The impact of geometrical variations on the transport properties of organic electronic ion pumps

LiU-ITN-TEK-A-13/021-SE

The impact of geometrical

variations on the transport

properties of organic

electronic ion pumps

Theresia Arbring

LiU-ITN-TEK-A-13/021-SE

The impact of geometrical

variations on the transport

properties of organic

electronic ion pumps

Examensarbete utfört i Medicinsk teknik

vid Tekniska högskolan vid

Linköpings universitet

Theresia Arbring

Handledare Björn Kronander

Examinator Daniel Simon

Upphovsrätt

Detta dokument hålls tillgängligt på Internet – eller dess framtida ersättare –

under en längre tid från publiceringsdatum under förutsättning att inga

extra-ordinära omständigheter uppstår.

Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner,

skriva ut enstaka kopior för enskilt bruk och att använda det oförändrat för

ickekommersiell forskning och för undervisning. Överföring av upphovsrätten

vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning av

dokumentet kräver upphovsmannens medgivande. För att garantera äktheten,

säkerheten och tillgängligheten finns det lösningar av teknisk och administrativ

art.

Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i

den omfattning som god sed kräver vid användning av dokumentet på ovan

beskrivna sätt samt skydd mot att dokumentet ändras eller presenteras i sådan

form eller i sådant sammanhang som är kränkande för upphovsmannens litterära

eller konstnärliga anseende eller egenart.

För ytterligare information om Linköping University Electronic Press se

förlagets hemsida

http://www.ep.liu.se/

Copyright

The publishers will keep this document online on the Internet - or its possible

replacement - for a considerable time from the date of publication barring

exceptional circumstances.

The online availability of the document implies a permanent permission for

anyone to read, to download, to print out single copies for your own use and to

use it unchanged for any non-commercial research and educational purpose.

Subsequent transfers of copyright cannot revoke this permission. All other uses

of the document are conditional on the consent of the copyright owner. The

publisher has taken technical and administrative measures to assure authenticity,

security and accessibility.

According to intellectual property law the author has the right to be

mentioned when his/her work is accessed as described above and to be protected

against infringement.

For additional information about the Linköping University Electronic Press

and its procedures for publication and for assurance of document integrity,

please refer to its WWW home page:

http://www.ep.liu.se/

Abstract

The organic electronic ion pump (OEIP) is an electrically controlled polymer-based device that has the capability to interact with biological systems down to a single cell level by mimicking neural signalling. This is accomplished by translation of an electrical signal into a chemical output, such as ions and neu-rotransmitters. Because of the combined spatial and temporal precision, this is a technology with a promising future as an advanced therapeutic device. De-pending on the application, the OEIP requires different geometries. Implants that will be used to control on a single cell level require very small dimensions, while for example extracorporeal mounted OIEPs, with only the delivery chan-nel penetrating the skin, require much longer chanchan-nels. Despite the application, it is necessary to have a good knowledge about the transport and delivery prop-erties and how they change due to the geometry. These propprop-erties have been observed as very varying and unstable in early unpublished results, and these findings motivate this project.

This project includes photolithographic fabrication and investigation of trans-port and delivery properties such as effective resistance, efficiency and stability of OEIPs with varying delivery channel lengths and widths. Shorter delivery channels show a consistent but relatively low efficiency. Delamination between different layers of the device is suspected as the cause. Initially, the longer delivery channels show a low functionality, most probably due to poor encap-sulation. It is suggested that a soft, water-permeable plastic best encapsulates OEIPs that will be used as a medical implant, while a material impermeable to water, for example a metal, could successfully encapsulate OEIPs operating in air.

Acknowledgments

First and foremost, I would like to thank Daniel Simon for introducing me to the field of organic bioelectronics, providing me this exciting project and, to-gether with my supervisors Bj¨orn Kronander and David Nilsson, gave me an outstanding guidance along the way. Your inspiration, support and encourage-ment have been invaluable. Big thanks also to Klas Tybrandt and Amanda Jonsson for many good ideas and long discussions through the ups and downs during these months and Erik Gabrielsson for the help you gave me during the fabrication. Neither should Nikolaos Felekidis be forgotten, thank you for being my opponent.

My appreciations also goes to the whole Organic Electronics group for the warm welcome, guidance in the lab and up-lifting lunches and coffee breaks. Finally, I want to thank my lovely friends, supporting family and especially my beloved Fredrik.

Contents

1 Introduction 1

1.1 The conducting polymer PEDOT:PSS . . . 1

1.2 The organic electronic ion pump . . . 3

1.2.1 Ion transport . . . 3

1.3 Objectives . . . 4

1.4 Scope . . . 5

2 Materials and methods 6 2.1 Fabrication . . . 6 2.2 Measurements . . . 10 2.2.1 Setup . . . 10 2.2.2 Electrical . . . 10 2.2.3 Chemical . . . 11 2.3 Transport efficiency . . . 14 3 Results 15 3.1 Effective resistance and resistivity . . . 15

3.2 Transported K+ _{vs Charge . . . . 16}

3.2.1 2 mm long ion pump channels . . . 16

3.2.2 10 mm long ion pump channels . . . 17

3.3 Efficiency vs Geometry . . . 17

3.4 Ion transport in 35 mm long ion pump channels . . . 18

3.4.1 Soaking methods . . . 18

3.4.2 Water cover . . . 19

3.4.3 Encapsulation materials . . . 19

4 Discussion and conclusions 20 4.1 Efficiency vs Geometry . . . 20

4.1.1 Reliability . . . 20

4.1.2 The low efficiency . . . 20

4.2 Water leakage through the encapsulation layer . . . 22

4.2.1 Cracks and pinholes . . . 22

4.2.2 Evaporation of water through the encapsulation layer . . 23

4.2.3 Suggested solution . . . 26

4.3 Future prospects . . . 26

References 27

1

Introduction

The field of bioelectronics opens the possibility to interact with biological sys-tems down to a single cell level, using electrically controlled devices [1, 2]. This can be accomplished by direct electrical stimulation, used e.g. in pacemakers or neuronal electrodes. Another possibility is to try to mimic the performance of the neurons found in our own bodies.

