the empirical equation (Equation 6.3; cos θS = k · cos θY − dS ) was found. Also in the data presented here, a specific region can be found where the slope changes to a value of approximate unity. The linear regression in this data region shows that the slope k is indeed close to unity and the two 154
9.1 Self-cleaning Effect: Single Versus Double Structures
0 1 0.5 cos θYflat-static & adv.
120° 90° 60° contact angle θ
cos θrough-rec 0 0.5 -0.5 -1
120° 90° 60° contact angle θ
cos θrough-static & adv. 0 1 0.5 -0.5 -1
contact angle θ 180° 120° 90° 60°
contact angle θ 180° 120° 90° 60°
0 0.5 cos θflat-rec
Figure 9.2: a.) cosine-cosine plot of static and advancing contact angles of the rough versus the flat surface (same type of plot as in Figure 6.4a). the dashed lines are linear regressions. Grey is the double GTMs and dark grey the single GTMs. b.) cosine – cosine plot of the receding data from the single and the double GTMs. surfaces differ only in the axis intercept dS (see Table 9.2). The double GTMs have a larger axis intercept dS , which corresponds to a higher pinning strength of the surface. This may indicate that both levels of GTMs, the small and the large pillars, remain in the Wenzel-state. The new finding of the linear relation following equation 6.3 is another experimental demonstration that important information can be gained from this kind of experiment, and that further investigations of the origin of the parameter k and dS are worthwhile. Figure 9.2b shows the cosine of the receding contact angle of the rough surfaces versus the cosine of the receding contact angle on the flat surface. The single GTMs are represented by dark grey symbols and the double GTMs by light grey symbols. Two details are striking. First, the single structure shows an increase in receding contact angle before the transition into the Cassie-state, which does not happen for the double GTMs, on which the receding angle remains zero until the transition. The reason for this different behaviour is still unclear. Second, the transition on the two surfaces with f1 of 7 % happens at the same surface energy as was observed for the advancing and static contact angle.
9.1.4 Conclusions and Outlook The data acquired during the hydrophobic recovery of the PDMS replicas of single and double GTM pillar surfaces show clearly that this is an approach which provides insight 155
9 Outlook and Future Directions in phenomena guiding wetting on multiscale-structured surfaces. In order to check that the hydrophobic recovery on the PDMS replicas was at the same rate for the rough and the flat surface, at least one set of experiments with another substrate-surface-chemistry system should be performed. The choice should be a system where the surface chemistry and with it the surface energy of the substrate can be closely controlled. An issue for the mixed-thiols-on-gold/epoxy system may be that the gold which is deposited by physical vapour deposition cannot be coated homogeneously between the second layer of the double GTMs. Other options should be evaluated, where no directional coating is involved. Two more surfaces with an f1 of 7 % should be examined: One surface consisting only of small GTM pillars (height 10 µm, diameter 10 µm) and a second surface of double GTMs, but with a very different combination of pitch distances and pillar diameters. In this way, size effects might be distinguished from effects solely originating from the second structure on top of the pillars. The linear region found in Figure 9.2a shows that the pinning region can also be observed by photolithographic model surfaces. By having clearly characterised surface features, it is possible to shed more light on the parameter dS . Another step would be to also have small pillars in-between the large pillars. This can be achieved by adding another step during the photolithographic preparation of the GTM pillars. First, a thin layer of SU-8 has to be spin-coated and illuminated through the mask for the small pillars. Then a thicker layer for the large pillars can be applied and be topped by another thin layer for the small pillars. To have the whole surface covered with pillars will probably increase the stability of the Cassie-state for impacting drops.
9.2 Visualisation of the Drop Base in the Cassie-state Many models and considerations of the stability of the Cassie-state involve assumptions about the contact angle of the drop at the side of the features, and its impact on the shape of f1 and f2 in general. Due to the Laplace pressure it is assumed that the liquidair interface f2 is flat underneath the drop. The quantity of f1 is usually unknown for two-scaled structures or structures with non-flat tops. On substrates with more than one scale of roughness, it is not always clear whether all levels of roughness are in the Cassie-state. Therefore it would be interesting to find an approach to make the drop base of a drop in the Cassie-state accessible for direct analysis. Priest et al.10 observed the composite interface with an optical microscope through the water phase. Ensikat et al.169 performed a cryo-SEM study on water-glycerol drops on a superhydrophobic leaf, freezing the drops and transferring them to the SEM prior 156
9.2 Visualisation of the Drop Base in the Cassie-state to imaging. They showed that indeed the liquid-air interface is not sagging between two papillae, but shows a certain waviness caused by the irregular relief of the leaf surface. This method also allowed an estimation of the shape and dimension of the liquid-substrate area fraction (f1 ). However, since the analysis was done on uncoated non-conducting ice-drops, electric charging by the electron beam induced diffuse dark and bright areas which were not due to topographical effects and made the interpretation of the SEM images challenging. In literature, a number of numerical simulations have been carried out, e.g. by the surface evolver165, 170 or FEM (finite element method) analysis.76 Thus, there is a need to gain experimental data to verify the outcome of the simulations. Choi et al analysed non-volatile liquids and curable polymers on micro-hoodoo surfaces in SEM to directly monitor the triple-phase contact line. SEM analysis and optical microscopy only provide 2D-information, and therefore these methods could be complemented by two further possible approaches: (1) gellation of the water droplet would render the drop stable enough to detach from the surface and examine its base by profilometry (WLP) or AFM, (2) confocal fluorescence analysis could give an in-situ view of the composite interface.
Gellation of the water droplet: The reliable determination of the three-phase contact angle on particles has been a challenge for many years. One of the many methods evolved is the gel trapping technique (GTT), which is based on spreading of the particles at the air-water interface with subsequent gelletion of the water phase with a nonadsorbing polysaccharide.171 This method could also be used to solidify a water drop (or a water layer) on a superhydrophobic surface and, after removal from the substrate, render the drop base analysable to techniques giving 3D-information of the surface (AFM, WLP). This approach has not been tested, yet, for this application. Thus many aspects would need to be determined: (1) whether it is possible to remove the gellated drop from the surface without breaking it; (2) whether the gellation proves to make the water a good replica material, thus if surface features and the air-water interface are well replicated. Most probably, the gellated drop would dry out quickly, thus it could serve as a mould to cast a PDMS replica. The PDMS replica would remain unchanged during analysis in ambient conditions.
Confocal fluorescence microscopy: A first attempt was made with confocal fluorescence microscopy of the air-water interface of the drop (see Figure 9.3). To perform the experiment, a fraction of a dried lotus leaf was taken and a drop containing a small amount of fluorescein was placed on the 157
9 Outlook and Future Directions lotus leaf sample. The wax of the lotus leaf has a self-fluorescence which leads to the blue colour in Figure 9.3. The water phase appears red due to the fluorescein and the air phase appears black. The objective used was a water-submersible objective. A stack of 30 images was taken, the focal point of each image 1 µm lower in z-direction than from the previous image. The fluorescence of the wax and the fluorescein were activated with the same laser. The reconstruction of the side-view shows that in principle it is possible to view the liquid-air interface. However, the liquid-air interface also shows a trace of the fluorescence of the lotus wax. This may be an artefact due to a reflection of the fluorescent light at the air-water interface. By creating a surface with a coating activated with a different laser than the fluorescence in the water phase, this artefact could probably be subtracted from the analysis. Thus, it may be worthwhile to explore this approach further with a photolithographic model surface, which has a coating or consists of a material which has a fluorescence activation different to the one in the water phase. With the well-known surface parameters on a model surface the technique could be validated and then also be used on more complex shapes; e.g. on surfaces consisting of round shapes such as particles.
