A Medical Microactuator based on an Electrochemical Principle · A Medical Microactuator based on...

A Medical Microactuator

based on an

Electrochemical Principle

The research described in this thesis was carried out at the group Micromechanics of theMESA Research Institute at the University of Twente, Enschede, the Netherlands.

These investigations have been supported by the Netherlands Technology Foundation(STW).

De promotiecommissie:

Voorzitter:Prof. Dr. A.F. Brand University of Twente

Secretaris:Prof. Dr. F.A. van Vught University of Twente

Promotor:Prof. Dr. M.C. Elwenspoek University of TwenteProf. Dr. J.J. Kelly University of Utrecht

Assistent Promotor:Dr. J.G.E Gardeniers University of Twente

Leden:Prof. Dr. M. Reed University of VirginiaProf. Dr. M. Rusu University of BucharestProf. Dr. J. Greve University of TwenteProf. Dr. P. Bergveld University of TwenteProf. Dr. A. van den Berg University of Twente

A Medical Microactuator based on an Electrochemical PrincipleCristina Rodica. Neagu, - [S.I.:s.n.]Thesis Twente University, Enschede.-With ref.ISBN 90 365 10910Subject headings: electrochemistry / active valve / telemetry / micromechanics

Cover Design: mArt in dEsign

Copyright 1998 by Cristina Neagu, Enschede, the Netherlands

A Medical Microactuator

based on an

Electrochemical Principle

PROEFSCRIFT

Ter verkrijging vanDe graad van doctor aan de Universiteit Twente,

Op gezag van de rector magnificus

Volgens het besluit van het College voor PromotiesIn het openbaar te verdedigen

Op vrijdag 28 august 1998 te 15:00 uur

door

Cristina Rodica Neagu

Geboren op 4 september 1966Te Boekarest, Roemania

Dit proefschrift is goedgekeured door de promotoren:Prof. dr. M.C. ElwenspoekProf. dr. J.J. Kelly

En de assistent promotor:Dr. J.G.E. Gardeniers

To my parents

“Alice: Would you tell me , please, which way I ought to go from here? That depends a good deal on where you want to get to, said the cat. I don’t much care where - says Alice. Then it doesn’t matter which way you go, said the cat.- so long as I get somewhere, Alice added as an explanation.”

(From Alice’s Adventures in Wonderland by Lewis Caroll)

Contents

i

1. Introduction 11.1 General introduction 11.2 Glaucoma project 31.2 Outline of the thesis 4References 5

2. General concept of an eye pressure regulator 72.1 Introduction 72.2 Current treatment of glaucoma 92.3 Medical requirements 112.4 An eye pressure regulating system: principle 122.5 Methods to regulate the eye flow resistance 13

2.5.1 The channel diameter variation method 142.5.2 The outlet spacing variation method 162.5.3 The channel length variation method 17

2.6 Actuation mechanisms 182.7 Energy sources 192.8 Adjusting system for the eye pressure 202.9 Eye pressure sensor 222.10 Connection silicone rubber tube - actuator 232.11 Biocompatibility, packaging and encapsulation 242.12 Conclusions 25References 25

3. Design considerations for an electrochemical actuator 293.1 Introduction 293.2 Electrochemical principles 313.3 Choice of the components of an electrochemical actuator 353.4 Properties of Nafion® as semi-permeable membrane 383.5 Life-time and power consumption of the cell 403.6 Alternative counter electrodes 41

3.6.1 Antimony oxide 423.6.2 Silver oxide 47

3.7 Conclusions 49References 50

CONTENTS

Contents

ii

4. Membrane microvalve 534.1 Introduction 534.2 Deflecting membrane 54

4.2.1 Theory 544.2.2 Fabrication process 564.2.3 Characterisation 58

4.3 Fluid flow in a channel 614.3.1 Analytical model 624.3.2 Numerical model 66

4.4 Conclusions 71References 71

5. The electrochemical microactuator 735.1 Introduction 735.2 Design layout 745.3 Electrochemical reactions 765.4 First prototype of an electrochemical actuator 80

5.4.1 Clean room processing and technology 815.4.2 Experimental 845.4.3 Results and discussion 865.4.4 Conclusion for the first design 92

5.5 Second design of the electrochemical microactuator 925.5.1 Clean room processing and technology 925.5.2 Experimental 955.5.3 Results and discussion 965.5.4 Flow measurements 99

5.6 A simple dynamical model for the electrochemical cell 1015.7 Conclusions 1075.8 Appendix 5A 1085.9 Appendix 5B 110References 111

6. Energy supply and feedback controller 1136.1 Introduction 1136.2 Telemetry 1156.3 Theory of a planar coil 119

6.3.1 Self-inductance 1206.3.2 Series resistance 1226.3.3 Parallel resistance 1236.3.4 Parasitic capacitance 1236.3.5 Quality-factor 1256.3.6 Mutual inductance 126

6.4. Design and fabrication of planar microcoils 127

Contents

iii

6.5 Results and discussion 1286.6 Feedback controller 134

6.6.1 Experimental 1346.6.2 Results and discussion 1356.6.3 Pt-O2-H2 reversible electrochemical actuator 137

6.7 Conclusions 1396.8 Appendix 6A 140References 141

7. General conclusions 1437.1 The active microvalve 143

7.1.1 Electrochemical cell 1447.1.2 Deflecting membrane 1447.1.3 Valve 145

7.2 Intraocular pressure sensor 1457.3 Electronical system 145

7.3.1 Microcoil 1457.3.2 Receiver, Transmitter 1467.3.3 Feedback controller 1467.3.4 Microprocessor 146

7.4 Biocompatibility 1467.5 Overall conclusion 1467.6 Suggestions for electrochemical actuator 147

Summary 149Samenvatting 150Rezumat 151Bibliography 152Epilogue 153Biography 156

1

Chapter 1

Imagine a display from which you can get information aboutthe condition of your body at any moment, transferredcordlessly from within the body to outside without involvementfrom the doctor or oneself. Imagine tiny robots injected in thebody for a special task, for example removing arterial plaque,or repairing neural interfaces. These might sound surreal orscaring but a lot of research is going on in the field of minutesensors and actuators. The source of inspiration for theworking principles is the wonderful nature, which has hadmillions of years and numerous species with which toexperiment.

1.1 General Introduction

In the last years, several medical instruments have been brought on the market,which were fabricated with Micro System Technology (MST) [Co9301, Bl9701,El9701, Pa9101]. This technology, which uses silicon micromachining techniques tocreate submillimetre mechanical structures, originated from IC technology.Micromachining techniques include, besides the basic processing steps of ICtechnology, like film formation, doping, lithography and etching, special etching,deposition, and bonding processes which allow for the creation of three-dimensionalmicrostructures.

For medical applications, Micro System Technology offers a number of technicaland commercial advantages: (i) miniaturisation; (ii) fabrication of identical, highlyuniform and geometrically well-defined sensor and actuator elements [Na9401]; (iii)large scale production, leading to low-cost devices [El9701]; (iv) possible integrationof interface electronics on the device [Wi7001]. Sensors, electronics and actuators(micropump, microvalves) can be integrated to result in complete analysis systems[En8301, Gu9301].

For applications in the biomedical sector, the microsystems meet the samecomplex and challenging problems which are common to most medical instruments,particularly those for implantation, such as: tissue response (e.g. mechanically induced

INTRODUCTION

1

Introduction

2

tissue trauma, chemically induced tissue toxicity), immune reaction, sterilisation,electric hazard, energy supply, heat dissipation, reliability, and life time.

The major applications of microsystems in the medical field are [Fu9201,Da9601, El9701] (fig. 1.1): • prosthesis and artificial organs (orthopaedics, ophthalmology, neurology)

[Al9701, Ba9701, La9501, Me9501, No9501, Sa9301]. • monitoring (glucose, urea, calcium, sodium, blood gases, disposable invasive

blood pressure sensors, dialysis control, disposable cartridges for point of caretesting, ).

• drug delivery systems (external and implanted micropumps). • cardiology (pacemakers, defibrillator, angioplasty catheters, intravascular

diagnostics). • analysis systems (intensive care). • minimal invasive surgery (diagnosis, therapy) [Br9701, Ca9701, Ik8801, Sl9501]. • biotechnology (DNA-chips, miniaturised and integrated devices for DNA

diagnostics).

Figure 1.1. An example of a modern ‘artificial man’

Chapter 1

3

1.2 Glaucoma project

About 5 % [Re7601] of people older than 40 suffer from an elevated intraocularpressure (IOP) and for 10 % of all blind people in the Netherlands glaucoma is thecause of blindness [Re7601, Th9201]. Glaucoma is a disease causing damage to theoptic nerve head due to a too high eye pressure. This damage will lead to visual fieldloss, and finally to blindness. The elevated eye pressure has different origins such ascongenital defects, eye trauma or ageing.

To determine the eye pressure an external method is currently used. A pressuresensor called a tonometer measures the external deformation of the eye and the internaleye pressure is related to it [Be9301, Sh9201]. This type of reading is discontinuousand may require local anaesthesia.

Glaucoma can be treated with medicine. When this is found to be ineffective,glaucoma is treated surgically. The objective of surgery is to improve the drainage ofthe eye fluid by introducing a draining device. The problem is that one cannot predictin advance the value of the eye pressure after surgery; this also differs from patient topatient. The drawback of the majority of the currently available drain implants is thatthey cannot be adjusted to the optimum eye pressure of the patient, after surgery.

A continuous adjustment of the eye pressure would simplify and improve thepresent treatment. This is in fact the aim of the research presented here.

The final goal of this Ph.D. project is to develop an eye pressure regulator whichcould be combined with existing glaucoma filter implants and would allow the eyepressure to be adjusted continuously. This system will be made with the use of siliconmicromachining and thin film deposition techniques.

The main subject of the research presented in this thesis is the design, fabricationand characterisation of a microactuator which adjusts the eye pressure. A feedbackcontrol system and an energy supply system for the microactuator, based oninductively coupled coils will also be described.

Introduction

4

1.3 Outline of the thesis

The background of glaucoma and the current medical solutions are presentedconcisely in chapter 2. An example of an eye pressure regulating system is alsodiscussed. An overview of different adjustment principles and actuation mechanisms incombination with the medical requirements for an implantable actuator are alsoincluded. Based on the requirements formulated at the start of the project, the mostfeasible actuator-power transmission combination is determined. The actuatordeveloped in this project, that acts as an active valve, will control the flow resistanceby changing the deflection of a flexible membrane, due to a gas pressure differencebetween front and back sides of the membrane.

One of the medical requirements is that the energy supply should be wireless.Preferably, power should not be dissipated if no work is done by the actuator. As isconcluded in chapter 2, a device that may meet these requirements is theelectrochemical actuator, which builds up a gas pressure by electrolysis of an aqueouselectrolyte. This type of actuator is considered in chapter 3. The design of theelectrochemical cell components, such as electrodes, electrolyte, chemical reactions ismainly determined by the compatibility with the silicon micromachining fabricationtechniques, which will be used, and by the medical requirements. The type ofelectrodes and ions in the aqueous electrolyte determine the kind of gases which aregenerated at the electrode(s). A general introduction to electrochemistry will be firstpresented followed by a description of electrode materials, that may be used in theelectrochemical cell.

A valve system that consists of a drainage flow channel and a deflectingmembrane are investigated in more detail in chapter 4.

Chapter 5 describes the design, fabrication and characterisation of twoelectrochemical microactuator prototypes. A problem is encountered with the long-term stability of the generated pressure inside a sealed electrochemical cavity (withoutusing any energy supply). Two designs which aim to minimise this decrease in pressureby protecting the counter electrode with a permselective membrane are discussed.

In chapter 6 the energy supply and a feedback controller are described. First, thewireless power supply by means of inductively coupled coils for implants is addressed.Then, an electronic feedback system to control automatically the pressure of anelectrochemical macrocell is presented.

Finally, chapter 7 summarises and concludes the research described in this thesis.

Chapter 1

5

References

[Al9501] B. Allota, P. Dario, N. Sorrentino, M. Marcacci, S. Larsen, Experiments towardthe design of mechatronic tools for orthopedic surgery, Proc. 9th World CongressTheory Machines Mechanisms, Milan, Italy, p. 2156-2160, 1995.

[Ba9701] F. Bartels, New implants for middle-ear surgery, MST News, no. 19, p. 14-15,1997.

[Be9301] K. van Besten, Sensor system for the measurement of intraocular pressure,Ph.D. thesis, University of Twente, Enschede, the Netherlands, 1993.

[Bl9701] P. Bley, U. Knapp, Microsystems technology in medical engineering, MST News,no. 19, p. 10-12, 1997.

[Br9701] B. Bröcher, Miniature laser endoscope, MST News, no. 19, p. 12, 1997.[Ca9701] M.C. Carrozza et al., The development of a microrobot system for colonoscopy,

Proc. CVRMED II - MRCAS, Grenoble, France, March 1997.[Co9301] Copeland Economics Group, Market Overview, February 1993.[Da9601] P.Dario et al., Micromechanics in medicine, IEEE/ASME Trans. On

Mechatronics, 1(2), p. 137-148, 1996.[El9701] J.C. Eloy, Microtechnologies, Microsystems: Medical applications, MST News,

no. 19, p. 7-9, 1997.[En8301] J.M.L. Engels, M.H. Kuypers, Medical Applications of silicon sensors, J. Phys.

E.: Sci. Instrum., 16, p. 987-994, 1983.[Fu9201] I. Fujimasa, Future medical applications of microsystem technologies, Micro

System Technologies’92, H. Reichl (ed), p. 43-49, 1992.[Gu9301] W. Gumbrecht, D. Peters, W. Schelter, W. Erhard, J. Henke, J. Steil, U. Sykora,

Integrated pO2, pCO2, pH sensor system for online blood monitoring, EurosensorsVII, Budapest, Hungary, 1993.

[Ik8801] K. Ikuta, M. Obama, F. Ozaki, K. Asano, Shape memory alloy servo actuatorsystem with electric resistance feedback and applications for active endoscopes,Proc. IEEE Conf. Robotic Automa., Tokyo, Japan, p. 427-430, 1988.

[La9501] R. Lazzarini at al., A tactile sensor layered in artificial skin, Proc. IROS ’95,Pittsburgh, USA, August 5-9, p. 114-119, 1995.

[Lö9701] J.C. Lötters, A highly symmetrical capacitive triaxial accelerometer, Ph.D. thesis,1997, University of Twente, Enschede, the Netherlands.

[Me9501] J-U Meyer et al., Perforated silicon dice with integrated nerve guidance channelsfor interfacing peripheral nerves, Proc. IEEE Micro Electrical MechanicalSystems, MEMS’95, 29 Jan-02 Feb., Amsterdam, the Netherlands, p. 358-361,1995.

[Na9401] K. Najafi, Solid-state microsensors for cortical nerve recordings, IEEE Eng.Medicine Biology Mag., June/July, p. 375-387, 1994.

[No9501] R.A. Normann, Visual neuroprosthetics, IEEE EMB Magazine, January/February1995, p. 77-83.

[Pa9101] Biomedical Technology Information Service, A.F. Pacela (ed), Nr. 7, 1991.[Re7601] J.W.G.A. Rens, Glaucoma simplex in de huisartsenpraktijk, Ph.D. thesis,

University of Utrecht, the Netherlands, 1976.[Sa9301] G. Sandini et al., Retina-like CCD sensor for active vision, in Robots and

Biological Systems, by P. Dario, G. Sandini, P. Aebischer (Eds), Springer-Verlag,p. 553-570, 1993.

[Sh9201] M.B. Shields, Textbook of glaucoma, 3rd ed., Williams & Wilkins, 1992.[Sl9501] B. Slatkin, J. Burdick, W. Grundfest, The development of roboticndoscope, Proc.

’95 Int. Conf. Robot. Automa., Nagoya, Japan, p. 162-171, 1995.[Th9201] J.V. Thomas, Glaucoma surgery, Mosby-Year Book Inc., St. Louis, 1992.[Wi7001] K.D. Wise, J.B. Angell, A. Starr, An integrated-circuit approach to extracellular

microelectrodes, IEEE Trans. Biomed. Eng., vol. BME-17, p. 238-247, 1970.

Chapter 2

7

This chapter describes the eye disease glaucoma. An eyepressure regulating system for the treatment of this disease ispresented. A brief review of different pressure adjustmentmechanisms and actuation principles, in combination with themedical requirements, is described. The most feasibleadjustment-actuation combination is selected for a moredetailed investigation in the following chapters.

2.1 Introduction

The anatomy of the eye with parts important for glaucoma research is shown infig. 2.1.1. The outer part of the eye consists of the sclera and the cornea. The cornea istransparent and is, in combination with the lens, the most important part of the opticalsystem of the eye. The sclera and cornea are kept under tension with the intraocularpressure (IOP) of the eye fluid. In this way the shape of the eye ball is preserved. Ontop of the sclera there is another transparent skin layer, called conjunctiva.

The eye fluid, called aqueous humour, is produced in the posterior chamber ofthe eye at the processus ciliares from the circulating blood, see fig. 2.1.1. It consists for99.1% of water. The exact mechanism of formation of aqueous humour and outfloware not yet well understood [Bi9201, Kr8801, Kr9201]. The eye fluid flows from theposterior chamber through the pupil to the anterior chamber and it is drained away viathe trabecular system into the channel of Schlemm which is located in the corner of thechamber. The channel of Schlemm passes the eye fluid through the sclera into thevenous system. An increase in the intraocular pressure occurs if too much eye fluid isproduced or if the eye fluid encounters an abnormal fluid resistance somewhere on theway to the venous system.

Intraocular pressure (IOP) elevations are considered potentially harmful whenocular hypertension is maintained for longer than 1 day. In that case the optic nervemay be damaged. This defect, which is called ‘glaucoma’, may ultimately lead toblindness. About 10 % of all blind people in the world [Th9201], and in theNetherlands 5 % of the people older than 40 suffer from an elevated IOP and haveglaucoma as the cause of their blindness.

GENERAL CONCEPT OF A GLAUCOMA

REGULATOR

2

8

General concept of a glaucoma regulator

Glaucoma is not a single disease process but rather a large group of disorderscharacterised by widely diverse forms of manifestations. This makes it difficult to givea general definition of glaucoma that includes all different types of disorders. Apossible definition is: “those situations in which the optic nerve head is damaged by atoo high IOP causing visual field loss” [Co8001]. However, in some cases a disorder inthe blood flow to the retina, like thrombosis, can induce glaucoma while the IOP canbe normal. Three common parameters: IOP (in most cases high IOP due to a decreasein aqueous outflow), optic nerve head damage and visual field loss represent thecommon pathway to blindness with all forms of glaucoma. It is not clear whether theinjury of the optic nerve head is a direct mechanical effect of elevated IOP (physicalalterations in the optic nerve head) or secondary to vascular changes (diminishing ofthe blood supply to the optic nerve).

An important factor that affects the IOP is the blood pressure in the vessels ofthe eye shell. The blood pressure can vary due to the cardiac cycle, the respiration, andthe position of the patient, leading to an ocular pulse wave, which is superimposed onthe IOP. The variation of the blood pressure causes the internal volume of the eye tovary. The reported mean amplitude of the daily fluctuation ranges from ca. 400 - 800Pa (3-6 mmHg). The average amplitude of the pulsating pressure is about 370 Pa innormal eyes, but in individual eyes the amplitude may have a range between 70 and1100 Pa. The IOP fluctuation due to respiration is 130 - 270 Pa and it is relatively slowcompared to the fast pulses caused by the cardiac cycles. The position of the patientgives an average difference of 400 Pa in the IOP between lying and sitting [Sh9201,Pe8101].

The normal IOP is in the range 1300 to 3000 Pa above atmospheric pressure andthe average is around 2250 Pa. An ill eye has a pressure of up to 3 times higher butthere are people who develop glaucoma at a lower IOP and some who do not developglaucoma at higher IOP. It is considered that for glaucoma the IOP has values in therange of 1800 to 9100 Pa with a standard deviation of 1500 Pa. The mean pre-surgicalIOP is of 4866 Pa. The success in surgical glaucoma treatment is achieved when thereis a reduction in IOP to less than 2900 Pa, and not lower than 700 to 1300 Pa. Thefailure of the surgical treatment is for an uncontrolled IOP higher than 2900 Pa. Theeye fluid production is about 1 - 3 µl.

Figure 2.1.1. Cross-section of the human eye and magnification of the limbus,composed from ref. [Th9201] and [Sh9201].

Chapter 2

9

2.2 Current treatment of glaucoma

The treatment of glaucoma consists of medication or surgery. In general,glaucoma simplex will be treated with medication (eye drops, orally). These drugseither reduce production of aqueous humour or enhance drainage from the eye.Depending on the form of the patient’s glaucoma, medicine can loose its effect over aperiod of time or can give little or no effect. If the medicine does not induce therequired effect, surgery can be an option. Depending on the kind of glaucoma, differenttypes of surgery can be done, such as operations to form paths for intraocular drainageof aqueous humour; to relieve intraocular block of aqueous humour; to reduceproduction of aqueous humour; to improve extraocular drainage of aqueous humour,etc. These types of operation are performed using laser therapy, “incision” surgery ormicrosurgery. Improvement of extraocular drainage of aqueous humour can be doneby the filtering operation, a specific class of microsurgery. Here, openings are made atthe limbus (trabeculectomie). These filtering procedures are characterised by thegrowth of a filtering bleb, an elevation of the conjunctiva at the surgical site. Now, theaqueous humour is caught by the bleb and filters through it. It is then mixed with tearfilm or absorbed by surrounding veins. The drawback of this procedure is that theincisions heal after a period of time, so repetition of the procedure is likely. Anotherdisadvantage of this treatment is that the IOP is not controllable, i.e. the IOP can stillbe too high or too low after the surgical treatment. A low IOP can lead to hypotony,which will lead to a flat anterior chamber with subsequent choroidal detachment or acataract formation (clouding of the lens).

In a later stage, implantation of a drain device is performed to prevent healing ofthe filtering operation (i.e. to prevent the closure of the openings made with the laser)or to enhance draining capability. Over the years, various devices have been developedand all of them have their drawbacks and limitations. Fairly good results have beenreported for a so-called seton or Glaucoma Filter Implant. A Glaucoma Filter Implantcreates an alternative aqueous pathway by channelling eye fluid from the anteriorchamber through a long silicone rubber tube to a plate, the ‘bleb promoting device’(BPD), acting as a reservoir, promoting the formation of the bleb (see fig. 2.2.1). Themost experience has been obtained with the Molteno device (fig. 2.2.1). The blebpromoting device, developed at the Academic Medical Centre in Amsterdam (AMC) issimilar to the Molteno device but instead of one plate it has two plates to increase thearea of the bleb, see fig. 2.2.2.

(a) (b)Figure 2.2.1. (a) Molteno implant, (b) Cross-section of a human eye where theMolteno device is implanted on top, the tube is led to the anterior chamber, picturesfrom ref. [Th9201]

10


These glaucoma filter implants work according to the ‘resistor’ principle, i.e. thedraining silicone rubber tube has a certain inner diameter and length which determinesthe fluid resistance ‘resistor’. In the case of glaucoma, the production of eye fluid, Φin,and its outflow, Φout, are not in balance. Therefore, by using a drain tube withappropriate dimensions, i.e. resistance, equilibrium at normal eye pressure may berestored, see fig. 2.2.3. The pressure difference between eye fluid and outlet is:

∆p p peye out= − (2.2.1)

Since the inlet and outlet flows have to be equal, the tube of resistance R controls theeye pressure:

Φin = Φout = ∆p/R (2.2.2)

One of the main problems with conventional drainage implants is the inability toprovide lasting aqueous draining due to fibrous tissue proliferation around the device.Therefore, to compensate for the body tissue encapsulation, the BPD can be made verylarge so the drain capabilities remain sufficient. The encapsulation can be slowed downby the use of fibrosis inhibitors but the problem is not solved. Moreover, the innerdiameter of most drainage tubes is too large and this can lead to hypotony (low IOP).Another disadvantage of the current glaucoma filter implants is that they cannot beadjusted after implantation. The optimal IOP is not known beforehand, which meansthat adjustment after implantation may be necessary.

Problems encountered with present-day implants include [Gr9601]: • although the tube is protected from direct obstruction by the collecting reservoir,

fibrosis remains a problem; • after surgery, the IOP is unpredictable, generally not stable in the postoperative

period, and not adjustable to the optimum IOP of the patient; • the passive implants can barely compensate the time varying IOP.

Figure 2.2.2. The glaucoma filterimplant used at the Academic CentreAmsterdam, group of Prof. Dr. Greve.

Figure 2.2.3. Sketch of the ‘resistor’principle to adjust the IOP with the useof glaucoma filter implants.

peyeΦin

pin

Φout

pout

implant

Chapter 2

11

2.3 Medical requirements

The eye pressure controller to be developed here should satisfy the followingmedical requirements, as defined by the Academic Medical Centre in Amsterdam[Gr9101]:

R1. after implantation of the eye pressure regulating device the IOP has to be adjustedto the optimum IOP of the patient since this is not known beforehand and variesfor different patients. The IOP range is from 700 Pa to 2700 Pa and theadjustment should have an accuracy of 130 Pa;

R2. a low IOP (hypotony) has to be avoided under all circumstances; otherwise,irreversible damage to the eye occurs within hours;

R3. the maximum size of the eye pressure regulator is 5 x 5 x 2 mm3;R4. the power supply to the eye pressure regulator has to be cordless;R5. the connection between the anterior chamber and the BPD (bleb-promoting

device) is a silicon rubber tube of 0.3 mm inner diameter and a length of about1.5 cm. The inner diameter should not be decreased permanently more than 0.03 -0.05 mm because of expected protein clogging. The conventional tubes have atube wall thickness of 0.15 mm. A thinner wall may make the tube too fragile;

R6. the device, after implantation, should not require any maintenance;R7. the device should be fail-safe; an external perturbation must have no influence on

the eye pressure adjustment; if such cases still occur, an alarm signal should warnthe patient;

R8. the device should operate correctly for a long period, preferably 10 years;R9. it should be waterproof and shockproof;R10. in an advanced model, the total regulating system should have an integrated

pressure sensor.

In the course of this project, requirement R5 changed. At the beginning of theproject it was thought that the inner diameter should not be decreased permanently,below 0.3 mm because after the surgical treatment of the eye, protein particles mayexist in the eye fluid with a diameter up to 0.3 mm. In the course of some years ofsurgical treatment of the eye using the AMC bleb design it was observed that a tube of0.3 mm inner diameter, produces too much outflow immediately after surgery and theeye pressure drops below the allowable value [Gr9601]. Therefore, a stainless steelwire of about 0.2 mm is inserted into the silicone rubber tube to prevent a lowpressure, and it is assumed that the spacing between steel wire and tube isapproximately 0.03 to 0.05 mm.

The maximum allowable space for the total implant is quite small (requirementR3), so the system is made using silicon micromachining techniques. These techniqueshave some advantages over conventional (fine) machining techniques; because theactuator is made on a silicon wafer, at least 100 actuators of 5 x 5 mm2 can beproduced on a 3” wafer, so batch processing is possible. This will greatly reduce theprice of the eye pressure regulator. The possibility of integration with ‘on chip’electronics is also an advantage, e.g. for a feedback and transmission system (see nextsection).

12


2.4 An eye pressure regulating system: principle

To overcome the problems experienced by the BPD, a regulating devicecombined with the BPD will give the possibility to monitor and control the intraocularpressure. A system that provides a continuous electronic regulation of the eye pressurearound a desired pressure value, is shown schematically in figure 2.4.1. The main partsof this system, which will be described in this thesis, are the microactuator, the IOPsensor, the feedback control system and the power supply.

The microactuator will be used to deflect a membrane which changes the flowresistance of the implant. The microactuator has to be reversible, i.e. it has to increaseand decrease the flow resistance as a function of the IOP pressure. In addition, as littleenergy as possible should be consumed in the stand-by state of the microactuator.

Without an implantable pressure sensor, the IOP can be measured externally (asis done currently) and the implanted actuator set to the desired pressure. It is assumedthat when the IOP has become more stable, i.e. after an extended period after theimplantation, the IOP has to be adjusted every week or every month, depending on thefluctuations of the IOP; this makes the external pressure reading and adjustmentinconvenient. A pressure sensor added to the system would be advantageous for thecontrol of the actual eye pressure. It is possible to have either an interface between thesensor and actuator and still have two separate devices or integrate them in a completeintraocular pressure control unit. A prototype of an implantable pressure sensor wasrealised and tested by Rosengren et al. at Uppsala University, Sweden [Ro9301].Different types of pressure sensors that may be used are discussed in a later paragraph.

A feedback controller and a microprocessor are necessary for informationexchange not only among the implanted parts of the regulating system but alsobetween the internal and external regulating systems. The doctor or the patient needsto have some indication of whether the transmitter is placed correctly and maximumenergy transfer is achieved. Furthermore, since powering of the internal regulatingsystem is cordless the transmitting system should receive some kind of informationfeedback. This could be done by the eye pressure regulator itself. As soon as thetransmitter is placed correctly, the regulator should send signals back forsynchronisation and for device identification. The feedback may be generated bypassive signalling. The major advantage of this type of signalling is that no internaltransmission circuitry has to be designed.

In the following sections some possible methods to regulate the eye flowresistance, actuation mechanisms and energy sources are discussed. At the end of thischapter, the most feasible principles will be chosen.

Figure 2.4.1. Schematicrepresentation of a total eyepressure regulation system.The whole regulator isinside of on the eye, exceptfor the remote coil.remote

coilhostcoil

k

inductivecoupling

actuator

Φoutpatm

flexiblemembranesilicone rubber tube

peye

Φin

Feedback controller+ microprocessor

PPressuresensor

Chapter 2

13

2.5 Methods to regulate the eye flow resistance

The flow resistance can be adjusted in three ways: the variation of the inner crosssection of the flow channel, the variation of the out-flow area at the end of the tube,and the variation of the flow channel length. The principles are shown in fig. 2.5.1.

To build up an appropriate IOP, it is necessary to know the pressure differenceacross the tube and that due to the restriction. First, the pressure difference across thetube is calculated. For a horizontal tube, the pressure drop across the tube, ∆p [Pa], isdirectly proportional to the volume flow, Φ [m3/s], for a fully developed, laminar flow[Wh9401]:

∆ Φp R= ⋅ (2.5.1)with R [Ns/m5] the channel resistance, a constant that depends on the tube geometry.For a circular tube R is given by:

Rl

d=

⋅⋅

1284

µπ

(2.5.2)

where µ [Pa⋅s] is the viscosity of the fluid, l [m] the tube length and d [m] the tubediameter. The pressure drop for a uniform, circular tube with laminar flow is:

∆ Φpl

d=

⋅⋅

⋅128

4

µπ

(2.5.3)

A laminar flow is developed when the Reynolds number Re is smaller than 2300:

Re=⋅ ⋅ρµv d

(2.5.4)

where ρ [Kg/m3] is the fluid density, v [m/s] is the average velocity of the fluid givenby, where A [m2] is the inner cross-sectional area of the tube:

v = =Φ ΦA d

42π

(2.5.5)

As an example, consider a typical silicone rubber tube, as is implanted in theanterior chamber of the eye. This tube has a length l = 2 cm, diameter d = 300 µm witha flow Φ = 2 µl/min = 3.3⋅10-11 m3/s of water with viscosity µ = 1⋅10-3 Pa⋅s, anddensity ρ = 1000 kg/m3. Since the Reynolds number Re = 0.14 << 2300, the flow in thetube is laminar. The pressure drop across the tube calculated with eq. 2.5.3 is ∆p = 3.3Pa (0.025 mmHg). The pressure drop across the silicone rubber tube is thereforenegligible compared to the normal IOP, ~2250Pa (17mmHg). This value demonstratesthat such a tube does not act as a significant flow restriction for eye fluid drainage.

(a1)

Figure 2.5.1. The flow resistance is adjusted (a) by changing the inner cross sectionarea of the tube, either deforming the tube or introducing a valve inside it; (b) bychanging the spacing at the end of the tube, and (c) by changing the channel length.2.5.1 The channel diameter variation method

Eye fluid

P

Eye fluid

(b) (c)

(a2)

14


There are two principles for varying the cross-section of the eye fluid flowchannel: (a) deformation of the silicone rubber tube, for example by buckling, bendingor squeezing, and (b) introduction of a valve into the tube.