The neural signalling system, which includes billions of neurons, communicates and control cells in our body in a very complex and precise manner [3,4] (Figure (1)). A neuronal dendrite reacts on a chemical substance (a neurotransmitter) that activates the cell. This activation gives rise to an action potential that propagates at high speed through the axon to the synapses where the action potential induces a release of a new neurotransmitter, exactly when and where the substance is supposed to be delivered. The strength of the signal is also modulated by the amplitude of the input stimuli, which determines the output release. In simpler words, the neuron translates a chemical input to an electrical signal and back again to a chemical output controlled in time, space and strength [3, 5, 6]. A device that mimics this behavior, even without the speed of a neural signal, can be very useful in several applications, for example to control cell growth, regulate cell communication and also for precise drug delivery [5, 6].

Figure 1: The neuron is able to translate chemical signals captured by the dendrites to electrical signals, send these signals through the axon and then translate them back to a chemical signal in the synapses [4].

To be able to fabricate a device capable of this functionality and also applicable in vivo, the device material must be able to conduct both electrons and ions or molecules. Another critical requirement is that it needs to be biocompatible to interact with living cells. One class of materials that are able to meet these requirements are conducting polymers [1, 7, 8], or in simpler words, conducting plastics. In addition, a device based on conducting polymers can also, for a medical device, be relatively cheap to produce [6].

1.1

The conducting polymer PEDOT:PSS

The polymer chosen in this study is the semiconducting polymer poly(3,4-ethylene dioxythiophene) (PEDOT), doped with the negatively charged

poly-capability of electric conduction depends on the PEDOT backbone of alternat-ing σ-bonds and π-bonds. Counter-ions to the negatively charged PSS is either small cations (such as H+_{, Na}+ _{or K}+_{) or a positive charge from PEDOT in}

its oxidized state (PEDOT+_{). In equilibrium, a positive charge appears}

ap-proximately every third EDOT molecule along the backbone [9]. These mobile carriers (also referred to as holes) make conduction of electrons very efficient and enhance the electrical conductivity by several orders of magnitude compared to un-doped PEDOT [5].

Figure 2: The structural formula of PEDOT:PSS. PSS−_{(bottom) can associate with}

the positive charges in the oxidized PEDOT+ _{(top) or to smaller positive molecules}

or ions (M+

).

When PEDOT:PSS is soaked in water it swells and forms a gel with resistivity close to metals and a Youngs modulus close to tissue [9], that makes this material perfect for bioelectronics. The ionic conduction is possible together with the absorbed water, where positively charged ions (cations) move from a shield of water to another by association with the negatively charged polyelectrolyte PSS−_{[6,10]. Consequently, the ionic conduction for this polymer-polyelectrolyte}

system is cation selective. It should be noted that even though the ions are transported in water, the water is not transported with the ions, resulting in a non-convective transport and delivery of ions [11].

In this application, both ionic and electronic conduction is necessary, but in selected regions only ionic conduction is wanted. In those regions the material needs to be modified from electronically conducting to electrically insulated. This can be achieved by either chemical or electrical over-oxidation, two methods which break the π-bonds in the PEDOT backbone [12], making transport of electrons impossible. Due to the overabundance of PSS in most PEDOT:PSS systems [13], once the material is hydrated, this over-oxidation process results in what is effectively a simple PSS hydrogel.

1.2

The organic electronic ion pump

The device of particular interest for this project is the organic electronic ion pump, seen in Figure 3. A positively charged ion or molecule is transported from one electrolyte (source) to another (target) through a channel. A voltage supply is connected between the source and the target, that will induce and control the transport of ions through an electrically insulated channel [14] and the release of the ion at the channel outlet. As delivery is diffusive (no liquid flow), and controlled by electrical signals, the ion pump is thus mimicking the synapse of a nerve cell (see Figure (1)).

Figure 3: The organic electronic ion pump seen from the top and from the side.

The ion pump is applicable for delivery of ions such as H+ _{[15], K}+_{, Na}+_{, and}

Ca2+_{[14] or bigger molecules and neurotransmitters such as acetylcholine (ACh)}

[16], γ-amino butyric acid (GABA), aspartate (Asp) and glutamate (Glu) [11]. The ion pump shows a good capability of transporting ions with both spatial and temporal precision, which makes it promising as a therapeutic device [11, 14]. 1.2.1 Ion transport

To induce the transport of a substance (cation M+_{), the source electrolyte}

(an-ode) is loaded with a solution of M+_{. In this initial state, the M}+ _{binds to}

the negatively charged PSS. When the voltage is applied between the source and target, as seen in Figure (4), an electron is removed and the PEDOT in the source is oxidized and then one M+ _{is released. The electrochemical}

half-reaction taking place in the anode is thus:

Figure 4: A voltage supply induces the ion transport through the channel.

The PEDOT in the target (cathode) is reduced when receiving the electron from the source oxidation, leaving the PEDOT:PSS negatively charged. Thus, the negatively charged target PEDOT:PSS is able to receive one M+ _{via the water}

in the ionically conducting channel. The electrochemical half-reaction taking place in the cathode (Eq. (1b)) is consequently reversed compared to Eq. (1a). PEDOT+_:PSS− + M+ + e− _→PEDOT0 + M+:PSS− (1b)

This M+ _{is released in the target electrolyte from the target electrode due}

to a steep concentration gradient for M+ _{over the vertical distance, since the}

height of the electrolyte (several millimeter) is much greater than the thin target electrode (∼250 nm). This release finalizes the transport of one ion. When the ratio between used electrons and transported ions is equal to one, the efficiency is said to be 100 %.

The described ion transport is possible as long as the PEDOT:PSS in the source electrode can be further oxidized from its initial state. When this is depleted, an irreversible over-oxidation is initiated. The reduction of the target electrode is not limiting since the PEDOT:PSS can be further reduced than oxidized, but also because of the fact that re-oxidization can occur by the oxygen in the surrounding air. Consequently, the capacity of the pump is decided by the amount of oxidizable PEDOT in the source electrode. In the ideal case, when there is no PEDOT left to be oxidized, the circuit will be broken due to the over-oxidation. In reality, the current decreases rapidly but the high voltages will still induce electrolysis, but most certainly it will be moved to the metal contacts and Eq. 1 will no longer be valid.

1.3

Objectives

How the efficiency is affected due to different circumstances is a property of high interest. In this geometrical study, when changing the length and/or the cross sectional area of the ion pump channel, the efficiency may be expected not to change at all, but early unpublished results shows very varying efficiency behavior.

To know how this behavior changes due to the geometry is of importance for several reasons and in several future applications. For example when the ion pump will be integrated in more complicated systems (with either several ion pumps connected together or with several ion pumps operating in parallel) it

will be necessary to have an understanding about the delivery properties, if or when the participating devices for any reasons have different geometries. Implants that will be used to control down to the single cell level require very small dimensions, while for example extracorporeal mounted pumps, with only the channel penetrating the skin, require much longer channels.