Figure 9.3: Confocal fluorescence microscopy of the liquid-air interface measured through a water drop containing fluorescein on a dried piece of a lotus leaf.
9.3 Smooth Gradient of Adhesion In Chapter 4.3 it was observed that that the receding contact angle measured on native PDMS may be influenced by the surface energy of the mould material it is cured against. This assumption can be tested by letting PDMS cure against a wettability gradient as described by Morgenthaler et al.172 Such wettability gradients are produced by slowly immersing a gold-coated silicon wafer into a strongly diluted ethanolic solution of CH3 terminated alkane thiols and subsequent backfilling with and OH-terminated alkane 158
9.4 Gradient Design by Means of Mathematical Algorithms thiols. By this approach, the surface roughness is constant over the full length of the gradient (equivalent to the roughness of the polycrystalline gold layer) and only the surface tension of the substrate varies. It would be expected that the advancing contact angle on the PDMS replica stays constant over the whole length of the substrate (∼118°), but the receding contact angle increases with increasing hydrophobicity of the mould (wettability gradient). If this behaviour is indeed observed it may be interesting to further examine the mechanisms causing it. Since the static and advancing contact angles are expected to stay constant, this kind of gradient could be employed as an analysis tool where the detachment of drops from the surface is systematically varied. For example, it is likely that a stream of a diluted particle suspension flowing perpendicularly to the PDMS adhesion gradient would result in a gradient of particles adhering to the surface. On the side where the hysteresis is large (low receding contact angle), the rate of particle attachment is expected to be higher than on the other side.
9.4 Gradient Design by Means of Mathematical Algorithms The mask design presented in Chapter 7 offers a quick and rather cost-effective possibility for testing a wide variety of patterns. This design approach is not limited to the use of graphic tools such as the Adobe Photoshop program. Patterns can for example also be programmed with custom-designed mathematical algorithms in a program such as Matlab (The MathWorks, USA). The advantage of choosing Matlab is greater control by designing patterns from scratch, all algorithms being precisely known, which is not the case in the commercial program Photoshop. It was recognised that at certain grey-scale values the black-to-white gradients prepared in Photoshop showed preferential directions in the distribution of the pixels (see Chapter 7.5.1). Thus, the placing of the pillars was only quasi-random, since clustering was inhibited by a built-in algorithm in the program Photoshop. Therefore, a black-to-white gradient was designed in Matlab with a truly random placing of pillars on the preset grid. By this approach a linear increase of black pixels along the gradient was achieved. The pixel distribution was free of preferential directions. After the preparation of the photolithographic masters, these gradients were replicated with PDMS investigated in the master project of Cathrein Hückstädt.173 No distinct differences were observed on these gradients or the Photoshop gradients except in the regime of very low values of f1 , Wenzel-state occurred at slightly lower f1 values than on the Photoshop gradients. This was assigned due to the fact that in this system, clustering of pillars was allowed 159
9 Outlook and Future Directions which led to considerably larger distances between the pillars than on the Photoshop gradients. Many more different patterns are conceivable with this method. The challenge is rather to decide which patterns make sense to investigate. One approach may be to design a gradient of adhesion as shown in Figure 9.4. In this type of gradient, single dots (pillars) become increasingly connected along the gradient, thus holes are formed once all four sides of a square are closed. f1 does not remain constant, but it varies a lot less than on the pillar-to-hole gradients used in Chapter 7. Thus, investigations would focus more on the feature change than on the change in f1 . Conclusions on what hole size or how many holes need to be present to prevent the drop from rolling off may be found.
Figure 9.4: Gradient designed in by custom-made mathematical algorithms in Matlab. The single pillars (dots) on the left side get more and more connected along the gradient, until it becomes a grid. In principle this is a different way to create a pillar-to-hole gradient. f1 will not be entirely constant over the length of the gradient, but the change will be much smaller than for the gradients in Chapter 7. Thus, the effect of the change from pillar to hole surface can be almost separated from the influence of f1 . Thus, the static contact angle may remain more or less constant, but the adhesion of the drop will change along the gradient.
9.5 Cassie-state in Oil and Under Pressure The development of ceramic replicas of photolithographic structures allows moving on to more exotic conditions to explore Cassie-state wetting. Ceramic substrates are resistant to heat and pressure. Perfluorinated silanes bind covalently to alumina replicas, and thus also the chemical functionalisation could be stable up to temperatures of 200 °C. By replacing air with an alkane phase or even with pressure-liquefied CO2 it could be investigated whether Cassie-state is also possible under these conditions. By tuning the density of the two liquid phases, e.g. of the alkane and the water, the effect of gravity 160
9.6 Raspberry-like Particles – A Move Into Applications? can be excluded. Thus, investigations on the Cassie-Baxter equation can be performed with drops undisturbed by gravity. The fit of the Cassie-Baxter equation may be even better under these conditions than under ambient conditions. Such conditions (water in oil, relatively high temperatures and pressures) are found in the oil recovery industry. Investigations under such conditions may help to find ways how to get more oil out of oil reserves.
9.6 Raspberry-like Particles – A Move Into Applications? Ming et al.26 presented a route to prepare raspberry-like silica particles which can be used to prepare self-cleaning surfaces. Two sizes of silica particles were prepared by the Stöber process.174 The smaller particles were bound to the larger particles by an amineepoxy linkage. In a master project Yannick Santschi175 managed to find experimental conditions to successfully prepare such particles in our lab (SEM image shown in Figure 9.5b). A glass slide was coated with an epoxy glue and the raspberry-like particles were strewn over the polymer. After curing the polymer, the substrate was subjected to a 30-second-air plasma and the particle-epoxy compound was functionalised with perfluorinated silanes in a desiccator. Despite the rather crude preparation of the final substrate the surface showed a strong self-cleaning effect, as can be seen in Figure 9.5a. Raspberry-like particles offer the possibility to create superhydrophobic surfaces with a hierarchical structure (small particles on larger ones) and re-entrant surface features (spheres). Since they are made of silica, they are also mechanically very stable, and the re-entrant structures will most probably provide Cassie-state wetting even with slightly hydrophilic surface chemistry. The use of spheres reduces the amount of sharp edges at which contact line pinning could occur. Thus, these are preconditions which promise a very stable self-cleaning effect. If these particles are to be used in industry, first an application has to be defined. The particles are rather large (about 1 µm) and will not provide a smooth and shiny surface. Outdoor applications such as a self-cleaning coating for high voltage insulators or in paints could be an option. Then approaches have to be sought to fix the particles onto a surface in an easy and cost-effective way. Thereby, two main issues have to be solved: at which step are the particles to be rendered hydrophobic and how are the particles to be fixed on the outermost layer of the surface. If these steps can be overcome, then a product containing raspberry-like particles has the potential to become a commercial success.
9 Outlook and Future Directions
Figure 9.5: a.) Microscope slide coated with a layer of epoxy polymer and perfluorinated raspberry-like silica particles. Through a syringe from the top left corner of the image, a jet of water is aimed onto the surface. Owing to the superhydrophobicity of the surface the water jet is reflected. Spherical water droplets from earlier water-spraying reside on the substrate. The large drop to the right is not in Cassie-state. b.) SEM image of the raspberry-like particles. On one of the large particles the coating with the small particles was not successful.