(a) In order to determine the regulation possibilities some parameters areconsidered. First, the diameter reduction of the tube will be calculated necessary toobtain a normal IOP, ~2250 Pa above the atmospheric pressure,. It is assumed that thepressure at the tube outlet equals the atmospheric pressure so that a pressure drop of2250 Pa along the deformed tube has to be achieved. Supposing that the channel iscircularly deformed over a length l = 5 mm, the required diameter is 40 µm (eq. 2.5.2).It is obvious that it will require large forces to decrease the inner diameter of the tubewith a 150 µm wall thickness from 300 µm to 40 µm; a deformed circular shape will bevery difficult to obtain. In fact, what will happen is that when an external pressure isapplied to the tube, larger than a critical pressure, the tube will buckle (fig. 2.5.2). Thedistributed buckling load qcr (force per unit perimeter) of a circular tube is given by[Ti6801, Ge8701]:

qEI

acr =3

3 (2.5.6)

where E [Pa] is the Young’s modulus of the tube material, a [m] is the mean radius ofthe tube, I [m4] the second moment of inertia given by:

Ilh

=3

12(2.5.7)

with h and l are the thickness of the tube wall and the length of the tube over which theload is applied, respectively. Thus the buckling pressure is:

pq

l

E h

acrcr= =

4

3

(2.5.8)

Conventional silicone rubber tube has an inner diameter of 300 µm, the wall thicknessh = 150 µm, outer diameter of 600 µm, so a = 225 µm, and Young’s modulus E = 1-4GPa. This results in a buckling pressure in the range 0.74⋅108 Pa < pcr < 3⋅108 Pa (eq.2.5.8); this is a very high pressure.

The sensitivity of the IOP to a change in the diameter of the tube as long as it iscircular (so before buckling) is given by:

∂∂

µπ

∆ Φp

d

l

d= −

⋅ ⋅ ⋅⋅

5125 (2.5.9)

The sensitivity in the operating point d = 40 µm, ∂∂∆p

d = -263 Pa/µm (1.6 mmHg/µm).

According to medical requirement R1 (chapter 2.3) for 130 Pa accuracy this means thediameter has to be controlled to within 0.6 µm. This will be very difficult to achieve,especially when a very high external pressure is necessary to deform the tube.

(b) The second method to change the innercross section area is the introduction of a valve into the flow channel (fig. 2.5.1 a2).The pressure drop across a valve can be estimated with a channel as schematically

Figure 2.5.2. The silicone rubber tubebuckles under an external pressure.

p > pcr

Chapter 2

15

shown in fig. 2.5.3. The channel has a square cross-section of side b [m], the step-likevalve has a length l [m] and a gap spacing yi [m]. Considering this geometry of thechannel, the pressure drop may be calculated by using [Sh7801]:

∆ Φpk l

D Ashape

h

=⋅ ⋅ ⋅

⋅⋅

22

µ(2.5.10)

where A [m] is the flow area, Dh [m] is the hydraulic diameter, defined as Dh = 4⋅A/P,the wetted perimeter of the flow cross-section area is P [m], kshape is a constant thatdepends on the shape of the flow cross section area (for a square area kshape = 14.2, fora circle 16, and for a rectangular area 24 (in our case the valve changes the squarecross-section of the flow)), and the rest of the terms are as defined before. Thepressure drop over the valve, is shown in fig. 2.5.4 as a function of gap spacing for twodifferent channels: (i) 300 µm width, length of the valve, l = 1 mm and 2 mm, and (ii)50 µm width, l = 1 mm. In the simulation, a fluid flow Φ ≈ 2 µl/min = 3.3⋅10-11 m3/s,and viscosity µ = 1⋅10-3 Pa⋅s are taken. From this figure it can be seen that a smallergap distance, i.e. a higher valve deflection, is necessary to obtain a pressure dropacross the valve equal to the eye pressure, 2250 Pa for a wider channel than for anarrow channel. Thus, this method is basically suited to adjust the eye fluid flow.

Figure 2.5.3. Schematic showing the change of the inner cross-sectional area of thetube by inserting a valve (restriction) into the tube.

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

0 2 4 6 8 10 12 14 16 18 20Gap spacing [µm]

Pre

ssur

e ac

ros

gap

[Pa]

Peye

b = 50 µml = 1 mm

b = 300 µml = 1 mm

b = 300 µml = 2 mm

Figure 2.5.4. Calculated pressure drop over the valve as a function of the gapspacing yi, for the channel geometry shown in fig. 2.5.3.

b

b

l

pEye flow

16


2.5.2 The outlet spacing variation

The flow resistance at the end of the tube can be changed by moving a stiff plateor by deflecting a flexible membrane, see fig. 2.5.5. The pressure drop across the flowresistance can be approximated by that between two parallel plates, which is:

∆ Φph

b l=

⋅⋅

123

µ(2.5.11)

where h [m] is the width of the plates (i.e. the tube wall thickness), b [m] is thedistance between the plates (i.e. the spacing between the end of the tube and the plate),l [m] is the width of the plates (i.e. the circumference of the mean radius of the innertube and outer tube, 2⋅π⋅a, with a the mean radius). For an IOP of 2250 Pa, Φ=3.3⋅10-

11 m3/s, wall thickness h = 150⋅10-6 m, eq. 2.5.11 yields a gap spacing, b = 2.65 µm.

The sensitivity of the IOP is obtained by differentiating eq. 2.5.11 with respect tothe gap spacing b:

∂∂

µ∆Φ

p

b

h

b l= −

⋅⋅

364 (2.5.12)

The sensitivity in the operating point, b = 2.65 µm, equals 1280 Pa/µm. According tomedical requirement R1 (chapter 2.3), an IOP accuracy of 130 Pa means that the platehas to be positioned with 0.1 µm precision. It is clear that a precise control of the plateis necessary to keep the IOP constant. The force necessary to move the plate is aboutthe eye pressure multiplied by the cross sectional area, i.e. 2250Pa⋅π⋅1502µm2=160µN.

In the case of a membrane, a pressure is used to deflect the membrane. For a flat,square membrane of thickness h [m], length 2a [m], Young's modulus E [Pa],Poisson's ratio ν, and initial stress σ [Pa], the relationship between the pressuredifference P [Pa] over the membrane and the corresponding centre deflection y [m] isdescribed by [Gi8401, Je9001, Pa9001]:

Ph

ay

Eh

ay= ⋅ + ⋅

⋅ −−

⋅3 41 198 1 0 295

12 43. . ( . )σ ν

ν(2.5.13)

As can be seen from eq. 2.5.13, the deflection is a highly non-linear function ofpressure. To have a deflection of 20 µm, for example, for a 1 x 1 mm2 flat siliconmembrane of 1 µm thick, a pressure P ~ 0.5⋅105 Pa (0.5 atm) is required. To decreasethis pressure a more flexible material (with a lower Young’s modulus) has to be usedfor the membrane. Another possibility is to use a membrane with corrugations[Gi8401, Je9001], which allows a larger deflection at the same applied pressure, or touse a membrane buckled to the operating point [Po9501], see fig. 2.5.6. Thesepossibilities are presented in more detail later. It can be concluded that this methodprincipally is suitable for the adjustment of the eye pressure.

Figure 2.5.5. Schematicdrawing of the end of the tube(a) by moving a plate and (b)by deflecting a flexiblemembrane [Ij9201].

(a) (b)

Chapter 2

17

Figure 2.5.6. A membrane with (a) corrugations and (b) a pre-deflection, may be usedto increase the deflection as function of the pressure [Ij9201].

2.5.3 The channel length variation method

As can be seen from eq. 2.5.3, the pressure drop across the tube, ∆P, isproportional to the length of the tube, l. This means that the flow can be regulated bychanging the channel length. The sensitivity of the IOP to a change in the length of thetube is a constant of the tube and flow parameters:

∂∂

µπ

∆ Φp

l d=

⋅ ⋅⋅

1284 (2.5.14)

It can be seen that the regulation sensitivity is lower than that of the channel diametervariation (eq. 2.5.9), and that the regulation sensitivity is independent of the operatingpoint, in contrast to the diameter variation method.

This principle has been used to develop a passive valve that regulates the IOP ina discrete manner. A manually operated micromechanical actuator was developed inour group by van Toor et al. [To9701] in collaboration with Prof. Dr. Greve, AMCAmsterdam. The basic idea is that the length of the channel can be adjusted by a slidingmechanism, shown in fig. 2.5.7. By moving the slider different resistance values can bechosen. The valve with dimensions of approximately 1.5 x 2.5 x 0.6 mm is made insilicon using silicon micromachining techniques. Fabrication, experimental results andmore details can be found in reference [To9701].

Disadvantages are the limited range of the resistance variation and the use ofonly a discrete instead of a continuous change in resistance. If it is manually adjustable,the possibility of integration with a feedback system (including a pressure sensor, datatransmission, and control of the regulator status) is excluded. These aspects make thisprinciple less suited for a complete eye pressure regulating system.

Figure 2.5.7. Schematic drawing of thechannel length regulator [To9601].

(a) (b)

18


2.6 Actuation mechanisms

Different actuation mechanisms may be used to decrease the cross-sectional areaof the tube or to move the plate or membrane at the end of the tube. Table 2-1 showssome possible actuators [Yn9201].

A magnetic actuator can be used to rotate parts of the eye pressure regulator bymeans of torsion generated by a magnetic field (drawing 1 in table 2-1).

An electrostatic actuator (drawing 2) involves two film electrodes attached toeach other at the ends, with an insulating layer in between. When a voltage is appliedbetween the electrodes, the electrostatic force will pull them together resulting in adecrease of the cross-sectional area of the tube.

Drawing 3 shows an electrochemical actuator that generates a gas pressure byelectrolysis of an electrolyte (e.g. water). The gas pressure is used to change thedeflection of a flexible membrane, which in turn can be used to move a part of the eyepressure regulator. When the pressure has to be reduced, the electrodes can be shortcircuited or the polarity of the applied current reversed.

An electro osmotic actuator is shown in drawing 4; when an electric current ispassed through a membrane, a flow of ions is induced. This leads to a pressuredifference across the membrane.

Drawing 5 shows schematically a micromechanical pump [Po9001], that may beused only when it can be made small enough for implantation in the eye.

An electrohydrodynamic pump [Ri9101, Ri9201] is shown in drawing 6. Thispump can be made small enough with the help of micromachining techniques but needshigh voltages (up to several hundreds volts) and it can only pump non-conductingliquids.

Drawing 7 shows an actuator which uses a liquid that boils at low temperature,e.g. 35°C. In this way a small temperature change will change the pressure in the box,deflecting a flexible membrane. For example, ethyl-ether C2H5OC2H5, with a boilingtemperature of 34.5°C has a pressure sensitivity ∆p/∆T = 4150 Pa/K [CR9301].

Drawing 8 shows an actuator in which a gas is heated to increase the pressureinside the box [Po9001].

Table 2-1. Some possible actuators that may be used to regulate the eye flowresistance [Ij9201].

Chapter 2

19

2.7 Energy sources

There are several ways to generate electrical or mechanical power in the eye, ascan be seen in table 2-2 [Yn9201].

The vibration of the eye to generate electrical energy is shown in drawing 1 and2 of table 2-3. These eye vibrations called tremors have a frequency of about 80 Hzand an amplitude of 1-2 µm. In drawing 1, a membrane with a fixed charge may vibratedue to the eye vibration. This membrane-charge vibration generates a current and avoltage, which can be accumulated in a capacitor. In drawing 2, a piezoelectric layergenerates a current and a voltage. The power produced by this type of generator is inthe order of 10-9 - 10-14 W [Ko9101].

A battery, drawing 3, may be used but it should operate at least as long as thelife-time of the eye pressure regulator, and it also has to be small enough to beimplanted.

Another way to transport energy to the eye regulator is by means of coupledcoils, drawing 4. This technique is reported to transport electrical power in the rangeof µW to mW, when the coils are close enough to each other (telemetry). We shalldiscuss this in detail in chapter 7.

Mechanical power can be cordlessly generated by means of torsion produced bya magnet in a magnetic field, drawing 5. It is possible to generate a torque of 5⋅10-4

Nm.

A solar cell in the eye may transform the light energy into electrical energy,drawing 6. The power generated may be of the order of 90 µW⋅mm-2 (power per areaof the photodiode) [Sz8501].

The mechanical power, produced by pressing with a finger a button in the eye,will close or open a valve [To9701].

Table 2-2. Possible electrical or mechanical energy sources which can be used for anactuator [Ij9201].

20


2.8 Adjustment system for the eye pressure

Some regulation principles based on the channel diameter variation method areshown schematically in table 2-3.

Table 2-3. Sketches of some methods to adjust the intraocular pressure (IOP)[Ij9201].

The possible combinations of adjustment method, actuator principle and energysource are presented in table 2-4 [Ij9201]. These combinations are discussed withrespect to the medical requirements and their feasibility as eye pressure regulator.Thermal actuation is rejected for all cases because of the need of continuous powersupply.

1. Mechanical rotation. One of the problems of the mechanical rotation principlecoupled with the magnetic field as actuator and energy source is the fixation of theposition. When an operating point has been adjusted, the permanent magnetic field hasto be removed without changing the position of the rotating device. This can be donein several ways. One is to create a friction in the centre of rotation or to use twodifferent magnetic fields which do not interfere with each other. The majordisadvantage is that of the accurate positioning.

2, 3. Mechanical sliding/translation. The mechanical power delivered by a finger maywork, as for example in the channel length variation method by pushing the slider.

4. Buckling. This principle in combination with an electrochemical and electro-osmoticactuator has the disadvantage that it is not continuous.

5. Pressure membrane. This principle in combination with an electrochemical andelectro-osmotic actuator does not have any disadvantage.

6. Pump. The disadvantage of these combinations is the high power consumption;more, the eye fluid is conducting so the electrohydrodynamic pump cannot be used.

Chapter 2

21

Adjustment principles(table 2-3.)

Actuator(table 2-1.)

Energy source(table 2-2.)

1, 3. Mechanical rotation 1. magnetic field, torsion 5. magnetic field2, 4. Mechanical sliding5, 9, 10. Mechanical translation

3. electrolysis4. electro osmosis7. thermal, boiling liquid8. thermal, heating gas

1, 2. eye movement3. battery4. magnetic induction6. light power

2, 4. Mechanical sliding5, 9, 10. Mechanical translation

7. mechanical power

6. buckling 3. electrolysis4. electro osmosis7. thermal, boiling liquid8. thermal, heating gas


11. pressure membrane 3. electrolysis4. electro osmosis7. thermal, boiling liquid8. thermal, heating gas


8. pump 5. micro mechanical pump6. electro-hydro dynamic pump


Table 2-4. Possible combinations of regulation methods, actuator principles andenergy sources.

The membrane may deflect when a pressure difference is applied between thefront and back sides of the membrane. This pressure difference can be due to a liquidor a gas. One way to obtain a fluid pressure is by heating it into a sealed cavity. Theincrease of the fluid pressure will increase the deflection of the membrane. Thedisadvantage of this thermal actuation principle is the need for a constant supply ofpower to keep the fluid at a certain temperature. Since the boundary condition requiresa cordless power supply (medical R4), an actuation mechanism which does notdissipate power when no work is done by the actuator is preferable. An actuator whichmay meet this requirement is the electrochemical actuator, which builds up a gaspressure by electrolysis of an aqueous solution.

22


2.9 Eye pressure sensor

Mainly because of their size, integrated sensors are suitable for application in thebiomedical world, and more precisely for “in vivo” determination of physical andphysiological parameters [Ch9301, Co6701, Fr7301, Ji9201, Pe8801, Te8701].

Currently, IOP is measured with applanation tonometry [Be9301, Bi9201]. Sincethese measurements can only be made intermittently and the IOP does not have aconstant value, a single measurement might not be a proper indication of the patientcondition. Thus, there is a strong demand for continuous cordless IOP monitoring.

For an eye pressure sensor, three specifications have to be considered in order tochoose a type: size of the sensor chip; ease of read-out, and power consumption.

The two main types of pressure sensors [Ea9701], piezoresistive and capacitivesensors, are briefly described in the following paragraph.

The piezoresistive effect is based on the variation of the resistance of a materialunder a (mechanical) deformation. This effect is very high in silicon [Ka9101]. Forexample, four p-type doped resistors are diffused in n-type silicon in which a thinsilicon membrane is realised. The resistors are connected in a Wheatstone-bridgeconfiguration. By positioning the resistors on the membrane at places where thebending stresses are largest (e.g. close to the sides of a membrane), a deflection of themembrane due to pressure results in a change in resistance [Ka9701]. This type ofsensor has a DC response and no extra electronic circuitry is needed for the detectionof the voltage change. The disadvantages are a high power consumption ( a permanentdirect current is required to perform continuous measurements), and a temperaturedependence of the piezoresisitive effect. Nevertheless, it has been difficult to make animplantable, medical pressure transducer mainly due to the critical step of thepackaging of the sensor.

A capacitive pressure sensor consists of a cavity covered by a pressure sensitivemembrane [Sm8601, Ro9201, Pe9701, Lö9701]. Electrodes are located on themembrane and on the bottom of the cavity, and they act as the two plates of a platecapacitor. The pressure-capacitance response of a membrane capacitive sensor is non-linear because of the inverse dependence of capacitance on the distance between theelectrodes, and thus the sensitivity depends on as: dC/dd ~ d-2. Due to this strongdependence of the capacitance on the distance, capacitive detection is well suited toextremely precise distance measurements, if the two electrodes can be positioned closeenough to each other (≤ 2 µm). However, the very small sensor capacitance valuesmay be of the same order as the parasitic capacitance introduced by the wireconnections of the sensor to the outside. This requires the electronic circuit on or veryclose to the sensor chip. Other disadvantage is its high output impedance. The mainadvantages are low power consumption, high sensitivity and low temperaturedependence.

A first solution to the cordless IOP measurement was proposed by C. Collins[Co6701]. He used a passive measurement system consisting of an implantablecordless passive sensor, inductively coupled to an outside detector. The implantedsensor is composed of a capacitive pressure sensor (thin glass of polyester asmembranes) and a coil (resonant LC system). There were some problems with the long

Chapter 2

23

term drift and temperature drift due to gas and moisture diffusion. The pressure sensorwas used in animal experiments.

Another prototype system for passive, cordless, continuous measurements of theIOP, based on Collins’ idea, was developed at the University of Uppsala, Sweden[Ro9401]. The capacitive pressure sensor implant is part of an oscillating LC circuit.As the surrounding pressure changes, the thin Si membrane deflects, changing thecapacitance and the resonance frequency of the circuit. This frequency is measuredremotely with an external, inductively coupled oscillator (grid-dip type). The sensorconsists of the capacitive chip and a hand-wound coil, glued onto a plastic fixture. Theentire resonator was covered with silicone rubber to ensure electrical insulation fromthe eye liquid. The overall size of the implant is 5 mm diameter by 2 mm thickness. Thesensitivity in vitro was 4 mV/mmHg. However, the prototypes suffered from theinfluence of parasitic impedances, both capacitive and resistive, which resulted in apoorly controlled sensitivity, and a low quality factor of the resonator (Q ~ 40 at aresonance frequency of 40 MHz).

It can be concluded that piezoresistive sensors generally consume more power,are temperature dependent, but are cheap to mass produce and easy to read-out. Thedimensions can also be made smaller compared to capacitive sensors that need a ratherlarge area for the capacitor. The capacitive sensors require on-chip read-out circuitry,but have a low power consumption, thus they are suitable for long term implantation,where the chip cost is of less concern than the load of the battery.

2.10 Connection silicone rubber tube - actuator

The most probable regulation method of the flow resistance which will be used inthe eye pressure regulator will be the diameter adjustment. This may be done as shownin fig. 2.5.1b, where a flexible membrane, inside the tube, deflects and varies thediameter of the flow channel. The microactuator and the flexible membrane will bemade in silicon with the use of silicon micromachining techniques. Thus, a possibledesign for the connection between the actuator and silicone rubber tube which iscurrently implanted is schematically shown in fig. 2.10.1. Some biocompatible gluemay be used to ensure the strength of the tube-silicon connection wafer.

Figure 2.10.1. A possible connection between the silicone rubber tube which isnormally implanted and the microactuator; the active valve, i.e. the actuator, the flowchannel and the connection are shown.

24


2.11 Biocompatibility, packaging and encapsulation

The issue of biocompatibility arises from a recognition of the difference betweenliving tissues and non-living materials, and the wide range of interactive behaviourbetween tissues and materials. The consensus definitions of biocompatibility are[Br9501]: “Biocompatibilty is the acceptance of an artificial implant by the surroundingtissues and by the body as a whole”; “Biomaterial is a synthetic material used toreplace part of a living system or to function in intimate contact with living tissue”. Thesuccess of a biomaterial in the body depends on factors such as the material properties,design, and biocompatibility of the material used, as well as other factors including thetechnique used by the surgeon and the health and condition of the patient.Biocompatible materials do not irritate the surrounding tissues, do not provoke anabnormal inflammatory response, do not incite allergic or immunological reactions, anddo not cause cancer.

Encapsulation and packaging of devices, such as sensors and actuators for long-term implantation in the body is of primary importance. It is difficult to meet therequirements for long-term survival and this appears to be the main reason why thedevelopment of biomedical micromachined devices has been delayed. Problems whichhave to be addressed include: • Development of a coating that should not only protect the device and its associated

circuitry from the body fluids, but should also allow for contact between the sensingsite and the tissue;

• The implant device should have adequate mechanical properties such as strength,stiffness and fatigue properties. Sterilisability, manufacturability, long-term storage,and appropriate engineering design have also to be considered.

• The human body tends to coat foreign bodies with fibrous connective tissueplaques, particularly over sharp edges. This could alter the sensitivity of the device.Thus, biocompatible coatings have to be developed that do not createimmunological reactions.

Biomaterials that have been used in the body up to now are: • polymers: polyamides (nylon), silicone rubber, polyvinylchloride (PVC),

polyethylene (PE), polypropylene (PP), polymethylmetacrylate (PMMA),polystyrene (PS), polytetrafuoroethylene (PTFE), polyurethane (PU).

• metals: Ti and its alloys, Co-Cr alloys, stainless steels, Au, Ag, Pt. • ceramics: aluminium oxide, calcium phosphates including hydroxyapatite, carbon. • composites: carbon-carbon, wire or fibre reinforced bone cement.

From the biocompatibility point of view, the biomedical polymers should notcause: uncontrolled thrombosis (thrombosis is the manifestation of those functions ofblood normally responsible for the stopping of bleeding); destruction of blood cellularelements; alteration of plasma proteins; destruction of enzymes; depletion ofelectrolytes; adverse immune reaction; damage to adjacent tissue; cancer ormutagenesis; they should not deteriorate in biological environment.

While in pacemakers biocompatible titanium coatings are well established,encapsulation with metals is not always possible for implantable systems (in the case ofelectromagnetic powering. Therefore, alternative materials and methods have to be

Chapter 2

25

investigated. Thin films of LPCVD silicon nitride and silicon dioxide, diamond,polyimide and Parylene C have been used to encapsulate both sensor/actuator andelectronics [Na8701]. These materials have been used in vivo for periods of only a fewmonths.

For electrical protection or passivation, the main packaging considerations are:electrostatic shielding, moisture penetration, interface adhesion, interface stress, andcorrosion of substrate material. Moisture penetration is one of the major problems.When moisture infiltrates the package and condenses on the electrical components andsubstrate, leakage currents may flow leading to an electrical failure. Glass, ceramic andmetal are materials considered to be impermeable to vapours at a thickness of 10micrometers or larger. Hermetic packaging using silicon-glass anodic bonding andsilicon-silicon bonding [Gu9701, Gu9702, Sa9701] is becoming more feasible as thesetechnologies mature and as new techniques for low-temperature bonding of glass tosilicon are being developed. Polymeric packages are not hermetic seals but are oftenused because they are easy to handle and less expensive [Be9201]. For organicmaterials, such as epoxies, silicones, and fluorocarbons, the moisture would penetrate apackage having a thickness of a millimetre in minutes to days.

2.12 Conclusions

At the moment, two methods to adjust the eye pressure by changing the fluidflow resistance are considered as feasible, (i) the flow channel diameter variation, and(ii) the change of the out-flow area at the end of the tube. In the device developedhere, this flow resistance adjustment will be achieved by changing the deflection of amembrane mounted inside the flow channel or at the end of the silicon rubber tube.

The membrane may deflect by applying a pressure difference between the frontand back side of the membrane. Since the boundary conditions require cordless powersupply (medical R4), an actuation mechanism which does not dissipate power when nowork is done by the actuator is preferable. An actuator which may meet thisrequirement and provide a continuous regulation around a desired pressure value is theelectrochemical actuator, which builds up a gas pressure by electrolysis of an aqueoussolution.

An electrochemical actuator with magnetic induction as power supply will bedescribed in the following chapters.

26


References

[Be9201] C. den Besten, R.E.G. van Hal, P. Bergveld, Polymer bonding of micro-machinedsilicon structures, IEEE Micro Electrical Mechanical Systems MEMS’92,Travemünde, Germany, 4-7 February, p. 104-109, 1992.

[Bi9201] L.Z. Bito, Glaucoma: a physiologic perspective with Darwinian overtones,Journal of Glaucoma 1, p. 193-205, 1992.

[Br9501] J.D. Bronzino (ed.), The Biomedical engineering handbook, CRC and IEEEPress, 1995.

[Ch9301] L. Christel, K. Petersen, A catheter pressure sensor with side vent using multiplesilicon fusion bonding, Proc. 7th Int. Conf. Solid-State Sensors and Actuators,Transducers’93, Yokohama, Japan, p. 620-623, 1993.

[Co6701] C. Collins, Miniature passive pressure transensor for implanting in the eye, IEEETrans. Biomed. Eng. BME-14(2), p. 74-83, 1967.

[Co8001] R. Collins, T.J. van der Werf, Mathematical models of the dynamics of thehuman eye, Springer Verlag, 1980.

[CR9301] CRC Handbook of chemistry and physics, D.R. Lide (ed) CRC Press, 1993/1994.[Ea9701] W.P. Eaton, J.H. Smith, Micromachine pressure sensors: review and recent

developments, Smart Materials and Structures 6(5), p. 530-539, 1997.[Fr7301] W. Frobenius, A. Sanderson, H. Nathanson, Microminiature solid-state capacitive

blood pressure transducer with improved sensitivity, IEEE Trans. Biomed. Eng.,p. 312, 1973.

[Ge8701] J.M. Gere, S.P. Timoshenko, Mechanics of materials, Van Nostrand Reinhold,1987.

[Gi8401] M. di Giovanni, Flat and corrugated diaphragm design handbook, MarcelDekker Inc., New York, U.S.A., 1982.

[Gr9101] E.L. Greve, Academical Medical Centre Amsterdam, the Netherlands.[Gr9601] E.L. Greve, Academical Medical Centre Amsterdam, the Netherlands, personal

communications.[Gu9701] C. Gui, M. Elwenspoek, J.G.E. Gardeniers and P.V. Lambeck, Present and future

role of CMP in wafer bonding, 4th Int. Symp. Semicon. Wafer Bonding: Science,Technology and Applications, Aug. 31 - Sept. 5, Paris, France 1997.

[Gu9702] C. Gui, H. Albers, J.G.E. Gardeniers, M. Elwenspoek, P.V. Lambeck, Fusionbonding of roughness surface with polishing technique for silicon micromachining,Microsystem Technologies 3(3), p. 122-128, 1997.

[Ij9201] D.J. Ijntema, Feasibility study for a micro machined eye pressure regulator forglaucoma patients, Report University of Twente, EL-TDM, Enschede, TheNetherlands, 1992.

[Je9001] H. Jerman, "The fabrication and use of micro machined corrugated silicondiaphragms", Sensors and Actuators A 21-23, p.988, 1990.

[Ji9207] J. Ji, S.T. Cho, Y. Zhang, K. Najafi, K.D. Wise, An Ultraminiature CMOSpressure sensor for a multiplexed cardiovascular catheter, IEEE Transactions onElectron Devices 39(10), p. 2260-2267, 1992.

[Ka9101] Y. Kanda, Piezoresistance effect of silicon, Sensors and Actuators A 28, p. 83-91,1991.

[Ka9701] Y. Kanda, A. Yasukawa, Optimum design considerations for silicon piezoresistivepressure sensors, Sensors and Actuators 62(1-3), p. 539-542, 1997.

[Ko9101] H.G. Kolk, M. Brambring, Haalbaarheidsstudie naar een micromechanischegenerator, Report University of Twente, EL-TDM, Enschede, The Netherlands,1991.

[Kr8801] T. Krupin, Manual of glaucoma, diagnosis and management, ChurchillLivingstone, New York, 1988.

Chapter 2

27

[Kr9201] K. Krieglstein, Glaucoma update IV, Archives of Ophthalmology 110(11), p.1540, 1992.

[Lö9701] J.C. Lötters, W. Olthuis, P.H. Veltink, P. Bergveld, Design, realization andcharacterization of a symmetrical triaxial capacitive accelerometer for medicalapplications, Sensors and Actuators, submitted for publication, 1997.

[Na8701] K. Najafi, J. Ji and K.D. Wise, Multichannel intracortical recording microprobes:scaling limitations, device characteristics, and circuit encapsulation, Tech. Digest,4th Int. Conf. Solid-State Sensors and Actuators Transducers ‘87, Tokyo, Japan,June 2-5, pp. 65-68, 1987.

[Pa9001] J. Y. Pan, P. Lin, F. Maseeh, S. D. Senturia, "Verification of FEM analysis ofload-deflection methods for measuring mechanical properties of thin films",Techn. Digest IEEE Solid-State Sensors Workshop, Hilton Head Island, U.S.A.,p.70, June 1990.

[Pe8101] T.S. Perkins, The ocular pulse, Curr. Eye Res. 1, p. 19-23, 1981.[Pe8801] K. Petersen et al., Silicon fusion bonding for pressure sensors, Proc. IEEE Solid-

State Sensors and Actuautors Workshop, Hilton Head Island, SC, USA, p. 144-147, !988.

[Pe9701] M. Pederson, W. Olthuis, P. Bergveld, Fabrication of IC-compatible capacitivesensors by polymer processing, Proceedings of SPIE Smart Structures andMaterials, San Diego, USA, March 1997.

[Po9001] F.C.M. van de Pol, H.T.G. van Lintel, M. Elwenspoek and J.H.J. Fluitman, Athermopneumatic micropump based on micro-engineering techniques, Sensors andActuators A21-A23, p.827-829, 1990.

[Po9501] D.S. Popescu, D.C. Dascalu, M. Elwenspoek, T. Lammerink, Silicon activevalves using buckled membranes for actuation, Tech. Digest, 8th Int. Conf. Solid-State Sensors and Actuators, Transducers’95, Stockholm, Sweden, 25-29 June,p. 309, 1995.

[Ri9101] A. Richter, A. Plettner, K.A. Hoffmann and H. Sandmaier, Electrohydrodynamicpumping and flow measurement, Proc. IEEE Micro Electrical MechanicalSystems, MEMS′91, Nara, Japan, p. 271-276, 1991.

[Ri9201] A. Richter and H. Sandmaier, An electrohydrodynamic micropump, Proc. IEEEMicro Electrical Mechanical Systems, MEMS′92, Travemünde, Germany, 4-7February, p. 99-104, 1992.

[Ro9201] L. Rosengren, J. Söderkvist, L. Smith, Micromachined sensor structures withlinear capacitive response, Sensors and Actuators, A31, 1992, p. 200-205.

[Ro9401] L. Rosengren, P. Rangsten, Y. Bäklund, B. Hök, B. Svedbergh, G. Sélen, Asystem for passive implantable pressure sensor, Sensors and Actuators, A43,1994.

[Sa9701] S. Sanchez, C. Gui, M. Elwenspoek, Spontaneous direct bonding of thick siliconnitride, Journal of Micromechanical and Microengineering 7(3), p. 111-113,1997.

[Sh7801] R.K. Shah, A.L. London, Laminar flow forced convection in ducts, AcademicPress, New York, 1978.

[Sh9201] M.B. Shields, Textbook of glaucoma, 3rd ed., Williams & Wilkins, Maryland,1992.

[Sm8601] M.J.S. Smith, L. Bowman, J.D. Meindl, Analysis, design and performance of acapacitive pressure sensor IC, IEEE Transactions on Biomedical Engineering,BME-33(2), p. 163-174, 1986.

[Sz8501] S.M. Sze, Semiconductor Devices, Ed. John Willey, 1985.[Te8701] L. Tenerz, B. Hök, 0.5 mm diameter pressure sensor for biomedical applications,

Proc. 4th Int. Conf. Solid-State Sensors and Actuators, Transducers ‘87, Tokyo,Japan, June 2-5, p. 312-315, 1987.

28


[Th9201] J.V. Thomas, Glaucoma surgery, Mosby-Year Book Inc., St. Louis, 1992.[Ti6801] S.P. Timoshenko, S. Woinowsky-Krieger, Theory of plates and Shells, McGraw-

Hill Inc., 1968.[To9701] M.W. van Toor, T.S.J. Lammerink, J.G.E. Gardeniers, M. Elwenspoek, D.

Monsma, A novel micromechanical flow controller, Journal of Micromechanicaland Microengineering 7(3), p. 165-169, 1997.

[Wh9401] F.M. White, Fluid mechanics, McGraw-Hill,Inc., 1994..