1.4

Scope

The set of ion pumps for this study (those seen in Figure 5) includes three ion channel widths; 50 µm, 300 µm and 1 mm, and four different channel lengths; 70 mm, 35 mm, 10 mm and finally 2 mm.

Figure 5: 12 pumps with varying channel geometry. Variations in the channel widths can be seen in the rows; 50 µm, 300 µm and 1 mm. The different channel lengths is seen in the columns; 70 mm, 35 mm, 10 mm and finally 2 mm.

This project includes both fabrication and measurement of the ion pumps. This project is divided into two bigger parts where the first part includes the shorter channel geometries (2 mm and 10 mm, see Figure 5). Ion transport of K+ _ions

was performed and samples were collected for concentration measurements. This yielded data to investigate and draw conclusions regarding how the efficiency of the ion pump changes due to the geometry of the ion channel.

In the second part of the project, the investigation of the pumps focused on longer channels (35 mm, see Figure 5) where the function of the devices was poor compared to the shorter channels. The focus was therefore shifted towards analysis and fabrication details regarding the encapsulation material.

2

Materials and methods

2.1

Fabrication

The fabrication was based on UV photolithography and was performed in a class 1000 cleanroom laboratory in the Link¨oping Nanofabrication Center, Norrk¨oping. UV photolithography creates desired patterns with photoresists, shadow masks and exposure to UV light. After the patterning, development of the photoresists and etching takes place. To fabricate a final ion pump device, several patterning steps with different shadow masks were performed. The fabrication steps and the materials and methods used in the fabrication are explained in more detail in the following sections.

As a starting material, we used Orgacon film (AGFA) comprising a thin film of PET foil pre-coated with PEDOT:PSS. The first fabrication step was to coat the PEDOT:PSS with the first photoresist, but to get better adhesion of that resist, the area was first coated with a thin film of poly(methyl methacrylate) (PMMA) dissolved in diethylcarbonate (DEC) [17], by spin coating. Spin coating uses rotation at high speed to spread a chemical uniformly over an area [5]. On top of the PMMA coated film, the photoresist Shipley S18XX was then spin coated over the surface. The result was three layers in the following order from bottom to top: PET foil, PEDOT:PSS, PMMA and photoresist, as seen in Figure 6.

Figure 6: Top view (left) and side view (right) of the first four layers; PET foil, PEDOT:PSS, PMMA and photoresist.

When the first photoresist was coated, the substrate was ready for the first UV exposure with the first shadow mask (Appendix Figure 26). Several pumps were fabricated at the same time on a 6-inch circular substrate where the ion pump channels had different lengths and/or widths. The shadow mask was aligned in a mask aligner, to cover and thereby protect the PEDOT:PSS and photoresist regions that would form the ion pump patterns. When the mask was properly aligned, the UV exposure took place in the mask aligner, with exposure time dependent on which photoresist and mask material was used. The UV light degrades the resist but does not affect the PEDOT:PSS layer, as shown in Figure 7. S18XX is a positive photoresist, which means that when it is exposed to UV light it becomes soluble by the developer [5] that is used in the next step.

Figure 7: UV light (shown as light blue arrows) degrades the photoresist in the areas that is not covered by the shadow mask.

After exposure of UV light, the developer dissolved the photoresist in the un-protected areas. The time needed for developing varied around 30-60 seconds dependent on the exposure time [17]. After development, the developer was rinsed away with water. A final visual inspection assured that the photoresist was properly dissolved. The result of the development is shown in Figure 8.

Figure 8: After development, only the protected and therefor insoluble areas of the photoresist remains as a pattern on the surface of the PEDOT:PSS layer.

The PEDOT:PSS layer was left unaffected by the UV exposure and the de-velopment steps, but selected patterns remained covered by the un-dissolved photoresist. To remove all the PEDOT:PSS not covered by the photoresist, a chemical reactive plasma was used in a process called reactive ion etching (RIE). The plasma used for etching polymers consisted of oxidizing O2 and a fraction

of CF4[5]. The etching removed the undesired parts of the PEDOT:PSS layer,

but not the photoresist layer entirely. The remaining layer of the photoresist and the PMMA underneath had to be removed in a separate step. PMMA is soluble by acetone, thus when cleaning the substrate with acetone the PMMA was dissolved and took the photoresist with it. The result (Figure 9) was a PET foil patterned with ion pump formations in PEDOT:PSS.

Figure 9: Etching leads to a PEDOT:PSS area shaped as the pre-designed ion pumps on the PET foil surface.

As mentioned earlier, PEDOT:PSS is both an electronically and an ionically conducting material. Since the purpose of the ion pump is to pump ions from the source to the target through the channel, it is important that the channel is only ionically conducting. This was achieved by chemical over-oxidation that breaks the π-bonds in the PEDOT, making it electronically insulating [5]. To preserve electrically conductive properties outside the channels, the same positive photoresist as earlier (Shipley S18XX) was used. The mask used and aligned in this step (Appendix Figure 27) exposed only the parts that would be over-oxidized. After alignment and UV light exposure, the photoresist was developed and the over-oxidation was then performed chemically with a chlorine solution (1% v/v NaClO in water) [17]. The result is shown in Figure 10 where the darker blue is electronically and ionically conducting PEDOT:PSS and the lighter blue is ionically conducting PSS (with over-oxidized PEDOT remnants).

Figure 10: The channel between the two electrolytes treated with over-oxidation to make it electronically insulating but still ionically conducting.

After the over-oxidation of the channel, the final step of the fabrication was to add a hydrophobic encapsulation layer. The encapsulation layer is a protective layer, and is placed everywhere except on areas for the source and target elec-trolyte chambers and where the electrodes should be contacted to the control hardware. Spin coating was used to coat the substrate with the highly viscous SU-8 (SU-8 2010, MicroChem). SU-8 is a negative resist that by cross-linking of monomers gets harder due to UV exposure. Thus, a negative photoresist has the opposite effect than the previous used positive photoresist [5]. Before UV exposure, a soft bake was performed that removes some part of the residual sol-vent. The final mask (Appendix Figure 28) was aligned in the mask aligner and

then exposed by UV light. During a post exposure bake, the SU-8 monomers cross-linked and created a rigid encapsulation layer. A final development took place to remove the un-exposed SU-8 over the electrolyte chambers and after that, the fabrication was complete. The final result can be seen in Figure 11.