Bibliography  http://www.buddhanet.net/e learning/history/b_lotus.htm. The symbol of the lotus, 2010.  W. Barthlott and C. Neinhuis. Purity of the sacred lotus, or escape from contamination in biological surfaces. Planta, 202(1):1–8, 1997.  K. Koch, B. Bhushan, and W. Barthlott. Diversity of structure, morphology and wetting of plant surfaces. Soft Matter, 4(10):1943–1963, 2008.  K. Koch, C. Neinhuis, H. J. Ensikat, and W. Barthlott. Self assembly of epicuticular waxes on living plant surfaces imaged by atomic force microscopy (afm). Journal Of Experimental Botany, 55(397):711–718, 2004.  Y. Yu, Z.-H. Zhao, and Q.-S. Zheng. Mechanical and superhydrophobic stabilities of two-scale surfacial structure of lotus leaves. Langmuir, 23(15):8212–8216, 2007.  R. N. Wenzel. Resistance of solid surfaces to wetting by water. Industrial And Engineering Chemistry, 28:988–994, 1936.  A. B. D. Cassie and S. Baxter. Wettability of porous surfaces. Transactions Of The Faraday Society, 40:0546–0550, 1944.  P. Roach, N. J. Shirtcliffe, and M. I. Newton. Progress in superhydrophobic surface development. Soft Matter, 4:224–240, 2008.  D. M. Spori, T. Drobek, S. Zürcher, and N. D. Spencer. Cassie-state wetting investigated by means of a hole-to-pillar density gradient. Langmuir, submitted.  C. Priest, T.W. J. Albrecht, R. Sedev, and J. Ralston. Asymmetric wetting hysteresis on hydrophobic microstructured surfaces. Langmuir, 25(10):5655–5660, 2009.  M. H. Jin, X. J. Feng, L. Feng, T. L. Sun, J. Zhai, T. J. Li, and L. Jiang. Superhydrophobic aligned polystyrene nanotube films with high adhesive force. Advanced Materials, 17(16):1977–1981, 2005.  Y. Lai, X. Gao, H. Zhuang, J. Huang, C. Lin, and L. Jiang. Designing superhydrophobic porous nanostructures with tunable water adhesion. Advanced Materials, 21(37):3799–3803, 2009. 163
Bibliography  X. Hong, X. Gao, and L. Jiang. Application of superhydrophobic surface with high adhesive force in no lost transport of superparamagnetic microdroplet. Journal of the American Chemical Society, 129(6):1478–1479, 2007.  X. J. Feng and L. Jiang. Design and creation of superwetting/antiwetting surfaces. Advanced Materials, 18(23):3063–3078, 2006.  J. Genzer and K. Efimenko. Recent developments in superhydrophobic surfaces and their relevance to marine fouling: a review. Biofouling, 22(5):339–360, 2006.  X. M. Li, D. Reinhoudt, and M. Crego-Calama. What do we need for a superhydrophobic surface? a review on the recent progress in the preparation of superhydrophobic surfaces. Chemical Society Reviews, 36(8):1350–1368, 2007.  T. L. Sun, L. Feng, X. F. Gao, and L. Jiang. Bioinspired surfaces with special wettability. Accounts of Chemical Research, 38(8):644–652, 2005.  L. Barbieri, E. Wagner, and P. Hoffmann. Water wetting transition parameters of perfluorinated substrates with periodically distributed flat-top microscale obstacles. Langmuir, 23(4):1723–1734, 2007.  M. Callies, Y. Chen, F. Marty, A. Pepin, and David Quéré. Microfabricated textured surfaces for super-hydrophobicity investigations. Microelectronic Engineering, 78-79:100–105, 2005.  C. Dorrer and J. Rühe. Advancing and receding motion of droplets on ultrahydrophobic post surfaces. Langmuir, 22(18):7652–7657, 2006.  D. Öner and T. McCarthy. Ultrahydrophobic surfaces. effects of topography length scales on wettability. Langmuir, 16(20):7777–7782, 2000.  D. M. Spori, T. Drobek, S. Zürcher, M. Ochsner, C. Sprecher, A. Mühlebach, and N. D. Spencer. Beyond the lotus effect: Roughness influences on wetting over a wide surface-energy range. Langmuir, 24(10):5411–5417, 2008.  J. Zimmermann, F. A. Reifler, U. Schrade, G. R. J. Artus, and S. Seeger. Long term environmental durability of a superhydrophobic silicone nanofilament coating. Colloids and Surfaces a-Physicochemical and Engineering Aspects, 302(13):234–240, 2007.  G. R. J. Artus, S. Jung, J. Zimmermann, H. P. Gautschi, K. Marquardt, and S. Seeger. Silicone nanofilaments and their application as superhydrophobic coating. Advanced Materials, 18(20):2758–2762, 2006.  P.-S. Tsai, Y.-M. Yang, and Y.-L. Lee. Fabrication of hydrophobic surfaces by coupling of langmuir-blodgett deposition and a self-assembled monolayer. Langmuir, 22(13):5660–5665, 2006.  W. Ming, D. Wu, R. van Benthem, and G. de With. Superhydrophobic films from raspberry-like particles. Nano Letters, 5(11):2298–2301, 2005. 164
Bibliography  H. J. Tsai and Y. L. Lee. Facile method to fabricate raspberry-like particulate films for superhydrophobic surfaces. Langmuir, 23(25):12687–12692, 2007.  C. Sun, L. Q. Ge, and Z. Z. Gu. Fabrication of super-hydrophobic film with dualsize roughness by silica sphere assembly. Thin Solid Films, 515(11):4686–4690, 2007.  Y. Kwon, N. Patankar, J. Choi, and J. Lee. Design of surface hierarchy for extreme hydrophobicity. Langmuir, 25(11):6129–6136, 2009.  H. Liu, J. Zhai, and L. Jiang. Wetting and anti-wetting on aligned carbon nanotube films. Soft Matter, 2:811–821, 2006.  A. Shastry, M. J. Case, and K. F. Bohringer. Directing droplets using microstructured surfaces. Langmuir, 22(14):6161–6167, 2006.  M. Reyssat, F. Pardo, and D. Quéré. Drops onto gradients of texture. Europhysics Letters, 87:36003, 2009.  C. J. Long, J. F. Schumacher, and A. B. Brennan. Potential for tunable static and dynamic contact angle anisotropy on gradient microscale patterned topographies. Langmuir, 25(22):12982–12989, 2009.  T. Young. An essay on the cohesion of fluids. Philosophical Transactions of the Royal Society of London, 95:65–87, 1805.  P. G. de Gennes, D. Quéré, and F. Brochart-Wyart. Capillarity and Wetting Phenomena: Drops, Bubbles, Pearls, Waves. Springer Science+Business Media, Inc., New York, 2004. ISBN 0-387-00592-7.  KrüssGmbH. Installation and operation, 2004.  F. M. Fowkes. Attractive forces at interfaces. Industrial And Engineering Chemistry, 56:40–52, 1964.  A. W. Adamson and A. P. Gast. Physical Chemistry of Surfaces. John Wiley & Sons, Inc., New York, sixth edition edition, 1997. ISBN 0-471-14873-3.  J. N. Israelachvili. Intermolecular and Surface Forces. Elsevier Academic Press, London, second edition edition, 1991. ISBN 0-120375181-0.  F. Bashforth and J. C. Adams. An attempt to test the theories of capillary action by comparing the theoretical and measured forms of drops of fluid. Cambridge University Press, Cambridge, 1883.  P. G. de Gennes. Wetting - statics and dynamics. Reviews of Modern Physics, 57(3):827–863, 1985.  J. Drelich, J. D. Miller, A. Kumar, and G. M. Whitesides. Wetting characteristics of liquid-drops at heterogeneous surfaces. Colloids and Surfaces a-Physicochemical and Engineering Aspects, 93:1–13, 1994. 165