Chapter 3

29

The basic idea is that reversible electrochemical reactions,driven by an external current source, lead to gas evolution orremoval, depending on the direction of current flow. In aclosed space the gas pressure is used to deflect a flexiblemembrane, which in turn can be used as a valve. In this way,the eye pressure can be adjusted by changing the resistance ofthe flow channel. Since this pressure will be achieved by usingan electrochemical actuator, a short review of theelectrochemical principles will first be given. This actuator willbe made with the use of silicon micromachining technologies.Thus, extra requirements related to micromechanics andelectrochemistry which have to be fulfilled by the electrodes-electrolyte system are considered. The working principle andthe choice of the electrochemical reactions are next explained.Different chemical and physical problems, such as electrodecorrosion, that may limit the long-term operation of the cell aredescribed. Some possible solutions such as electrode protectionby an ion-selective membrane or the use of a metal/metal oxideelectrode are considered.

3.1 Introduction

Depending on the application, certain principles to actuate a valve, i.e. to deflecta flexible membrane, may be preferable [El9401, Ze9401, Ba9501]. For large strokes,for example, coil valves, shape memory alloys and thermally excited valves can beused; large forces can be obtained with electrostatic devices; for high speed operation,thermally operated valves are too slow, so piezovalves are preferred. Most of thesetypes of actuation need either high voltages (electrostatic, piezoelectric) or have quitehigh power consumption with low efficiency (heat dissipation in the case of thermalactuation), which limits the applications. For medical applications and particularlywhen a wireless power supply is required, an actuator which does not dissipate powerwhen no work is being done is preferable. For a valve, operated by a differential gaspressure, an electrochemical principle, as described in chapter 2, is attractive. Theadvantage of this type of actuation is a low power consumption and a possiblediscontinuous use of power.

DESIGN CONSIDERATIONS FOR AN

ELECTROCHEMICAL ACTUATOR

3

30

Design considerations for an electrochemical actuator

The electrochemical actuator (fig. 3.1.1) is based on the electrolysis of anelectrolyte solution to generate gas pressure. In a closed system the build-up andreduction of gas pressure can be used to change the deflection of a membrane. Themembrane deflection can subsequently serve to close or open a fluid channel.

Electrochemical actuators were reportedfor the first time in 1988 by Janocha [Ja8801].These actuators, with a size of at least 5x5x5cm3, were made with conventional techniquesfrom a steel case. Janocha used two systems.One consisted of platinum electrodes andphosphoric acid as electrolyte, and a semi-permeable membrane (Nafion) to separate thegases (oxygen and hydrogen) which evolve atthe anode and cathode. The Nafion membranecan transport positive ions but rejects negativeions and non polar molecules [duPont]. Theoxygen is accumulated in a chamber and thehydrogen is used as pressurised gas to move amembrane. The other type of actuator usedsilver as anode and platinum as cathode in analkaline electrolyte. Hydrogen gas evolves at

the cathode and silver oxide is formed at the anode. Nafion is again used to separatethe silver oxide from the hydrogen gas. When hydrogen contacts silver oxide, it reactsspontaneously to give water (H2O) and silver (Ag). The latter principle was alsodescribed by Kempe in 1990 [Ke9001].

In 1996 a patent describing an electrochemical actuator produced withconventional techniques was issued [Ba9601]. The actuator consists of a series of cellsmounted in a stack. The working electrode is made from an oxidizable material such assilver or nickel hydroxide. The counter electrode is made from carbon with a binder,such as PTFE. The electrolyte is absorbed into a matrix of a material such as a porousceramic oxide. By passing a current through the cell the working electrode is oxidizedwhile hydrogen is formed at the counter electrode. When a reverse current is applied,the metal oxide of the working electrode is reduced and hydrogen is oxidized at thecounter electrode,. The pressure increase or decrease in the gas compartment can beused for generating movement. The actuator can regulate or control processes ordevices such as, for example, control valves on radiators.

This chapter will start with some electrochemical principles necessary tounderstand the cell operation. The next section explains the working principle and thechoice of the electrochemical system. An essential requirement for the electrochemicalactuator is that the gas, produced during electrolysis, remains and does not react in anyway in the cell. In practice this ideal situation is very difficult to achieve. Several waysto improve this situation will be discussed.

Figure 3.1.1. A schematicdrawing of a singlecompartment electrochemicalactuator having two electrodes.

Chapter 3

31

3.2 Electrochemical principles

The actuator is based on an electrochemical cell. Gas is evolvedelectrochemically by an oxidation or a reduction reaction at one of the electrodes in thecell. Such “redox” reactions can be represented schematically by:

On+ + ne- ↔ R (3.2.1)

where On+ and R refer to the oxidized and reduced forms of the redox couple,respectively, and n is the number of electrons involved in the reaction. At equilibrium,the potential of the electrode EO/R, measured with respect to that of a referenceelectrode, is given by the Nernst equation [Ba8001, Bo9301, Ri8701]:

E ERT

nFO/ R O/ R0= + ⋅ ln

C

C

O

R

(3.2.2)

where EO R/0 is the standard redox potential, and CO and CR are the concentrations of

oxidized and reduced components of the redox system, respectively. The referenceused by electrochemists is the normal hydrogen electrode (NHE), whose potential isarbitrarily attributed a value of 0V. For convenience, other reference electrodes areoften employed; in this work the silver/silver chloride electrode (Ag/AgCl) (0.22V withrespect to NHE) or saturated calomel electrode (SCE) (0.24V with respect to NHE)were used.

Figure 3.2.1 shows a typical electrochemical cell containing three electrodes. Thereaction of interest occurs at the working electrode (WE); its potential is measuredwith respect to that of the reference electrode (RE). The potential of the workingelectrode can be varied by the voltage source connecting the working electrode and thecounter electrode (CE). Generally, a potentiostat is used to regulate the potential ofthe WE. If the potential of the working electrode, E, differs from the equilibrium valueEO/R, a current, i, may flow through the cell. We can distinguish two cases:

(i) E < EO/R

A net reduction reaction occurs at the WE:

On+ + ne- ↔ R (3.2.3)

Electrons for this “cathodic” process are supplied by anoxidation reaction at the CE. The driving force forreaction (3.2.3) is the overpotential η, describing thedeviation from equilibrium:

η = E - EO/R (3.2.4)

If the reaction is determined solely by electron transferbetween the electrode and oxidized species, thereaction rate and thus the ‘cathodic’ current density, jc,is generally an exponential function of η:

j jnF

RTc = −

0 exp

αη (3.2.5)

where j0 is the exchange current density, and α is thetransfer coefficient.(ii) E > EO/R

Powersupply

WE RE

CE

i

V

Figure 3.2.1. Schematicarrangement of a three-electrode cell, showing theworking electrode, WE, thecounter electrode, CE, andthe reference electrode, RE.

32


A net oxidation reaction occurs at the working electrode:

R ↔ On+ + ne- (3.2.6)

The electrons generated at the WE are used for a reduction reaction at the CE. As inthe previous case, eq. 3.2.5, the ‘anodic’ current density is an exponential function ofthe overpotential:

j jnF

RTa =−

0

1exp

( )αη (3.2.7)

Figure 3.2.2 shows schematically the potential dependence of jc and ja. The totalcurrent density (the sum of jc and ja):

j jnF

RT

nF

RT=

−

− −

0

1exp

( )exp

αη

αη (3.2.8)

is indicated by the solid line. Equation 3.2.8 is the Butler-Volmer equation.

At high overpotential the rate of reaction 3.2.3 or 3.2.6 may be so large thatmass transport, i.e. supply of O or R to the electrode surface from the bulk solution,determines the reaction rate. In this case the current density depends on thehydrodynamics of the system and is independent of the applied potential (fig. 3.2.3).

To illustrate some of these factors current-potential curves are shown in fig.3.2.4 for a platinum rotating disk electrode (RDE) (electrode area ~ 7 mm2) measured

Figure 3.2.2. Current-potentialrelationship corresponding toeq. (3.2.8). The dotted linesrepresents the componentcurrents ic and ia.

Figure 3.2.3. Anodiccurrent-potential curvesshowing the influence ofkinetic control (dotted line)and mass transport (solidline); ilim is the masstransport current.

Chapter 3

33

in an aqueous 1M KCl solution in contact with air. The potential of the Pt workingelectrode was measured with respect to a saturated calomel electrode (SCE). Thepotential was scanned at a constant rate of 25 mV/s in these experiments. Platinum wasalso used as a counter electrode.

In the scan to negative potentials a large cathodic current due to hydrogenevolution:

2 2 22 2H O H g OH+ → +− −e ( ) (3.2.9)

is observed below -0.8 V. This current increases exponentially with decreasingpotential. When the scan direction is reversed at -1.0 V, an anodic peak indicates theoxidation of hydrogen formed cathodically (the reverse of reaction 3.2.9). On scanningfurther in the positive direction anodic features are observed due to oxide formation onthe electrode ( > 0 V). At still more positive potential ( > 0.8 V) an exponentiallyincreasing anodic current is found, due to oxygen evolution:

2 4 42 2H O O g H→ + ++ −( ) e (3.2.10)

Since the measurements were performed in chloride solution, some chlorine can beformed (when oxygen is formed):

2 22Cl Cl g− −→ +( ) e E0 = 1.36V vs NHE (3.2.11)

At 1 V the scan direction is again reversed. In the return scan to the negative potentialthe current becomes cathodic due to reduction of oxide on the metal and reduction ofoxygen dissolved in the solution (reverse of reaction 3.2.10); the solution is inequilibrium with air.

The role of oxygen reduction becomes clear when the electrode is rotated(curves 1, 2, 3 in fig. 3.2.4). A cathodic current plateau is found, from -0.3 to -0.8 V.

Figure 3.2.4. Cyclic voltammograms of aPt RDE in a 1M KCl solution for differentrotation rates: (1) 0 rpm; (2) 400 rpm; (3)2500 rpm; the solution was in equilibriumwith air at room temperature; V is vs. SCE(0.24V vs. NHE).

34


The limiting current depends strongly on the rotation rate of the electrode. The currentdrops to a very low value when oxygen is purged from the solution by nitrogenbubbling. These results suggest that diffusion of oxygen is important. The limitingcurrent density, j lim, for a diffusion-controlled electrochemical reaction at an RDE isgiven by the Levich equation [Gi9301]:

j nFD Cbulklim/ / /.= −0 62 2 3 1 6 1 2µ ω (3.2.12)

where µ [cm2/s] is the kinematic viscosity (viscosity/density) of the electrolyte solution,D [cm2/s] is the diffusion coefficient of the species studied, and ω [s-1] is the angularvelocity. That oxygen reduction at platinum is completely diffusion-controlled can beseen from fig. 3.2.5; a plot of the limiting current versus the square root of the rotationrate gives a straight line through the origin yielding a value for the diffusion coefficientof oxygen of 1.8⋅10-5 cm2/s, which is in reasonable agreement with the literature(2.9⋅10-5 cm2/s).

We repeated the experiment with a Cu electrode instead of Pt. The resultingRDE result is also shown in fig. 3.2.5. The reduction of oxygen at copper RDE is alsodiffusion controlled; as for the platinum, the limiting current is directly proportional tothe square root of the rotation rate.

The results given in fig. 3.2.4 show that various gases, e.g. oxygen andhydrogen, can be generated electrochemically at a platinum electrode. The currentdensity and thus the rate of gas evolution depends on the electrode potential. Highrates can be readily achieved. It is also clear from figs. 3.2.4 and 3.2.5 that oxygen canbe electrochemically reduced at platinum. The maximum rate in this case is limited,since the reaction is diffusion-controlled (Levich eq.) and the solubility and thus theconcentration of oxygen in aqueous solution is limited.

Figure 3.2.5. Rotating disk electrode data for oxygen reduction at Pt and Cu in 1 MKCl solution in equilibrium with air at room temperature.

Chapter 3

35

3.3 Choice of the components of an electrochemical actuator

• The cell

The electrochemical cell used for the actuator could, in principle, contain threeelectrodes, as shown in fig. 3.2.1. However, such a configuration is impracticalbecause the cell volume of a micromachined actuator is limited.

A two electrode cell containing WE and CE is favoured. The cell can be drivengalvanostatically by controlling the current flowing in the external circuit.Alternatively, the cell voltage (the potential difference between WE and CE) can beregulated.

• The gas-evolving electrode

The essential feature of the cell is its ability to generate a gas causing the build-up of pressure. This reaction should be reversible allowing, if necessary, the pressureto be decreased. A compatible reaction is required for the counter electrode. Animportant requirement is that the pressure built up by electrolysis is maintained underopen-circuit conditions.

Various gas-evolving reactions could be considered for use in the cell. However,most of these have obvious disadvantages. We shall briefly consider some possibilities.(a) Hydrogen: can be produced by reducing water at an inert electrode such asplatinum or gold (reaction 3.2.9). The reaction is reversible. However, hydrogendiffuses easily through many materials making the construction of a gas-tight celldifficult. In addition, hydrogen in the presence of oxygen is unstable and potentiallyexplosive.(b) Chlorine can be generated by the oxidation of chloride ions (reaction 3.2.11).Chlorine is highly corrosive.(c) Carbon dioxide is not easy to form and remove electrochemically.(d) Oxygen (reaction 3.2.10) does not have the disadvantages of the gases mentionedabove and will therefore be chosen for further study.

• The counter electrode

If oxygen is evolved at the WE in an aqueous electrolyte solution not containingan added redox system, hydrogen will be produced at a Pt CE. The characteristics ofthe cell are shown in fig. 3.3.1. Curve (a) is the current-potential curve for O2

evolution at the Pt WE, curve (b) for H2 evolution at the Pt CE. A cell voltage ∆Ecauses a current to flow through the cell (ia = ic). The redox potentials for the O2/H2O(reaction 3.2.10) and H2O/H2 (reaction 3.2.9) couples are indicated on the potentialaxis, as well as the corresponding overpotentials. It is clear that:

( )∆E E EO H O H O H a c= − + +2 2 2 2/ / η η (3.3.1)

Figure 3.3.1. Partial anodic (ia)and cathodic (ic) current as afunction of potential for oxygenand hydrogen evolution,respectively; Ecell is the cellvoltage, EH

2O/H

2 and EO

2/H

2O are the

redox potentials.

36


We assume that the conductivity of the electrolyte solution is high so that ohmic lossescan be disregarded.

As described above, a reaction involving hydrogen is not a good choice for theactuator. To avoid hydrogen evolution an oxidizing agent, O, can be added to thesolution. If the redox potential of the couple O/R is more positive than that of H2O/H2

(0 V (NHE) at pH = 0) and ∆E is chosen correctly then O can be reduced withoutreducing water. A possible choice of redox system is the Fe3+/Fe2+ couple:

Fe3+ + e- → Fe2+ EFe3+/Fe2+ = 0.771 V (NHE) (3.3.2)

This combination complies with two of the requirements for the actuator.

• Pressure reduction

When pressure reduction is desired, the cell could be short-circuited, so that thereverse reactions occur. However, if the anode and cathode compartments of the cellare not separated, the reverse reaction occurs spontaneously in the bulk solution, forexample:

4Fe2+ + O2 + 4H+ → 4Fe3+ + 2H2Osince EO

2/H

2O is more positive than EFe3+/Fe2+ i.e. the combination of O2 and Fe2+ is

thermodynamically unstable (the change in free energy ∆G for the reaction is < 0). Thisproblem of thermodynamic instability is inherent to the electrochemical cell concept.The products formed when the cell is driven to evolve O2 react under open or short-circuit conditions. This holds for the first example involving O2 and H2 production,although in this case the rate is slow; the rate is enhanced at the surface of a catalystsuch as Pt or Pd (see chapter 6).

• Preventing the back reaction of O2

It is not practical to have a microcell with two compartments, one for the anode,the other for the cathode separated by a membrane to prevent products reacting. Forthe microactuator described in this work a single compartment cell was used. The CEconsisted of a metal (Cu) in contact with its ions (Cu2+) in solution. An attempt wasmade to protect the metal (the reduced form of the redox couple) from oxygen bymeans of a membrane (Nafion). In the last section of this chapter some suggestions aremade for an alternative counter electrode.

• The copper/copper ion electrode

The half reaction for this situation is:

Cu2+ + 2e- ↔ Cu ECu2+/Cu = 0.34 V (NHE)

The reduced form is a solid. The combination of this couple at the CE with the O2/H2Ocouple at the WE gives the same problem of thermodynamic instability as the Fe3+/Fe2+

system. Under open-circuit conditions O2 reacts with Cu. At lower pH the followingreaction occurs:

2Cu + O2 +4H+ → 2Cu2+ + 2H2O

However, in this case, it may be possible to protect the solid CE from O2 by depositingon it a membrane permeable to cations but impermeable to O2. This concept isconsidered in more detail later in the chapter. The various reactions occurring can besummarised as follows:

(i) pressure build-up

Chapter 3

37

By applying the correct voltage to the cell O2 is evolved at the Pt WE:

2 4 42 2H O O g H→ + ++ −( ) e

and Cu is deposited at the Cu CE:

2Cu2+ + 4e- → 2Cu

(ii) open circuitA semipermeable Nafion membrane prevents contact between Cu and the O2 formedduring pressure build-up.

(iii) pressure reductionIf the cell is short-circuited the reverse reactions occur.

At the Cu CE: 2Cu → 2Cu2+ + 4e-

At the Pt WE: O g H H O2 24 4 2( ) + + →+ −e

The electrochemical characteristics are shown schematically in fig. 3.3.2. Thepotential of the short-circuited electrodes with respect to the reference electrode isdenoted by Esc. As already pointed out the rate at which the pressure can be decreasedis considerably lower than that for pressure build-up. The O2 which is reduced at thePt/solution interface, must be supplied by diffusion from the bulk solution.

Figure 3.3.2. Schematic current-potential curves for reactions which take placeduring the pressure reduction state (short-circuit): copper dissolution and oxygenreduction. Esc is the potential of the short-circuited electrodes.

38


3.4 Properties of Nafion as semi-permeable membrane

For the design of the electrochemical cell, information about the diffusion ofoxygen through the Nafion membrane is needed. An overview of the properties ofNafion based on the literature is given in [As9301, Bü9501, Gr8601, La7701,Lu7901, Nafion, Og8401, Og8501, Sa8601, So8601, Ye8101, Ye8301].

Nafion is a perfluorinated ion-exchange polymer, which is used in variousapplications, for example, as a membrane separator in electrolytic cells (e.g. as gasseparator membrane (for O2 and H2) in fuel cells; as electrolyte in solid polymerelectrolyte (SPE) water electrolysis cells; as separator between the two porouselectrodes.); as a gas diffusion membrane; and as a protection film for electrodes inelectrocatalysis [So8601, Wa8701, Ye8301]. The structure of Nafion is based on apolytetrafluoroethylene (PTFE) back-bone with pendant side chains shown in fig. 3.4.1[Lu7901]. Nafion is permeable to many positive ions and to polar compounds[Gr8601], but it is impermeable to negative ions and nonpolar compounds such asoxygen gas. Charge carriers in the membrane are hydrated hydrogen ions (H+ · xH2O,where x ~ 3.5 to 4 molecules) which move through the solid electrolyte by passingfrom one fixed sulphonic acid group to an adjacent one [La7701]. A new model forNafion morphology has been proposed by Litt, in which it is shown that Nafionbehaves like most other polymers [Li9701]. It is known [As9301] that the permeabilityof oxygen through a Nafion membrane in the dry state is much lower than in the wetstate; the gas permeation is therefore closely related to the water uptake of themembrane, which directly affects the mass transfer rate of gases to the electrode. Theincrease in the thickness/volume ratio due to immersion is about 65% of the dry bulkvolume.Some relevant Nafion properties, from the literature, can be summarised as follows:

The oxygen diffusion coefficient, D:• is dependent on the Nafion sample and pre-treatment history.• in dried Nafion is close to the value in PTFE.• in hydrated Nafion, is independent of the pressure, in the range (4 - 13)·105 Pa.• in hydrated Nafion is 20 times greater than in the dried Nafion.

The oxygen solubility, S:• is almost independent of the sample pre-treatment.• in dried Nafion is close to that in PTFE.• in hydrated Nafion is half that in the dried polymer.

It can be concluded that when Nafion is hydrated, the solubility decreases a littleand the diffusivity increases markedly. The properties of Nafion are listed in table 3.4-1.

In an electrochemical cell with a Nafion membrane between the two electrodesthe resistance is mainly determined by the Nafion thickness. The proton conductivity ofNafion is σ = 0.0316 Ω-1⋅cm-1 (almost independent of temperature). Nafion has thelowest diffusion coefficient for oxygen (among the commercially available materials)and still allows the (positive) ions to diffuse.

Chapter 3

39

(a) (b)

Figure 3.4.1. (a) The general chemical formula of a perfluorinated polymer; (b) the three zonechemical structure of Nafion as suggested by Yeager [Ye8101]: bulk polymer (A), intermediatephase (B), and electroactive phase (C).

Nafion type 115*

wet117 H+ [Sa8601]

dry wet

120 [Og8401]

wet

125 H+ [Sa8601]

dry wet

equivalent weight [g/eq] 1100% water content(based on dry weight )

37

density [g/cm3] 1.6O2 permeability coefficientPm [cm3⋅cm/cm2⋅s⋅cmHg]

30⋅10-12

[mol/cm⋅s]2.7⋅10-10

(37°C)2.7⋅10-9

(37°C)1.43⋅10-10

(30°C)2.15⋅10-10

O2 diffusion coefficientD [cm2 / s]

3⋅10-6 1.7⋅10-7

(37°C)2.8⋅10-6 (25C)4⋅10-6

2.4⋅10-7 (20°C)2.9⋅10-7 (30°C)4.4⋅10-7 (40°C)

8.8⋅10-8 7.6⋅10-8

1⋅10-7 (37°C)

O2 solubility [M=mol/dm3] 10⋅10-6

[mol/cm⋅s]1.7⋅10-3

(37°C)1⋅10-3

(25°C)7.2⋅10-3 (20°C)6.5⋅10-3 (30°C)5.3⋅10-3 (40°C)

1.8⋅10-3

(cm3/cm3⋅cmHg)

2.7⋅10-3

Table 3.4-1. Physical-chemical properties of some types of Nafion.

[-(CF2- CF2)n- CF- CF2-]m

O CF- CF2-O- CF2-SO3-M3

CF3

40


3.5 Life-time and power consumption of the cell

The life-time of the actuator depends on the life-time of its critical parts: theelectrodes, the electrolyte, the deflecting membrane, the material composing the bodyof the actuator, and the Nafion membrane. The stability of electrodes and electrolytedepends on the electrochemical reactions which occur.

Platinum is a noble metal and is chemically very stable under the conditionsconsidered in this thesis.

As mentioned above, oxygen dissolved in the electrolyte causes a parasiticreaction at the copper electrode, corrosion of the metal and reduction of oxygen towater. This unwanted reaction decreases the oxygen gas pressure and therefore theefficiency of the actuator. To reduce this effect, the copper electrode has to beprotected against oxygen, using e.g. an ion-exchange polymer membrane Nafion. Thelife-time of copper depends mainly on the diffusion of oxygen through the Nafionmembrane, which will be investigated later (chapter 5).

The flexible membrane and the body of the actuator are made with the aid ofmicromachining technologies using materials such as silicon, silicon nitride, siliconoxide, etc. which are not affected by the copper sulphate solution or oxygen. Theproperties of these materials do not change under the working conditions: a bodytemperature of 36 - 41 °C; oxygen gas pressure of a few bars, and contact with eyefluid. Diffusion of oxygen through these materials under such conditions has not yetbeen tested for extended periods of time but is not expected to give problems.

In the fabrication of the electrochemical cell, bonding of the components may benecessary. Leakage of oxygen out of the actuator will very likely be through or alongthe bonding area. This is considered and estimated in chapter 5.

In the ideal case, gas will not leak out of the electrochemical actuator andoxygen gas will not diffuse through the Nafion membrane, so that a parasitic reactiondoes not occur at the copper electrode. The theoretical power consumption, P,depends on the cell voltage, Ecell, necessary for the electrolysis and on the currentwhich flows through the cell, i,:

P = Ecell ⋅i (3.5.1)

The current, i, necessary to increase the pressure in time, t, is:

idQ

dte

dn

dte= = ⋅ (3.5.2)

where the number of electrons is ne. To generate 1 molecule of O2, 4 electrons arerequired, so to increase the pressure in the electrochemical actuator with ∆p, thenumber of electrons ne needed is ne = 4⋅NO2⋅NA, where NO2 [mol] is the number of

moles of oxygen, and NA is the Avogadro’ number (6.023⋅1023 molecules/mol). Thegeneral law of gases is,

p⋅V = NRT (3.5.3)

Chapter 3

41

with p [Pa] the pressure in the electrochemical actuator (relative to the atmosphericpressure), the general gas constant, R = 8.31 J/mol⋅K, and the absolute temperature, T[K]. Thus,

dn N N Np V

RTgas

e O A A2= ⋅ ⋅ =

⋅4 4

∆(3.5.4)

The power consumption is given by:

PFV p

RT

E

tgas cell=

⋅⋅

4 ∆

(3.5.5)

where F = NA·e = 96484.6 C/mol is the Faraday constant. With Ecell = 0.89 V, Vgas = 2mm3 = 2⋅10-9 m3 (the whole cell volume), ∆p = 1 bar = 1⋅105 Pa, T = 310 K, t = 10 min= 600s, the power consumption is P ~ 40 µW. Different resistive losses are not takeninto account. The cell voltage will be in the range 1-2V. This will lead to P ≤ 100 µW.

3.6 Alternative counter electrodes

One way to prevent oxygen from reacting at the counter electrode is to use asemi-permeable membrane like Nafion® as described in section 3.4. Another approachwould be to use a metal electrode passivated by e.g. a coherent oxide layer, which doesnot react with oxygen gas. This involves the formation and reduction of a passiveoxide on a metal. The general reaction for a divalent metal can be denoted by:

M + H2O → MO + 2H+ + 2e-

The metal oxide could, in principle, be formed by a method other than anelectrochemical method. The three stages of cell operation are then:

(i) Pressure build-up: oxygen evolution at the Pt WE

2 4 42 2H O O g H→ + ++ −( ) e

is accompanied by metal oxide reduction

2MO + 4H+ + 4e- → 2M + 2H2O

under influence of an applied voltage. It is essential that not all the metal oxide isreduced in this stage.

(ii) Open-circuit: the passivating oxide present on the metal prevents access of oxygento the metal; the pressure is therefore maintained.

(iii) Pressure reduction: the polarity of the cell is reversed. O2 is reduced at the Pt WEand oxide is grown on the metal.

Oxide formation must meet the following requirements:• the voltage at which this process occurs has to be in the same range as that available

for operating the cell (wireless transmission), even if the oxide is formed previouslyat higher voltages;

• the oxide layer has to be a conductor (to allow current flow);• the oxide has to be chemically and electrochemically stable in the electrolyte

solution.

Two possible systems will be considered Sb/Sb2O3 and Ag/Ag2O.

42


3.6.1 Antimony oxide∗∗∗∗

The Pourbaix diagram representing the electrode potential as a function of pHgives the conditions of thermodynamic equilibrium. However, these are valid only inthe absence of complexing agents. Two oxide reactions are important for our system[Pi5701]:

2Sb + 3H2O = Sb2O3 + 6H+ + 6e- E0 = 0.152 - 0.0591⋅pH(3.6.1)

Antimony trioxide is thermodynamically stable in the presence of water and aqueoussolutions free from reducing agents. The higher oxide is formed at more positivepotentials:

Sb2O3 + 2H2O = Sb2O5 + 4H+ + 4e- E0 = 0.671 - 0.0591 pH(3.6.2)

The pentoxide Sb2O5 is thermodynamically stable in the presence of oxygen, but it isvery soluble in water and aqueous solutions (in particular in alkaline solutions).

From the literature [Am7201], it is known that the anodic formation of antimonytrioxide follows the high field approximation. The steady state film thickness, δ, isdirectly proportional to the voltage drop over the oxide, Vf:

δ = c⋅Vf

(3.6.3)

where c is a constant. The current density associated with film growth is given by:

j = A⋅exp(B⋅Vf /δ)(3.6.4)

where Vf /δ is the electric field across the oxide, and A, B are constants. Equation 3.6.3implies that at a constant potential scan rate (dVf /dt), the oxide thickness increases (dδ/dt) at a constant rate, i.e. the current density is constant. The growth of an antimonyoxide film, prepared electrochemically is influenced by the pH of the solution whichdetermines the solubility of the oxide, and may also affect its structure, e.g. porosity(see literature for antimony oxides at the end of this chapter).

When O2 is formed at the Pt electrode during pressure build-up, Sb2O3 isreduced at the CE:

Sb electrode: Sb O 6 H 6 e 2 Sb 3 H O2 3 2+ + ⇔ ++ −

(3.6.5)

The oxygen will not be reduced at the Sb electrode if the electrode remains passivated,i.e. if the oxide layer is not completely removed during reaction 3.6.5. Furthermore, theoxide should be chemically stable in the solutions used in the actuator, and have arelatively low electrical resistance in order to limit the voltages required to operate theactuator.

To study the suitability of antimony oxide cyclic voltammetric measurementswere performed in aqueous solution. These electrodes were made from 99.99% Sb

∗ This section is based on the paper ‘An electrochemical active valve’ C. R. Neagu, J.G.E. Gardeniers,M. Elwenspoek, J.J. Kelly, Electrochimica Acta 42(20-22), p. 3367-3374, 1997.

Chapter 3

43

rods supplied by Highways International. The electrodes had a surface area of about 7mm2 in contact with the test solution. Before the measurements, the surface waspolished mechanically. Measurements were performed in 0.1 M Na2SO4 solution atroom temperature without stirring. A Pt plate served as counter electrode andAg/AgCl as reference electrode (0.222V vs. NHE). All experiments were done at ascan rate of 25 mV/s, using an EG&G 366A potentiostat.

Figure 3.6.1. Cyclic voltammogram of an Sb electrode in the oxide formation range;curves are shown for successive scans starting from 0 V to increasing anodicpotentials. 0.1 M Na2SO4, Pt counter electrode, Ag/AgCl reference electrode, scanrate 25 mV/s.

Three consecutive voltammograms are shown in figure 3.6.1. In the first scan,from 0 to 2V, the anodic current is almost independent of the applied potential untilthe scan direction is reversed. In the second sweep, from 0 to 5V, the current remainslow until the limit of the first scan (2V) is passed. At this point the current rises to avalue close to the limiting value of the first scan. In the third sweep, the limit of thesecond scan (5V) has to be passed before the current again reaches its potentialindependent value. When the scan direction is reversed, the current drops in all cases.The fact that in the second and third runs the potential limit of the previous run has tobe exceeded before the limiting current is observed, points to the growth of an oxide(the reverse of reaction 3.6.5), whose thickness is dependent on potential. This, andthe constant current in the forward scan direction at constant scan rate indicate field-dependent growth of the oxide (eq. 3.6.3). If the potential is kept constant or the scandirection is reversed, the current (and the oxide growth rate) drops. Antimony oxideclearly remains on the surface when the potential is kept constant or when the voltageis scanned back to 0V.

In the third experiment of fig. 3.6.1, the metal is anodised to 10V. Considerablyhigher potentials (up to 50V) can be used to give a thicker oxide. However, the qualityof antimony oxides obtained in this way is, in general, rather poor: extremely roughand porous surfaces may result [La9101, Bo9401]. For the present applicationantimony oxides formed by other methods may be more practical.

Figure 3.6.2. shows results relating to the cathodic reduction of the oxide film.An Sb electrode was anodised for 1 hour at +5V and the potential was scanned fromthis value to -1.5V (curve 1). The current in the greater part of the scan is very low. At-0.5V a small cathodic current plateau is observed after which the cathodic currentincreases strongly before levelling off. At more negative potentials (not shown) thecurrent increases further and gas evolution is observed; hydrogen is evolved at theelectrode. When the potential is scanned from -1.5V in the positive direction, the

44


current becomes anodic, showing two peaks before returning to a value typical foroxide growth. When the potential is returned to -1.5V, curve 1 is again retraced. Thesmall cathodic current plateau (in curve 1 of fig. 3.6.2) is absent when nitrogen isbubbled through the solution; it is very likely due to oxygen reduction.

Figure 3.6.2. Successive cyclic voltammograms of Sb anodized at +5V for 1 hourshowing the relation between anodic and cathodic peaks; curve 1: +5V ⇒ -1.5V ⇒+5V ⇒ -1.5V, curve 2: -1.5V ⇒ 0V ⇒ -1.5V, curve 3: : -1.5V ⇒ -0.5V ⇒ -1.5V.

The large cathodic current peak results from the reduction of the anodic oxide. Wefound that the charge under the peak increases as the limit of the potential scan in thepositive direction is made more positive i.e. as the oxide becomes thicker. The anodicmaxima in the return scan are related to nucleation and initial growth of the oxide layer[Me9401]. The first maximum is sensitive to oxygen in the solution. If the scan from -1.5V in the positive direction is halted at 0V (curve 2, fig. 3.6.2), the subsequentcathodic current in the reverse scan is greatly reduced. This effect is even morepronounced when the limit of the positive scan is -0.5V. Obviously to grow an oxidefilm with a significant thickness it is necessary to go to considerably more positivepotentials.