Figure 11: Fabricated ion pumps ready to transport ions or molecules.

To get better contact with the electrodes, and to protect the hydrated PE-DOT:PSS from scratching, a thin layer of carbon paint was added over the area where the electrode probes should be placed.

2.2

Measurements

2.2.1 Setup

The measurement setup comprised a few preparation steps. The pump was first soaked in deionized water for at least a few hours before the measurement, to enable ion transport carried out in water. After soaking, the pumps were dried with N2 gas before placing the electrolytes and the probes from the electrical

source.

The electrolytes, salt dissolved in deionized water, were added to the source and target regions. The source electrolyte contained the ion or molecule that should be transported and the target electrolyte is any other dissolved salt. In these measurements, KCl was used as source electrolyte (that gives a transport of K+_{) and NaCl was used as target electrolyte. Since the electrolytes were}

water based, evaporation occured during the measurements. To minimize this evaporation, the pumps were placed under a plastic lid with beakers filled with water, between the two pumps that run simultaneously. The setup is shown in Figure 12.

Figure 12: The setup used during measurements. The top ion pump is placed at placement A and the bottom ion pump is placed at placement B.

Finally, a voltage between the source electrode and the target electrode was ap-plied and the ion transport initiated. The voltage was apap-plied with a Keithley 2602 source-meter controlled by a LabVIEW script. With this setup, the sourc-ing could be either constant voltage or constant current. In this study, constant current was used. The measurements took place in the class 10000 cleanroom laboratory at the Link¨oping Nanofabrication Center, Norrk¨oping. To get a data set big enough to see trends and draw conclusions, at least four pumps of each geometry with channel lengths of 2 mm and 10 mm were tested, all transporting the same transport substance, K+_.

2.2.2 Electrical

Electrical measurements comprised current vs time and voltage vs time data. An example of such data can be seen in Figure 13. The current (I) in the circuit is directly dependent on the applied voltage (U) and the effective resistance (REff) in the channel according to Ohms law (Eq. (2))

REff= U

Figure 13: Raw data from the first 1.5 mC during a measurement of a pump where the channel length is 2 mm and the channel width is 1 mm, sourced with a constant current of 1 µA (red data, right y-axis). The voltage (black data, left y-axis) is adjusted by the control hardware and software to give the preset constant current. A higher resistance in the beginning is a phase seen in almost every pump, before it finds a steady state with constant resistance.

Theoretically, resistance should be directly dependent on the channel geometry. The longer and thinner the channel is, the higher resistance. The resistivity (ρ) (decided by Eq. (3)) on the other hand is a material property, and is therefore a good control parameter to see that the pumps are working properly.

ρ = REff

A

l (3)

The amount of transported ions vs time is also necessary to decide the pumps efficiency. This was done by taking 5 samples with constant charge intervals of 1.5 mC, up to 7.5 mC, where the charge is related to the number of electrons involved in the PEDOT:PSS redox reactions (see Eq. 1). This charge should be well within the range of the pumps total capacity. The control software cuts the circuit and removes the voltage after approximately 1.5 mC. In reality it is a little bit later, so the actual charge must be decided for all runs. This actual charge is calculated as the integral of the current vs time curve (Eq. (4)).

Q = Z t

t0

Idt (4)

Eq. (4) is simplified to Q = I · t in curves with constant, steady current. 2.2.3 Chemical

The strategy used for collecting chemical samples is shown in Figure 14. The target electrolyte was initially 120 µl of 0.1 M NaCl. With intervals of 1.5 mC, samples with a volume of 80 µl were taken and the target electrolyte was replenished with 80 µl clean 0.1 M NaCl. The rest of the fluid from the previous step (approximately 40 µl) was left on the electrolyte and contributed to the

Figure 14: Samples of 80 µl are taken every 1.5 mC.

Electrolytes evaporation during the measurements also effected the concentra-tion of ions and had to be adjusted for. Two different evaporaconcentra-tion experiments were performed, with two placements of the pumps (A and B) for each exper-iment. The first experiment (Exp. 1) was performed close to a scale and the second (Exp. 2) at the measurement station located in another room, both with the setup seen in Figure 12, with the only difference that the cables were not connected in Exp. 1. The pumps were weighed without and with electrolytes and the difference gave the initial mass (m0). During one day, the loss in mass

(m) was monitored by measuring the weight with ∼1 h intervals. Assuming density (ρ) of water to be 1 g/ml, the volume (V) was decided by V = m/ρ. This data was collected as a function of volume vs time, where the slope (kn)

from the linear fit is the evaporation rate. The results of these experiments can be seen in Figure 15, where Figure 15a shows the result for placement A and Figure 15b shows the results from placement B.

(a) Exp. 1 and 2 for placement A (b) Exp. 1 and 2 for placement B Figure 15: Exp. 1 and Exp. 2 from the two placements. Exp. 2 shows a more rapid evaporation rate than Exp. 1.

The two different experiments (Exp. 1 and Exp. 2) showed two quite differ-ent evaporation rates. This can be explained by differdiffer-ent air flow at the two measurement sites, the fact that the cables in Exp. 2 allowed more air flow under the lid than in Exp. 1 or that water was lost during the walk from the measurements site to the scale. To find an evaporation rate that is as close to the actual case as possible, information from the experiments was used. When

the electrolyte had evaporated too much (when it was hard to collect the 80 µl sample), additional volume of clean 0.1 M NaCl was added (VExtra) and noted.

These notes from the experiments gave an indication of how well the evapora-tion rate corresponded to the actual case. Three alternatives existed from the evaporation experiments: the rate from Exp. 1, the rate from Exp. 2 and a mean value of the two. For pumps at placement A, the mean value of the two evaporation rates correlated best compared to VExtra, while VExtrafor pumps at

placement B correlated best with the fastest evaporation rate (Exp. 2). Also note in Figure 15 that the difference between Exp. 1 and Exp. 2 is greater for placement A than for placement B. With the results from these experiments (kAand kB), the actual volume of the target electrolyte for sample 1 (V1T) was

determined by Eq. (5)

V1T = VStart+ knt1 (5)

where k was the evaporation rate with the index n = A or B and t was time in seconds. For samples 2-5 the volumes (VT

i ) were determined by Eq. (6)

VT i = V

T

(i−1)+ VExtra+ kn(ti+ tPause) (6)

where i = 2, 3, ..., 5 and VExtra were extra volumes added due to too much

evaporation and tPause was the estimated time for taking samples, estimated to

be 120 seconds.