Bibliography  C. Priest, R. Sedev, and J. Ralston. Asymmetric wetting hysteresis on chemical defects. Physical Review Letters, 99(2):026103–4, 2007.  A. B. D. Cassie. Contact angles. Discussions of the Faraday Society, 3:11–16, 1948.  J. N. Israelachvili and M. L. Gee. Contact angles on chemically heterogeneous surfaces. Langmuir, 5:288–289, 1989.  M. V. Panchagnula and S. Vedantam. Comment on how wenzel and cassie were wrong by gao and mccarthy. Langmuir, 23(26):13242–13242, 2007.  D. Quéré. Wetting and roughness. Annual Review of Materials Research, 38:71– 99, 2008.  C. Dorrer and J. Rühe. Some thoughts on superhydrophobic wetting. Soft Matter, 5:51–61, 2009.  M. Nosonovsky and B. Bhushan. Superhydrophobic surfaces and emerging applications: Non-adhesion, energy, green engineering. Current Opinion in Colloid & Interface Science, 14:270–280, 2009.  D. Quéré. Rough ideas on wetting. Physica A-Statistical Mechanics And Its Applications, 313(1-2):32–46, 2002.  E. Martines, K. Seunarine, H. Morgan, N. Gadegaard, C. D. W. Wilkinson, and M. O. Riehle. Superhydrophobicity and superhydrophilicity of regular nanopatterns. Nano Letters, 5(10):2097–2103, 2005.  C. W. Extrand, Sung In Moon, P. Hall, and D. Schmidt. Superwetting of structured surfaces. Langmuir, 23(17):8882–8890, 2007.  R. H. Dettre and R. E. Johnson. Contact angle hysteresis ii. contact angle measurements on rough surfaces. Advances in Chemistry Series, 43, 1964.  R. E. Johnson and R. H. Dettre. Contact angle hysteresis .3. study of an idealized heterogeneous surface. Journal of Physical Chemistry, 68(7):1744–1750, 1964.  L. Gao and T. J. McCarthy. How wenzel and cassie were wrong. Langmuir, 23(7):3762–3765, 2007.  C. W. Extrand. Contact angles and hysteresis on surfaces with chemically heterogeneous islands. Langmuir, 19(9):3793–3796, 2003.  L. Gao and T. J. McCarthy. Wetting 101Â°. Langmuir, 25(24):14105–14115, 2009.  V. R. Gray. Surface aspects of wetting and adhesion. Chemistry and Industry, 23:969, 1965.  L. Gao and T. J. McCarthy. An attempt to correct the faulty intuition perpetuated by the wenzel and cassie "laws". Langmuir, 25(13):7249–7255, 2009. 166
Bibliography  L. Gao and T. J. McCarthy. Reply to "comment on how wenzel and cassie were wrong by gao and mccarthy". Langmuir, 23(26):13243–13243, 2007.  A. Marmur and E. Bittoun. When wenzel and cassie are right: Reconciling local and global considerations. Langmuir, 25(3):1277–1281, 2009.  G. McHale. Cassie and wenzel: Were they really so wrong? 23(15):8200–8205, 2007.
 M. Nosonovsky. On the range of applicability of the wenzel and cassie equations. Langmuir, 23(19):9919–9920, 2007.  C. G. Furmidge. Studies at phase interfaces. 1. sliding of liquid drops on solid surfaces and a theory for spray retention. Journal of Colloid Science, 17(4):309– 324, 1962.  M. Reyssat and D. Quéré. Contact angle hysteresis generated by strong dilute defects. The Journal of Physical Chemistry B, 113(12):3906–3909, 2009.  J. F. Joanny and P. G. Degennes. A model for contact-angle hysteresis. Journal of Chemical Physics, 81(1):552–562, 1984.  J. F. Oliver, C. Huh, and S. G. Mason. Resistance to spreading of liquids by sharp edges. Journal of Colloid and Interface Science, 59(3):568–581, 1977.  C. W. Extrand. Modeling of ultralyophobicity: Suspension of liquid drops by a single asperity. Langmuir, 21(23):10370–10374, 2005.  Y. V. Kalinin, V. Berejnov, and R. E. Thorne. Contact line pinning by microfabricated patterns: Effects of microscale topography. Langmuir, 25(9):5391–5397, 2009.  D. Hull. Introduction to Dislocations, volume 16 of International series on materials science and technology. Pergamon Press, Oxford, 2nd edition, 1975.  Z. Yoshimitsu, A. Nakajima, T. Watanabe, and K. Hashimoto. Effects of surface structure on the hydrophobicity and sliding behavior of water droplets. Langmuir, 18(15):5818–5822, 2002.  J. Bico, C. Marzolin, and D. Quéré. Pearl drops. Europhysics Letters, 47(2):220– 226, 1999.  D. Quéré and M. Reyssat. Non-adhesive lotus and other hydrophobic materials. Philosophical Transactions of the Royal Society A, 366:1539–1556, 2008.  N. J. Shirtcliffe, G. McHale, S. Atherton, and M. I. Newton. An introduction to superhydrophobicity. Advances in Colloid and Interface Science, In Press, Corrected Proof. doi: DOI: 10.1016/j.cis.2009.11.001. 167
Bibliography  G. McHale, M. I. Newton, and N. J. Shirtcliffe. Dynamic wetting and spreading and the role of topography. Journal of Physics-Condensed Matter, 21:464122, 2009.  W. Choi, A. Tuteja, J. M. Mabry, R. E. Cohen, and G. H. McKinley. A modified cassie-baxter relationship to explain contact angle hysteresis and anisotropy on non-wetting textured surfaces. Journal of Colloid and Interface Science, 339(1):208–216, 2009.  A. Tuteja, W. Choi, M. L. Ma, J. M. Mabry, S. A. Mazzella, G. C. Rutledge, G. H. McKinley, and R. E. Cohen. Designing superoleophobic surfaces. Science, 318(5856):1618–1622, 2007.  A. Ahuja, J. A. Taylor, V. Lifton, A. A. Sidorenko, T. R. Salamon, E. J. Lobaton, P. Kolodner, and T. N. Krupenkin. Nanonails: A simple geometrical approach to electrically tunable superlyophobic surfaces. Langmuir, 24(1):9–14, 2007.  J. L. Liu, X. Q. Feng, G. F. Wang, and S. W. Yu. Mechanisms of superhydrophobicity on hydrophilic substrates. Journal of Physics-Condensed Matter, 19(35):12, 2007.  J. C. Love, L. A. Estroff, J. K. Kriebel, R. G. Nuzzo, and G. M. Whitesides. Selfassembled monolayers of thiolates on metals as a form of nanotechnology. Chemical Reviews, 105:1103–1169, 2005.  A. Ulman. Formation and structure of self-assembled monolayers. Chemical Reviews, 96(4):1533–1554, 1996.  D. Barriet and T. R. Lee. Fluorinated self-assembled monolayers: composition, structure and interfacial properties. Current Opinion in Colloid & Interface Science, 8(3):236–242, 2003.  M. K. Chaudhury. Self-assembled monolayers on polymer surfaces. Biosensors & Bioelectronics, 10(9-10):785–788, 1995.  T. M. Willey, A. L. Vance, T. van Buuren, C. Bostedt, L. J. Terminello, and C. S. Fadley. Rapid degradation of alkanethiol-based self-assembled monolayers on gold in ambient laboratory conditions. Surface Science, 576(1-3):188–196, 2005.  S. M. Morgenthaler, S. Lee, and N. D. Spencer. Submicrometer structure of surface-chemical gradients prepared by a two-step immersion method. Langmuir, 22(6):2706–2711, 2006.  G. M. Whitesides. The origins and the future of microfluidics. 442(7101):368–373, 2006.