Information about the stability of the oxide film under open-circuit conditions isgiven in fig. 3.6.3. A 2V oxide film, was held at open-circuit potential for two hours.The potential was subsequently scanned from the open-circuit value (about -0.75V) to

Chapter 3

45

Figure 3.6.3. Curve 1: cyclic voltammogram of Sb anodized at 2V, held for 2h atopen circuit potential (-0.75 V) and scanned from -0.75V ⇒ -1.5V ⇒ -0.75V. Curve(2) is the subsequent scan from -0.75V ⇒ +2V ⇒ -1.5V ⇒ -0.75V.-1.5V and back to open-circuit potential. A cathodic current (curve 1, fig. 3.6.3)typical of oxide reduction is observed (compare with curve 1, fig. 3.6.2). The oxide isobviously stable under open-circuit conditions (up to 24 hours). In order to reproducethe cathodic current-potential curve, the potential has to be scanned to +2V (curve 2,fig. 3.6.3).

The growth rate can be evaluated from the charge used to form the oxide,knowing the scan rate and the applied potential; the charge density is qox = j⋅t. For ascan rate, ν = 25 mV/s and a potential, V = 10 V, the time is, t = V/ν = 400 s, and witha formation current of i = 40 µA, electrode area of 0.07 cm2, the charge density is, qox

= 0.23 C/cm2. The thickness of the antimony oxide layer can be estimated from thecharge density stored in the oxide and the number of moles of antimony oxide per unitarea, noxide:

noxide = qox / 6F (3.6.6)

and noxide = ρox⋅hox / Mox (3.6.7)

hq M

Foxox ox

ox

=6 ρ (3.6.8)

where F is the Faraday constant, ρox and Mox are the density and the atomic weight ofthe oxide, respectively, and 6 electrons are involved in the reaction (reaction 3.6.5).For example, with a density, ρ Sb O2 3

= 5 g/cm3, and a molar mass,M Sb O2 3= 291.5 g/mol,

the oxide thickness is about 230 nm.

The number of moles of oxygen is related to the number of moles of antimonyoxide which are reduced according to eqs. 3.2.10 and 3.6.5:

2Sb O 3O 4Sb2 3 2→ + (3.6.9)

N NO ox2

32= ⋅ (3.6.10)

The charge density required to build-up O2 pressure can also be estimated. The amountof oxygen created during electrolysis can be estimated from the gas law:

NP V

R TO2=

⋅⋅

∆ ∆(3.6.11)

where ∆V is the change in the cell volume (the deflection of the membrane) due tooxygen gas generation, ∆P. The charge density required to give O2 is obtainedknowing the number of moles of oxygen per unit area, nO2 = NO2 /(area):

q F nF P w

RTO O2 24

4 0= ⋅ =⋅ ⋅∆

(3.6.12)

where w0 is the height of the centre deflection of the membrane. As an example: for T= 300 K, R = 8.31 J/mol⋅K; for ∆P = 2 bar = 2⋅105 Pa, and w0 = 50 µm = 50⋅10-6 m, thecharge density is qO2 = 0.15 C/cm2. This value of charge density required to obtainoxygen is comparable with the charge density stored in antimony oxide, qox = 0.23C/cm2, so the charge stored in the oxide is sufficient to give sufficient oxygen.

46


The thickness of the oxide layer required to build up pressure can also beestimated from the change in the cell volume, knowing the deflection of the membranedue to oxygen gas generation. The number of moles of antimony oxide is:

NV

Moxox ox

ox

=⋅ρ

(3.6.13)

and together with eqs. 3.6.10 and 3.6.11, the thickness of the oxide layer is given by:

hM P V

A R Toxox

ox ox

= ⋅⋅ ⋅

⋅ ⋅ ⋅2

3

∆ ∆ρ (3.6.14)

The shape form of the deflected membrane can be expressed by [Ti6801]:

( )W w xa

yasq1 0 2 2

= ⋅ ⋅ ⋅ ⋅

cos cosπ π (3.6.15)

and the volume is:

∆V w xa

ya

dxdy a b wsqb

b

a

a

1 0 2 02 216

= ⋅ ⋅

⋅ ⋅

= ⋅ ⋅

− −∫ ∫ cos cosπ π

π(3.6.16)

The shape form can be also expressed by a spherical function:

W wx

a

y

bsq2 0 1 12 2 2 2

= ⋅ −

⋅ −

(3.6.17)

and the volume is:

∆V wx

a

y

ba b wsq

a

a

b

b

2 0 1 1256

225 0

2 2 2 2

= ⋅ −

⋅ −

= ⋅ ⋅

−−∫∫ (3.6.18)

As an example with density :ρ Sb O2 3= 5 g/cm3 = 5000 kg/m3, molar mass:M Sb O2 3

=

291.5 g/mol = 291.5·10-3 kg/mol, electrode area:ASb O2 3= 0.8 mm2 = 0.8·10-6 m2,

membrane side 2a = 2b = 1.2 mm = 1.2⋅10-3 m, T = 300 K, R = 8.31 J/mol K; for ∆P =2 bar = 2⋅105 Pa and w0 = 50 µm = 50⋅10-6 m, then ∆Vsq1 = 30⋅10-12 m3, and ∆Vsq2 =20⋅10-12 m3. Therefore hsq1 = 117 nm and hsq2 = 78 nm, which, as the measurementsshow, can be formed electrochemically.

These results show that a considerable charge can be stored by forming ananodic oxide on Sb so that oxide reduction can allow the build-up of the requiredpressure during oxygen evolution at the WE. The oxide can be reduced cathodically;the system is reversible. The oxide seems to be quite stable in the electrolyte solution atopen-circuit on a time scale of hours. However, further experiments are necessary todetermine the long-term stability. The first results indicate that this system might besuitable as a replacement for the Cu electrode in the microactuator.

Chapter 3

47

3.6.2 Silver oxide

Certain metals such as silver and nickel form multivalent oxides or hydroxides.These systems find use in rechargeable battery applications [Na7101, Du..01]. Suchelectrodes might be interesting as counter electrode in the microactuator. Here wereport briefly on the properties of the silver electrode [Po6601, Di9001, Po9501].

Silver can be anodically oxidized to give Ag2O and AgO. The oxides are stable athigh pH, as can be seen in the Pourbaix diagram of fig. 3.6.4. At pH 14, the reactionscan be represented by:

2Ag OH Ag O H O 2e2 2+ − ⇔ + + −2 E = 0.35V vs.NHE (3.6.19)

Ag O OH 2AgO H O 2e2 2+ − ⇔ + + −2 E = 0.59V vs.NHE (3.6.20)

Fig. 3.6.4 refers to equilibrium conditions. Line ‘a’ represents the H2O/H2

equilibrium and line ‘b’ the O2/H2O equilibrium.

A cyclic voltammogram of Ag electrode (~ 0.15 cm2) measured in 1 M NaOHsolution is shown in fig. 3.6.5. The counter electrode is a Pt plate and Ag/AgCl is usedas reference electrode. The potential range is between - 0.2 to 1V at a scan rate of 25mV/s, without stirring. In the positive potential scan, the anodic peaks are related tooxidation of Ag to Ag2O (reaction 3.6.19) in the potential range 0.25 - 0.55 V vs.Ag/AgCl (anodic peak A1), and its further oxidation to AgO, reaction (3.6.20), athigher potentials (anodic peak A2) [Po9501, Ji9401, Po6601]. Gas evolution isobserved for potentials higher than 1.7 V.

During the reverse potential scan, the cathodic current peaks are associated withthe reduction of AgO to Ag2O (cathodic peak C1), and from Ag2O to Ag (cathodicpeak C2) at about 0.32 V and 0.035 V vs. Ag/AgCl, respectively. The charge densityunder the anodic peaks A1 and A2 is about the same, q ~ 0.11 C/cm2, and under thecathodic peaks, q(C1) ~ 0.064 C/cm2, q(C2) ~ 0.14 C/cm2.

Figure 3.6.4. Simplified Pourbaixdiagram for the Ag/H2O system.

48


These results show that sufficient charge can be stored in Ag2O so that oxidereduction:

Ag2O + H2O + 2e- → 2Ag + 2OH-

can allow the build-up of the required pressure during oxygen evolution at the workingelectrode. Again, it is essential that not all the oxide is reduced, i.e. that the bare metalis not exposed to oxygen in solution.

Clearly more work is needed to determine the stability of Ag2O under theworking conditions of the cell. It is also worthwhile to consider other rechargeableelectrodes such as Ni(OH)2/NiOOH [Du8401].

Another interesting system that may be considered is the Ag/AgCl couple[Sm5901], which is used as a reference electrode due to its high reversibility [Ba8001].At one electrode the Ag/AgCl reaction occurs and at the other electrode the Cl2 gasevolution:

AgCl(c) +e- ⇒ Ag(c) + Cl-(aq) E0 = 0.222V vs NHE

Cl2(g) + 2e- ⇒ 2Cl-(aq) E0 = 1.358V vs NHE

The problems which may occur are [Ol9001]:• during Cl2 gas evolution perhaps an interference of oxygen evolution may happen

(when aqueous solutions are used).• the reaction of Cl2 gas with Ag/AgCl during open circuit may take place, so the same

problem occurs as with a Cu electrode.

Figure 3.6.5. Cyclicvoltammogram of Ag, measuredin a 1M NaOH solution with a Ptcounter electrode and Ag/AgClreference electrode, scan rate 25mV/s; scan begins at -0.2V.

Chapter 3

49

3.7 Conclusions

The working principle and the feasibility of a new actuator based onelectrochemical processes are discussed in this chapter. The microactuator consists ofan electrochemical cell and a membrane that deflects because of the pressure of oxygengas generated by electrolysis. The actuator has three states: the electrolysis state, inwhich the pressure is built up; the passive state, in which the circuit is open and thepressure is maintained, and the pressure reduction state, in which the electrodes areshort-circuited in order to reverse the electrolysis reaction.

The life time of the components, which determines the long-term operation ofthe actuator should not be a problem. The power consumption is estimated to be lessthan 100 µW.

Under open circuit conditions the gas pressure built up in a previous electrolysisstep should, in principle, remain constant; in practice this ideal situation is difficult toachieve; the gas produced at the Pt electrode reacts at the Cu electrode. One way toprevent this is to protect the second electrode with a semi-permeable membrane likeNafion®; this approach will be described in this thesis. Another approach would be touse a metal electrode passivated by e.g. a coherent oxide layer, which does not reactwith oxygen gas. Experiments were performed and results obtained with differentmetal/metal-oxide electrodes, such as Ag and Sb; the oxide layer can be reducedelectrochemically. At open circuit, the oxide is stable on a time scale of a few hours.However, further experiments are necessary to determine the long-term stability.

The performance of the actuator is determined by the efficiency of gasproduction; this depends not only on how gas-tight the cavity can be sealed but also onhow well oxygen reduction can be suppressed. The results of this research show thatthere are no fundamental limits which make it impossible to fabricate the actuator.

50


References

[Am7201] A. Ammar and A. Saad, Anodic oxide film on antimony: II.Parameters of filmgrowth and dissolution kinetics in acid media, J. Electroanal. Chem., 34, 159-172, 1972.

[As9301] T. Asaoka, The water uptake and properties of perfluorinated polymer, ExtendedAbstracts Electrochemical Soc. Spring Meeting 1993.

[Ba8001] A. J. Bard and L. R. Faulkner, Electrochemistry methods: fundamentals andapplications, John Wiley & Sons, 1980.

[Ba9501] P.W. Barth, Silicon microvalves for gas flow control, Proc. Int. Conf. Solid-StateSensors and Actuators, Transducers’95, & European Conf. on Solid-StateTransducers, Eurosensors IX, 25-29 June, Stockholm, Sweden, p. 276-279,1995.

[Ba9601] H. Bauer, F. Derisavi-Fard, U. Eckoldt, R. Gehrmann, H. Kickel, from FriwoSilberkraft Gesellscaft fuer Batterietechnik mbH, Duisburg, Germany,Electrochemical actuator, German patent document DB 41 16 739 C1, Issued 22Oct., 1996.

[Bo9301] J.O`M. Bockris, Surface electrochemistry, Plenum Press 1993.[Bo9401] Bojinov and M. Bojinov, The antimony/klebelsbergite electrode, J. Electroanal.

Chem., 367, 195-204, 1994.[Bü9501] F.Büchi, M. Wakizoe, S. Srinivasan, Microelectrode investigation of the oxygen

permeation in different proton exchange membranes, The Electrochem.Soc.Spring meeting, Reno, Nevada, 1995, vol. 95-1, Abstract no. 465, 714-715.

[Di9101] T.P. Dirkse, Open circuit voltages of electrolytically prepared AgO,Electrochimica Acta 36(10), p. 1533-1536, 1991.

[Du8401] J.D. Dunlop, Nickel-hydrogen batteries, in Handbook of batteries and fuel cells,D. Linden (ed), McGraw-Hill, 1984.

[El9401] M. Elwenspoek, T.S.J. Lammerink, R. Miyake and J.H.J. Fluitman, Towardsintegrated microliquid handling systems, J. Micromech. Microeng. 4, pp. 227-245, 1994.

[Gi9301] E. Gileadi, Electrode kinetics, VCH Publishers, 1993.

[Gr8601] W. G. F. Grot, Nafion as a separator in electrolytic cells, The Electrochem.Soc. meeting, Boston, May, 1986.

[Ja8801] H. Janocha, Proc. Int. Technology-Transfer Conf., Actuator’88, Bremen,Germany, June 9-10, 1988, p. 389.

[Ke9001] W. Kempe and W. Schapper, Electrochemical actuators, Proc. Int. Conf. on NewActuators, Actuator’90, June, Bremen, Germany, 1990, p. 162.

[La7701] A.B. LaConti, A.R. Fragala, J.R. Boyack, Solid polymer electrolyteelectrochemical cells: electrode and other materials consideration, Proc.Symposium on Electrode Materials and Processes for Energy Conversion andStorage, The Electrochem. Soc. 1977, p.354.

[La9101] T. Laitinen, H. Revitzer, G. Sundholm, J.K. Vilhunen, D. Pavlov, M. Bojinov,Electrochemical behaviour of the antimony electrode in sulphuric acid solutions-I,II, III; Electrochimica Acta 36(14), 2081-2102, 1991.

[Li9701] M.H. Litt, A reevaluation of Nafion morphology, Polymer preprint 38(1), p.80-81, 1997.

[Lu7901] P.Lu, S. Srinivasan, Advances in water electrolysis technology with emphasis onuse of the solid polymer electrolyte, J. Appl. Electroch. 9, 1979, 269-283

[Nafion] Nafion is a E. I. du Pont de Nemours & Co. Inc. registered trademark; productcatalogue no. 27,470-4.

[Na7101] G.D. Nagy, Zn - Ag oxide batteries, (ed.) A. Fleischer, p. 136, 1971.

Chapter 3

51

[Og8401] Z. Ogumi, Z. Takehara, S. Yoshizawa, Gas permeation in SPE method, I. Oxygenpermeation through Nafion and NEOSEPTA, J. Electrochem. Soc. 131(4), 1984,769-773.

[Og8501] Z. Ogumi, T. Kuroe, Z. Takehara, Gas permeation in SPE method, II. Oxygenand hydrogen permeation through Nafion, J. Electrochem. Soc. 132(11), 1985,2601-2605.

[Ol9001] W. Olthuis, Iridium oxide based coulometric sensor-actuator systems, Ph.D.thesis, University of Twente, Enschede, the Netherlands, 1990.

[Po6601] M. Pourbaix, Atlas of electrochemical equilibria in aqueous solutions, PergamonPress, Oxford, 1966.

[Po9501] G.S. Popkirov, M. Burmeister, R.N. Schindler, Electrode potential redistributionduring silver oxidation and reduction in alkaline solution, J. ElectroanalyticalChemistry 380, p. 249-254, 1995.

[Pi5701] A.L. Pitman, M. Pourbaix and N. de Zoubov, Potential-pH diagram of theantimony-water system. Its applications to properties of the metal, its compounds,its corrosion, and antimony electrodes. J. Electrochem. Soc., 104, 594-600, 1957.

[Ri8701] P. H. Rieger, Electrochemistry, Prentice-Hall International, 1987.[Sa8601] T. Sakai, H. Takenaka, E. Torikai, Gas diffusion in the dried and hydrated

Nafions, J. Electrochem. Soc. 133(1), p. 88-92, 1986.[So8601] S.J. Sondheimer, N.J. Bruce, C.A. Fyfe, Structure and chemistry of Nafion-H: a

perfluorinated sulfonic acid polymer, JMS-Rev. Macromol. Chem. Phys. 26(3), p.353-413, 1986.

[Sm5901] D.M. Smyth, The silver/silver chloride/chlorine solid electrolyte cell, J.Electrochem. Soc. 106(8), p. 635-639, 1959.

[Wa8701] F.J. Waller, R.W. Scoyoc, Catalysis with Nafion, Chem. Tech. July 1987, p.438-441.

[Ye8101] H.L. Yeager, A. Steck, Cation and water diffusion in Nafion ion exchangemembranes: Influence of polymer structure, J. Electrochem. Soc. 128(9), p. 1880-1884, 1981.

[Ye8301] R.S. Yeo, Clustering and proton transport in Nafion membranes and itsapplications as solid polymer electrolyte, J. Electrochem. Soc. 130(3), p. 533-538, 1983.

[Ze9401] R. Zengerle, W. Geiger, M. Richter, J. Ulrich, S. Kluge, A. Richter, Applicationof micro diaphragm pumps in microfluid systems, Proc. Int. Conf. on NewActuators, Actuator’94, June, Bremen, Germany, p. 25-29, 1994.

Chapter 4

53

A deflecting membrane is used in a microvalve to adjust apressure difference within a range of 1000 to 4000 Pa overatmospheric pressure for very low liquid flow rates, 1 to 3µl/min. The membrane deflection and the drainage fluidchannel that are used to characterise the electrochemicalmicroactuator are described in this chapter. The realisation andthe experimental results of various membranes, such as 1 µmthick flat and corrugated LPCVD low-stress silicon nitride, and7 µm thick flat polyimide are described. The fluid flow througha 50 µm high channel is estimated by analytical equations, andthe pressure drop across the valve is modelled and simulated byfinite element analysis.

4.1 Introduction

The intraocular pressure (IOP) can be adjusted either by varying the flow channeldiameter or by changing the flow resistance of the eye fluid at the end of the implantedtube where it is connected to the bleb, as was described in chapter 2. In both methods,we want to use the deflection of a flexible membrane as a valve to adjust the eye fluidflow. The basic problem is: what valve geometry is needed to obtain the desiredpressure drop, for a given flow rate, fluid properties and tube geometry? The valvedoes not have to close the flow channel completely which implies a less severerestriction on the design of the membrane. Furthermore, (i) the membrane has to begas-tight and liquid-tight so that the life time of the microcell is not influenced by it. Itmust also separate the eye fluid from the electrolyte solution; (ii) since the membranewill be in contact with the eye fluid the biocompatibility has to be guaranteed; (iii) thepressure necessary to deflect the membrane should preferably not be very high.

In the following sections, the application of membranes used as pressure-to-deflection transformation are studied. A simple model of the pressure-deflectionrelationship for flat and corrugated membranes will be presented, followed by the

∗ This chapter is partly based on the papers: “An electrochemical microactuator: principle and firstresults” C. Neagu, J.G.E. Gardeniers, M. Elwenspoek, J.J. Kelly, J. Microelectromech. Syst. 5(1),1996, and “Coupled fluidics and mechanics simulations using finite element analysis” J. van Kuijk,C.R. Neagu, J.R. Gilbert, Proc. Mechanics in MEMS, Symposium at the 1996 ASME, DSC-vol.59,Nov. 17-22, Atlanta, USA, 1996.

MEMBRANE MICROVALVE ∗∗∗∗

4

Membrane microvalve

54

fabrication process and results of measurements. To investigate the electrochemicalmicroactuator, we used two types of flat membrane: silicon nitride for the dynamiccharacterisation of the electrochemical behaviour [Ne9601], and polyimide [Ne9602]for the study of the valve behaviour (liquid flow measurements of the valve). In thecase of some flat silicon nitride membranes, a metal cross was deposited on top of themembrane for an easier positioning during deflection measurements by an opticalmethod.

In the second part of this chapter, the fluid flow behaviour through a channel isdiscussed. The flow parameters, such as channel resistance and pressure drop along thechannel are calculated for a simple geometry with the use of analytical equations. Thepressure drop across the valve is calculated by numerical models: coupled fluidics andmechanics simulations using finite element analysis [Ku9601]. In the numericalsimulation a simplification is made: the flow is considered to be fully developed andlaminar.

4.2. Deflecting membrane

4.2.1 Theory

The good performance of membranes is important for the success of the valve-actuator. In active valves, for example, membranes capable of large deflection can beused to transform pressure into displacement. The effect of applied pressure onmembrane deflection was modelled using the theory for small and large deflections offlat and corrugated membranes [Je9001, Gi8201, Ma9501]. For small deflections, therelationship between pressure and deflection is linear, only bending of the membraneoccurs. In classical mechanics, the linear theory for membranes made of isotropic,homogeneous, linear elastic materials can be derived based on the following simplifiedassumptions: (1) the thickness, h, is small compared with the other dimensions of themembrane, (2) deflection, y, of the membrane is small compared with the thickness.The non-linearity which occurs for deflections more than 25 % of the membranethickness, regardless of its radius, is the result of tensile stress caused by the stretchingof the membrane. If large initial stress is present in the membrane, due to depositionprocess parameters and packaging, the effect of bending can be ignored, and a new,first order term is introduced [Vo8401].

Flat membranes

Due to the fabrication process, some stress may be present in the membrane.When a high initial stress, σ [Pa], is present in the membrane, the relationship betweenthe pressure difference P [Pa] over the membrane and the corresponding centredeflection y [m] for a flat, square membrane of thickness, h [m], side length, 2a [m],Young's modulus, E [Pa], and Poisson's ratio ν, is described by [Pa9001]:

Pa

Eh

a

Eh

y

h

y

h

4

4

2

2

3341 198 1 0 295

1= ⋅

+

⋅ −−

⋅

. . ( . )σ

νν (4.2.1)

The first term on the right side represents the resistance to bending introduced by theinitial stress. The second term shows the geometrical non-linear effect for largerpressures as a result of stretching of the membrane. In the following calculations, the

Chapter 4

55

equation of Pan will be used because his FEM-computations and measurements arewithin 5% [Ma9501].

In microtechnology, the materials often used as membranes are silicon-likematerials, such as silicon [Je9001, Bo9201], silicon nitride [Sc9401, Sp9301];polymers, mylar [Ho8401], polyimide [Mu9101]; metals, titanium [Sc9201], titaniumnickel [Be9701]; or rubbers. Silicon has the Young’s modulus of about 170 GPa andthe initial stress zero; silicon nitride, depending on the fabrication varies E = [200-300]GPa, σ = [-50-170] MPa, and polyimide E = [3-10] GPa, σ = [20-80] MPa.

Corrugated membranes

A way to increase the deflection, for the same size of the membrane, is to usecorrugations instead of a more flexible material. Experimentally it has been shown thata corrugated membrane has a larger linear range than a flat one because of thereduction of the radial stress in the membrane [Ha5701, Je9001, Mu9101].

With the introduction of corrugations, the situation can be changed essentially:the cross-sectional length of the membrane increases and the coefficients of the linearand cubic terms are altered. This effect increases both: (1) the transition between linearand non-linear deflection and (2) the deflection range in the large deflection region forequivalent loads. For shallow, sinusoidally corrugated membranes (see fig. 4.1.1) thedeflection is given by [Gi8201]:

Pa

EhA

y

hB

y

hp p

4

4

3

= ⋅ + ⋅

(4.2.2)

with the coefficients: Aq q

v

q

p =+ +

−

2 3 1

3 12

2

( )( )(4.2.3)

Bq q qp =

−−

−− +

32

9

1

6

3

32

νν( )( )

(4.2.4)

where the corrugation quality factor, q, for sinusoidal profiles is:

Figure 4.1.1. Schematic cross-sectional view of a circular corrugated membrane andits characteristic parameters: radius, a; thickness, h; corrugation depth, H, profile, s,and frequency, l, respectively.

qs

l

H

h2

2

21 15= +

. (4.2.5)

Membrane microvalve

56

where s is the corrugation profile, l is the corrugation frequency and H is thecorrugation depth. Thus q varies from 1 (flat membrane: H = 0 and s = l) to

qH

h= +1 15

2

2. (shallow profiles: s = l). The corrugation profile can be calculated by:

s R

l

R= ⋅

⋅

ππ

24

2

arcsin

/sinusoidal profile (4.2.6a)

s la NH

a=

+ 2rectangular profile (4.2.6b)

where N is the number of corrugations, and R H l H= +( ) ( )4 162 2 .As the number and frequency of corrugations increase, the deflection increases,

indicating an increase in mechanical sensitivity of the membrane. The presence ofcorrugations reduces initial stress [Sp9302], and for large corrugation depths themechanical sensitivity of the membrane only weakly depends on initial stress. Theeffects of internal stress cannot be treated as easily as for planar membranes, where asimple term is added to the pressure-deflection equation. Therefore, analytical analysis[Sc9401] or a finite element method become necessary in modelling the effects ofdifferent corrugation profiles and membrane thickness on pressure-deflectioncharacteristics.

4.2.2 Fabrication process

Silicon nitride membranes

The deflecting membranes are made of 1 µm thick low-stress LPCVD siliconnitride and are either flat or corrugated. The fabrication process of flat membranesconsists of three steps: growth of 1 µm LPCVD low-stress silicon nitride on a siliconwafer (see fig. 4.2.3 step 1); photolithography on one side of the wafer and siliconnitride etching by reactive ion etching (RIE) in an CHF3/O2 plasma to make openingsin silicon nitride for the next step: KOH etch of silicon to obtain the silicon nitridemembranes (fig. 4.2.3, step 2). In the case of the membranes with a metal cross, anextra fabrication step is added: before the Si KOH etch, a metal, Cr, is deposited andpatterned by lift-off.

The corrugated membranes are made as follows, see fig. 4.2.1: deposition on a Siwafer of a mask (e.g. SiO2, Ni, Cr, Al, photoresist). The mask is patterned and thesilicon wafer is etched by RIE in an SF6/O2/CHF3 plasma to form the corrugations; themask is removed and a short RIE etch with SF6 plasma for smoothing the corners ofthe corrugations may be performed. We omitted this step, but to avoid breakage of themembranes due to sharp corners this step should be included. Standard cleaning anddeposition of 1 µm low-stress LPCVD silicon nitride on both sides of the wafer isdone, followed by photolithography of the backside and etching of the silicon wafer inKOH solution. Corrugations with a depth of 9 µm, 11 µm and 15 µm have beenetched. An optical photo of a silicon nitride membrane with 8 corrugations is shown infig. 4.2.2.

Chapter 4

57

Figure 4.2.1. Schematic fabrication process of corrugated membranes with flat zone.

Figure 4.2.2. Optical photo of a silicon nitride membrane with 8 corrugations; 1.2mm side length, and 1 µm thick.

Polyimide membranes

We use a photosensitive polyimide (negative resist), Probimide HTR3-200 fromOCG Microelectronic Materials N.V., that it is a PMDA/ODA type of polyimide, withits own developer HTR-D2. Some tests were done to check the quality of its adhesionto different surfaces: on chromium the adhesion is relatively good but on silicon oxideand aluminium the adhesion is poor, so that a primer is necessary before spinning. Theprimer is made of 19 ml methanol, 1 ml H2O and 1 droplet APS (commerciallyavailable together with the HTR3-200).

The process starts with the growth of 1µm LPCVD low-stress silicon nitride,step 1 in fig. 4.2.3. The silicon nitride from the backside of the wafer is patterned byRIE in an CHF3/O2 plasma, then the silicon is etched in KOH until the silicon nitridemembrane is formed (step 2). On the front side, 20 nm Cr is sputtered as adhesionlayer and the primer is spun (as an extra sureness), and on top of it polyimide is spun.Depending on the spinning velocity and the time different thicknesses can be obtained.We chose a thickness of 7 µm, which was obtained by spinning at 3000 rpm for 30 s.After that, the layer is pre-baked at 90°C for 25 min, exposed and patterned to UV for30 s, baked at 90°C for 10 min, developed in HTR-D2 for 3 min, and rinsed with

Mask deposition andpatterning

Si etch by RIE

Mask removal; smoothingthe corrugations by RIE

Silicon nitride deposition;Si etch by KOH

1.

2.

3.

4.

Membrane microvalve

58

isopropanol (IPA). The polyimide is patterned to enable an easy dicing of themembranes. For the same reason, extra breaking-grooves are incorporated into thesilicon. The last step is the thermal curing at 300°C for 1 hour in a vacuum oven (step3). The last step is the silicon nitride etch on the backside of the wafer to release thepolyimide membrane (step 4).

Figure 4.2.3. Technological process of a polyimide membrane.

4.2.3 Characterisation

The mechanical behaviour of different types of membrane was tested. Flat andcorrugated silicon nitride membranes with 1.2 mm side length and 1 µm thickness weremade in square and circular geometry. The corrugation profile influences theperformance of the membrane: sharp corners will lead to stress concentration[Po9101], so a desirable shape would be a groove with a rounded bottom point androunded edges. For this, an etch process like RIE with controllable anisotropy isnecessary [Ja9401, Ja9601].

The deflection of all membranes was measured with the use of air pressure. Thesample was mounted on a holder with a pressure inlet. The pressure was supplied andmeasured by a pressure regulator and was increased gradually until membrane fractureoccurred. Membrane deflection was recorded by a Sloan DEKTAK 3030 profiler,which is a mechanical stylus profilometer. The profile of the deflection was scanned onthe diagonal of the membrane to be sure that the centre deflection is measured. Astylus force of 2x10-5 N was used for the flat membranes; for the corrugatedmembranes the stylus force was 4x10-5 N.

A set of measured pressure-deflection curves for a 1 µm thick, silicon nitride, flatsquare membrane, and for three different circular corrugations: 1 corrugation with adepth of 15 µm, 8 corrugations of 11 µm deep, and 13 corrugations of 9 µm deep areshown in fig. 4.2.4. The flat membrane deflection indeed depends on the pressure as p~ y3 whereas, the corrugated membrane deflection has a longer linear range withrespect to pressure, p ~ y.

Si 1 µm Si3N4

pattern Si3N4 by RIEKOH etch of Si

20 nm Cr deposition7 µm polyimide

backside Si3N4 etchby RIE

1.

2.

3.

4.

Chapter 4

59

0

10

20

30

40

50

60

70

0 0.1 0.2 0.3 0.4 0.5Pressure [10

5 Pa]

Def

lect

ion

[µ µµµ

m]

flat

1 corrugation

8 corrugations

13 corrugations

1

10

100

0.01 0.1 1Pressure [bar]

Def

lect

ion

[µ µµµm

]

P ~ y3

P ~ y

flat

1 corrugation

8 corrugation

13 corrugation

Figure 4.2.4. Pressure-deflection measurements for different types of silicon nitridemembrane, 1.2 x 1.2 mm2 , 1µm thick: flat; 1 corrugation with corrugation depth,H = 15 µm; 8 corrugation, H = 11 µm; 13 corrugation, H = 9 µm.

The stress in the membrane is very dependent on the specific fabrication process.Our flat square membranes were made of low-stress silicon nitride and of polyimide.From the theoretical plots of the pressure-deflection relationship for stress-free andinitial stress membranes, our experimental data are closer to those with initial stress.Therefore, equation 4.2.1 is used as the analytical expression for the comparison withthe experimental measurements. The stress parameters of the polyimide membranes,processed in our laboratory were obtained experimentally by fitting the experimentalpressure-deflection curve with theory: Young's modulus E = 5 GPa and initial stress σ= 33 MPa. Silicon nitride membranes have initial stress σ = 100 MPa and Young'smodulus E = 300 GPa. Analytical simulations for flat 1.2 mm side, silicon nitride of 1µm thick and of 8 µm polyimide with Poisson's ratio ν = 0.3 are shown in fig. 4.2.5. Inthe same figure, the experimental results for both membranes are presented. Goodagreement was found between calculated and experimental deflection-pressurerelations.

Summarising for the three types of membrane, we conclude that (1) the flatsilicon nitride cannot reach the required deflection of 50 µm, unless the side of themembrane is increased, which cannot be done due to the required limited size of theactuator; (2) the corrugated silicon nitride can easily deflect 50 µm for a relatively lowpressure (0.3⋅105 -0.5⋅105 Pa) but until now the membranes are very brittle and break

Membrane microvalve

60

easily. The corrugation form is not good enough due to the existence of sharp corners.(3) flat polyimide membranes deflect 50 µm, at a relatively higher pressure, 2⋅105 Pa,than corrugated silicon nitride; this is because polyimide is thicker, 8 µm).