To decide the amount of K+ _{substance (n}S

i) in each sample, concentration

measurements of the samples were performed with flame atomic emission spec-troscopy (Perkin-Elmer AAnalyst 300) at Tema Vatten, Link¨oping. The output from these measurements was the intensity from the emission that is directly dependent on the concentration of ions in the sample. A calibration curve was made of samples with known concentration (ci), and the trend curve was later

used to decide concentration through ci = Emission/k. The amount was then

calculated as

nSi = ci·Vi (7)

for every sample i = 1, 2, ... , 5. Eq. (8) was then used to calculate the total amount of substance in the target electrolyte (first term of Eq. (8)) (nT

i) and

simultaneously adjust for the amount left from the previous run (second term of Eq. (8)). nT i = VT i VS i nS i − VT i₋₁−V S i₋₁ VT i₋₁ nS i₋₁ (8) Finally, all nT

i was summed up (Eq. (9)) to give five data points of transported

ions vs charge. ni= i X 1 nT i (9) where i = 1, 2, ... , 5.

2.3

Transport efficiency

Calculations from section 2.2.2 and 2.2.3 were used to represent the results as transported K+ _{vs charge. Information of the pumps efficiency was found in}

the slope (k) calculated by linear regression. The amount of substance was dependent on the amount of K+ _{and Avogadros number (N}

A) and the charge

is dependent on the amount of electrons (e−) and the elementary charge (e) as

shown in Eq. (10).

k = n Q =

K+NA

e−e (10)

Since efficiency (ε) is the fraction K+_/e−, Eq. (10) can be rewritten as

ε =k · e NA

3

Results

3.1

Effective resistance and resistivity

Figure 16: The effective resistance as a function of channel width to length ratio.

Figure 16 showed that the effective resistance decreases with increased width to length ratio, as suspected. The fit of the power trend line can be used to calculate the expected resistance for any length to width ratio by the expression in Eq. (12)

REff≈ 6 · 10 5

Width to length ratio (12) when K+ _{ions is the transport substance. Other transport substances are}

as-sumed to show the same behavior, but with a shifted numerator.

Figure 17: The resistivity as a function of channel width to length ratio.

In Figure 17, a spread of the resistivity with a mean value of 0.14 Ωm was seen, but the linear trend showed a fairly straight line with the expression seen in Eq. (13).

ρ = 0.03 · (Width to length ratio) + 0.14 (13) In an optimal scenario, the slope would be equal to zero since the same material induces the resistance.

3.2

Transported K

vs Charge

The amount of transported K+_{as a function of charge for at least four ion pumps}

per geometry combination is shown in the following figures together with their combined linear fit. The first shown results are from the ion pumps with 2 mm long channels (Figures 18a-c) followed by the ion pumps with 10 mm long channels (Figures 19a-c).

The ion pump channels was initially filled with ions (usually H+ _{or Na}+ _from

the fabrication steps) that had to be transported before the transport of K+

was started. This starting point can be seen in Figures 18-19 as the intersection between the linear trend and the x-axis. Naturally, the larger channel area, the more charge is required to fill the channel with K+ ions. This trend can be seen in these experiments, with an exception in Figure 19b where transport seem to occur earlier than in Figure 19c. When the line from the linear trend cutted the negative x-axis, a false value of transport start was received. This can be seen in Figure 18c. In this case, when deciding efficiency, the line from the linear regression was forced to cut the x-axis at the origin.

3.2.1 2 mm long ion pump channels

(a) Channel length: 2 mm Channel width: 1 mm

(b) Channel length: 2 mm Channel width: 300 µm

(c) Channel length: 2 mm Channel width: 50 µm

Figure 18: Transported K+_{vs Charge from pumps with channel length of 2 mm. For}

every geometry combination, the result from at least four pumps are shown together with their combined linear fit.

3.2.2 10 mm long ion pump channels

(a) Channel length: 10 mm Channel width: 1 mm

(b) Channel length: 10 mm Channel width: 300 µm

(c) Channel length: 10 mm Channel width: 50 µm

Figure 19: Transported K+_{vs Charge from pumps with channel length of 10 mm. For}

every geometry combination, the result from at least four pumps are shown together with their combined linear fit.

3.3

Efficiency vs Geometry

From the slope of the linear fit, the transport efficiency vs geometry was decided (as explained in section 2.3). The mean value of the four pumps are collected in Table 1 and the spread within each geometry can be seen in Figure 20.

Table 1: The mean efficiency for all geometries with 2 mm and 10 mm long channels.

Length: Width: Efficiency: Length: Width: Efficiency: 2 mm 1 mm 61% 10 mm 1 mm 65% 2 mm 300 µm 57% 10 mm 300 µm 59% 2 mm 50 µm 59% 10 mm 50 µm 64%

Figure 20: The efficiency of the individual pumps as a function of channel width to length ratio.

The linear fit to the data from Figure 20 gave the expression seen in Eq. (14) that showed a slope close to zero.

ε = −0.02 · (Width to length ratio) + 0.61 (14)

3.4

Ion transport in 35 mm long ion pump channels

Initially, the pumps with the longer channels, regardless the channel width, gave unreasonable high resistance or no current at all. In this section, different tests that are performed to investigate these results are described and listed together with observations from these tests. Statements regarding reasonable or unreasonable resistances are made by comparison with results from Eq. (12). 3.4.1 Soaking methods

The pumps were soaked by two different methods, all covered or partially cov-ered by water. After the soaking, the pumps were dried with N2gas, electrolytes

were placed and a voltage was applied, exactly as usual. Soaking with both methods was performed in parallel, with durations from one day to one week. The results (seen in Table 2) were independent of the soaking duration. The observations regard the resistance in the circuit and the ability to transport ions when the ion pumps are connected to the sourcing device.

Table 2: The reactions observed with two different soaking methods.

Soaking method: Observation:

Pump all covered by water Initially reasonable current level with rapid decrease down to no current at all Only electrolyte chambers No current at all

3.4.2 Water cover

A test was also made where water was added over the channel. In all cases the resistance decreased down to reasonable low or even lower level than expected. To see if there were any leakage of ions, H+ _{ions were transported and a pH}

indicator solution were placed to cover the channel area. In many, but not all cases, the pH indicator showed a decrease in pH.