 M. Charnley, M. Textor, A. Khademhosseini, and M.P. Lutolf. Integration column: microwell arrays for mammalian cell culture. Integrative Biology, 1:625–634, 2009. 168
Bibliography  J. F. Schumacher, M. L. Carman, T. G. Estes, A. W. Feinberg, L. H. Wilson, M. E. Callow, J. A. Callow, J. A. Finlay, and A. B. Brennan. Engineered antifouling microtopographies - effect of feature size, geometry, and roughness on settlement of zoospores of the green alga ulva. Biofouling, 23(1):55–62, 2007.  J. Genzer and K. Efimenko. Creating long-lived superhydrophobic polymer surfaces through mechanically assembled monolayers. Science, 290(5499):2130– 2133, 2000.  M. Morra, E. Occhiello, R. Marola, F. Garbassi, P. Humphrey, and D. Johnson. On the aging of oxygen plasma-treated polydimethylsiloxane surfaces. Journal of Colloid and Interface Science, 137(1):11–24, 1990.  D. J. Graham, D. D. Price, and B. D. Ratner. Solution assembled and microcontact printed monolayers of dodecanethiol on gold: A multivariate exploration of chemistry and contamination. Langmuir, 18(5):1518–1527, 2002.  D. Bonn, J. Eggers, J. Indekeu, J. Meunier, and E. Rolley. Wetting and spreading. Reviews of Modern Physics, 81(2):739–805, 2009.  M. Taylor, A. J. Urquhart, M. Zelzer, M. C. Davies, and M. R. Alexander. Picoliter water contact angle measurement on polymers. Langmuir, 23(13):6875–6878, 2007.  D. Y. Kwok, C. N. C. Lam, A. Li, A. Leung, R. Wu, E. Mok, and A. W. Neumann. Measuring and interpreting contact angles: a complex issue. Colloids and Surfaces a-Physicochemical and Engineering Aspects, 142(2-3):219–235, 1998.  D. Y. Kwok and A. W. Neumann. Contact angle measurement and contact angle interpretation. Advances in Colloid and Interface Science, 81(3):167–249, 1999.  M. Ruths and J. N. Israelachvili. Surface forces and nanorheology of molecularly thin films. In Nanotribology and Nanomechanics, pages 417–515. 2008. ISBN: 978-3-540-77607-9.  R. Fetzer and J. Ralston. Dynamic dewetting regimes explored. Journal of Physical Chemistry C, 113(20):8888–8894, 2009.  T. D. Blake. The physics of moving wetting lines. Journal of Colloid and Interface Science, 299(1):1–13, 2006.  C. W. Extrand. Contact angles and their hysteresis as a measure of liquid-solid adhesion. Langmuir, 20(10):4017–4021, 2004.  D. M. Spori, N. V. Venkataraman, S. G. P. Tosatti, F. Durmaz, N. D. Spencer, and S. Zürcher. Influence of alkyl chain length on phosphate self-assembled monolayers. Langmuir, 23(15):8053–8060, 2007. 169
Bibliography  S. Tosatti, R. Michel, M. Textor, and N. D. Spencer. Self-assembled monolayers of dodecyl and hydroxy-dodecyl phosphates on both smooth and rough titanium and titanium oxide surfaces. Langmuir, 18(9):3537–3548, 2002.  L. C. Gerhardt, A. Schiller, B. Müller, N. Spencer, and S. Derler. Fabrication, characterisation and tribological investigation of artificial skin surface lipid films. Tribology Letters, 34(2):81–93, 2009.  F. Walther, P. Davydovskaya, S. Zürcher, M. Kaiser, H. Herberg, A. M. Gigler, and R. W. Stark. Stability of the hydrophilic behavior of oxygen plasma activated su-8. Journal of Micromechanics and Microengineering, 17(3):524–531, 2007.  H. W. Fox and W. A. Zisman. The spreading of liquids on low-energy surfaces. ii. modified tetrafluoroethylene polymers. Journal of Colloid Science, 7(2):109–121, 1952.  D. K. Owens and R. C. Wendt. Estimation of the surface free energy of polymers. Journal of Applied Polymer Science, 13(8):1741–1747, 1969.  C. J. VanOss, M. K. Chaudhury, and R. J. Good. Interfacial lifshitz-vanderwaals and polar interactions in macroscopic system. Chemical Reviews, 88(6):927–941, 1988.  C. Della Volpe and S. Siboni. Some reflections on acid-base solid surface free energy theories. Journal of Colloid and Interface Science, 195(1):121–136, 1997.  N. V. Venkatamaran, S. Zürcher, and N. D. Spencer. Order and composition of methyl-carboxyl and methyl-hydroxyl surface-chemical gradients. Langmuir, 22(9):4184–4189, 2006.  A. Amirfazli, D. Y. Kwok, J. Gaydos, and A. W. Neumann. Line tension measurements through drop size dependence of contact angle. Journal of Colloid and Interface Science, 205(1):1–11, 1998.  R. Tadmor and P. S. Yadav. As-placed contact angles for sessile drops. Journal of Colloid and Interface Science, 317(1):241–246, 2008.  R. Maoz and J. Sagiv. On the formation and structure of self-assembling monolayers. 1. a comparative atr-wetability study of langmuir-blodgett and adsorbed films on flat substrates and glass microbeads. Journal of Colloid and Interface Science, 100(2):465–496, 1984.  P. Silberzan, L. Leger, D. Ausserre, and J. J. Benattar. Silanation of silica surfaces. a new method of constructing pure or mixed monolayers. Langmuir, 7(8):1647– 1651, 1991.  G. Shafrin E. and A. Zisman W. Upper limits to the contact angles of liquids on solids. In Contact Angle, Wettability, and Adhesion, Advances in Chemistry, pages 145–157. American Chemical Society, Washington, D.C., 1964. 170
Bibliography  C. Chen, J. Wang, and Z. Chen. Surface restructuring behavior of various types of poly(dimethylsiloxane) in water detected by sfg. Langmuir, 20(23):10186–10193, 2004.  C. W. Extrand and Y. Kumagai. Contact angles and hysteresis on soft surfaces. Journal of Colloid and Interface Science, 184(1):191–200, 1996.  M. E. R. Shanahan and A. Carré. Viscoelastic dissipation in wetting and adhesion phenomena. Langmuir, 11:1396–1402, 1995.  M. E. R. Shanahan and P. G. De Gennes. Equilibrium of the Triple Line Solid/Liquid/Fluid of a Sessile Drop. Adhesion 11. Elsevier Applied Science, New York, 1987. ISBN 1851660798.  M. R. Dusseiller. Micro- and Nanoengineering the 3-Dimensional Environment of Cells in Culture. PhD thesis, ETH Zurich, 2005. doi:10.3929/ethz-a-005206909.  H. Hillborg, M. Sandelin, and U. W. Gedde. Hydrophobic recovery of polydimethylsiloxane after exposure to partial discharges as a function of crosslink density. Polymer, 42(17):7349–7362, 2001.  R. Heeb. Surface Modifications for Improved Aqueous Lubrication under LowContact-Pressure Conditions. PhD thesis, ETH Zurich, 2009. Diss. ETH No. 18431.  J. N. Lee, C. Park, and G. M. Whitesides. of poly(dimethylsiloxane)-based microfluidic devices. 75(23):6544–6554, 2003.