Silicon nitride is a relatively good gas-tight and liquid-tight material since it isused as mask for silicon oxidation at high temperature (~1100°C) and humidity.However, long-term tests for gas and liquid diffusion have not yet been done. Thebiocompatiblity of both types of material for eye implant has yet to be checked. Asolution may be to cover these membranes with a thin film of a biocompatible material,such as titanium, Teflon, or silicone rubber. The properties of this material can bechosen in such a way that this thin film will not influence the efficiency of themembrane.

Figure 4.2.5. Comparison of calculated and experimental deflection for 1.2 mm side,square flat membrane of 8 µm thick polyimide and of 1 µm thick Si3N4.

Chapter 4

61

4.3 Fluid flow in a channel

We want to use the deflecting membrane in a microvalve to adjust a pressuredifference between 1000 and 4000 Pa over atmospheric pressure for very low liquidflow rates, 1 to 3 µl/min. The microvalve may be, for example, designed in such a waythat the membrane closes an opening in a plate above the membrane, or the membraneis situated inside the channel, fig. 4.3.1.

(a) (b)

Figure 4.3.1. Two possible designs to regulate the resistance of the fluid flow; (a)opening, (b) channel.

The valve system will consist of the silicone rubber tube implant (with inner andouter diameter of 0.3 mm and 0.6 mm, respectively), the deflecting membrane (valveitself with 1.2 mm side length) and a flow channel which makes the connectionbetween the silicone rubber tube and the valve (that is 50 µm high), as is sketched infig. 4.3.2a. For the measurements and flow simulation a plastic tube of 1.5 mmdiameter is used. The experimental set-up is shown in fig. 4.3.2b.

In order to get an impression of the characteristics of such a valve, we estimatethe pressure drop across different components of the valve system under the followingassumptions: laminar flow and fully developed velocity profile; the pressure drop dueto the inlet/outlet, bending of the flow channel and the losses introduced by them werenot considered.

(a)

Membrane microvalve

62

(b)

Figure 4.3.2. Sketch of a (a) valve system composed of the silicone rubber tubeimplant, transition channel, and the valve itself (deflecting membrane); (b) Set-upused for simulation and flow experiments.Since the pressure difference to be adjusted is very low and the real situation in thevalve will be different and much more complex, a more exact simulation is necessary.This situation can only be modelled accurately with numerical methods. Coupled fluidand mechanical finite element calculations, which were developed in our group byKuijk [Ku9601], showed that membrane deflections differ more than 10% from thenon-coupled calculations so, for accurate calculation this coupled calculations can beused.

In the following paragraphs, analytical and numerical models will be described.Some basic concepts of fluid mechanics will first be defined. Then, an analytical modelfor a simple geometry such as that of the channel is introduced in order to calculate theresistance of the channel and the loss in fluid pressure (pressure drop) due to the lengthof the channel. The losses due to the entrance effects and changes in the flow channeldiameter are only roughly estimated. For a more complex geometry, such as that of avalve, a two-dimensional numerical simulation will be presented. This model is used toestimate the pressure drop over the valve.

4.3.1 Analytical model

Consider a flow through a long, horizontal, circular pipe, as shown in figure4.3.3. As fluid uniformly enters the pipe, the flow will evolve into a distribution whichis independent of distance. This means that the fluid axial velocity, u(r,x) in the xdirection does not change with x, so u ≈ u(r). This distribution is called a fullydeveloped profile, where the velocity profile is constant with increasing distance andthe pressure decreases linearly with x.

Chapter 4

63

Figure 4.3.3. Flow in a circular pipe: schematic description of developing velocityprofiles and pressure changes in the entrance of a pipe to form a fully developed fluidflow [Wh9401].

For a laminar flow, the volume flow rate through a channel, Φ [m3/s], and thepressure drop along the channel, ∆p [Pa] are linearly related:

∆ Φp R= ⋅ (4.3.1)

where R [Ns/m5] is the channel resistance, a constant that depends on the channelgeometry. For a circular pipe R is given by [Wh9401]:

Rp l

d= =

∆Φ

1284

µπ

(4.3.2)

with fluid viscosity, µ [Pa⋅s], length of the flow channel, l [m], and channel diameter, d[m]. Formula (4.3.2) can be applied only for a Hagen-Poisseuile flow, i.e. when (i) theflow is laminar; (ii) the flow is fully developed, u ≈ u(r); and (iii) the pressure drop inthe inlet and the outlet region of the channel is small compared to the viscous pressuredrop in the fully developed region; this means the channel is long compared to theentrance and exit length of the flow into and out of this channel.

Since silicon micromachining techniques are used to construct the actuator andthe micromechanical etch processes can make only specifically defined cross-sections,it is necessary to use other models to describe the relationship between pressure andvolume flow than those that describe circular geometry. In the case of channels with anarbitrary cross section, two parameters are introduced: the hydraulic diameter, Dh [m](that replaces the diameter):

DA

h =4

P(4.3.3)

Membrane microvalve

64

where A [m2] is the cross section of the channel and P [m] is the wetted perimeter ofthe channel. The second parameter is the shape constant, kshape, that depends only onthe cross section.

A general formula for the flow resistance of a channel of arbitrary shape, in thecase of a laminar, fully developed, incompressible flow can be written as [Sh7801]:

Rk l

D A

shape

h

=⋅ ⋅

⋅

22 µ (4.3.4)

In the Reynolds number given by:

Re=⋅ ⋅ρµv d

(4.3.5)

its characteristic diameter, d, should be replaced by Dh in case of a general channel:

ReDh = ⋅ ⋅ρµv Dh (4.3.6)

where ρ [Kg/m3] is the fluid density, v [m/s] is the average velocity of the fluid.

Some resistance formulas for a circular, rectangular, square and triangular cross-section are presented in the table 4-1 [Wh9401].

ChannelShape

Parameters

Shape constant

Kshape

16 14.2 15.5 24 13.3

HydraulicDiameter, Dh

[m]

2⋅a 2⋅a 4a b

a b

⋅+

4⋅b 1.04⋅a

Area

A [m2]π⋅a2 4⋅a2 4⋅a⋅b 4⋅a⋅b 1.41⋅a2

Entrance length

Le [m]

equation

(4.3.9)0.09⋅Dh⋅Re 0.085⋅Dh⋅Re > 0.01⋅Dh⋅Re 0.04⋅Dh⋅Re

Resistance

R [Ns/m5]

84

⋅⋅

l

aπµ 178

4

. ⋅ l

aµ 31

64

2

3 3

( )a b l

a b

+ ⋅ µ3

4 3

⋅⋅l

a bµ 17 4

4

. ⋅ la

µ

Table 4-1. Shape constant, hydraulic diameter, area, entrance length and (flow)resistance for different types of channel cross section, with channel length, l [m], andfluid viscosity µ [Ns/m2] [Sh7801, Ta9401].

An example of a typical silicone rubber tube was considered in chapter 2: theflow in the tube will be laminar (Re << 2300), and the pressure drop along the siliconetube is smaller than 5 Pa, which is negligible compared to the normal IOP ~ 2000 Pa.The pressure drop along a straight channel, ∆p, is a loss in fluid pressure due to thefriction and it is called friction-head loss. This loss in pressure in the fully developedregion dominates the pressure drop due to other effects, such as the entrance and exiteffects [Bl8401, Wh9401]. A fully developed flow condition actually means that the

2a 2a

2a 2a/2b = 2

2b 2b

2a a>>b

2b

2a

55°

Chapter 4

65

flow channel should be long compared to the entrance length, Le [m]. The entrance(inlet) length is the distance between the inlet and the point at which an essentiallyfully developed flow is established. For the uniform inlet flow through a uniformcircular pipe, the accepted relation for the entrance length is [Sh7801, Ch7301]:

L De h=⋅ +

+ ⋅

0 6

0 035 10 056

.

. Re. Re laminar (4.3.7)

In the case of the implanted, silicone rubber tube of 300 µm inner diameter where Re <1, the entrance length is circa 180 µm, which is very small compared to the total lengthof about 2 cm. Thus, the pressure drop along the tube and the resistance of the tubecan be calculated considering its total length, l = 2 cm.

As was mentioned above, other losses are due to the entrance and exit effects,bends in pipes, sudden or gradual expansions or contractions, valves, open or partiallyclosed. These losses may be quite big: a partially closed valve can create a greaterpressure drop than a long pipe, as we will see in the case of our valve. The measuredminor loss is usually tabulated as a loss coefficient, K, from which the pressure dropcan be calculated as [Wh9401]:

( )∆ Φp K K Aminor2v= ⋅ = ⋅1

212

2ρ ρ (4.3.8)

However, if the channel/tubing size changes so that v2 changes, each of the minorlosses has to be summed separately.

In the case of our tubing and valve design, we may have the following types ofentrances (see fig. 4.3.2 and 4.3.4): the fluid entrance into the silicone rubber tube canbe seen as a sharp-edged entrance for which, KSEE = 0.5 (while for a well-rounded inletK is negligible); the transition from the silicon rubber tube of 300 µm diameter to theconnection channel of 50 µm square side length may be considered a suddencontraction, KSC ~ 0.42⋅(1-d2/D2); ). The transition from the connection channel to thevalve of 1 mm side length can be seen as a sudden expansion, KSE = (1-d2/D2)2. For alltypes of channel outlets that have the exit larger than the channel diameter, Kout = 1.0.The pressure drop due to all these types of fluid entrances are negligible (< 1 Pa)compared to pressure drop that is achieved due to the valve, ~ 2000 Pa.

The pressure drop that has to be regulated by the whole actuator (includingtubing and valve) is between 1000 and 4000 Pa for a flow of 1-3 µl/min. Therefore,these values of the pressure drop are much higher than those due to the friction andminor losses; this implies that regulation of the pressure is done mainly by the valve asrequired for the device.

So far, we have described the fluid flow in a horizontal tube. When the tube isnot horizontal, gravity may influence the variation in pressure. In the case of aninclined tube the head loss includes, besides the pressure drop along the tube, also thepressure due to the difference in the height, ∆p g z z+ −ρ ( )2 1 . The length of the siliconerubber tube implant is about 2 cm, and if we consider that the maximum difference in

Membrane microvalve

66

height is ∆z ~ 4 cm, the loss in fluid pressure due to the gravity is 400 Pa. Such avalue is large compared to the 132 Pa accuracy with which the IOP has to be adjusted.This variation in pressure due to the position of the implant can be compensated, forexample, by shortening the tube, that can not be made smaller than 1 cm (for surgeryreasons), or by using an electronic feedback control.

v v

(a) (b) (c)

Figure 4.3.4. Several types of minor losses that may be seen in the fluid drainagechannel: (a) sharp-edged entrance, (b) sudden expansion;(c) sudden contraction.

KSEE ~ 0.5

d D dD

KSE ~ (1-d2/D2)2 KSC ~ 0.42(1-d2/D2)

Chapter 4

67

4.3.2 Numerical model

The geometry of the types of valve shown above are not standard, so modelswhich allow an analytical calculation of the pressure drop across the membrane are notavailable. In order to get an impression of the characteristics of such a valve, weestimate the pressure drop with the following assumptions: laminar flow, and fullydeveloped velocity profile; the pressure difference due to the bending and the lossesintroduced by it were not considered. Since the pressure difference to be adjusted isvery low and the real situation in the valve will be different and much more complex, amore exact simulation is necessary.

For the modelling and simulation of the liquid flow through the valve, a design ofthe channel similar as that in fig. 4.3.2b is used. The channel is 50 µm high, 1.5 mmwidth and the flow is from right to left. The valve is centred under the outlet.

Two types of pressure are related to this valve, the pressure generated by theactuator (PA) and the pressure drop over the valve (∆p). As the membrane partiallycloses the outlet the local flow velocity over the membrane at the outlet increases. Inthis design, it is not necessary to completely close the outlet but just to control thepressure drop between inlet and outlet ∆p.

In the following, the variation of the ∆p as a function of the fluid flow issimulated with the coupled calculation tools [Ku9701].

The actuator pressure is presented in dimensionless form (PA* ) and is related to

the membrane by:

P = A

4

4* P a

E h

⋅⋅

(4.3.9)

in which a, E and h are the radius, Young’s Modulus and thickness of the membrane,respectively. The pressure drop over the valve is related to the flow through thedevice:

Rp

= ∆Φ

(4.3.1)

The resistance of the flow channel changes with flow, determined by the height of thedeflection of the valve. Clearly a non linear behaviour. The resistance, R, can benormalised using the following equation:

R = 3Rh

µ (4.3.10)

in which h is the height of the channel.

In the coupled simulation of the active valve the geometric non linear behaviourof the membrane is modelled in the structural finite element calculations. The nodesand elements of the membrane finite element mesh correspond to the fluid finiteelement mesh. A non-deflected membrane was chosen as the starting mesh and thepressure PA (that is kept constant during each flow simulation) was applied. A meshwhich is used for the eye pressure regulator is shown in fig. 4.3.5.

Membrane microvalve

68

Figure 4.3.5. Finite element mesh of the flow channel. A zoom in of the mesh at theoutlet of the flow channel is shown on the right. The flexible membrane is centredbelow the outlet.

The height of the channel is 50 µm, the membrane is centred underneath the outlet, andthe flow is from right to left with the outlet extending upwards. The inset on the rightin fig. 4.3.5 shows a close up of the outlet of the valve at the point where themembrane of the actuator is deflected upwards.

The fluid pressure is found by calculating the complete Navier-Stokes equationfor a steady state incompressible fluid which generates the velocity and pressure field.From this the viscous stresses on the flexible part (the membrane) are calculated. Theinlet velocity is kept constant and the output of the fluid calculation is the pressure andthe velocity field, so the pressure drop can be calculated. Mechanical parameters werevaried with the actuator pressure. Fluid parameters were varied by changing theReynolds number with the flow rate (eq. 4.3.5). So, the Reynolds number represents,in our case, a measure for the flow: The channel area is A = h⋅w, where the channelwidth, w, for 2-D simulation equals 1. The pressure drop can be written as function ofReynolds number:

Rew

=⋅ρ

µΦ (4.3.11)

∆pw

R Re=⋅

⋅µ

ρ (4.3.12)

The numerical modelling and the simulation tools are described else where [Ku9701].

Chapter 4

69

0

10

20

30

40

50

60

70

80

90

0 2000 4000 6000 8000 10000

Actuator Pressure [Pa]

Max

imum

Def

lect

ion

[ µ µµµm

] lin non-lin

Re=1 Re=2

Re=4 Re=8

Figure 4.3.6. Simulation of a 10 µm polyimide membrane deflection as a function ofactuator pressure and Reynolds numbers (that are equivalent to different flow rates).

The results of the simulation may be presented by, for example, comparing themaximum deflection of the membrane. This is done for different actuator pressures andflow velocities (that are introduced into the Reynolds numbers, eq. 4.3.5) in fig. 4.3.6.The actuator pressure ranges from 103 to 104 Pa. These simulations were done with apolyimide membrane of 10 µm thick, 1.5 mm side length, E = 6 GPa. Large deflectionsare readily obtained at these lower pressures.

Another way to present the results is by calculating the pressure drop betweeninlet and outlet i.e. over the valve system. In this way the results can be compared withthe measurements and the complete device can be characterised. The pressure drop canbe presented as fluid normalised resistance (eq. 4.3.10) or as pressure (eq. 4.3.12). Thesimulation of an 8 µm polyimide membrane with air as fluid is shown in fig. 4.3.7, andfor water flow over a 1 µm silicon nitride in fig. 4.3.8. In both cases, the actuatorpressure was varied from 0 Pa to the limit of the coupled solver, PA > 1⋅105 Pa cannotbe simulated. These simulations involve about hundred runs. Figures 4.3.7a and 4.3.8ashow the normalised flow resistance of the valve system function of the normalisedactuator pressure for different Reynolds numbers (i.e. different flows). The pressuredrop over the valve system corresponding to the normalised flow resistance, shown infigs. 4.3.7a and 4.3.8a, are presented in figures 4.3.7b and 4.3.8b as function of theactuator pressure. As can be seen, the pressures obtained across the valve due to themembrane deflection (actuator pressure) are not in the range of 2000 Pa for low flowrates (Φ ≤ 3⋅10-11m3/s), what we would like to have, since actuator pressures higherthan 1⋅105 Pa cannot be simulated. When the actuator pressures are higher, the valvedoes close the channel more effectively, a 2000 Pa pressure drop across the valve isobtained, as we will see in the flow experiments described in chapter 5.4.4. From thesetwo figures it can be seen that the shape and the pressure drop over the flexiblemembrane vary strongly with Reynolds number and material properties.

Membrane microvalve

70

(a)

Air flow over Polyimid diaphragm

0

50

100

150

200

250

300

350

400

450

0 0.0002 0.0004 0.0006 0.0008 0.001

Normalised Actuator Pressure

Nor

mal

ised

Res

ista

nce

[103 ]

(b)

Air flow over Polyimide diaphragm

0

1

2

3

4

5

6

0 0.2 0.4 0.6 0.8 1Actuator pressure [10 5Pa]

∆ ∆∆∆p [k

Pa]

Φ = 1.54.10-

11

Φ = 3.10-11

(Re=2)

Φ = 8.5.10-11

(Re=5.5)

Figure 4.3.7. Coupled simulation on a 7 µm polyimide membrane with air flow. Theactuator pressure is varied and the pressure drop across the valve system is given as:(a) normalised resistance (normalised to membrane parameters, 0.001 ≈ 1 bar) asfunction on the normalised actuator pressure, for different Reynolds numbers, and (b)pressure drop across the system function on the actuator pressure, for different flowrates.

It is obvious that the results can only be compared with the analytical formulas ifthe experimental point of maximum deflection and shape of the membrane do notdeviate much from the theoretical behaviour. For example at Reynolds numbers over 8the flow causes the membrane to deflect downwards, which is difficult to compare toanalytical membrane theory.

Another prediction of the coupled solution is the shift of the maximum deflectionpoint away from the inlet, when the flow pushes harder than the pressure whichdeflects the membrane [Ku9601].

Also, it shows that the pressure decreases linearly over the right hand side of themembrane when it is still not deflected. When the membrane deflects, the flow channelis cut off and therefore the pressure drop increases. The highest pressure gradient isacross the top part of the membrane or the first 100 µm in this example.

Chapter 4

71

(a)

Water flow over Silicon Nitride diaphragm

50

55

60

65

70

75

80

0 0.00004 0.00008 0.00012 0.00016

Normalised Actuator Pressure

Nor

mal

ised

Res

ista

nce

[103 ]

Re=1

Re=2

Re=3

Re=4

Re=5

(b)

Water flow over Silicon nitride diaphragm

0

0.5

1

1.5

2

2.5

3

3.5

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9Actuator Pressure [10

5Pa]

∆ ∆∆∆p [k

Pa]

Φ = 1.10-12

(Re=1)

Φ = 2.10-12

Φ = 3.10-12

Φ = 4.10-12

Φ = 5.10-12

(Re=5)

Figure 4.3.8. Coupled simulation on an 1 µm silicon nitride membrane with waterflow. The actuator pressure is varied and the pressure drop across the valve system isgiven as: (a) normalised resistance (normalised to membrane parameters, 0.001 ≈ 1bar) as function on the normalised actuator pressure, for different Reynolds numbers,and (b) pressure drop across the system function on the actuator pressure, fordifferent flow rates.

Membrane microvalve

72

4.4 Conclusions

Two types of flat membrane used in a valve system and the fluid flow across thevalve were examined. The mechanical behaviour of 1.2x1.2 mm2, 1 µm flat andcorrugated low-stress LPCVD silicon nitride and 7 µm flat polyimide was investigated.Their fabrication process is also presented. Large deflection performance ofmembranes is strongly influenced by the introduction of corrugations into themembrane. In the large deflection range and at equivalent pressures, the deflection ofcorrugated membranes is increased compared to that of flat membranes. In the smalldeflection range, the corrugations extend the linear range of the pressure-deflectionbehaviour. The performance of the corrugated membranes depend, among otherfactors, on the shape of the corrugation, i.e. sharp corners will make the membranemore brittle. From the results presented here, the polyimide membrane is the bestcandidate to be used in the valve system due to its high deflection, 50 µm.

The fluid flow in a valve system was analysed with the use of analyticalequations, for a simple channel geometry, while the pressure drop across the valve wasestimated using a numerical model. The results show the influence of the membranedeflection on the flow resistance, as well as on the pressure across the valve system.The volume flow rates are small, so the membrane is not deformed due to the fluidpressure. Clearly the pressure drop across the valve system can be adjusted in thedesired range, 2000 - 4000 Pa, when the actuator pressure has a higher value than1⋅105 Pa.

The whole valve system is in contact with the eye fluid, thus biocompatibility is avery important issue that has to be taken into consideration. Materials used so far aresilicon, silicon nitride and polyimide that are not biocompatible. Therefore, a coverwith a thin layer of Teflon®, silicone rubber or titanium might be sufficient. Besides thisrequirement, gases and liquids should not diffuse through the membrane in order toobtain a tight electrochemical cell. Silicon and silicon nitride are known to have a quitelow diffusion coefficient for gases and liquids.

References

[Be8801] C.W. Bert, J.L. Martindale, An accurate, simplified method for analyzing thinplates undergoing large deflections, J. Am. Inst. Aeronautics Astronautics, 26, p.235, 1988.

[Be9701] W. Bernard, H. Kahn, Heuer, M. Huff, A titanium-nickel shape-memory alloyactuated micropump, Transducers’97, June 16-19, 1997, Chicago, USA

[Bl8401] R.D. Blevis, Applied fluid dynamics handbook, 1984.[Ch7301] R-Y. Chen, Flow in the entrance region at low Reynolds numbers, J. Fluids Eng.

95, p.153-158, 1973.[Ch8701] Hin-Leung Chau and K. D. Wise, Scaling limits in batch fabricated silicon

pressure sensors, IEEE Trans. Electron Devices, ED-34, p. 851, 1987.[Gi8201] M. di Giovanni, Flat and corrugated diaphragm design handbook, Marcel

Dekker Inc., New York, U.S.A., 1982.[Ha5701] J.A. Haringx, Design of corrugated diaphragms, ASME Trans. 79, p.55-64, 1957.[Ho8401] D. Hohm, R. Gerhard-Multhaupt, Silicon-dioxide electret transducer, J. Acoust.

Soc. Am. 75, p. 1297-1298, 1984.

Chapter 4

73

[Ja9401] H.V. Jansen, M.J. de Boer, R. Legtenberg, and M.C. Elwenspoek, The blacksilicon method: an universal method for determining the parameter setting of afluorine-based reactive ion etcher in deep silicon trench etching with profilecontrol. Journal of Micromechanics and Microengineering 5, p. 115-120, 1995.

[Ja9601] H.V. Jansen, M.J. de Boer, M. Elwenspoek, The black silicon method VI: highaspect ratio track etching for MEMS applications, Proc. IEEE Micro ElectricalMechanical Systems MEMS '96, p.250-257, February, San Diego, USA, 1996.

[Je9001] H. Jerman, "The fabrication and use of micro machined corrugated silicondiaphragms", Sensors and Actuators A21-23, 1990, p.988.

[Ku9601] J. van Kuijk, C.R. Neagu, J.R. Gilbert, Coupled fluidics and mechanicssimulations using finite element analysis, Mechanics in MEMS symposium at the1996 ASME, DSC-vol.59, Nov. 17-22, Atlanta, USA, 1996.

[Ku9701] J. van Kuijk, Numerical modelling of flows in Micro Mechanical Systems, Ph.D.Thesis, University of Twente, Enschede, The Netherlands, 1997.

[Ma9501] D. Maier-Schneider, J. Maibach, E. Obermeier, A new analytical solution for theload-deflection of square membranes, J. MEMS 4(4), p. 238-241, 1995.

[Mu9101] C.J. van Mullem, K.J. Gabriel and H. Fujita, Large deflection performance ofsurface micromachined corrugated diaphragms, Proc.Transducers ‘91, SanFrancisco, USA, June 24-27, p. 1014 - 1017, 1991.

[Ne9601] C. Neagu, J.G.E. Gardeniers, M. Elwenspoek and J.J. Kelly, An electrochemicalmicroactuator: principle and first results, J. Microelectromech. Syst. 5(1), p.2-9,1996.

[Ne9602] C. Neagu, J.G.E. Gardeniers, M. Elwenspoek and J.H.J. Fluitman, Anelectrochemical actuated microvalve, Actuator ‘96, June 19-21, 1996, Bremen,Germany.

[Pa9001] J. Y. Pan, P. Lin, F. Maseeh, S. D. Senturia, "Verification of FEM analysis ofload-deflection methods for measuring mechanical properties of thin films",Techn. Digest IEEE Solid-State Sensors Workshop, Hilton Head Island, U.S.A.,p.70, June 1990.

[Po9101] F. Pourahmadi, D. Gee and K. Petersen, The effect of corner radius of curvatureon the mechanical strength of micromachined single-crystal silicon structures”,Dig. Tech. Papers Transducers ‘91, San Francisco, U.S.A., 1991, p. 197-200.

[Sc9201] W.K. Schomburg, B. Scherrer, 3.5 mm thin valves in titanium membranes, J.Micromec. Microeng., 2, p. 184-188, 1992.

[Sc9401] P.R. Scheeper, W. Olthuis, P. Bergveld, “The design, fabrication, and testing ofcorrugated silicon nitride diaphragms”, J. Microelectromech. Syst. 3(1), pp. 36-50, 1994.

[Sh7801] R.K. Shah, W.E. A.K. London, Laminar flow forced convection in ducts,Academic Press, 1978.

[Sp9301] V. L. Spiering, S. Bouwstra, J. F. Burger, M. Elwenspoek, "Membranesfabricated with a deep single corrugation for package stress reduction and residualstress relief", J. Micromech. Microeng., 3, p.243, 1993.

[Sp9302] V. L. Spiering, S. Bouwstra, J. Burger, M. Elwenspoek, Package stress reductionand low initial stress thanks to deep corrugations, 7th Int. Conf. Solid-stateSensors and Actuators (Transducers ’93), Yokohama, Japan, June 6, 1993.

[Vo8401] J.A. Voorthuyzen and P. Bergveld, “The influence of tensile forces on thedeflection of circular diaphragms in pressure sensors, Sensors and Actuators 6,pp. 201-213, 1984.

[Wh9401] F.M. White, Fluid mechanics, McGraw-Hill,Inc. 1994, chapter 6.[Zh9401] Y. Zhang, K.D. Wise, Performance of non-planar silicon diaphragms under large

deflections, J. Microelectromech. Syst. 3(2), p. 59-68, 1994.

Chapter 5

73

An electrochemical microactuator has been developed as anactive valve. The microactuator consists of an electrochemicalcell with a deflecting membrane, which is used for current-to-pressure-to-deflection transformation. The microactuator witha size of 2 x 2 x 1 mm3 was made with silicon micromachiningand thin film techniques. When the microvalve was actuated at1.6 V and currents below 50 µA, pressures of 2 bar could beeasily obtained, causing membrane deflections in the range of30 to 70 µm. The microactuator may be used as active valve inmicro-liquid systems or biomedical applications.

5.1 Introduction

Micro System Technology is a broad area with many applications in e.g.biomedicine, chemical analysis and liquid handling. The use of silicon microtechnologyhas increased rapidly during recent years. This technology has led to a growing interestin research on micro-liquid handling systems [Gr9301]. Basic components such aspumps, passive/active valves and flow sensors using for example electrostatic,piezoelectric or thermo-pneumatic principles have already been realised [Po9001,El9401]. The active valves reported in the literature differ in supply voltage, powerconsumption and dimensions, as table 5.1-1 shows. An extensive comparison can befound in [Ze9401, Ba9501, Ku9701].

Actuation principle[Ze9401, Ba9401]

Pressure[105 Pa]

Stroke[µm]

Voltage[V]

Power consumption[mW]

Electrostatic < 0.5 < 10 100 - 200 < 1Thermopneumatic 1< P < 100 30 < d < 100 5 - 20 2000 - 5000Electromagnetic < 0.5 > 100 20 - 50 50 - 300Piezoelectric < 1 10 < d < 50 100 - 200 < 5

Table 5.1-1. Different micro-valves made with silicon micromachining techniques.

∗ This chapter is based on the paper: “An electrochemical microactuator: principle and first results”C. Neagu, J.G.E. Gardeniers, M. Elwenspoek, J.J. Kelly, J. Microelectromech. Syst. 5(1), p.2-9, 1996,and presentation “An electrochemical actuated microvalve” C. Neagu, J.G.E. Gardeniers, M.Elwenspoek, J.H.J. Fluitman, Proc. Actuator ‘96, p.41-44, June 19-21, 1996, Bremen, Germany.

THE ELECTROCHEMICAL MICROACTUATOR ∗∗∗∗

5

The electrochemical microactuator

74

The goal of this research was to develop a micromachined active valve thatprovides a continuous regulation around a desired pressure value. To obtain a lowenergy consumption and to have the possibility of discontinuous supply of power, wechose for electrochemical actuation, which is based on the electrolysis of an aqueouselectrolyte solution [Ba8001, Ri8701]. The reversible chemical reactions, which aredriven by an external current source, lead to gas evolution or gas removal, dependingon the direction of current flow. In a closed system the generation and reduction of thegas pressure [Ja8801, Ke9001, Ha9501, Ne9601] can be used to change the deflectionof a membrane, which in turn can close or open a liquid channel [Ne9602]. If such anelectrolytic cell is operated under open-circuit conditions, the pressure and thus thedeflection state of the membrane will, ideally, be maintained. This means that noenergy is required to maintain the state of the valve. The advantage of this type ofactuation is a possible discontinuous use of power. Relatively large pressures (up totens of bar) can be reached in the electrochemical actuator with a low energyconsumption.

At the beginning of this project there were numerous questions concerning thefeasibility of the electrochemical eye pressure actuator based on silicon technology.These included the choice of the electrodes-electrolyte system, technologicaluncertainties concerning the fabrication (for example, the adhesion of Nafion), and thedemands from medical application. Obviously, it was not possible to provide answersimmediately to all these questions. However, in the course of the project it seemed as ifmore questions were raised than were answered. This is what science is about: anendless chain of problems and solutions and more problems.

In this chapter, the design, the technology and the experimental results of twotypes of electrochemical actuator are described. Both microactuators are made withthe use of micromechanical fabrication techniques. The fabrication of and the tests onthe first prototype gave insight into the operating principle, the feasibility of actuation,and an indication of the complexity of the technological and physical problems due tothe scaling down and the long-term operation (such as bonding, leakage and unwantedchemical reactions). The main drawback of the first prototype was the decrease of thegas pressure in the open circuit mode. Improvements of the cell design, primarilyaimed at protection of the counter electrode, were introduced in the second prototype.Although a slower decay of the pressure during open circuit was obtained, the actuatorwas still not satisfactory. The following step was a simple integration of themicroactuator into a drainage channel, described in chapter 4. The microactuator actsas an active valve to adjust the fluid resistance and consequently the pressure over thevalve. For the first flow experiments, presented here, air was used as fluid.

5.2 Design layout

In chapter 4, the (eye) fluid drainage channel and the deflecting membrane weredescribed and characterised. With these data a first microactuator model was designedto gain insight into the practical problems concerning the microfabrication andoperation of the device.

The basic idea is to etch a cavity in a silicon wafer with a KOH solution and tobond this wafer to another wafer in order to get a closed cavity. Figure 5.2.1 shows a

Chapter 5

75

schematic drawing of an electrochemical microactuator, having a size of about 2 x 2 x1 mm3. In order not to be limited by the technological restrictions, the actuatorprototypes were fabricated with a combination of clean room processing and muchmanual handling.

The electrochemical cell contains two electrodes, platinum and copper in a 1molar (1M) CuSO4⋅5H2O solution. In the construction of the micromechanical actuatorthere are several important design and production steps:

1. the anode: the platinum electrode should have a relatively large active area becausethe oxygen gas has to react at the platinum electrode to form water when theactuator is in the pressure reduction state.

2. the cathode: the copper electrode should be thick enough to ensure the requiredlife time: during electrolysis copper precipitates on the copper electrode; when theactuator pressure is reduced, copper ions redissolve in the electrolyte solution. Inthis cycle some copper may be lost due to poor adhesion of the deposited metal.Copper released in this way is no longer available for the electrolysis cycle.

3. the perm-selective membrane, Nafion: should be produced by a method compatiblewith the micromechanical techniques, and after deposition it should maintain itsproperties during the subsequent steps.

4. the electrolyte: electrolysis should yield only one type of gas (oxygen). Thesolution should be compatible with other materials in the actuator.

5. the deflecting membrane: this is used to convert the gas pressure into mechanicalmovement. It should have a small area, not bigger than 2 x 2 mm2; its deflectionrange is from 0 to at least 50 µm. It should also have a low diffusion coefficient foroxygen, to guarantee the life-time of the actuator. Since it is in contact with the eyefluid, it is preferable to use a biocompatible material. Alternatively, it may becovered with a very thin layer of a biocompatible material (glass, metal, polymers).