3.4.3 Encapsulation materials

Additional or replaced encapsulation material was tested to see if that would im-prove the situation. The tested materials are SU-8, a silicon based glue called Medical Adhesive (Dow Corning) and a metal encapsulation where gold was evaporated over the channel area. The observations presented in Table 3 regard the transport of ions and reactions on humidity changes. To investigate the re-action to humidity changes, the plastic lid (see Figure 12) was simply removed and added again.

Table 3: The observed performances of ion pumps encapsulated with different encap-sulation materials.

Encapsulation material(s): Observations:

SU-8 2010 Initially reasonable current level with rapid decrease down to no current at all Some reaction on humidity changes Medical Adhesive Initially reasonable current level with

rapid decrease down to no current at all Strong reaction on humidity changes SU-8 2010 & Medical Adhesive Initially reasonable current level with

rapid decrease down to no current at all Some reaction on humidity changes SU-8 2010 & Gold Two of three pumps showed reasonable

resistance and a pH indicator paper indicated transport of ions to the target electrolyte

4

Discussion and conclusions

4.1

Efficiency vs Geometry

Over all, the results shows an even and constant delivery of ions and steady efficiency of ∼60% independent of geometry. The independence of geometry is a desired result since theoretically, the efficiency should not depend on geometry. But, even if the efficiency is steady, the efficiency of ∼60% is very low compared to ∼100% showed previously with K+ _{ion transport by Ref. [18].}

4.1.1 Reliability

There are a few reasons that could affect the reliability of these results. One problem is the difficulty to know the exact evaporation coefficient. A high evap-oration rate will give a low efficiency and vice versa. This mostly affects the measurements of the ion pumps with the highest resistivity, that consequently take longer time to measure up to ∼7.5 mC than ion pumps with lower resis-tance. In practice, the only geometry concentration that was really sensitive to this effect was the pump with channel length of 10 mm and channel width of 50 µm. If the data points from this geometry combination are excluded, the linear fit still shows a very straight line (ε = 0.01(Width to length ratio) + 0.60). This indicates that the overall results are not too sensitive to the evaporation rate for this study, but would increase if the study was expand to a broader field of geometrical dimensions with lower width to length ratio. A deeper investigation of evaporation behavior with additional experiments and/or operation in even higher humidity would decrease this problem.

Another fact that could effect the measurements is the fact that many of the samples were not mixed before they were collected, which would lead to a non-representative sample. Since all calculations were based on homogeneous so-lutions, one badly collected sample would affect all following calculations. In early experiments, we were possibly too concerned about the sensitivity of the hydrated PEDOT:PSS surface and did not focus sufficiently on standardized sampling, leading to inhomogenous samples. The technique of collecting sam-ples was changed for the later samsam-ples. The electrolyte fluid was removed and placed again several times (i.e., pumped in and out with the pipette) before the final sample volume was collected. After concentration measurements, the results from the well-mixed samples were compared with the early samples and no clear difference could be seen. This indicates that the sample homogeneity of samples was not a big problem for the overall results, even though occa-sional samples could be misleading. The fact that at least four pumps of every geometry combination were tested also helps to even out errors of this type. 4.1.2 The low efficiency

There are two ionic transport processes than could contribute to a current in the circuit (Figure 4): either a positive ion could be transported from the source electrode to the target electrode (that is, the intended case) or a negatively

charged ion could be transported from the target electrolyte to the source elec-trolyte. This second case should not be possible since the ion pump channel with over-oxidized PEDOT:PSS is cation selective. But, if the encapsulation layer is partially released from the channel and regions of water channels occur, the transport of negative ions is possible. An experiment was performed with NaCl in the source electrolyte and NaOH in the target electrolyte. Additionally, a pH indicator paper was placed in the source electrolyte that would reveal if the negative OH− ions were transported backwards, since that would lower the

pH in the source. The pH indicator paper is yellow in neutral solution and is shifted towards a blue colour with decreasing pH. The result of this experiment can be seen in Figure 21.

Figure 21: The pH paper shift to blue color that indicates that negative OH− _is

transported from the target electrolyte (right) to the source electrolyte (left).

An evident colour shift indicates that there is some transport of negative ions from the target electrolyte to the source electrolyte. The amount of ions cannot be decided by this experiment since this test is only qualitative. However, the results shown in section 3.3 indicate that transport of negative ions contributes up to ∼40% of the total ionic flow.

The reason why the efficiency presented by Ref. [18] is much higher than in this study is not clear, but probably those earlier devices did not transport these negative ions from target to source. The only known difference regarding fabrication is that the devices in this study are fabricated with a new Orgacon film. One possibility could be that adhesion of SU-8 to this new Orgacon is poorer than to the older one, leading to delamination and possible water-filled channels.

4.2

Water leakage through the encapsulation layer

All results presented in section 3.4 indicate that the problems with poor func-tionality of the pumps with longer channels originate in some kind of water loss. A dry channel looses the ability to transport ions, which would cause the lack of current. The channel could be drying out due to poor encapsulation (pin-holes or cracks in the encapsulation surface) or diffusion of water through the encapsulation. Of course, a combination of these reasons is also a possibility. 4.2.1 Cracks and pinholes

A microscope was used in an attempt to investigate if there was any visible ev-idence of poorly-covering encapsulation. A lot of ion pumps were investigated and in a few, findings similar to those presented in Figure 22 were found. Al-though, in the majority of the ion pumps, the SU-8 surface looked even and well covering.

Figure 22: During investigations under a microscope findings like a) radial cracks b) possible pinholes and c) linear cracks were found in a minority of the total amount of the devices.

The origin of these pinholes and cracks could possibly be explained by the process parameters while adding the SU-8 layer. As described in section 2.1 the process step of SU-8 includes a spin coating of SU-8, a soft bake, an exposure of UV light and a final post exposure bake.

Since SU-8 is highly viscous, an even spread over the surface is hard to achieve. Any bubbles that occur during the spin coating can give rise to pinholes. Hope-fully though, during the soft bake, these bubbles are evened-out. Pinholes could also occur by an uneven UV exposure. In this project, a plastic mask was used, that is not as permeable to UV light as for example a glass mask. This could give small areas of shadows under the mask that if left unexposed and thus not cross-linked and soluble by the developer. A change from the plastic mask to a glass mask could reduce this problem.