Solvent compatibility Analytical Chemistry,
 A. F. Stalder, G. Kulik, D. Sage, L. Barbieri, and P. Hoffmann. A snake-based approach to accurate determination of both contact points and contact angles. Colloids and Surfaces a-Physicochemical and Engineering Aspects, 286(1-3):92– 103, 2006.  J. Zimmermann. Silicone Nanofilaments as Functional Coatings:Properties, Applications and Modifications. PhD thesis, University of Zurich, 2008.  T. P. Kunzler, T. Drobek, C. M. Sprecher, M. Schuler, and N. D. Spencer. Fabrication of material-independent morphology gradients for high-throughput applications. Applied Surface Science, 253(4):2148–2153, 2006.  M. Wieland, B. Chehroudi, M. Textor, and D. M. Brunette. Use of ti-coated replicas to investigate the effects on fibroblast shape of surfaces with varying roughness and constant chemical composition. Journal Of Biomedical Materials Research, 60(3):434–444, 2002.  P. Wagner, R. Furstner, W. Barthlott, and C. Neinhuis. Quantitative assessment to the structural basis of water repellency in natural and technical surfaces. Journal of Experimental Botany, 54(385):1295–1303, 2003. 171
Bibliography  Y. N. Xia and G. M. Whitesides. Soft lithography. Annual Review of Materials Science, 28:153–184, 1998.  Y. Zhang, C. W. Lo, J. A. Taylor, and S. Yang. Replica molding of high-aspectratio polymeric nanopillar arrays with high fidelity. Langmuir, 22(20):8595–8601, 2006.  O. D. Gordan, B. N. J. Persson, C. M. Cesa, D. Mayer, B. Hoffmann, S. Dieluweit, and R. Merkel. On pattern transfer in replica molding. Langmuir, 24(13):6636– 6639, 2008.  M. Wieland, P. Hanggi, W. Hotz, M. Textor, B. A. Keller, and N. D. Spencer. Wavelength-dependent measurement and evaluation of surface topographies: application of a new concept of window roughness and surface transfer function. Wear, 237(2):231–252, 2000.  T. Kunzler. Surface Morphology Gradients. doi:10.3929/ethz-a-005420464.
PhD thesis, ETH Zurich, 2007.
 M. Schuler, T. P. Kunzler, M. de Wild, C. M. Sprecher, D. Trentin, D. M. Brunette, M. Textor, and S. G.P. Tosatti. Fabrication of tio2-coated epoxy replicas with identical dual-type surface topographies used in cell culture assays. Journal Of Biomedical Materials Research, 88A(1):12–22, 2009.  S. Lee and S. Lee. Shrinkage ratio of pdms and its alignment method for the wafer level process. Microsystem Technologies, 14(2):205–208, 2008.  P. Bariani, L. De Chiffre, H. N. Hansen, and A. Horsewell. Investigation on the traceability of three dimensional scanning electron microscope measurements based on the stereo-pair technique. Precision Engineering, 29(2):219–228, 2005.  H. N. Hansen, K. Carneiro, H. Haitjema, and L. De Chiffre. Dimensional micro and nano metrology. CIRP Annals - Manufacturing Technology, 55(2):721–743, 2006.  E. A. Turi, editor. Thermal characterization of polymeric materials. Academic Press, San Diego, second edition, 1997. ISBN 0-12-703783-7.  R. Hofer. Surface Modification for Optical Biosensor Applications. PhD thesis, ETH Zurich, 2000. doi:10.3929/ethz-a-004040300.  C. Cremmel, D. M. Spori, S. Zürcher, and N. D. Spencer. On the way to chemically stable superhydrophobic surfaces. Technical report, LSST - ETH Zurich, 2010.  M. Heule, S. Vuillemin, and L. J. Gauckler. Powder-based ceramic meso- and microscale fabrication processes. Advanced Materials, 15(15):1237–1245, 2003.  Christian and J. A. Kenis. Fabrication of ceramic microscale structures. J. Am. Ceram. Soc., 90(9):2779–2783, 2007. 172
Bibliography  C. Marzolin, S. P. Smith, M. Prentiss, and G. M. Whitesides. Fabrication of glass microstructures by micro-molding of sol-gel precursors. Advanced Materials, 10(8):571–574, 1998.  U. P. Schönholzer and Ludwig J. Gauckler. Ceramic part patterned in the micrometer range. Advanced Materials, 11(8):630–632, 1999.  M. Heule and L. J. Gauckler. Miniaturised arrays of tin oxide gas sensors on single microhotpolate substrates fabricated by micromolding in capillaries. Sensors and Actuators B, 93(2003):100–106, 2003.  M. Heule and L. J. Gauckler. Microfabrication of ceramics based on colloidal suspensions and photoresist masks. Journal of Polymer Science and Technology, 14(3):449–452, 2001.  M. Heule, U. P. Schönholzer, and L. J. Gauckler. Patterning colloidal suspensions by selective wetting of microcontact-printed surfaces. Journal of the European Ceramic Society, 24:2733–2739, 2004.  C. R. Martin and I. A. Aksay. Microchannel molding: A soft lithography-inspired approach to micrometer-scale patterning. J. Mater. Res., 20(8):1995–2003, 2005.  C. Cremmel, D. M. Spori, S. Zürcher, and N. D. Spencer. On the way to chemically and mechanically stable superhydrophobic surfaces. Technical report, LSST - ETH Zurich, 2009.  G. Csucs, T. Kunzler, K. Feldman, F. Robin, and N. D. Spencer. Microcontact printing of macromolecules with submicrometer resolution by means of polyolefin stamps. Langmuir, 19(15):6104–6109, 2003.  U. P. Schönholzer, N. Stutzmann, T. A. Tervoort, P. Smith, and L. J. Gauckler. Micropatterned ceramics by casting into polymer molds. J. Am. Ceram. Soc., 85(7):1885–1887, 2002.  U. P. Schönholzer, R. Hummel, and L. J. Gauckler. Microfabrication of ceramics by filling of photoresist molds. Advanced Materials, 12(17):1261–1263, 2000.  S. Morgenthaler. Surface-Chemical Gradients. PhD thesis, ETH Zurich, 2007. doi:10.3929/ethz-a-005415853.  A. del Campo and C. Greiner. Su-8: a photoresist for high-aspect-ratio and 3d submicron lithography. Journal of Micromechanics and Microengineering, 17(6):R81– R95, 2007.  B. Bhushan and Y. C. Jung. Wetting study of patterned surfaces for superhydrophobicity. Ultramicroscopy, 107(10-11):1033–1041, 2007.  R. Blossey. Self-cleaning surfaces - virtual realities. Nature Materials, 2(5):301– 306, 2003. 173
Bibliography  M. Callies and D. Quéré. On water repellency. Soft Matter, 1(1):55–61, 2005.  N. A. Patankar. On the modeling of hydrophobic contact angles on rough surfaces. Langmuir, 19(4):1249–1253, 2003.  M. Wieland, M. Textor, N. D. Spencer, and D. M. Brunette. Wavelength-dependent roughness: A quantitative approach to characterizing the topography of rough titanium surfaces. International Journal Of Oral & Maxillofacial Implants, 16(2):163–181, 2001.  J. Bico, C. Tordeux, and D. Quere. Rough wetting. Europhysics Letters, 55(2):214– 220, 2001.  S. Morgenthaler, C. Zink, and N. D. Spencer. Surface-chemical and -morphological gradients. Soft Matter, 4:419–434, 2008.  W. Li and A. Amirfazli. Microtextured superhydrophobic surfaces: A thermodynamic analysis. Advances in Colloid and Interface Science, 132(2):51–68, 2007.  C. W. Extrand. Model for contact angles and hysteresis on rough and ultraphobic surfaces. Langmuir, 18(21):7991–7999, 2002.  A. Steinberger, C. Cottin-Bizonne, P. Kleimann, and E. Charlaix. High friction on a bubble mattress. Nat Mater, 6(9):665–668, 2007.  S. Wang and L. Jiang. Definition of superhydrophobic states. Advanced Materials, 19(21):3423–3424, 2007.  A. Tuteja, W. Choi, J. M. Mabry, G. H. McKinley, and R. E. Cohen. Robust omniphobic surfaces. Proceedings of the National Academy of Sciences of the United States of America, 105(47):18200–18205, 2008.  P. S. H. Forsberg, C. Priest, M. Brinkmann, R. Sedev, and J. Ralston. Contact line pinning on microstructured surfaces for liquids in the wenzel state. Langmuir, 26(2):860–865, 2010.  M. Nosonovsky and B. Bhushan. Hierarchical roughness optimization for biomimetic superhydrophobic surfaces. Ultramicroscopy, 107(10-11):969–979, 2007.  C. Lee and C. J. Kim. Maximizing the giant liquid slip on superhydrophobic microstructures by nanostructuring their sidewalls. Langmuir, 25(21):12812–12818, 2009.  L. Gao and T. J. McCarthy. The "lotus effect" explained: Two reasons why two length scales of topography are important. Langmuir, 22(7):2966–2967, 2006.  H. J. Ensikat, A. J. Schulte, K. Koch, and W. Barthlott. Droplets on superhydrophobic surfaces: Visualization of the contact area by cryo-scanning electron microscopy. Langmuir, 25(22):13077–13083, 2009. 174