Figure 5.2.1. Schematic layout of a micromachined electrochemical actuator; thedeflecting membrane and the electrochemical cell are shown; the size is about 2 x 2 x1 mm3.


76

6. bonding: the wafers should be bonded to each other to create a gas-tight cavity.This cannot be done with conventional high temperature bonding techniques,because this would destroy the Nafion membrane. Another alternative has to befound. Polymer bonding may be a good solution, although the swelling of thepolymer when brought into contact with water can be a problem.

7. sealing: the filled cavity has to be gas-tight; the electrolyte can be enclosed in thecavity during the bonding or after bonding in a vacuum chamber, after which thecavity may be sealed with a polymer or an electrochemical plug [Ga9401].

5.3 Electrochemical reactions

Depending on requirements and applications, several electrochemical reactionscan be chosen [Ja9201]. The choice of the electrodes and ions in the solutiondetermines the kind of gas which evolves at the electrode(s) (see chapter 3). The build-up time of the pressure depends on different variables which affect the electrodereaction rate [Ba8001]: electrode material, surface area and surface condition, cellvolume and geometry; ionic and gas diffusion, convection; electrolyte concentration(some oxygen will dissolve in water depending on the pressure of the oxygen); externaleffects (temperature, pressure); and finally electrical parameters (current density,overpotential, impedance).

The electrolysis experiment can be performed either by controlling the currentexternally and measuring the resulting changes in the cell potential, or by controllingthe voltage and measuring the resulting current [Ba8001, Ri8701]. Under controlledpotential conditions (potentiostatic) the potential of the electrode with respect to areference electrode can be set at a value that ensures that only one chemical processcan take place. The surface concentration of electroactive species is held constant andthe mass flux (the current density is proportional to it) related to the specific rateconstant of the reaction decreases with time. On the other hand when using constantcurrent (galvanostatic), the rate of the reaction is controlled externally, and thus thegradient of concentration (flux) of reactant at the electrode surface is constant (but thesurface concentration decreases with time). Since the current is constant, the rate ofproduct generation can be calculated, without any specification of the rate constant ofthe reaction. The choice between these methods depends on the application.

The measurements described in this thesis are conducted at constant current. Inthis way, the number of moles of gas formed and thus the gas pressure may beestimated. The current flowing in the cell is stepped from zero to a finite value, and thecell voltage is measured as a function of time. When a current step is applied, threeprocesses take place: the ionic double layer between electrode and electrolyte charges,charge-transfer reactions start, and concentration gradients are set up near theelectrode surfaces. Charging of the double layer is a relatively fast process (about 10µs), and therefore should not interfere with the measurements. When the current pulsehas a long duration [Ba8001, Gi9301], mass transport by convection occurs, and thecompeting reactions give a complex response, which is generally difficult to model.

Reversibility is an important requirement for the actuator, and therefore thechemical reactions must be reversible. The electrolyte has to be compatible withforward and back reactions at both electrodes. We chose only one type of gas

Chapter 5

77

evolution, oxygen, a solution of copper sulphate, and platinum (Pt) and copper (Cu)electrodes. The three states of the actuator are:

A. Pressure building state: during electrolysis, water is oxidised and oxygen gasevolves at the Pt anode:

2H2O(l) → O2(g)+ 4H+(aq) + 4e- (5.3.1)

with equilibrium potential E0O2/H2O = 1.23 V vs. NHE at standard conditions (T = 298

K, p0 = 1 bar, M = 1), while copper is deposited at the Cu cathode:

Cu2+(aq) + 2e- → Cu(s) (5.3.2)

with E0Cu2+/Cu = 0.34 V vs. NHE. The overall cell reaction for O2(g) production is:

2H2O(l) + 2Cu 2+(aq) ⇔ O2(g) + 4H+(aq) + 2Cu(s) (5.3.3)

with E E E Vcell O H O Cu Cu0 0 0

2 22 089= − =+/ /

. . For conditions different from the standard

conditions, the potential can be computed with the Nernst equation (3.2.2):

E ERT

F

C

C pcell cellCu

H O

= + ⋅+

+ ⋅

0 22

4

24

ln( )

( )(5.3.4)

where pO2 [bar] is the numerical value of the oxygen partial pressure at the electrode

surface. The dependence of partial anodic and cathodic currents on potential at bothcopper and platinum electrodes is shown schematically in fig. 5.3.1. The total currentpassing through the circuit is the sum of all partial currents. The evolution of oxygengas is represented by line (d), and the deposition of copper by line (b1).

If the cell voltage has a value higher than that necessary for proton reduction,hydrogen evolution will occur at the copper electrode (curve (a) fig. 5.3.1):

2H+(aq) + 2e → H2(g) (5.3.5)

The evolution of hydrogen has a negative effect on the adhesion of Cu during itsdeposition, and therefore causes a decrease of the lifetime of the electrode. Thisreaction depends primarily on the nature of the metal used as electrode e.g. on copperthe exchange current density for reduction of hydrogen ions is about 800 times lowerthan on platinum [Ri8701].

Figure 5.3.1. Qualitative description of partial anodic and cathodic current, ia and ic,as a function of potential (vs NHE) during electrolysis. At the copper electrode:hydrogen evolution (curve a), copper deposition (curve b1), oxygen reduction (curvec); and at the platinum electrode: oxygen evolution (curve d).


78

The presence of oxygen dissolved in the electrolyte causes another reaction atthe copper electrode; the reduction of oxygen to water (the reverse of reaction 5.3.1)takes place because its equilibrium potential is more positive than that for copperdeposition (line (c) fig. 5.3.1). The solubility of oxygen in water is low and the limitingdiffusion current for oxygen is reached quickly (the reduction of oxygen is diffusionlimited). This unwanted reaction decreases the oxygen gas pressure and therefore theefficiency of the actuator. To reduce this effect, the copper electrode has to beprotected against oxygen, for example, using a semipermeable membrane. Thismembrane should have the following properties: good ionic conductivity, lowelectronic conductivity (to prevent short-circuiting of the electrodes), thermal andchemical stability, sufficient mechanical strength, good adhesion to the electrodesurface, and low gas permeability. A suitable choice might be an ion-exchange polymermembrane called Nafion (see chapter 3). Nafion is permeable to many positive ionsand polar compounds, so that the current loop during electrolysis is closed [Gr8601],but it is impermeable to negative ions and nonpolar compounds such as oxygen gas. Inthe literature an exceptional chemical stability is reported [Gr8601]. It is known[As9301] that the oxygen permeability of a Nafion membrane in the dry state is muchlower than in the wet state, and therefore the gas permeation is closely related to thewater content of the membrane.

During electrolysis gas bubbles are generated by the following steps [Si8301,Wu9301, Vo9601]: initially, electrolytically evolved gas dissolves in the electrolyte.Low solubility in the liquid and small diffusion coefficients of the gas will lead tosupersaturation of the electrolyte in the vicinity of the electrode. At very low currentdensities the dissolved gas will be transported to the bulk by means of diffusion and(natural) convection. As a consequence, a stationary state will be reached: thesupersaturation near the electrode surface will become constant in time. At highercurrent densities the supersaturation at the electrode surface will increase and willexceed a value necessary for the formation of gas bubbles. Finally, the detachment ofgas bubbles occurs when the Archimedes’ force on the bubble overcomes the surfaceforces of adhesion.

The performance of the electrochemical cell might be affected by generated gasbubbles in two ways [Wu9301, Si8301]: i) the resistance of the gas-liquid electrolytemixture increases at higher gas production rates, and ii) the mass transfer to and fromelectrodes of the ions and the gas varies in time.

B. Passive state: Under open circuit conditions, the pressure built up should bemaintained but this "steady-state" can be disturbed by leakage from the system or byparasitic reactions at the copper electrode:

2 Cu + O2 + 4 H+ → 2 Cu2+ + 2 H2O (5.3.6a)

2 Cu + O2 → 2 CuO (5.3.6b)

The combination O2/Cu is thermodynamically unstable (EH2O/O2 is more positive than

ECu/Cu2+) and the metal corrodes.

The corrosion of Cu is shown schematically in fig. 5.3.2. The electrons thatbecome available from oxidation of the metal (curve b2) are used to reduce oxygen atthe Cu (curve p1). At the open-circuit or corrosion potential the partial anodic and

Chapter 5

79

cathodic currents are equal (ia = ic). It is clear that for a higher oxygen concentration insolution or more efficient transport of oxygen to the metal (curve p2) the corrosionpotential is shifted to a more positive value and the rate of dissolution of the metalincreases. The metal dissolution products depend on the electrolyte solution, inparticular on the pH. At low pH Cu2+ ions are formed (reaction 5.3.6a) while at higherpH oxides may be formed (reaction 5.3.6b).

Figure 5.3.2. Schematic current-potential curves for the copper electrode at opencircuit: copper dissolution (curve b2), oxygen reduction (curve c), for two limitingcurrents (p2 > p1 either because of a higher oxygen concentration in solution or amore efficient transport of oxygen to the Cu).

C. Pressure decreasing state: In this case generated oxygen is reduced to water andcopper is oxidised/dissolved as copper ions. This can be achieved by reversing thecurrent or short-circuiting the electrodes. The partial currents of these reactions areshown qualitatively in fig. 5.3.3. The reduction of oxygen by short-circuiting the cell,shown in fig. 5.3.3a occurs at both electrodes, and copper is oxidised at the copperelectrode. Reversing the current direction, oxygen and copper ions are reduced at theplatinum electrode, and solid copper is oxidized to copper ions at the copper electrode,as shown in fig. 5.3.3b.

(a) (b)

Figure 5.3.3. Schematic current-potential curves during pressure reduction state. a)short-circuit conditions: copper oxidation at the Cu electrode (curve b2), with oxygenreduction at both electrodes (curve c); b) reversed polarity conditions: copperoxidation at the Cu electrode (curve b2), with copper ion reduction (curve b1), andoxygen reduction (curve c) at the platinum electrode.5.4 First prototype of an electrochemical actuator


80

The actuator is made by standard micromachining technology. It consists of anarrangement of two electrodes in a sealed cavity filled with electrolyte. The actuator ismade from two processed silicon wafers and consists of two main parts[Ha9501,Ne9601]: the electrochemical cell and the deflecting membrane (fig. 5.4.1). One wafercontains the electrolyte cavity and the membrane. The two electrodes, Pt and Cu, andthe filling holes are located on the second wafer.

The Cu electrode is protected against oxygen with a layer of the perm-selectivemembrane, Nafion® (gel). The Nafion normally used is a foil (solid film), approximately100 µm thick. It is difficult to bond such a layer on a silicon wafer because, when theNafion is wetted in the electrolyte, it swells which results in a poor adhesion to anddetachment from the silicon substrate. In our case a Nafion film could be produced byspin-coating, dip-coating or plasma polymerisation. Plasma polymerisation [Ya9401]was not investigated because at MESA no expertise was available in this field. Testswhere done to spin-coat a solution of 5%wt Nafion, commercially available from theAldrich Chemical Company, on different substrates (Si, SiO2, Si3N4, Al) and resultswere obtained with Si3N4 (Al as sacrificial layers) and a curing temperature of 300°C[Ij9201]. It is possible to pattern Nafion with an Al mask and etch the Nafion with anoxygen plasma. The Nafion ‘structures’ with dimensions of 0.5 mm survive organicsolvents, ultrasonic cleaning, strong acids such as H3PO4 (Al etch) and 1% NaOHsolution [Ij9201]. The use of Nafion as stand alone membrane requires three wafers:one for the anode, one for the cathode and one for

(a)

(b)

Figure 5.4.1. (a) Cross-section, and (b) top view of the first design of themicromachined electrochemical actuator; the area of the platinum electrode is 1.5mm2, that of the copper electrode is about 0.8 mm2, and the volume of the cell cavityis around 1.3 mm3.the Nafion membrane to be bonded between the other two. Instead, we used anotherdesign with only one chamber with two electrodes, in which the Cu electrode is

Chapter 5

81

constructed as a perforated standing table in an attempt to solve the adhesion problem.The Nafion membrane is made by applying the Nafion solution to the Cu grid-table.

The most important process steps are discussed in the following paragraph.Besides, the complete technological process is described in detail in Appendix 5A.

5.4.1 Clean room processing and technology

a. Electrochemical cellOn the front side of one wafer, the two electrodes (Pt and Cu) and two access

holes for the filling of the cavity are required. On the backside of this wafer electricalcontacts for the electrodes are deposited, as well as wiring to seal the filling holes bymeans of electroplating. First, the backside is processed in order to preventcontamination of the electrodes in the following process steps.

A layer of 1 µm low-stress silicon rich nitride (Si3N4) is deposited by lowpressure chemical vapour deposition (LPCVD) on the silicon wafer to act as electricalinsulator between the electrodes and the silicon substrate. The silicon nitride from thebackside is patterned by reactive ion etching (RIE) in a CHF3/O2 25:5 sccm plasma,chamber pressure 10 mTorr, power 75 W, support temperature 25°C, etch rate ofabout 80 nm/min. The silicon is then etched in 25% KOH solution, 70°C at about 1µm/min up to the nitride on the front side, see fig. 5.4.2, step 1. In this way, the accessholes are made and also the openings for the electrical contacts to the electrodes on thefront side. The electrical contacts through wafer and the wiring are produced bysputtering copper through a shadow mask on the wafer’s backside, with a 100 nm Cradhesion layer, fig. 5.4.2, step 2. The shadow mask is made as follows: positivephotoresist is spin-coated on the front side of a wafer and patterned. Then the silicon isetched by RIE to the desired thickness, about 30 µm.

Figure 5.4.2. The important process steps to make the wafer containing the electrodeson the front side of the wafer, the electrical contacts on the backside, and the fillingholes for the electrolyte.The silicon nitride on the backside is patterned by RIE to allow Si to be etched in KOHsolution up to the silicon nitride on the front side. As a final step, the protection layer

1.

2.

3.

4.

5.


82

on the backside is etched away by RIE to create the openings in the shadow mask.Deposition through a shadow mask with high precision into deep KOH etched groovesof has been developed in our group by G.J. Burger [Bu9601].

In the next step, the electrodes on the front side of the wafer are processed. Afterpatterning of the silicon nitride by RIE, the planar electrodes are patterned by lift-off; 1µm of Cu (step 3) and 0.5 µm of Pt (step 4), are deposited by dc magnetronsputtering, with 100 nm Ti as adhesion layer. For the magnetron sputtering, a Balzerscryo pump was used with a base pressure when starting the process of ca. 2·10-7 mbar,Ar flow during deposition creating a pressure of 8·10-4 mbar, 200W dc power; thedeposition rate for Cr was ca. 2.2 Å/s, for Cu, 5.7 Å/s, for Ti, 1.3 Å/s, and for Pt, 4.1Å/s. For the first actuators, Cr instead of Ti was used as adhesion layer. Unfortunately,during electrolysis, Cr under the Pt electrode dissolved electrochemically and the Ptelectrode was lifted from the substrate.

The final step is the etching in KOH solution of the silicon under the Cu to formthe copper table-electrode, fig. 5.4.2, step 5. The Nafion gel-membrane is made byapplying the Nafion solution (5% wt. Nafion® solution) to the Cu grid-table, andslowly evaporating the alcohol solvent. During the drying process the Nafion filmshrinks forcing the Cu grid to bend, which sometimes results in failure. In fig. 5.4.3 aphoto is shown of the Cu grid-table electrode, U-shape Pt electrode (to obtain anoptimum distribution of the electrical field between Cu and Pt), the electrical contactson the electrode, and the filling holes. Pt has a width of 320 µm, a total length of 4.8mm, so an area of 1.54 mm2; Cu has a side width of 1.1 mm and an effective area, dueto the grid form, of about 0.8 mm2; the area of the electrical contact on the Ptelectrode is 100 x 100 µm2, on the Cu 45 x 45 µm2, and the size of the opening for theaccess holes is about 170 x 170 µm2.

Figure 5.4.3. Optical photo showing the U-shape of the Pt electrode, Cu grid-tableelectrode, the electrical contacts on Pt and Cu electrodes, and the filling holes (thetwo squares on Si). Area of Pt is 1.54 mm2, the effective area of Cu is about 0.8 mm2;the area of the electrical contact on the Pt electrode is 100x100 µm2, on the Cu is45x45 µm2, and the size of the opening for the access holes is about 170x170 µm2.A SEM photo of the same Cu grid-table after the KOH etch of the Si under the table,is shown in fig. 5.4.4. Some stress, caused by the multilayer structure Si3N4/Cr/Cu inthe grid can be seen.

Chapter 5

83

Figure 5.4.4. SEM photo of the Cu grid-table electrode released after the KOH etchof the Si under the electrode. Stress in the table leads to bending.

b. Deflecting membraneThe second wafer contains the deflecting membranes. The fabrication and

characterisation of the membranes are described in detail in chapter 4. Briefly, thecorrugated membranes are made of 1 µm thick low-stress LPCVD silicon nitride. Thefabrication process of corrugated membranes consists basically of three steps: frontside etching of corrugations in the silicon wafer by RIE in an SF6/O2/CHF3 plasma;deposition of 1 µm low-stress LPCVD silicon nitride on both sides of the wafer, andbackside etching of the silicon wafer in KOH solution to form the membrane.

c. BondingIn principle, the two processed wafers can be bonded above 300 °C to form a

sealed cavity. However, at these temperatures the properties of Nafion® deteriorate[DuPont]. Tests were done in order to lower the bonding temperature [Ko9501] andsome results were obtained at about 175°C but these were not easily reproducible.Different commercial glues were tested for gas leakage in the following way. A smallmetal cylinder filled with helium and sealed with the glue to be tested was placed in avacuum chamber and the helium leakage rate was measured. The background leakagerate of helium measured in the vacuum chamber without the cylinder was 1.8·10-9

mbar·l/s that is 4.5·10-12 mol/min. The experimental values obtained for five types ofglue are given in table 5.4-1. For the work presented here, the bonding was done atroom temperature with “High Super 5”, a Japanese two-component epoxy, which wasfound to be the most suitable for this purpose.

Glue leakage rate [mbar·l/s]Adolf Würth GmbH & Co. KG 2.2 x10-8

Araldit Ciba-Geigy B.V. 4.7 x10-9

High Super 5 Semedain Co.Ltd. 4 x10-9

Silastic Dow Corning GmbH 3.1 x10-8

Hysol Dexter Electronic 5 x10-9

Table 5.4-1. Experimental data for the gas leakage rate of some types of glue usingthe helium leakage test.

The cavity is filled by immersing the wafer package in the electrolyte solution ina beaker, and placing in a vacuum jar. In this way the cavity is completely filledwithout any gas bubbles.


84

The filling holes were sealed by glue. Attempts to seal them by Cu electroplatingwere promising but as yet unsatisfactory: they did not withstand pressure differencebigger than 1 bar.

Since a 1 µm silicon nitride membrane is transparent, an optical microscope maybe used to check how well the cavity is filled. The distance between the Cu table-electrode and the silicon nitride membrane is measured before and after inserting theelectrolyte. Since the refractive index of the electrolyte and Nafion are known, anestimate of the cavity filling can be done.

The two processed wafers bonded together form a cell with a volume of about1.3 mm3.

For the measurements, the microactuator was connected to a printed circuitboard (PCB) by wire bonding (Ag wires of 20 µm) between the electrical contact padsof the two electrodes and the copper pads of the PCB.

5.4.2 Experimental

The device was characterised in two ways:i) The mechanical properties of different types of membrane were tested. As wasshown in chapter 4, flat and corrugated silicon nitride (radial corrugations) membraneswere made with 1 mm side length and 1 µm thickness. The deflection of all membraneswas measured with the use of air pressure. The sample was mounted on a holder with apressure inlet. The pressure was measured by a regulator and was increased graduallyuntil fracture occurred. Membrane deflection was recorded by a Sloan DEKTAK 3030profiler, which is a mechanical stylus profilometer.

For the actuators described here, a flat silicon nitride membrane was used. Itsdeflection as a function of the air pressure is shown in fig. 5.4.5. The analyticalsimulation equation (4.2.1):

Figure 5.4.5. Comparison of simulated and measured deflection of a 1 µm thick,square (1.2 mm side), flat silicon nitride membrane as a function of air pressuredifference.

Ph

ay

Eh

ay= ⋅ + ⋅

⋅ −−

⋅341 198 1 0 295

12 43. . ( . )σ ν

ν (4.2.1)

Chapter 5

85

with E = 3 GPa, ν = 0.3, and σ =100 MPa is also plotted in fig. 5.4.5, as a solid line.The experimental data are in good agreement with theory. From this "calibrationcurve", the volume of the cavity and measurements of the deflection during oxygenpressure build-up, the amount of gas produced could be calculated.

ii) The performance of the electrochemical cell was characterised by measuring themembrane deflection due to the oxygen gas production for the three states A, B, Cdescribed in 5.3. Electrochemical measurements were conducted at constant current.

All deflections measured are in the linear range of the pressure-deflection curve,so the oxygen pressure is proportional to the deflection. However, the DEKTAKprofilometer used for the experiments of fig. 5.4.5 cannot be used here because it doesnot allow measurement of the deflection at one particular point on the membrane as afunction of time; it only scans the whole surface, and the scanning time is shorter thanthe measurement time. The deflection of the membrane was sensed by an atomic forcemicroscope set-up in order to investigate dynamic behaviour. The deflection of themembrane causes displacement of the force sensor, which is detected with an opticalbeam deflection technique, schematically shown in fig. 5.4.6. The atomic forcemicroscope is described in more detail elsewhere [We9301]. The maximum deflectionthat could be measured with the atomic force microscope is only a few micrometers.For this reason we did not measure at pressures higher than 30 mbar. For largerdeflections of the membrane other techniques have to be used, e.g. opticalautofocusing, position-sensing photodetectors or phase stepping interferometer.

The variation of the voltage across the electrochemical cell (Ecell) and the voltagefrom the AFM for the deflection of the membrane were measured simultaneously witha 1 MΩ input impedance recorder.

Figure 5.4.6. Principle of the atomic force microscopy

5.4.3 Results and discussion

As was shown in section 5.3, three states of the actuator can be distinguished.During pressure build-up, the Pt electrode acts as an anode at which oxygen gas


86

evolves; Cu is deposited at the copper cathode. During the pressure reduction stage, inwhich the deflection is decreased, the electrodes can be short-circuited so that O2 isreduced at the Pt electrode and the Cu electrode dissolves. Ideally under open circuitconditions the oxygen pressure and thus the deflection of the membrane should bemaintained. In fact, it is the parasitic reaction at the copper electrode which mainlylimits the performance of the electrochemical cell until now.

It was observed that for a cell voltage ≤ 2V and a current of about 1 mA, a hugeproduction of oxygen gas (bubbles) occurred, so that flat membranes were blown upwithin a few seconds. Therefore, currents below 50µA were used to ensure slower andmore controlled gas production; in fig. 5.4.7 the oxygen gas bubble generation at theplatinum electrode is shown. The photo was made with an optical microscope throughthe transparent silicon nitride membrane. The dynamic behaviour of the membranedeflection is shown in fig. 5.4.8: the current generates an oxygen gas pressure, whichcauses membrane deflection (curves AB, CD); on short-circuiting the electrodes thepressure decreases (curves BC, DE), and by reversing the current (-50µA) the pressuredrops faster (curve EF).

Figure 5.4.7. Image of the actuator seen through the silicon nitride membrane of 1µmthickness. The arrow shows the oxygen gas bubbles which evolve at the Pt electrode;bar is 0.1 mm.

Figure 5.4.8. Time dependence of the deflection of the membrane: the curves AB, CD:the gas pressure builds up when a current of 50 µA is applied; the curves BC, DE: theelectrodes are short-circuited and the pressure decreases; curve EF: the pressuredrops faster when a negative current (-50 µA) is applied.

For a constant DC current of 2 µA, the voltage across the actuator cell increasedasymptotically to a value of 1.1 V within a few minutes. However, after 15 minutes, nooxygen pressure could be detected, although the voltage was larger than the value of

Chapter 5

87

∆E0 (0.89 V) (for reaction 5.3.3). The oxygen production rate can be calculated fromthe applied constant current. The number of moles N formed is given by:

NQ

nF

i t

nF= =

⋅(5.4.1)

where n = 4, is the number of electrons required to obtain 1 molecule of O2, and F =96485 C/mol, the Faraday constant. Thus, the theoretical oxygen production rate is:

dN

dt

i

Fth

=

4(5.4.2)

For a current of 2 µA, an O2 production rate of 5.2·10-12 mol/s was expected.Assuming that the cell is absolutely gas-tight, we may explain the absence of a pressureincrease by the diffusion of O2 through the Nafion, and its subsequent reduction at theCu electrode. Under steady-state diffusion conditions, a linear concentration gradient isestablished and the oxygen diffusion flux through the Nafion membrane to the copperelectrode is given by Fick’s law:

J DdC

dx= − (5.4.3)

The total oxygen diffusion flow rate is I = A·J, this represents the loss of oxygen due tothe diffusion through the Nafion membrane, and is given by:

dN

dtI D

A

dC

lossON

ON N

NO

≡ =

2 2 2(5.4.4)

where DON2

is the oxygen diffusion coefficient in Nafion, AN and dN are the area and

thickness of Nafion membrane, C is the concentration of oxygen at the outer edge ofthe Nafion membrane.

Values reported for the diffusion coefficient of O2 through Nafion® films rangefrom 8·10-12 to 2·10-10 m2/s [La8801, Og8501]. If the Nafion film thickness is at least 1µm, the Nafion area 1 mm2, the diffusion coefficient of oxygen is taken to be 1·10-11

m2/s, and the maximum O2 concentration difference between the two sides of theNafion membrane is considered to be the solubility of oxygen in water, CO2

≈ S = 1.26

mol/m3 (in equilibrium with air at 760 mTorr and t = 25°C,), we estimate an oxygenloss rate of about 12.6·10-12 mol/s. Since this is higher than the expected productionrate, 5.2·10-12 mol/s, it is not surprising that we did not observe pressure build-up. Ithas to be mentioned that these are just rough estimates, since we do not know exactlythe area or the thickness of the Nafion film.


88

For a current of 5µA, the deflection of the membrane due to the generation ofoxygen (curve AB) and its reduction by short circuiting the cell (point B) are shown asa dashed line in figure 5.4.9a. The variation of Ecell is also shown by a continuous line.Figure 5.4.9b shows the open circuit mode under the same conditions. Open circuitwas achieved by switching off the constant current supply. Curve AC gives the buildup of pressure, and at point C the circuit is opened. For a deflection of 1.25 µm, thepressure obtained is 20 mbar, and the build up time is 100 s at 5 µA. At points B andC, the pressure is 27 mbar for 1.71 µm, and the build up time is 180 s.

The theoretical (expected) oxygen production rate for a current of 5µA,assuming no losses, is 13·10-12 mol/s (eq. 5.4.2). The experimental (measured) valuemay be estimated from the law of ideal gases:

P·V = N·R·T (5.4.5)is given by:

dN

dt

dP

dt

V

RTcell

=

exp exp

(5.4.6)

Figure 5.4.9. The voltage across the electrochemical cell (Ecell) and the deflection ofthe membrane vs. time at a current of 5µA; (a) in point B the cell is short-circuited,and (b) in point C the circuit is opened.

(a)

(b)

Chapter 5

89

From the measured deflection of the membrane and the deflection vs. air pressure‘calibration curve’ (fig. 5.4.5) we are able to estimate the pressure in the cell; further,knowing the time to obtain the deflection, the rate of pressure increase can becalculated. From the experimental data, we derive a pressure increase rate of 15 Pa/s(for 1.71 µm corresponds to a pressure of 27 mbar, obtained in 3 min). As the volumeof the cell is about 1.2 mm3, the experimentally observed oxygen production rate isabout 7.2·10-12 mol/s.

However, in eq. 5.4.6 the assumption was made that the extra volume obtaineddue to the membrane deflection is very small compared to the volume of the cell. Forthe modelling and the calculation of the extra volume due to gas pressure, sphericalfunctions were used to approximate the shape-functions of the square membranes:

( ) ( )y y a b y xa

yb0 0

2 2 2 2

1 1, , = −

⋅ −

(5.4.7)

with centre deflection, y0, and membrane side length, 2a = 2b. The extra volume can beestimated by:

∆V y y a b a b ya

a

b

b

= = ⋅ ⋅−−∫∫ ( , , )0 0

256225

(5.4.8)

Clearly, the relationship between pressure and volume change is not simple since thedeflection is in the non-linear range, P ≈ (y0)

3, and the amount of gas present in thecavity before electrolysis is not known precisely. This deflection depends on thepressure so the volume is a function of the pressure; it is not simply the cell volume aswas assumed.

Since the difference between this and the theoretical one can be considered to bethe oxygen removal rate by diffusion through the Nafion and reaction at the copperelectrode (see above):

dN

dt

dN

dt

dN

dt

i

F

dP

dt

V

RTloss th

cell

=

−

= −

exp exp4(5.4.9)

we now have a possibility to estimate the diffusion coefficient of O2 through the Nafionfilm, eq. 5.4.4:

Dd

A C

dN

dtON N

N O loss2

2

=⋅

⋅

(5.4.10)

With the assumptions mentioned before, we arrive at a diffusion coefficient of about4.6 10-12 m2/s, which is within the range of the reported values.

When the current is increased to 10 µA, the reactions show the same generaltrends but are faster. This can be seen in fig. 5.4.10. As in fig. 5.4.9, curve AB is forthe build-up of pressure, in point B the pressure is reduced by short-circuiting theelectrodes, and in point C the circuit is opened. For the same deflection of 1.25 µm,the pressure obtained is 20 mbar, but the build up time is 42 s (as compared to 100 sfor 5 µA). At points B and C, the pressure is 33.4 mbar for a deflection of 2 µm, and


90

the time is 60 s. Similar calculations give a theoretical oxygen production rate of 26 10-

12 mol/s, an experimental value of 12.3 10-12 mol/s; the loss rate is therefore 13.710-12 mol/s.

Figure 5.4.10. The time dependence of the voltage across the electrochemical cell(Ecell) and the deflection of the membrane for a current of 10µA; (a) in point B thecell is short-circuited, and (b) in point C the circuit is opened.

The theoretical and experimental values of the production and loss rates foroxygen are summarised in table 5.4-2: the loss rate is calculated from the differencebetween theoretical and experimental data and represents gas lost during pressurebuild-up.

I [µA] O2 production rate [mol/s]theoretical experimental

O2 loss rate[mol/s]

2 5.2 x10-12 ---- > 5 x10-12

5 13 x10-12 7.8 x10-12 5.2 x10-12

10 26 x10-12 12.3 x10-12 13.7 x10-12

Table 5.4-2. Theoretical and experimental data of oxygen production and loss rate.

(a)

(b)

Chapter 5

91

Table 5.4-3 shows the time constants τ for the decrease of deflection d for thetwo values of current, where time dependence is assumed to have the following form:

( )d d t= −0 exp τ (5.4.11)

At open circuit, oxygen reduction at the copper electrode is diffusion limited, and thusthe pressure decreases slowly (slow decay fig. 5.4.9b, 5.4.10b). At short circuit, thereduction of oxygen takes place at both electrodes, copper and platinum, andconsequently the pressure drops faster than at open circuit (fast decay fig. 5.4.9a,5.4.10a). The difference between the decay constants for fig. 5.4.9a and fig. 5.4.10a isprobably due to inaccuracy in the measurements. During the build up of pressure,oxygen may also be reduced at the copper electrode, which might explain the lowpressures obtained experimentally.

The decay time, as has been mentioned above, for the open circuit state, isdetermined by the oxygen permeability of the Nafion membrane and leakage from thecell. For the pressure reduction state, the decay time is mainly determined by the rateof the electrochemical reaction. The decay rate at open circuit is ca. 0.1 µm/min, forshort circuit 0.3 µm/min, and for current reversal (-5 µA) about 0.46 µm/min.

Oxygen created at Pt electrode diffuses through the electrolyte and Nafion to getto the Cu electrode. The diffusion of oxygen through the Nafion membrane can beestimated by:

I DC

dAO O

N O

NN2 2

2=∆

(5.4.12)

As an example with DON

2 ≈ 1⋅10-11 m2/s, ∆CO2 ≈ 1.26 mol/m3, area and thickness of

Nafion AN ≈ 1 mm2, dN ~ 2 µm, respectively, the flow rate is I ON

2 = 6.3⋅10-12 mol/s. For

an amount of oxygen created by electrolysis N = 2.33⋅10-9 mol (eq. 5.4.1, i = 5 µA, t =180 s, or i = 10 µA, t = 90 s, n = 4), the time of oxygen diffusion through Nafion is ca.370 s:

t = N/IO2 (5.4.13)

The electrochemical experiments (RDE) indicate that the reaction of O2 with Cu issignificantly faster than O2 diffusion, thus the diffusion of oxygen through Nafion is thedetermining step for the oxygen pressure decay.