The temperature and duration of the soft bake includes a trade-off regarding the fraction of residual solvent left on the substrate, before UV exposure and post exposure bake. More residual solvent (low temperature and/or short duration of the soft bake) gives increased mobility of the monomers during the poly-merization, that results in a higher cross-link density and therefore a stronger encapsulation layer [19]. But, too much residual solvent could also cause an ad-hesion to the mask in the mask aligner during UV exposure. One attempt with a more careful soft bake was made, but some cracks were still found. Another theory is that when the substrates are baked in an oven with a fan, the substrate is heated unevenly and would leave more residual solvent in the lower regions and less at the surface. This would give a layer with an uneven strength. This problem would possibly be solved by the use of a plate oven.

All types of cracks and pinholes would definitely cause an evaporation of the water in the channel, that would give an increase in resistance and the devices would also be very sensitive to humidity. After these findings, an additional layer of Medical Adhesive was added on top of the SU-8, to close these cracks and pinholes. But, this did not improve the performance. In a few pumps, only Medical Adhesive was used as encapsulation layer to see if that would improve the results. Unfortunately, those devices showed the same increase in resistance and even more sensitive to humidity levels.

4.2.2 Evaporation of water through the encapsulation layer

Despite cracks and pinholes in the encapsulation surface, a possible diffusion through the SU-8 layer could also give the same or similar behavior as observed in section 3.4. Diffusion is a process where the flux (J) is driven by the concen-tration gradient (∇c) according to Fick’s law (Eq. 15) where D is the diffusion coefficient.

J = −D∇c (15)

According to Ref. [20], diffusion of water into SU-8 is possible, and they present a diffusion coefficient for water in SU-8 that is DSU-8= 3.0(±0.5) · 10−13m2s−1

and the saturation of water in SU-8 is ∼3.3%.

A two-dimensional steady state model of the ion pump incorporating water diffusion through the SU-8 encapsulation was implemented in COMSOL Multi-physics, where Ficks law (Eq. 15) was used to simulate the water concentration in the channel with PEDOT:PSS and in the encapsulation layer of SU-8. The boundary conditions that were set up for this model includes water concentra-tion at the boundaries to the electrolytes that matches saturated PEDOT:PSS (∼ 50% water [10]), a concentration that matches the concentration of water in air (at 30% RH and 25 °C) at the air boundary and no flux at the bottom boundary. DSU-8 was set according to Ref. [20] and scaled to ensure saturation

at 3.3%. A starting guess for the diffusion coefficient of PEDOT:PSS (DP:P) was

set to 3.0·10−12 m2s−1 according to Ref. [21] that investigated water transport

in PSS. The model represents a cross-section of the channel and an overview can be seen in Figure 23. The thickness of the two layers was set to 250 nm for PEDOT:PSS and 10 µm for SU-8. Finally, the length parameter was swept over the interval from 2 mm to 35 mm (matching most of the geometries studied

Figure 23: Schematic view of the model’s domains and geometries.

With the stated starting guess of DP:P, no channels, regardless of length, were

soaked at the steady state. This is not a realistic result, but still not surprising since the water flow through the channel is not restricted to diffusion alone, but also to surface tension and capillary action. Additionally, when a voltage is applied and the ions move towards an area that is not soaked, they drag a hydration sheath of water with them and consequently contribute to keeping the channel hydrated.

To show that these first results were not true, an experiment with an unsoaked pump was made. The electrolytes were added to the dry ion pump and then instantly connected to the sourcing device. The ion pump (2 mm long, 1 mm wide) was sourced with constant voltage of 30 V. As soon as the channel is soaked, ions can be transported and the current is induced. In Figure 24, it is clear that this occured after ∼190 s. It is therefore clear that soaking the channels only from the sides is possible, at least when the channel is short enough. At a certain channel length though, the evaporation rate across the SU-8 layer is assumed to be too large compared to the distance from the electrolytes, which will cause a drying at the channel center. To find a more realistic view

Figure 24: A voltage was added over an ion pump instantly after the electrolytes were added. As soon as the channel is soaked the current is induced in the circuit, as seen after ∼190 s.

of the water concentration and distribution in the channel, a scaling factor was added to DP:P to represent the capillary action and the contribution from the

ion transport. This scaling factor was then swept to see where the model shows the same behavior as observed in the unsoaked pump experiment. A good result was found with a scaling of 105_{, that is a big contribution to the water flow.}

Figure 25: Results from simulation of water concentration distribution at the ion pump channel and encapsulating SU-8 layer. Simulations for different channel lengths are displayed in increasing order from 2 mm to 35 mm (bottom). The color scale represents the concentration of water where blue shows maximum concentration and red shows zero concentration. The two domains (channel and encapsulation layer) has separated color scales, 0 - 27777.5 mol/m3

for PEDOT:PSS and 0 - 1833.3 mol/m3

for the encapsulation layer.

With this approach of modeling, the results should be used more as a demon-stration of the phenomena rather than an explanation of the real situation, since the scaling factor comprises a significant amount of unknown information and is tuned by observations instead of physical theory. However, these results give a hint about the possibility for SU-8 to encapsulate different channel lengths. Of course, by adding more complexity to the model, more reliable predictions could be gained. For example, the influence and reaction to humidity changes as seen in section 3.4 could be investigated further.

4.2.3 Suggested solution

The two methods from section 3.4 that resulted in functioning long channel ion pumps was i) channel covered by water and ii) metal encapsulation. However, some unwanted leakage of ions was observed with method i. Leakage of ions through cracks and pinholes could also explain the unexpected low resistance that was observed in a few devices with water covered channels, since the ions could be transported in high speed through the water and skip parts of the channel.

In applications where the ion pump is implanted as a medical device, water will always be surrounding the device but there will be a high demand on the flexibility of the device, that neither SU-8 or a metal encapsulation might fulfil. Thus, in this application a soft plastic (softer than SU-8) that allows penetration of water but not of ions could be used. For applications where the device operates in air, the encapsulation must be impermeable to water to retain the water in the channel. For this reason, metal encapsulation might be a good alternative for this application. Although, in all applications, leakage of the transport substance from cracks and pinholes must be avoided to ensure delivery of the transport substance at the desired target site.

4.3

Future prospects

This project is a starting point regarding the impact of geometry of the or-ganic electronic ion pumps. The results so far indicate that the efficiency is independent of geometry, but a wider study, especially with longer channels, would be desirable. The ion pumps with longer channels are of special interest, since channel lengths up to 70 mm could be required for extracorporeal mounted pumps for precise drug delivery, for example, delivery of neurotransmitters to the spinal cord.