Bibliography  C. Dorrer and J. Rühe. Drops on microstructured surfaces coated with hydrophilic polymers: Wenzel’s model and beyond. Langmuir, 24(5):1959–1964, 2008.  V. N. Paunov. Novel method for determining the three-phase contact angle of colloid particles adsorbed at air-water and oil-water interfaces. Langmuir, 19(19):7970–7976, 2003.  S. Morgenthaler, S. W. Lee, S. Zürcher, and N. D. Spencer. A simple, reproducible approach to the preparation of surface-chemical gradients. Langmuir, 19(25):10459–10462, 2003.  C. Hückstädt, D. M. Spori, and N. D. Spencer. Exploring superhydrophobic surfaces using photolithography. Technical report, LSST - ETH Zurich, 2009.  W. Stöber, A. Fink, and E. Bohn. Controlled growth of monodisperse silica spheres in micron size range. Journal Of Colloid And Interface Science, 26(1):62–&, 1968.  Y. Santschi, D.M. Spori, T. Drobek, and N. D. Spencer. Lotus effect with raspberrylike particles. Technical report, LSST - ETH Zurich, 2006.
Acknowledgements Many people have supported me during these years I have been working on my PhD thesis. I am glad to find the opportunity here to thank them for their support. In the second part, the acknowledgements to the single chapters are added. Nic Spencer. I am very grateful that he offered me a PhD position in his lab. His clarity of thoughts, his experience and knowledge were inspiring and a driving force to perform at the highest level of which I was capable. By giving a lot of freedom to his PhD students to define their own research, he allows them to have the satisfaction of making their own contribution to science. Despite the freedom, one can be sure that if one was going in the completely wrong direction, he would intervene. I have always been impressed how he managed to constantly fly around the world and still find enough time for monthly meetings with all his students. I very much appreciated the quick and useful answers to e-mails. In this way, it did not matter whether he was in India or next door. I probably enjoyed most the times, when we had meetings just the two of us and we were brushing up a text that I had written. It was amazing in what a short time my thoughts got a shape and clarity, as if there had never been any doubts. Tanja Drobek. When I started I was excited and curious to work for the first time with a woman as supervisor. And I think, it has worked out really well. It is very easy to awake her interest in a research topic and she can transfer her enthusiasm on her co-workers. She has this special gift of finding the red line in a story. I will always be thankful for the long distance support from Munich during the writing of the work on the Photoshop gradients. It helped me a lot during times when I had troubles to find sense in my data. Stefan Zürcher. Ever since I did my second master’s project with him I knew he was a brilliant chemist, a clear thinker and a good friend. From the moment Tanja Drobek left for Munich, it was clear that I would choose him as my future supervisor. He would always find time to give me an answer to my questions. E-mails were often short, but he replied fast and with all the information I needed. I will never forget how we tried to find correlations in graphics equipped with pencil and ruler. He could always motivate me in situations when I did not see how it could go on. I have been very lucky with my supervisors. I believe friendship has grown between all of us. 177
Anthony B. Brennan. I would like to thank Tony Brennan for accepting my invitation to be my external co-examiner. Ludwig Gauckler. My acknowledgements also go to Ludwig Gauckler for being the representative of the department at my PhD defense. Ciba is acknowledged for generously funding this thesis and the useful and friendly discussions during the regular meetings with Andreas Mühlebach, Werner Rutsch and Niklaus Bühler. Despite the project taking quite a fundamental turn in the middle, the Ciba crew remained very supportive of me and my work.
Life in the lab and about One of the best things about working in LSST is that in an international group such as ours, we have the whole world in our lab. This makes work a lot more interesting, entertaining, surprising, sometimes challenging, and it definitely adds a fascinating experience to one’s life. I would like to thank the whole LSST for being the way it is. Each member of the group was in one way or other important to me. Nevertheless I would like to thank a few people especially: Eva Beurer: for jogging over lunchtime and becoming a true friend. I have seldom met someone who can listen so closely and draw conclusions as clearly as she can, be it in scientific or in personal matters. Whitney Hartung: for impressing me with her energy to do things, organise parties and being a friend inside and outside the lab. Christoph Mayer and Torben Gillich: for sharing my passion for good food, beer and wine and for organising a few magnificent barbecues in the bamboo forest. Both are excellent chemists who willingly share their knowledge with their fellow PhD students. Christian Zink: for going with me to the Nanoparticles Conference in Florida, organising the legendary offsite group meeting in Arosa and for keeping the group going. Filippo Mangolini, Maura Crobu and Clément Cremmel: for occasional invitations for dinner and joining me on sportive events like the Monday Night Skates through Zurich, hiking, and swimming. Karthik Kumar: for being a dedicated philosopher. He initiated many absorbing discussions in the lab and at parties. My students Yannick Santschi, Cathrein Hückstädt and Clément Cremmel for doing master projects with me. I was very lucky that these three excellent students chose to work on my project. All the users of the lab E 465 who supported me in my task of being a lab manager. 178
Writing the thesis Stefan, Tanja and Nic did a great job correcting the chapters I wrote, in style, logic, scientific input and the short time they needed to make corrections. Martin Fopp and Christoph Mayer helped me in getting used to the program Lyx, which I used to write the thesis, and helped type-setting the book. Sara Morgenthaler’s and Raphael Heeb’s thesis often served as a guide when I was unsure how to write a chapter. I would like to thank Mr. Anderegg, my optician, for prescribing me Essilor® AntiFatigue™ glasses. My eyesight is good and I do not need glasses to drive, for example, but I get quickly tired when working in front of the computer. With these glasses, my maximum working hours were no longer determined by my eyes, but by my brain. The band OneRepublic for their gorgeous song “Good Life” which would always cheer me up, when writing this book was starting to get on top of me.