One way to increase the decay time would be to increase the thickness of theNafion membrane:

τ ~ dNafion (5.4.14)

This idea will be tried in the next prototype by modifying the design for the copperelectrode.

Figure i [µA] τ [sec]

5.4.9a (short-circuit) 5 965.4.9b (open circuit) 5 3485.4.10a (short-circuit) 10 725.4.10b (open circuit) 10 348

Table 5.4-3. Time constants τ for the oxygen gas pressure decay.


92

5.4.4 Conclusion for the first design

As can be seen from the results, the characterisation of the microactuator hasrevealed some problems. The most serious drawback of the first design is the relativelyfast decrease of the pressure during the open circuit state. We attribute this to the poorprotection of the copper electrode: the Nafion membrane is thin so oxygen diffusesrapidly through it. However, a very thick Nafion will have a higher electrical resistanceincreasing the total cell resistance; a higher voltage will be necessary to run theactuator. Thus a compromise has to be made. Another problem is the fact that the Cugrid-electrode was not stiff enough and bends when Nafion shrinks during the dryingprocess, leading to cracks in the Nafion film. To ameliorate these problems a re-designof the actuator might be required or the choice of a different material for the perm-selective membrane.

By an optimisation of the protection layer, the performance can be improved inorder to achieve the desired characteristics. A second design of the microactuator willprobably show more clearly the advantages and the limits of this sensor technology.

5.5 Second design of the electrochemical microactuator

Based on the experience with the first prototype, a new electrochemicalmicroactuator was developed, fabricated and characterised. The operating principle ofthe microactuator remains the same. The aim is to optimise the performance of thedevice, in particular the Nafion protection of the copper electrode. One way to achievethis is to make a flat copper electrode instead of a grid-table and to use a mesh on topof the copper. This mesh will improve the adhesion of the Nafion and will allow asmaller area and a thicker membrane to be used. Furthermore, the second prototypefocuses on a simple integration of the microactuator into a flow channel, for example,as shown in chapter 2.10, fig. 2.10.1, to measure the pressure drop across thedeflecting membrane (valve) due to the oxygen gas pressure.

5.5.1 Clean room processing and technology

In order to improve the adhesion of Nafion to the Cu electrode and to allow athicker Nafion layer with a smaller area, a polyimide mesh was used. Tests have shownthat the mechanical stability of this polymer is sufficient for our application.

Two types of deflecting membrane were used. For the dynamic experiments,optical measurements were performed. A flat silicon nitride membrane with a metalcross on top was used for a better positioning and also to be able to see what washappening in the cell. For the flow measurements, where the pressure drop across thevalve (membrane) due to the deflection of the membrane was determined, a polyimidemembrane was used because larger deflections could be obtained for the samepressures.

a. Deflecting membraneOne wafer contains the cavity and the deflecting membrane, which is made either

of low-stress LPCVD silicon nitride or polyimide. The process of polyimidedeposition, as described in chapter 4.2.2, is briefly: a silicon wafer with a layer of 1 µmLPCVD silicon nitride is etched from the backside in KOH solution; on the front side a20 nm Cr film is sputtered as adhesion layer for the polyimide layer, that is spun and

Chapter 5

93

cured. Finally, the silicon nitride layer is removed by RIE in an O2/CHF3 plasma, whilethe Cr layer is removed by wet etching.

For the flat silicon nitride membranes with a metal cross only one extra step isadded to the process described in chapter 4.2.2: after RIE patterning of the siliconnitride on one side of the wafer, 20 nm of Cr is deposited on the other side by dcsputtering and patterned by lift-off; the wafer is then etched in KOH solution to obtainthe membranes having the Cr cross on top. The crossing point has an area of ca. 20 x20 µm2 (larger than that of the laser beam).

b. Electrochemical cellThe second wafer, which contains the two electrodes, Pt and Cu, is processed as

follows (all steps are described in detail in Appendix 5B): an insulating layer of 500 nmsilicon oxide is grown by wet oxidation at 950°C for ca. 4 hours, see fig. 5.5.1 step 1.Then 200 nm Cu is deposited by magnetron sputtering on a 20 nm Ti adhesion layer.Positive photoresist (S1813) is spun on the copper and patterned to form the mask.The Cu electrode is obtained by wet etching in a solution of 1HNO3, 4H3PO4,4CH3COOH, 1H2O (also etches Al); the etch rate is 20 nm/s (step 2). The Ti adhesionlayer is etched by dipping in 2% HF. The Pt electrode is then deposited by dcmagnetron sputtering and patterned by lift-off (step 3). To make the polyimide mesh, a1.5 µm thick aluminium sacrificial layer is evaporated and patterned by lift-off (step 4).The polyimide used is a photosensitive Probimide HTR3-200 (OCG MicroelectronicMaterials NV) which can easily be patterned by lithography.

Figure 5.5.1. The process steps for the second design of the electrochemical cell; thepolyimide mesh provides a better adhesion and thicker Nafion membrane.

1.

2.

3.

4.

5.

6.


94

The polyimide is spun, patterned and thermally cured, forming a 10 µm thick mesh(step 5). The following step is the selective etching of the Al sacrificial layer in asolution of Na2CO3 (37.5g) + Na3PO4⋅12H2O (17.5g) + K3Fe(CN)6 (8g) in 250 mlwater, at 25°C for ca. 6 hours (~ 130 nm/min), and then the Nafion layer is formed bydepositing a 5%wt Nafion® solution on the mesh and letting it dry (step 6).

Figure 5.5.2a is a SEM photo showing the U-shaped Pt electrode, the polyimidemesh and the underlying Cu electrode. As can be observed from the photo, thepolyimide layer (dark colour) in fact covers the whole surface and only openings aremade in this layer for the Pt electrode, for the holes of the mesh, and for the electricalpads of the electrodes. There is no danger of short-circuiting the electrodes since thepolyimide layer has very good insulating properties. In fig. 5.5.2b a detail of the meshis shown, where the Cu electrode under the mesh can be seen. There are two differentsizes of approximately square holes in the mesh; one with 75 µm side and a distancebetween holes of 55 µm, and a less dense mesh with holes of 50 µm side and distanceof 25 µm. The relevant area of the Cu electrode is 0.18 mm2 and 0.225 mm2,respectively; the area of the Pt electrode is about 1.1 mm2.

(a)

(b)

Figure 5.5.2. SEM photo of (a) the U-shape Pt electrode, the polyimide mesh on topof the Cu electrode; a polyimide layer covers the electrical contacts of the electrodeson the bonding region; (b) a detail of the mesh.

ElectricalPads

Pt

Cu

Polyimide

Chapter 5

95

c. BondingAs in the first design, glue is used to assemble the two wafers to form the

electrochemical microactuator, fig. 5.5.3. In order to reduce the processing time andallow a fast control of the new design’s feasibility, filling holes where not speciallydesigned. These can be made by not glueing a small area between the two wafers, thatcan be sealed after the cavity is filled with electrolyte.

Figure 5.5.3. Cross section of the electrochemical microactuator. The area of theplatinum electrode is 1.5 mm2, that of the copper electrode is about 0.8 mm2, and thevolume of the cell cavity is around 12 mm3.

5.5.2 Experimental

(i) The mechanical properties of the membranes were tested in the same way as for thefirst prototype; the membrane deflection due to air pressure was recorded by a SloanDEKTAK 3030 profiler.

ii) The performance of the microactuator was characterised by cyclic voltammetry toevaluate the electrochemical operation of the cell, and by measuring the membranedeflection due to the oxygen gas production for the dynamic behaviour, i.e. the threestates of the actuator.

Cyclic voltammetric measurements were performed with a three-electrodeconfiguration in the same electrolyte solution, 1M CuSO4⋅5H2O. The potential wasmeasured between the Pt electrode and an auxiliary Pt wire, which served a ‘pseudo’reference electrode; the Cu electrode was used as counter electrode. The experimentswere done at a scan rate of 25 mV/s, using an EG&G 366A potentiostat.

Dynamic measurements were conducted at constant current to be able toestimate the oxygen gas pressure from the number of moles of gas formed. The currentflowing in the cell is stepped from zero to a finite value, and the cell voltage ismeasured as a function of time. As constant current source a Keithley 237 high voltagesource was used. The DEKTAK profilometer used in the experiments of fig. 5.4.8could not be used here because it does not allow measurement of the deflection at oneparticular point on the membrane as a function of time. Therefore, to investigate thedynamic behaviour, the deflection of the membrane was measured optically with aLaser Nano Sensor LNS (Sensors 95 B.V.) based on an autofocusing principle, shownschematically in fig. 5.5.4.


96


A cyclic voltammogram measured in an open microactuator (without a siliconnitride membrane) is shown in fig. 5.5.5, for the potential range of -0.05V to +1.4V.Both reactions 5.3.1 and 5.3.2 are clearly observed in fig. 5.5.5; the cathodic andanodic peaks at ca. -0.05 V are related to copper deposition and dissolution at the Ptsurface. Oxygen gas is formed at potentials more positive than 1.2 V. For a closedmicroactuator with the silicon nitride membrane, the cyclic voltammogram has thesame form as is shown in fig. 5.5.6. The sweep range is between -0.1V and 2V, and theexperimental conditions were the same as for the open cell. In fig. 5.5.6, it can beobserved that for potential values higher than 1.6V, the anodic current becomes moreirregular. This effect is due to a higher rate of oxygen gas evolution and to huge gasbubble formation. The gas bubbles reduce the effective area of the electrode surfaceeither directly by masking or indirectly by causing non-uniformity in the currentdistribution [Si8301, Vo9601].

Figure 5.5.5. Cyclic voltammogram for the open electrochemical cell (without asilicon nitride membrane), with the potential range -0.05V to +1.4V; Pt is the workingelectrode, Cu the counter electrode, and a Pt wire as ‘pseudo’ reference electrode, in1M CuSO4 solution, scan rate 25 mV/s.

Figure 5.5.4. The optical set-up for measurement ofthe membrane deflection, based on the autofocusingprinciple

Chapter 5

97

Figure 5.5.6. Cyclic voltammogram for the closed electrochemical cell (with siliconnitride membrane), between -0.1 and +2V; the same experimental conditions as in fig.5.5.5.

When a constant current was passed through the electrochemical microactuator,the voltage across the actuator cell increased asymptotically to a value of 1.6 - 1.7 V.This value is slightly higher than that observed for the first prototype (1.5V).Twoeffects may be responsible: the decrease of the effective (working) area of the Cuelectrode caused by the electrical shielding of the polyimide mesh (a higher potential isrequired for the same current) and a thicker Nafion membrane than in the firstprototype giving a higher cell resistance. A SEM photo showing the shielding effectintroduced by the polyimide mesh is presented in fig. 5.5.7; deposition of the copperthrough the Nafion membrane on the copper electrode occurs only through theopenings in the mesh during electrolysis.

For a constant current of 10 µA, the deflection of the silicon nitride membranedue to generation of oxygen is 13 µm in 9 min. Subsequently, the electrodes weredisconnected. The decrease of the deflection at open circuit is presented in fig. 5.5.8.

Figure 5.5.7. SEM photo of copper electrodeposited on the Cu electrode. Depositionoccurs through the Nafion membrane only via the openings in the polyimide mesh dueto the shielding effect.


98

I = 10 µA, open circuit

0

1

2

3

4

5

6

7

8

9

10

0 20 40 60 80 100time [min]

defle

ctio

n [ µ

m]

Figure 5.5.8. The time dependence of the membrane deflection due to oxygen gasgeneration, for a current of 10 µA and subsequent decrease under open-circuitconditions.

The deflection-time relationship exhibits the same general characteristics as for the firstprototype, fig. 5.4.10. The decay time is now markedly longer, ca. 90 min.

When the current is increased to 50 µA, the reactions show the same generalfeatures but are faster. A deflection of 11.5 µm is obtained in 0.9 min, and thecorresponding pressure for this deflection is about 300 mbar. The circuit issubsequently opened, fig. 5.5.9a. The decay time is τ = 93 min. Pressure reduction byshort circuiting the cell (fig. 5.5.9b) is much faster, about 15 min.

The build-up time is mainly determined by the current. For the pressure reductionstate, the decrease of pressure depends on the rate of the electrochemical reactions butalso on the diffusion rate of oxygen through the Nafion layer. Therefore, the pressuredecrease for short circuit conditions is much faster than for the open circuit case, ascan be seen in fig. 5.5.9, where it depends only on diffusion of oxygen through theNafion. The long decay time, for the open circuit state, is determined mainly by thegood adhesion, the area, the thickness and the oxygen permeability of the layer. Withthe use of the polyimide mesh, the area of Nafion is decreased and the thickness isincreased, which leads to an increase of the decay time in open circuit state from 6 minto 93 min.

I = 50 µA, open circuit

0

2

4

6

8

10

12

0 20 40 60 80 100 120 140time [min]

defle

ctio

n [ µ

m]

Figure 5.5.9. The time dependence of the membrane deflection due to oxygen gasgeneration for a current of 50 µA, and subsequent decrease after switching off thecurrent for (a) open circuit, and (b) short circuit case.

τ ~ 93 minτ ~ 15 min

τ ~ 90 min

(a) (b)

Chapter 5

99

The diffusion of oxygen through the Nafion membrane can be estimated in amanner similar to that for the first prototype:

I DC

dAO O

N O

NN2 2

2=∆

(5.4.12)

and the time of oxygen diffusion through Nafion is:

t = N/IO2 (5.4.13)

In the case of the second design, for a thicker Nafion membrane, at least 5 µm, and asmaller area, ca. 0.2 mm2, the diffusion time of diffusion through Nafion is longer than77 min, which is in good agreement with the measurements.

5.5.4 Flow measurements

As described in chapters 2 and 4, the aim is to use the electrochemicalmicroactuator as an active microvalve to adjust a pressure difference within a range of1000 to 4000 Pa over atmospheric pressure for low liquid flow rates of 1 to 3 µl/min.The microvalve may, for example, be designed in such a way that the membrane closesan opening in a plate above the membrane, or the membrane is situated inside thechannel (fig. 4.3.1, ch.4).

In order to get an impression of the behaviour of such a valve, we estimate thepressure drop across the membrane with the following assumptions: laminar flow and afully developed velocity profile; losses in fluid pressure due to the bending andchanging in flow channel size are not considered. The pressure difference across thevalve that has to be adjusted is very small and the real situation in the valve will bemuch more complex. This situation can only be modelled accurately with numericalmethods. Coupled fluid and mechanical finite element calculations, which weredeveloped in our group [Ku9601], are used to estimate the change in flow pressuredue to the membrane deflection (see chapter 4.3.2).

The first pressure-flow rate measurements were performed using air. The flowchannel used is shown in fig. 5.5.10. The minimum diameter of the outlet is 50 µm; theheight of the channel is 50 µm but the presence of glue increases the height to circa 75µm. A 7 µm polyimide membrane of 0.5 mm radius was used to obtain deflections inthe 50 µm range. A constant volume flow is generated by the syringe,

Figure 5.5.10. Essential parts for the flow measurements: electrochemical actuatorand the flow channel.


100

that is pushed by a step motor. The fluid flow is directed through the valve and theresulting pressure across the valve system is measured by a differential pressure sensor(26 PCA Honeywell) with respect to the atmospheric pressure, with maximum value 1bar (100 mV). The pressure sensor is placed just before the inlet in the flow channel.The tubing was kept as short as possible to ensure a very small pressure drop over thetubes compared to the pressure drop over the valve. The flow range that can begenerated depends on the area of the cross section of the syringe and the speed atwhich the syringe is pushed by the step motor. We use a 100 µl syringe with a crosssection area of 1.7⋅10-6 m2. The step motor can generate 1 - 500 steps/s. As each stepis 1 µm, a volume flow rate of 1.7⋅10-12 m3/s to 8.5⋅10-10 m3/s can be obtained.

At a constant volume flow rate the pressure across the valve system wasmeasured for a few membrane deflections caused by the generated gas pressures of theelectrochemical cell. The pressure drop over the flow channel was measured for twovolume flow rates, 1 µl/min (17.7⋅10-12 m3/s) and 2 µl/min, as a function of the oxygengas pressure obtained in the electrochemical cell. Only a few experimental points weremeasured (due to the instability of the membrane deflection as result of actuatorpressure decrease during open circuit state) and they are listed in table 5.5-1. Tocompare the experimental data with the results from the numerical modelling, theactuator pressure is normalised (according to eq. 4.3.14), column 2. The flowresistance is normalised for the type of fluid, air, (eq. 4.3.15) and shown in column 4and 6. The experimental results are shown in fig. 5.5.11. It can be seen that the eyefluid pressure can be adjusted: for an actuator pressure of up to 1⋅105 Pa, a pressuredifference of 2000 Pa across the valve system is easily obtained. The results can becompared with the theoretical results (fig. 4.3.7 and 4.3.8) for only one point (becauseof a limitation in the numerical model software, actuator pressures higher than 1⋅105 Pacannot be simulated). This is the zero actuator pressure point which in the simulationgave a normalised flow resistance of 60 and in the measurements 73. Due to the factthat the actuator pressure was decreasing during flow measurements (the cell was inopen circuit condition) we were unable to obtain reliable results on the time stabilityand accuracy in adjusting the valve pressure. It is clear from these results that the finalsystem needs a control unit to maintain constant eye pressure.

ActuatorPressure

[105Pa]

NormalisedActuatorPressure

[103]

Measured∆Pvalve [Pa]

Φ = 17⋅10-12m3/s

NormalisedFlow

Resistance[103]

Measured∆Pvalve [Pa]

Φ = 34⋅10-12m3/s

NormalisedFlow

Resistance[103]

0 0 200 72 400 731.0 1.04 4600 1690 940 17301.5 1.56 7240 2660 14900 27402.0 2.08 10400 3700 20250 3720

Table 5.5-1. Measured pressure drop across the active valve as a function of actuatorpressure.

Chapter 5

101

0

5

10

15

20

25

0 0.5 1 1.5 2

Pcell [bar]

∆ ∆∆∆P

[kP

a]

1 µl/min

2 µl/min

Figure 5.5.11. Pressure drop over the flow channel due to the oxygen gas pressureobtained in the electrochemical cell.

5.6 A simple dynamic model for the electrochemical cell

In the dynamic measurements of the membrane deflection as function of theoxygen gas pressure generated by electrolysis (described above) a delay in themembrane deflection was observed (see fig. 5.6.1). This delay was of about 42-48s for5 µA, 24s for 10 µA. The build-up of the oxygen gas pressure is nearly a linearfunction of time. After the current is stopped, the gas pressure decreases exponentiallyat open circuit. The decay time is 348 s for both 5 µA and 10 µA.

A qualitative model for the dynamic properties of the electrochemical cell ispresented to simulate its behaviour. In this model it is assumed that (i) the oxygenproduced diffuses instantaneously and therefore homogeneously in the whole volumeof the electrolyte; (ii) no convection occurs after the oxygen gas bubbles are detachedfrom the Pt electrode; and (iii) at the start (t=0) the cell is filled completely withelectrolyte solution. The electrolyte solution is introduced in the electrochemical cell(cavity) with the use of vacuum, after which the filling holes were closed (glued) in arelatively short time (< 3 min). So, we can assume that at the start of the current cycle,there is no oxygen gas dissolved in the solution.

Figure 5.6.1. The time dependence of the deflection measured with theelectrochemical cell for i = 5 µA.


102

When the current is switched on, oxygen is produced by electrolysis at the Ptelectrode. Since the current is constant the production of oxygen is constant, so thequantity of oxygen produced Nprod increases linearly with the time. Oxygen diffusesfrom the Pt electrode in the bulk electrolyte forming the dissolved oxygen, Ndis. Let usfirst suppose that the diffusion process is fast, so the surface of Nafion is reached atonce by the dissolved oxygen and there are no concentration gradients in the water.Due to a concentration gradient across Nafion some oxygen diffuses through Nafion(concentration at Cu electrode is zero) creating a diffusion current of oxygen through

Nafion, N naf

•. The amount of oxygen lost through Nafion increases with time, causing

the dissolved oxygen concentration to increase less rapidly. As a result, both Nnaf(t) andNdis(t) will have an exponential behaviour. In order for the nucleation of gas bubbles totake place, the liquid has first to be supersaturated with dissolved gas. Until thesaturation value is reached, no oxygen bubbles are formed and therefore no pressureincrease is generated and no deflection of the membrane can be observed. Therefore,the sum of Nnaf and Ndis is the produced oxygen, Nnaf + Ndis = Nprod.

After the electrolyte becomes saturated Ndis is constant with time, Ndis = S⋅Vcell,

and N dis

•= 0 . The concentration gradient across Nafion becomes constant, thus the

lost oxygen through Nafion increases linearly. Since the solution is saturated, thedifference between the oxygen production and loss through Nafion is used to form theoxygen bubbles (gas phase), leading to the gas pressure that is observed by thedeflection of the membrane. The minimum delay time, until the solution is saturatedcan be estimated by:

N S Vi t

Fsat cell= ⋅ ≈⋅

4(5.6.1)

For a current of 5 µA, the solubility of oxygen in water S = 1.26 mol/m3, and thevolume of the cell Vcell ~ 1.2 mm3, we calculate a delay of 116 s. For a current of 10 µA, the delay time is 58 s. Below we will see that these times decrease when we takediffusion of oxygen in the water into account.

The relationship between the gas (bubble) formation and current can beexpressed as a function of the rates:

N N Ni

FNgas prod naf naf

• • • •= − = −

4(5.6.2)

where the rate of oxygen gas production is proportional to the applied current,

Ni

Fprod

•=

4, and the rate of Nnaf, N naf

•, is constant for different currents because it

reaches the maximum value:

N S DA

dnaf O

N N

N

•= ⋅ ⋅

max

2(5.6.3)

We can distinguish two cases:

(i) N Nprod naf

• •≤

max

⇒ no oxygen bubbles (5.6.2a)

Chapter 5

103

(ii) N Nprod naf

• •>>

max

⇒ N Ngas prod

• •≈ (5.6.2b)

Equation 5.6.2a indicates that no gas will be generated because the saturation valuewill not be reached. For a high current (eq. 5.6.2b), the total current will be used tocreate gas, i.e. the gas production is almost directly proportional to the applied current.

From the experimental data for 5 µA and 10 µA, it was observed that the gasformation rate for 10 µA was approximately three times that for 5 µA:

N A N Agas gas

• •≈( ) ( )10 3 5µ µ (5.6.4a)

10

43

5

4

µ µA

FN

A

FNnaf naf− = ⋅ −

• •(5.6.4b)

so, N AF

naf

•= ⋅

5

2

1

4µ (5.6.4c)

This means that for a current smaller than about 2.5 µA (that corresponds to eq.5.6.2a) no gas bubbles are formed. Equating eq. 5.6.4c to the maximum rate ofdiffused oxygen (eq. 5.6.3), the oxygen diffusion coefficient through Nafion can beestimated by:

DA

F S

d

AON N

N2

5

8=

⋅⋅

µ(5.6.5)

For a Nafion area and thickness, AN = 1 mm2 and dN = 2 µm, respectively, the oxygendiffusion through Nafion is 2⋅10-12 m2/s, which is within the range of the reportedvalues.

When the current is switched off, no oxygen is formed. The oxygen exists in thecell in two phases (i) dissolved and (ii) in the gas phase. The oxygen from the gasphase is assumed to dissolve instantaneously in the electrolyte. We assume that thegas-water interface does not form a barrier for transport of oxygen. Consequently, theconcentration at the interface will be S. The oxygen concentration between the bulkelectrolyte and the electrolyte layer next to the gas phase varies because of the

continued diffusion of dissolved oxygen through Nafion. N naf

• remains constant with

time because Ndis stays constant until all the oxygen gas phase is consumed. Since

N naf

•= constant N gas

•= constant. The diffusion of oxygen through Nafion continues

until no oxygen remains in the cell, so until Nnaf = Nprod. The consumption of theoxygen gas phase is observed by the decrease of the pressure. When all the oxygenfrom the gas phase is dissolved, the measured pressure reaches the zero value.

Although this simple model describes reasonably the behaviour of the cell, themodel needs some refinement because (i) the decrease of the gas pressure isexperimentally exponential and slower than estimated above; (ii) the delay time untilbubble formation is experimentally shorter (ca. 50 s (24 s) instead of 116 s (58 s)).These differences might arise from the fact that it was supposed that the diffusionprocess takes place instantaneously and homogeneously in the whole solution.


104

During the electrolysis process, the concentration of dissolved gas is alwayshigher at the electrode-electrolyte interface than in the bulk liquid because the gas isgenerated at the electrode surface and the diffusion process is relatively slow.Therefore, there is a bigger chance for gas bubbles to start growing at the electrodeinterface. This probability is even bigger due to the higher concentration of nucleationcentres at the electrode surface. Thus, at the electrode surface the saturation will befirst reached and gas bubbles will be created before the bulk solution will reachsaturation. The delay will therefore be shorter than that calculated with eq. 5.6.1.

The following diffusion model is thought to describe the growth and change ofthe concentration profile before the formation of gas at the electrode. Solving thediffusion problem for the general case is very difficult, therefore some simplificationshave to be made. No migration and convection will be considered. Diffusion of oxygenis perpendicular to the surface of the electrode. The diffusion problem is governed byFick’s second law (one-dimensional case):

∂∂

∂∂

C x t

tD

C x t

x

( , ) ( , )=

2

2 (5.6.6)

where C(x,t) [mol/m3] is the concentration of dissolved oxygen at the distance x and attime t, and D [m2/s] is the diffusion coefficient of the dissolved oxygen molecules. Asan initial condition, at t = 0, it is assumed that the concentration of dissolved oxygen inthe whole solution, x ≥ 0, is zero:

C(x,t) = 0 x ≥ 0, t = 0 (5.6.6a)

A boundary condition relates to the concentration gradients at the electrode surface.Since the current is the controlled quantity, the flux of oxygen molecules J(x,t)[mol/m2⋅s] gives the boundary condition (i.e. Fick’s first law):

J x ti

nFAD

C x t

x x( , )

( , )= =

=

∂∂ 0

(5.6.6b)

where i is the applied current, A is the area of the Pt electrode. The solution of the onedimensional case in the x direction is [Ba8001]:

C x ti

nFAD

Dt x

Dtx erfc

x

Dt( , ) exp= ⋅ −

− ⋅

24 2

2

π(5.6.7)

A concentration profile of dissolved oxygen, corresponding to eq. 5.6.7, isshown in fig. 5.6.2, with i = 5µA, oxygen diffusion coefficient in water, D = 3⋅10-9

m2/s, area of Pt electrode, Apt= 1.5 mm2, number of electrons involved in the reaction,n = 4, and Faraday constant, F = 96500 C/mol. The variation of concentration asfunction of the distance from the electrode (Pt) to the bulk solution is shown in fig.5.6.2a for different times. The same concentration profile but as function of the time atthree distances from the electrode is presented in fig. 5.6.2b. From this figure it can beseen that the concentration at the electrode surface (x = 0) reaches the saturation value(S) after ca. 50s, which corresponds well with the experimental value.

Chapter 5

105

(a)0 200 400 600 800 1000

1

2

t1=1s, t2=60s, t3=120s, t4=180s2.5

0

C (x,t)

1.26

x [µm]

(b) 0 50 100 150

1

2

x1=0µm, x2=200µm, x3=500µm2.5

0

C (x,t)

1.26

t [s]

Figure 5.6.2. Concentration profile for the diffusion of oxygen into water (a) functionof the distance from the electrode for different times, (b) function of the time for threedistances.

The pressure decrease after the current is switched off is determined mainly bythe dissolution of the oxygen gas bubble into the solution (water) and the slowdiffusion through the electrolyte and Nafion. The time of oxygen gas dissolution is afunction of the oxygen concentration gradient between the gas phase and solution. Thenon-linear decay of the gas pressure may be explained as following. After the current isswitched off, the gas bubbles are probably at the top of the solution, or in any case faraway from the Nafion. Therefore, the oxygen gas has first to dissolve in the solution,to diffuse through the electrolyte and then through Nafion. During this process, the gasbubble volume will shrink, and the contact area between gas and solution will decreaseas well. Since the dissolution of oxygen gas is proportional to this contact area, thedissolution rate will slowly decrease in time.

t1

t2t3 t4

x1

x2

x3


106

As a conclusion, it can be said that this simple electrochemical model predicts (i)a delay in the gas pressure generation, for small currents; (ii) a non-linear decay in gaspressure during the open circuit condition; (iii) the gas pressure does not necessarilyscale with the current. An improvement of the electrochemical cell can be obtained byusing a thicker Nafion membrane having a smaller surface area, and by finding otherelectrode material to replace the copper electrode.

For long-term functioning, there are also other factors which might account forthe oxygen pressure drop, such as (i) diffusion of oxygen through the glue, used for thebonding of the wafers which form the electrochemical actuator; (ii) diffusion of oxygenthrough the silicon nitride membrane; (iii) if some hydrogen gas is formed duringelectrolysis, then the catalytic reaction of oxygen and hydrogen gases at Pt electrodewill take place; (iv) if the potential of the cell is much higher than the reversible cellpotential, an expansion of the electrolyte and the gas might take place due to the Jouleeffect. We shall briefly discuss only this thermal effect. This effect will create an extrapressure that overlaps those generated by electrolysis. Since heat dissipates to thesurroundings, cooling will lead to a fast decrease of the extra pressure. An estimationof this effect can be done as follows: The electrical energy produced due toelectrolysis, for a current, i = 10 µA, voltage, V = 1.7 V and electrolysis time telectrol ~500 s is:

∆Wel = i⋅V⋅telectrol = 8.5⋅10-3 J (5.6.8)

The temperature variation of water and Si due to this energy is:

∆∆

T =W

C + C el

H O p(H O Si p(Si2 2m m) )

(5.6.9)

With the data, Cp(H2O) = 760 J/kg⋅K, Cp(Si) = 4180 J/kg⋅K, ρSi = 2330 Kg/m3, ρwater =1000 kg/m3, and considering that the volume of silicon around the water (electrolyte)is 20 times bigger (the mass is 20⋅ρSi/ρwater = 46 times bigger), the temperaturevariation is ∆T ~ 0.4 K. The variation of the water volume due to this temperature is:

∆VH2O = α⋅Vcell⋅∆T = 1⋅10-5

mm3

(5.6.10)

where α = 0.21⋅10-3

K-1

is the thermal coefficient of volumetric expansion for water at

25°C at 1 atm and Vcell ~ 1.2 mm3, the volume of the cell. The variation of the gas

volume due to this temperature is:

∆Vgas/Vgas = ∆T/T ~ 0.13% (5.6.11)

It can be seen that this effect is negligible with respect to the diffusion throughNafion.

Chapter 5

107

5.7 Conclusions

For the re-designed microactuator, a polyimide mesh was used to increase theadhesion of Nafion. With the use of this mesh the area of Nafion is decreased, whileits thickness is increased, leading to an increase of the decay time in open circuitconditions from 6 min. in the first prototype to 93 min. for the second type. With anactuation voltage of 1.6 V and currents below 50 µA, pressures of 2 bar are easilyobtained in seconds to minutes. The membrane of the microvalve is deflected in therange of 30 to 70 µm.

In the above experiments the build-up time is mainly determined by the chargepassed through the cell, while the decrease of pressure in the short-circuited statedepends on the rates of the electrochemical reactions and on the electrical impedanceof the cell. The rate of decrease in pressure under open circuit conditions very likelydepends on the rate at which oxygen diffuses through the Nafion layer and is reducedat the copper electrode. When the copper and platinum electrodes are short-circuited,oxygen can also be reduced at the noble metal and the rate of Cu corrosion isenhanced; as expected, the pressure drops faster in this case.

The operation of the actuator as a valve was tested: The fluid flow in a valvesystem was analysed with the use of analytical equations, for a simple channelgeometry, while the pressure drop across the valve was estimated using a numericalmodel. With the numerical simulation, the influence of the membrane deflection on theflow resistance, as well as on the pressure across the valve system could be seen. Inour case, the volume flow rates are small, so the membrane is not deformed due to thefluid pressure. From the experimental results it can be concluded that for an actuatorpressure up to 1⋅105Pa, the fluid pressure across the valve can be easily adjusted in therequested range (1000 - 3000 Pa).

As a general conclusion, it can be stated that the performance of theelectrochemical actuator depends on the efficiency of gas production; this efficiencydepends not only on how gas-tight the cavity can be sealed but also on how effectivelyoxygen reduction at the copper electrode can be prevented.

Advantages of the microactuator are a low power consumption and thepossibility of discontinuous energy supply. Moreover, a regular adjustment of the flowresistance is possible. From the present results it may be concluded that a microvalvecan be actuated, using an electrochemical driving principle.


108

Appendix 5A

Process 1st design electrochemical cell

1. Pre-processing1a Standard clean wafer1b Have 1µm Silicon rich low-stress Nitride grown.

2. Processing of holes on the backside2a Patterning of nitride on backside

- photolithography with mask bord1, standard positive resist- nitride etch by RIE: CHF3-O2 (25:5 sccm) 10mTorr, 75W, ~12.5 min/1 µm- resist stripping by O2 plasma, 10-15 min.