However, a study with even higher priority is the investigation and choice of suitable encapsulation material. The desired encapsulation is soft and flexible, impermeable to ions (and possibly water), without cracks or pinholes, but with a good adhesion to the channel material that does not allow any transport of negative ions. When this material is found, the performance of the organic elec-tronic ion pump will increase significantly to an even higher level than today.

References

[1] M Berggren and A Richter-Dahlfors. Organic Bioelectronics. Adv Mater, 19(20):3201–3213, September 2007.

[2] R M Owens and G G Malliaras. Organic Electronics at the Interface with Biology. MRS bulletin, 35:449–456, 2010.

[3] Daniel T Simon, Karin C Larsson, Magnus Berggren, and Agneta Richter-Dahlfors. Precise Neurotransmitter-Mediated Communication with Neu-rons In Vitro and In Vivo Using Organic Electronics. Journal of Biome-chanical Science and Engineering, 5(3):208–217, 2010.

[4] Karin C Larsson, Peter Kj¨all, and Agneta Richter-Dahlfors. Organic bio-electronics for electronic-to-chemical translation in modulation of neuronal signaling and machine-to-brain interfacing. Biochimica et Biophysica Acta (BBA) - General Subjects, December 2012.

[5] Klas Tybrandt. Ionic Circuits for Transduction of Electronic Signals into Biological Stimuli. PhD thesis, Link¨oping University, Physics and Electron-ics, The Institute of Technology, 2012.

[6] Joakim Isaksson. Organic Bioelectronics : Electrochemical Devices using Conjugated Polymers. PhD thesis, Link¨oping University, Department of Science and Technology, The Institute of Technology, 2007.

[7] Nathalie K Guimard, Natalia Gomez, and Christine E Schmidt. Conducting polymers in biomedical engineering. Progress in Polymer Science, 32(8-9):876–921, August 2007.

[8] Sarah M Richardson-Burns, Jeffrey L Hendricks, Brian Foster, Laura K Povlich, Dong-Hwan Kim, and David C Martin. Polymerization of the con-ducting polymer poly(3,4-ethylenedioxythiophene) (PEDOT) around living neural cells. Biomaterials, 28(8):1539–1552, February 2007.

[9] David C Martin, Jinghang Wu, Charles M Shaw, Zachary King, Sarah A Spanninga, Sarah Richardson-Burns, Jeffrey Hendricks, and Junyan Yang. The Morphology of Poly(3,4-Ethylenedioxythiophene). Polymer Reviews, 50(3):340–384, July 2010.

[10] Tongwen Xu. Ion exchange membranes: State of their development and perspective. Journal of Membrane Science, 263(1-2):1 – 29, 2005.

[11] Daniel T Simon, Sindhulakshmi Kurup, Karin C Larsson, Ryusuke Hori, Klas Tybrandt, Michel Goiny, Edwin W H Jager, Magnus Berggren, Bar-bara Canlon, and Agneta Richter-Dahlfors. Organic electronics for precise delivery of neurotransmitters to modulate mammalian sensory function. Nature Materials, 8(9):742–746, July 2009.

[12] Payman Tehrani, Nathaniel D Robinson, Thomas Kugler, Tommi Remo-nen, Lars-Olov Hennerdal, Jessica H¨all, Anna Malmstr¨om, Luc Leenders, and Magnus Berggren. Patterning polythiophene films using electrochem-ical over-oxidation. Smart Mater Struct, 14(4):N21–N25, June 2005.

[13] X Crispin, S Marciniak, W Osikowicz, G Zotti, A W Denier van der Gon, F Louwet, M Fahlman, L Groenendaal, F De Schryver, and W R Salaneck. Conductivity, morphology, interfacial chemistry, and stability of poly(3,4-ethylene dioxythiophene)-poly(styrene sulfonate): A photoelectron spec-troscopy study. Journal of Polymer Science Part B: Polymer Physics, 41(21):2561–2583, October 2003.

[14] Joakim Isaksson, Peter Kj¨all, David Nilsson, Nathaniel D Robinson, Mag-nus Berggren, and Agneta Richter-Dahlfors. Electronic control of Ca2+

signalling in neuronal cells using an organic electronic ion pump. Nature Materials, 6(9):673–679, September 2007.

[15] Joakim Isaksson, David Nilsson, Peter Kj¨all, Nathaniel D Robinson, Ag-neta Richter-Dahlfors, and Magnus Berggren. Electronically controlled pH gradients and proton oscillations. Organic Electronics, 9(3):303–309, September 2008.

[16] Klas Tybrandt, Karin C Larsson, Sindhulakshmi Kurup, Daniel T Simon, Peter Kj¨all, Joakim Isaksson, Mats Sandberg, Edwin W H Jager, Agneta Richter-Dahlfors, and Magnus Berggren. Translating Electronic Currents to Precise Acetylcholine-Induced Neuronal Signaling Using an Organic Elec-trophoretic Delivery Device. Adv Mater, 21(44):4442–4446, November 2009. [17] David Nilsson. Process for Orgacon substrates, 10.1 edition, August 2005. [18] Daniel T Simon, Edwin W H Jager, Klas Tybrandt, Karin C Larsson,

Sindhulakshmi Kurup, Agneta Richter-Dahlfors, and Magnus Berggren. An organic electronic ion pump to regulate intracellular signaling at high spatiotemporal resolution. In TRANSDUCERS 2009 - 2009 International Solid-State Sensors, Actuators and Microsystems Conference, pages 1790– 1793. IEEE, 2009.

[19] Kristof Wouters and Robert Puers. Diffusing and swelling in SU-8: in-sight in material properties and processing. J Micromech Microeng, 20(9):095013, August 2010.

[20] Chuanjun Liu, Ying Liu, Mordechai Sokuler, Daniela Fell, Stephan Keller, Anja Boisen, Hans-J¨urgen Butt, G¨unter K Auernhammer, and Elmar Bonaccurso. Diffusion of water into SU-8 microcantilevers. Physical chem-istry chemical physics : PCCP, 12(35):10577–10583, September 2010. [21] Pilar Aranda, Wen-Janq Chen, and Charles R Martin. Water transport

across polystyrenesulfonate/alumina composite membranes. Journal of Membrane Science, 99:185–195, 1995.

Figure 26: The first shadow mask that is used in the first exposure of UV light to protect the regions where the ion pumps will be placed.

Figure 27: The second shadow mask that expose the channels of the ion pumps during the over-oxidation step.