Personal My parents: I would like to give special thanks to my parents. Without their love and support over all the years I would certainly not have reached such a high level of education. Martin Fopp: He has always been a great supporter of whatever I was doing. Having a PhD student as a girlfriend is not always easy. But he never complained when it took me half the holidays to switch off thoughts about work and he was always listening with interest to my gabbling on about science. Additionally, he has also always been my real-time computer support and he introduced me into the use of the Adobe programs Photoshop and Illustrator.
Chapter 1 The cryo-SEM images of the Lotus leaves were imaged with the help of Stefan Handschin from the EMEZ.
Chapter 4 I would like to thank Eva Beurer for the PMIRRAS measurements of the thiol SAMs on gold, which were used in chapter 4.1. 179
Chapter 5 The greatest thanks go in this chapter to Christian Zink (fellow Ph.D. student). We cosupervised Daniel Daniel (2 months internship: Cambridge summer placement), who did most of the experimental work in chapter 5.4. Additionally, Chris had the initial idea of creating ceramic replicas and showed to Clément Cremmel and me how to prepare, handle and sinter ceramic slurries. The Polymer Technology group allowed Daniel and me to use the DSC and Jan Giessbrecht gave us the introduction to the instrument. Urs Gonzenbach from the Nonmetallic Inorganic Materials group always found time for us, when Clément and I needed advice in dealing with ceramics or high-temperature ovens. The Nonmetallic Inorganic Materials group also provided the alumina powder used in chapter 5.5. Clément Cremmel, coming from the ESPCI in Paris, did an internship with me and was mostly working with the two-scale photolithographic structures and ceramic replicas, as presented in chapter 5.5. He continued his excellent work in a master project. The EMEZ, especially Karsten Kunze and Philippe Gasser, helped me to carve the FIB structure into the silicon wafer. The machine shop always provided us with the right tools and samples.
Chapter 6 The authors thank the Botanical Garden of Zurich for providing the Lotus leaves and Electron Microscopy ETH Zurich, EMEZ, for their support.
Chapter 7 The authors thank Electron Microscopy ETH Zurich, EMEZ, for their support. Additionally, we thank René Tölke for help with the photolithography, Cathrein Hückstädt for fruitful discussions and Martin Elsener for building the roll-off angle goniometer.
Chapter 9 Three of the outlook chapters were based on master projects performed by Yannick Santschi, Cathrein Hückstädt and Clément Cremmel.
Doris Madeleine Spori Date of Birth:
November 27, 1979
Swiss, citizen of Boltigen (BE)
Bergacker 34, CH-8046 Zürich, Switzerland
2005 - 2010
Doctor of Sciences Laboratory for Surface Science & Technology Department of Material Science, ETHZ, Switzerland
1999 - 2005
Dipl. Werkstoff-Ing. ETH ETH Zurich, Switzerland
1994 - 1999
Matura, Type E Wirtschaftsgymnasium Bern Neufeld, Switzerland 181
Practical Experience 2005 - 2010
Supervision of three master projects and three internships. Teaching of two semesters of lab courses.
06/2003 - 09/2003
Internship at Ian Wark Research Institute Adelaide, South Australia, Australia
02/2003 - 04/2003
Internship at ABB Corporate Research Dättwil, Aargau, Switzerland
D. M. Spori, T. Drobek, S. Zürcher, and N. D. Spencer. "Cassie-State Wetting Investigated by Means of a Hole-to-Pillar Density Gradient." Langmuir (submitted).
D. M. Spori, T. Drobek, S. Zürcher, M. Ochsner, C. Sprecher, A. Mühlebach, and N. D. Spencer. "Beyond the Lotus Effect: Roughness Influences on Wetting over a Wide Surface-Energy Range." Langmuir 24, no. 10 (2008): 541117.
D. M. Spori, N. V. Venkataraman, S. G. P. Tosatti, F. Durmaz, N. D. Spencer, and S. Zürcher. "Influence of Alkyl Chain Length on Phosphate SelfAssembled Monolayers." Langmuir 23, no. 15 (2007): 8053-60.
Oral Presentations 2009
Kontaktwinkelmessungen auf „idealen“ und realen Oberflächen Doris M. Spori Krüss Seminar (Dresden, Germany)
Exploring Superhydrophobic Surfaces with „Budget“-Photolithography Doris M. Spori, Tanja Drobek, Stefan Zürcher, Nicholas D. Spencer Material Science Colloquium, (ETH Zurich, Zurich, ZH, Switzerland)
Wettability Effects on Lotus Leaf Replicas Doris M. Spori, Tanja Drobek, Stefan Zürcher, Andreas Mühlebach, Nicholas D. Spencer 2nd MRC Graduate Symposium (Zurich, ZH; Switzerland)
Biomimetic Self-cleaning Surfaces: Surface Chemical Gradients on Lotus Leaf Replicas Doris M. Spori, Tanja Drobek, Andreas Mühlebach, Nicholas D. Spencer Empa PhD Student’s Symposium (St.Gallen, SG, Switzerland)
Poster Presentations 2009
What Keeps Drops in Shape? Doris M. Spori, Tanja Drobek, Stefan Zürcher, Nicholas D. Spencer Materials Day 2009 (Zurich, ZH, Switzerland)
Contact Angle Study of Morphological Gradients Doris M. Spori, Tanja Drobek, Stefan Zürcher, Nicholas D. Spencer Gordon Conference: Chemistry at Interfaces (Waterville, NH, USA)
Self-Cleaning Surfaces by the Use of Silica Nanoparticles Doris M. Spori, Stefan Zürcher, Andreas Mühlebach, Nicholas D. Spencer Particles 2008 (Orlando, FL, USA)
Beyond the Lotus Effect... Roughness Influences on Wetting over a Wide Surface Energy Range D. M. Spori, T. Drobek, S. Zürcher, M. Ochsner, C. Sprecher, A. Mühlebach and N. D. Spencer SAOG: Liquid meets Solid (Freiburg, FR; Switzerland)
Wettability Effects on Lotus Leaf Replicas Doris M. Spori, Tanja Drobek, Stefan Zürcher, Andreas Mühlebach, Nicholas D. Spencer Gordon Conference: Chemistry of Supramolecules & Assemblies (Lucca, Italy)
Biomimetic Self-cleaning Surfaces Doris M. Spori, Tanja Drobek, Andreas Mühlebach, Nicholas D. Spencer 1st MRC Graduate Symposium (Zurich, ZH; Switzerland)
Variation of Surface Chemistry on Lotus Replicas Doris M. Spori, Tanja Drobek, Andreas Mühlebach, Nicholas D. Spencer SAOG: Functional Surfaces and Interfaces (Freiburg, FR; Switzerland)