2b KOH etch backside, required depth 380µ-50µ=330µm,etch speed with 25% KOH at 70°C about 1µm/min -> 330 min = 5h 30 min.

2c. Transfer of alignment marks to front side- photolithography with mask bord8, standard positive resist- nitride etch by RIE: CHF3-O2 (25:5 sccm) 10mTorr, 75W, 5 min/~ 400nm.- resist stripping by O2 plasma, 10-15 min.

2d. Etching of nitride mask on backside- nitride etch by RIE: CHF3-O2 (25:5 sccm) 10mTorr, 75W, 12.5 min.

2e. KOH etch backside down into nitride membrane on front side, depth 50µ,etch speed with 25% KOH at 70°C about 1µm/min -> 50 min.

2f. Post-KOH cleaning. with H2SO4:H2O2:H2O (1:1:5) solution, 80ºCmake new solution every time because H2O2 disappears within half an hour.

2g. Standard Clean wafer.2h. 100 nm Si rich low-stress nitride (insulation layer for electrodes on silicon).

3. Copper contacts for electrodes through wafer and wiring at the backside- Cr/Cu sputtering with shadow mask (bord mask5) in cryo-sputter.

Cr: Cryo, 14K, Pressure: 8.0 10-4, Ar flow 60, DC power: 200W,Deposition speed about 2.2 Å/s, thickness 100nm -> 8 min

Cu: Cryo, 14K, Pressure: 8.0 10-4, Ar flow 60, DC power: 200W,Deposition speed about 4 Å/s, thickness 500nm -> about 20 min.

4. Patterning of nitride on front side- photolithography with mask bord2, positive resist- Nitride etch with RIE: CHF3-O2 (25:5 sccm) 10mTorr, 75W, 2*12.5=25 min

- Resist stripping by O2 plasma, 10-15 min.

5. Copper electrode- photolithography with mask bord4, positive resist S1828 (12s illumination)- 5 min. chlorine-benzene dip, rinse with demi-water and develop in 60 s. No post bake


Cu: Cryo, 14K, Pressure: 8.0 10-4, Ar flow 60, DC power: 200W,Deposition speed about 5 Å/s, thickness 500nm -> about 20 min

- Lift off etching of Cu in acetone.

6. Platinum electrode- photolithography with mask bord3, pos. resist S1818 (8s illum.)- 5 min. chlorine-benzene dip, rinse with demi-water and develop in 60 s. No post bake


Chapter 5

109

Pt: Cryo, 14K, Pressure 8.0 10-4, Ar flow 60, DC power 200W,Deposition speed about 5.3 Å/s , thickness 500nm -> 16 min (no substrate

rotation !)- Lift off etching of Pt in acetone.

7. Etching of cavity under Cu electrodeDepth that is necessary to under-etch the copper electrode (and nitride support) is about

100µm.etch speed with 25% KOH at 70°C about 0.8µm/min -> 100µm=115 min = 1h 55 min

8. Removal of oxide from copper and deposition of NafionDrop a small drop of Nafion on each copper electrode with a pipette.


110

Chapter 5

111

Appendix 5B

Process 2nd design electrochemical cell

1. Pre-processing1a. Standard cleaning wafer1b. SiO2 wet oxidation: oven at 950°C for 4 hours, ca. 500 nm thick oxide.

2. Copper electrode2a. Ti/Cu sputtering in Sputterkje: base pressure 2 10-7 mbar, sputtering pressure 510-3, Ar flow 40, dc power 200W,

Ti: deposition thickness 20nm -> 3 min;Cu: deposition thickness 200nm -> about 10 min.

2b. photolithography, mask ‘Ti/Cu’, positive resist S1813 (6s illumination), no postbake.2c. - Cu etch, ca. 20 nm/s in 1HNO3, 4H3PO4, 4CH3COOH, 1H2O solution (etchesalso Al);

- Ti etch in 2% HF (just dip in, the change of colour);- remove photoresist with acetone.

3. Platinum electrode3a. photolithography, mask ‘Ti/Pt’, positive resist S1813 (6s exposure),

5 min. chlorine-benzene dip, no post bake.3b. Ti/Pt sputtering in Sputterkje: base pressure 2 10-7 mbar, sputtering pressure 5 10-

3, Ar flow 40, dc power 200W,Ti: deposition thickness 20nm -> 3 min;Pt: deposition, thickness 200nm -> about 10 min.

3c. lift off etching of Pt in acetone.

4. Al sacrificial layer for the mesh4a. photolithography, mask ‘Al’, positive resist S1818 (12s exposure),

5 min. chlorine-benzene dip, no post bake.4b. 1.5 µm Al evaporation in Varian.4c. lift off etching of Al in acetone.

5. Polyimide mesh (on top of Cu electrode)- spin APS primer, 4000 rpm, 30s,- pre-bake 90°C, 10s,- spin HTR3-200 solution, 1800 rpm, 30s, thickness of 11 µm,- pre-bake 90°C, 25 min.,- mask ‘PI’, exposure 30s,- bake 90°C, 10 min.,- develop in HTR-D2, ca. 3 min., rinse in IPA,- post-bake 300°C, 1 hour in vacuum oven.

6. Al etch awayAl etch in a solution of Na2CO3 (37.5g), Na3PO4⋅12H2O (17.5g), K3Fe(CN)6 (8g)

dissolved in 250 ml water, at 25°C, for 5-6 hours with stirring (ca. 130 mn/min).


112

References

[As9301] T. Asaoka, The water uptake and properties of perfluorinated polymer, ExtendedAbstracts Electrochemical Soc. Spring Meeting 1993.

[Ba8001] A. J. Bard and L. R. Faulkner, Electrochemistry methods: fundamentals andapplications, John Wiley & Sons, 1980.

[Ba9501] P.W. Barth, Silicon microvalves for gas flow control, Proc. Int. Conf. Solid-StateSensors and Actuators, Transducers’95 & European Conf. on Solid-StateTransducers, Eurosensors IX, 25-29June, Stockholm, Sweden, p. 276-279, 1995.

[Bu9601] G.J. Burger, E.J.T. Smulders, J.W. Berenschot, T.S.J. Lammerink, J.H.J.Fluitman, S. Imai, High-resolution shadow-mask patterning in deep holes and itsapplication to an electrical wafer feed-through, Sensors and Actuators 54(1-3), p.669-673, 1996.

[El9401] M. Elwenspoek, T.S.J. Lammerink, R. Miyake and J.H.J. Fluitman, Towardsintegrated microliquid handling systems, J. Micromech. Microeng. 4, pp. 227-245, 1994.

[Gi9301] E. Gileadi, Electrode kinetics, VCH Publishers, 1993.[Gr8601] W. G. F. Grot, Nafion as a separator in electrolytic cells", The Electrochem.

Soc. meeting, Boston, May, 1986.[Gr9301] P. Gravesen, J. Branebjerg, O.S. Jensen, Microfluidics - A review, J. Micromech.

Microeng. 3, p. 168-182. 1993.[Ha9501] M.W. Hamberg, C. Neagu, J.G.E. Gardeniers, D.J. Yntema and M. Elwenspoek,

An electrochemical microactuator, Proc. IEEE Workshop on MEMS, Amsterdam,The Netherlands, p. 106-110, 1995.

[Ij9201] D.J. Ijntema, Feasibility study for a micro machined eye pressure regulator forglaucoma patients, Report University of Twente, EL-TDM, Enschede, TheNetherlands, 1992.

[Ja8801] H. Janocha, Neue Aktoren, Proc. Actuator 88, June, Bremen, p.389, 1988.[Ja9201] H. Janocha, Aktoren, grundlagen und anwendungen, Springer-Verlag, 1992.[Ke9001] W. Kempe, W. Schapper: Electrochemical actuators, Proc. Int. Conf. on New

Actuators, Actuator’90, June, Bremen, Germany, p.162, 1990.[Ko9501] H. Kok, Closure of liquid filled cavities for an electrochemical eye-pressure,

M.Sc. Thesis, University of Twente, Enschede, the Netherlands, 1995.[Ku9701] Joost van Kuijk, Numerical modelling of flows in Micro Mechanical Devices,

Ph.D. thesis, University of Twente, Enschede, The Netherlands, 1997.[La8801] D. R. Lawson, L. D. Whiteley, C. R. Martin, M. N. Szentirmay, J.I. Song,

Oxygen reduction at Nafion film-coated platinum electrodes: Transport andkinetics, J. Electrochem. Soc. 135, p. 2247, 1988.

[Ne9601] C. Neagu, J.G.E. Gardeniers, M. Elwenspoek and J.J. Kelly, An electrochemicalmicroactuator: principle and first results, J. Microelectromech. Syst. 5(1), p.2-9,1996.

[Ne9602] C. Neagu, J.G.E. Gardeniers, M. Elwenspoek and J.H.J. Fluitman, Anelectrochemical actuated microvalve, Proc. Int. Conf. on New Actuators,Actuator’96, June 19-21, Bremen, Germany, p. 41-44, 1996.

[Og8501] Z. Ogumi, T. Kuroe, Z. Takehara, Gas permeation in SPE method. II. Oxygenand hydrogen permeation through Nafion, J. Electrochem. Soc. 132, p. 2601,1985.

[Pl9301] D. Pletcher, S. Sotiropoulos, “A study of cathodic oxygen reduction at platinumusing microelectrodes”, J. Electroanal. Chem. 356, pp. 109-119, 1993.

[Po6601] M. Pourbaix, Atlas of electrochemical equilibria in aqueous solutions, PergamonPress, Oxford, 1966.

[Ri8701] P. H. Rieger, Electrochemistry, Prentice-Hall International, 1987.

Chapter 5

113

[Si8301] C.W.M.P. Sillen The effect of gas bubble evolution on the energy efficiency inwater electrolysis, Ph.D. thesis, University of Eindhoven, Eindhoven, TheNetherlands, 1983.

[Sm8001] W.H. Smyrl, in Comprehensive treatise of electrochemistry, vol. 4, J.O’M.Bockris, B.E. Conway, E. Yeager, R.E. White, Eds., 1980.

[Vo9601] A. Volanschi, Dynamic surface tension measured with single nucleation siteelectrodes, Ph.D. thesis, University of Twente, The Netherlands, 1996.

[We9301] K. O. v.d. Werf, C. A. J. Putman, B. G. de Grooth, F. B. Segerink, E. H.Schipper, N. F. van Hulst, J. Greve, Compact stand-alone atomic forcemicroscope, Rev. Sci. Instrum. 64, p. 2892, 1993.

[Wi9001] E. Wilkins, W. Radford, Biomaterials for implanted closed loop insulin deliverysystem: A review, Biosensors & Bioelectronics, Elseviers, p. 167-213, 1990.

[Wu9301] W.S. Wu, G.P. Rangaiah ”An experimental study of oxygen evolution and masstransfer at microelectrodes”, J. Chemical Engineering of Japan 26(6), pp. 620-626, 1993.

[Ya9401] K. Yasuda, Y. Uchimoto, Z. Ogumi, Z.-i Takehara, Preparation of thinperfluorosulfonate cation-exchanger films by plasma polymerization, J.Electrochem. Soc. 141(9), p. 2350-2355, 1994.

[Ze9401] R. Zengerle, W. Geiger, M. Richter, J. Ulrich, S. Kluge, A. Richter, “Applicationof micro diafragm pumps in microfluid systems”, Proc. Int. Conf. on NewActuators, Actuator’94, June, Bremen, Germany, p. 25-29, 1994.

Chapter 6

113

This chapter presents two components of the electrical system,included in an total eye-pressure system: the energy supply anda feedback controller. First, the modelling, design andcharacterisation of planar microcoils, to be used in telemetrysystems that supply energy to miniaturised implants isdescribed. Parasitic electrical effects that may becomeimportant at AC-frequencies of several MHz are evaluated.The fabrication process and electrical characterisation of planarreceiver microcoils will be described, and it will be shown thata power of a few mW is feasible.In the second part, an electronic feedback system to control thepressure of an electrochemical macrocell automatically ispresented. A value of the pressure is set and the regulatoractuates the electrochemical cell to maintain the desiredpressure.

6.1 Introduction

Over the last few years, interest in and use of implants in the human body hasincreased. Permanent implants can be used for continuous monitoring of certainmedical parameters (e.g. insulin level, intraocular pressure) or for actuation of specificnerves [Pu9501, Na9501]. As the implants get smaller, much attention is directedtowards down scaling of the electronics, data transmission circuitry, and power supply.Considering the latter, the use of a battery is not always feasible or desirable because ofsize and lifetime limitations [Ka6901], while in general wire connections cannot beplaced for long periods of time. An alternative is the transmission of energy byinductive coupling of a pair of coils; a transmitter coil and a receiver coil implanted inthe human body as found in figure 6.1.1. The receiver coil for this type of applicationhas two important requirements: a very small size and a high efficiency.

∗ This section is based on the paper: “Characterisation of a planar micro coil for implantablemicrosystems” C.R. Neagu, H.V. Jansen, A. Smith, J.G.E. Gardeniers, M. Elwenspoek, Sensors &Actuators A 62(1-3), p.599-611, 1997.∗∗ This section is based on the paper: “The electrolysis of water: an actuation principle for MEMSwith a big opportunity:” C.R. Neagu, H.V. Jansen, J.G.E. Gardeniers, M. Elwenspoek, accepted at the6th Int. Conf. on New Actuators, Actuator’98, Bremen, Germany, June, 1998.

ENERGY SUPPLY∗∗∗∗ and FEEDBACK CONTROLLER ∗∗∗∗∗∗∗∗

6

Energy Supply and Feedback Controller

114

The small size of the receiver coil might lead to the need for a dense time-varyingmagnetic field through this coil, in order to transfer a sufficient amount of energy.However, such a field induces eddy currents in a conductive material, such as humantissue, resulting in resistive heating or actuation of nerves. It has been reported that thisheating can cause damage to human tissue [Le8201]. So, in general dense magneticfields are not desirable. In order to be able to transfer sufficient energy at a less densefield, the design of the receiver coil should be optimised.

In this paper we will focus on different configurations of planar receiver coilsmade with the use of micromachining technologies such as thin film deposition andelectroplating. With this technology small coils are relatively easy to fabricatereproducibly. Insight has been gained into the importance of parasitic effects, whichstart dominating at frequencies of several MHz. First, a model for the electricalcharacteristics of planar coils will be discussed, followed by a discussion of designconfigurations and fabrication procedures. The parameters and performance offabricated coils will be evaluated by electrical measurements, and first results of theenergy transfer from a conventional coil to the planar coil will be presented. However,before starting we will take a closer look at the energy transfer needed to feed thereceiver-electronics; i.e. the telemetry.

Figure 6.1.1. Cross-section of a transmitter-receiver set-up showing magnetic fieldlines.

Chapter 6

115

6.2 Telemetry

Currently, the transport of energy by inductive coupling between a set of coils isfound in almost any electronic apparatus. The energy transfer is accomplished by usingtwo coils; a transmitter coil which supplies the energy in the form of a magnetic fieldand a receiver coil to collect as much energy as possible from this field. Generally,when the coils are allowed to be in close contact, a transformer is used. In thetransformer, the magnetic field is directed from the transmitter coil into the receivercoil by way of a magnetically conducting medium like iron. At a greater distance, it isno longer possible to use iron to concentrate the magnetic field. Therefore, the coilshave to be aligned in order to couple enough energy.

When the frequency of the field, generated by the transmitter coil, is higher thana few kHz, the transmission of electromagnetic energy by electromagnetic waves (EM-waves) becomes efficient. EM-waves have their application in a broad frequencyspectrum from several kHz up to many GHz. For example, in a microwave ovenworking at a few GHz, energy is transferred to water molecules to heat-up a meal. Theadvantage of EM-waves is that they can be directed with the use of special antennas.Due to this characteristic, it is possible to direct energy over quite a large distancewithout any problem. Moreover, it is possible to modulate the EM-wave, i.e. thecarrier signal, with data in order to transmit, wireless, information. This is the field oftelecommunication, without doubt one of the most important pillars of the modernsociety. It is found in applications such as radio and television, but it is also used tocontrol an object at a distance, as in case of a remote controller to zap your televisionset or a small battery-powered transmitter to steer a model aeroplane flying throughthe sky.

Frequently, it is necessary to transport information from a sensing or measuringunit to a remote controller and backwards; this is the field of telemetry. Data is nowgoing in both directions, the remote system is not only receiving its control signals butalso transmits its measured data. For this reason, such systems are generally calledtransceivers, a combination of transmitter and receiver. Exciting examples of this fieldare the voyager space crafts which explore the universe.

In some special cases, the remote transceiver does not have the possibility tosupply the energy needed to feed its own electronics. These electronics are necessaryto prepare the data to be sent to the transmitter or host coil. In such cases, the host coilis not only used to transmit the control signals but it is also used to transport theenergy for the electronics of the remote transceiver. In such cases, after stripping anEM-wave from its control signals, the carrier signal is rectified in order to supply theelectronics with this energy. It is obvious that this type of communication is one of themost challenging technologies in telemetry, because the carrier signal not only carriesthe energy but also the data. In case of human implants, this technology seems to beone of the few possibilities to control functions in a “body friendly” way.

In the remaining part of this section we will therefore, concentrate on how theenergy is received by a remote receiver coil and how this energy is directed to itselectronics. Although the system has also to communicate by way of control signalsand sensed data, this subject is not discussed here. We do not mean to imply that thissubject is not important, but mainly that it is not of concern in this chapter.


116

In order to find the equations which describe the transfer of energy and efficiencyin a receiving coil, we will start with some aspects dealing with energy transfer andefficiency. Figure 6.2.1 presents a schematic drawing of the transceiver part of atelemetric system. It consists of a host system which transmits an energy Etrans and aremote unit which receives only a part of this energy Erec due to the distance betweenboth units and the divergence of the EM-field. The received energy, in turn, is partlydissipated as Joule heat in the resistance of the receiver coil Edis and the energy left isstored as EM-energy EEM in the self-inductance L and capacitance C of this coil; theLC-circuit. Finally, a part of the EM-energy stored can be used to supply theelectronics for the requested energy Eelec, that is, it will be dissipated in a loadresistance. Now, when we look at the receiving remote unit it is possible to define aquality factor Qrec as (fig.6.2.1):

QE

ErecEM

dis

= (6.2.1a)

The Q-factor is a measure for the efficiency of the receiver coil: the higher the Q-factor, the better the performance with respect to the energy transfer. Note that, formaximum energy transfer from the stored EM-energy to the load resistance, we haveto optimise the load resistance. This can be achieved by matching the load resistance tothe characteristic impedance of the LC-circuit, but this subject will be treated furtheron.

Another way to describe the efficiency of the receiver coil is by looking at theplot of the impedance or output voltage of the coil as a function of the frequency. Inthis case the operation quality-factor, Qo, turns out to be (fig.6.2.1):

Qoo=

ωω

2

(6.2.1b)

In this formulation ωo is the working or operating resonance frequency and ω√2 denotesthe width of the resonance curve at 1/2√2~0.707 times the maximum of the outputvoltage as found in figure 6.21. Note that, when the output function is the transferredpower, this point is taken at half maximum because P = U2⋅R. So, a sharp peak in theresonance curve is indicative of a high efficiency.

Figure 6.2.1. Schematic drawing of the energies involved in a transceiver part of atelemetric system. The receiver coil has to have a high efficiency, i.e. sharp resonancepeak; this is not necessary for the transmitter coil (see text); ωo is the working oroperating resonance frequency and ω√2 denotes the width of the resonance curve athalf maximum of the output voltage.

Chapter 6

117

If both the transmitter and receiver coils would have a sharp resonance peak, it wouldbe difficult to tune both circuits at the same frequency. However, it is not necessarythat the transmitter coil has a high efficiency because it does not have to be implantedand therefore the supply of energy for this coil is not restricted. Therefore, the Q-factoror efficiency of the transmitter coil is chosen to be low, to allow for easy tuning of bothcoils.

We now have two ways to determine the efficiency of the remote receiving unit.There is still another practical definition found in the literature which uses the intrinsiccharacteristics of the receiving coil, i.e. the self-inductance L, parasitic capacitanceCpar, and series resistance RS. When the coil with these three intrinsic components isleft on its own (i.e. no extra circuitry such as a load is connected to the coil) and thecoil is activated by an external alternating field, the circuit starts to oscillate at aspecific frequency; the intrinsic resonance frequency ωi [rad/s]:

ωipar

S

parLC

R

L LC= −

1 12

2 ~ (6.2.2a)

The approximation on the right-hand side of eq. 6.2.2a is only allowed when RS

<< ωiL. Generally in electronics, the receiving coil is not left on its own to determinethe resonance frequency. Instead, an extra tuning capacitor, Ctune, is placed in parallelwith the coil to be able to tune for the desired operating resonance frequency, ωo. Insuch cases, the resonance frequency can be found by adjusting eq. 6.2.2a for thiscapacitance:

( )ωo

par tuneL C C~

1

+(6.2.2b)

Thus, in many practical situations, the real resonance frequency is lower than theintrinsic resonance frequency. Now we return to eq. 6.2.1a. In this equation, theenergy for the electronics is supplied by the energy stored in the magnetic field of theself-inductance of the coil and the dissipated energy is due to the series resistance.When there is no load connected to the LC-circuit, the intrinsic Q-factor Qi (which isthe ratio between the energy which can be used for the load and the energy dissipatedin the coil) becomes:

QE

E

L

R

L

R C R

L

Ci

L

R

i

S par S parS

= = = −ω2

11

~ (6.2.1c)

The intrinsic characteristics of the coil mainly depend on geometrical factors, so,with the help of this formulation we are able to calculate the efficiency of the systembeforehand and not, as in case of the first two definitions, only afterwards. This givesus the possibility to optimise the efficiency of the remote receiver coil with respect tothe geometry, if and only if the intrinsic equations are known.

There is still remains one important issue about efficiency: What happens withthe efficiency when a load resistance and/or tuning capacitor are connected to thereceiver coil in fig. 6.2.1? In this drawing, the equivalent of the receiver coil is given asan ideal self-inductance L in series with a series resistance Rs to account for Ohmiclosses. Both components are placed in parallel with a parasitic capacitance C caused by


118

voltage differences between closely-spaced conducting surfaces. The load is connectedin parallel with the parasitic capacitance. There are at least two ways to look at thiscircuit. One is to replace the receiver coil by three individual components; a self-inductance, series resistance, and a parasitic capacitance. The second is to treat the coilas a distributed system. In the following it will be shown that there is quite aremarkable difference between the two approaches with respect to optimising the loadresistance.

In the non-distributed approach, the self-inductance with its series resistance canbe replaced by an equivalent parallel circuit of a new self-inductance L’ with resistanceR’. The behaviour of this purely parallel circuit is identical to the original one when thefollowing conditions are fulfilled:

L LR

Land R

L

R Cs

s/ / / /= +

=1

2

ω(6.2.3a)

resulting in eq. 6.2.1c and 6.2.2a. With this parallel equivalent circuit it is easy toobtain the state of maximum power transfer of the LC-circuit to the load:

R RL

R Cload //s

= = (6.2.3b)

When a load capacitance is connected to the output terminals of the receiver coil,the resonance frequency will decrease in the same way as found in eq. 6.2.2b.

In the distributed approach, we look at the coil as a transmission line. Atransmission line, such as a coaxial cable, can be represented by N distributed circuitsconsisting of a self-inductance in series with a resistance, both placed in parallel with acapacitance. For such a distributed system, the characteristic impedance Zo is foundwith the help of the so-called telegraph equations. In the case of maximum powertransfer the load should match with Zo:

R ZR j L

G j C

L

Cload oS

P

= =++

ωω

~ (6.2.3c)

The approximation at the right-hand-side of this equation is only allowed whenRS << jωL and Gp = 1/Rp << jωC which is quite often the case. Clearly, equations6.2.3b and 6.2.3c are not the same. The formulation that should be used depends onthe resonating system we are looking at: Is it distributed or not?

Chapter 6

119

6.3 Theory of a planar coil

In this section a theoretical model of theintrinsic characteristics of a planar microcoil, asshown in figure 6.3.1, will be discussed. This willgive a better understanding of the influence ofparasitic effects and gives us the possibility tooptimise the coil’s design with respect to energytransfer. We will concentrate on such a planarcoil, because this type can be readilymanufactured in a small size by means of micro-machining techniques. For the coil that has to beimplanted in the human body, the small size is aprerequisite. The simplified equivalent electricalcircuit of the transceiver system is shown in figure6.3.2. It consists of two, almost identical,inductively coupled circuits; the transmitter andreceiver units. Both units have an LC-circuit, which can be driven in resonance. Due toimperfections in the conductors and insulators, the coils do have a finite (small) seriesresistance, Rs, and a (big) parallel resistance Rp (to the capacities), respectively. Theseresistors dissipate energy (Joule heat) and, therefore, they will lower the efficiency ofthe resonance circuits as expressed by eq. 6.2.1a. So, the dissipated energy is directlyinfluenced by the series resistance which depends on geometrical factors. The receivedenergy is mainly dependent on the self-inductance of the coil which also depends onthe dimensions of the coil. A theoretical analysis is therefore, necessary to find criteriafor an optimum performance of the receiver coil.

Different complex classical formulas and approximations for coil parameters arein use, depending on the geometry of the coil and the desired degree of accuracy[Re9301, Ka8401, Gr7401, Bu2601, Gr7301, Du6101, Lo9201, Wh2801, Pe8801,La8801, Ca8301, Te4301]. In this study, we shall only consider copper coils,deposited on a silicon substrate covered with an insulating silicon oxide film. Theelectrical properties of such a planar microcoil depend on the inductance Lr, theconductor series resistance Rs, the parasitic capacitance, Cpar, and the insulator(substrate) resistance Rp. In the following, these parameters will be treated one-by-one.

Figure 6.3.2. The simplified equivalent electrical circuit of the transceiver system;the transmitter system coupled with the non-ideal receiver coil.

6.3.1 Self-inductance

Figure 6.3.1. Simplifiedrepresentation of a planarreceiver coil with some electriccomponents.


120

The self-inductance of a coil, L, is defined as the magnetic flux linkage per unitcurrent in the coil itself. Alternatively, it is defined as twice the energy stored in themagnetic field divided by the square of the current through the coil. In 1943, Termanarrived at some expressions for the low-frequency, i.e. no skin-effect, self-inductanceof differently shaped coils [Te4301]. We will start with expressions for a single turn orloop and we will then increase the number of turns to increase the self-inductancedrastically. The self-inductance of a rectangle of sides s1 [m] and s2 [m] and diagonal g[m] of wire of diameter d [m] is [Te4301, eq. 33]:

( ) ( ) ( )L [H] s ss s

ds s g s s g d g s s[] = ⋅ +

⋅− + − + + + − −

µπ 1 2

1 21 1 2 2 1 2

42 2 2ln ln ln

(6.3.1a)

where µ = 4π⋅10-7 [H/m] the relative permeability of air. In the case of a square, s1 = s2

= D and g = s√2, this expression becomes [Te4301, eq. 32]:

L [H]D D

d

d

D.[] =

⋅⋅ + −

2 2

20 774

µπ ln (6.3.1b)

Generally in micro-machining, the shape of the wire of a planar microcoil is notround but rectangular of thickness b and height h as shown in figure 6.3.3. Then,equation 6.3.1b becomes:

L [H]D D

b h.

b h

D.[] =

⋅⋅

++ ⋅

+−

2 40894

40 660

µπ ln (6.3.1c)

When we take D = 4.5mm and h = b = 10 µm, the self-inductance is approximately 20nH. Making h = b = 1 µm, the self-inductance only slightly increases to 30 nH. So, theinfluence of the cross section of the wire on the self-inductance of a loop is relativelysmall in our situation.

Figure 6.3.3. Schematic cross-section of a square, planar microcoil showing thegeometrical parameters and the relation between them.

Examination of the loop inductance equations for only one turn shows them tobe of the same general form regardless of the shape. If l [m] is the total perimeter of

Chapter 6

121

the figure and d [m] the diameter of the wire, then the inductance at high frequenciescan be written as [Te4301, eq. 35]:

L Hd

[ ] ln=⋅

⋅ −

µπ

θ

2

4(6.3.1d)

The quantity θ is a constant that depends upon the shape of the loop. The value of θfor a circle is 2.451, for a square it is 2.853, and for a straight line θ = 0.75. In the caseof a square with D >> d, the expression 6.3.1d, which is of course identical to eq.6.3.1b, becomes:

L [H]D D

d.

D D

. d[] =⋅

⋅ −

=⋅

⋅⋅

2 162 853

8

10

16

17 34

µπ ln ln (6.3.1e)

and the expression for a circular loop is [Te4301, eq. 29]:

L [H]D D

d.

D D

. dO =⋅

⋅ −

=⋅ ⋅

⋅⋅

µ π π π2

42 451

2

10

4

116ln ln (6.3.1f)

When the diameter of the winding is much bigger than the diameter of the wire, theratio of the self-inductance of the square loop with respect to the circular loop isapproximately:

L

L[] []

Ο Ο

≈ =

4π (6.3.1g)

Quite surprisingly, the ratio of the perimeter of the square and circular loop isidentical to the ratio of their areas. Up to this point, already two ways are found tooptimise the self-inductance of a loop covering a given area in a telemetric system. Thefirst possibility is decreasing the height of the loop as expressed by eq. 6.3.1c and theother one is increasing the perimeter of the loop, i.e. making it square (eq. 6.3.1g) or,even better, star- or meander-shaped. There is also a third interesting method: Thenumber of turns can be increased by winding turns to the centre of a coil whichincreases the self-inductance per unit area enormously. The geometrical parameters ofa rectangular planar microcoil, which is not completely filled with turns, the wirehaving a diameter d [m], and having a pitch of winding p = x1 + b [m], are shown infigure 6.3.3. For such a coil the inductance is, see eq. 6.3.1c, [Te4801, eq. 48]:

L Hs

Ns

N p

N p

s

A B

N

sN

s

N p

N p

s

A B

N

[] [ ] ln . .

ln . .

=⋅

⋅+

⋅+ −

+

=⋅

⋅+

⋅− −

+

20 2235 0 726

2 40894

40 660

2

2

µπ

µπ

(6.3.1h)

with A and B constants depending upon the wire spacing and number of turnsrespectively. For 0.3⋅p < b < 0.8⋅p, which is normally the case, the constant A isapproximately A = 2[(b/p) - 0.6]. The constant B increases rapidly when the number ofwindings is still small but it saturates to B ~ 0.336 for the higher N values (N > 10). Ingeneral, when N > 10 the influence of the factors A and B can be neglected. When wesubstitute s = (D+Di)/2, N = (D-Di)/2p, α = Di/D, and neglecting A and B, eq. 6.3.1hbecomes:


122

( )( ) ( )( )

( )( )L H

D

p[] [ ] ln . .=⋅⋅

⋅ − − ⋅+−

+−+

+

µπ

α ααα

αα

3

22

41 1

1

10 2235

1

10 726 (6.3.1i)

For a completely filled square coil α = 0 and we arrive at:

L HD

p[] [ ] ~ .0 954

3

2

µπ⋅⋅

(6.3.1j)

6.3.2 Series resistance

The series resistance of a coil, Rs, can be divided into two parts; one isindependent of and the other one is dependent on frequency. The frequencyindependent part is the DC resistance of a wire which is found by using Ohm’s law.Considering a square coil in which the spiral does not completely extend to the centre(see figure 6.3.3). When we introduce the mean radius of the spiral as a = (D + Di)/4 =s/2 = (1 + α)D/4, the total length is:

( ) [] [m] a N=D

p= −8 1

22α (6.3.2)

and the total series resistance is then:

RSCu Cu

hb h bp[] [ ]Ω = ⋅−ρ ρ α

=D2 1 2

(6.3.3a)

where ρCu [Ωm] is the resistivity of the wire’s material and, h [m], b [m] and l [m] arethe height, width and the total length of the wire, respectively. For a completely filledsquare coil α = 0 and we arrive at:

RSCu

bph[] [ ]Ω =ρ D2

(6.3.3b)

The frequency dependent part of the series resistance is caused by strong time-varying magnetic fields produced by an alternating current which passes the conductor.This field produces eddy currents which are local currents normal to the magnetic fluxand opposite to the direction of the applied current. As the frequency increases, thecurrent tends to shift to the surface of the wire, resulting in an non-uniform currentredistribution in the inner wire leading to an increase in the series resistance. Thisphenomenon of current concentration on the outer ‘skin’ of the wire is known as theskin-effect. Moreover, the magnetic field causes eddy currents in adjacent wires andtherefore Ohmic power losses due to Joule heating. So, the high-frequency current notonly alters the local current distribution but also changes the current distribution in therest of the coil. The skin depth δ [m] (or penetration depth), is given by [Ch8901]:

δρ

ωµ[ ]m =2

(6.3.4)

This effect becomes important only when the penetration depth is small,compared to the diameter of the wire, and the DC resistance has to be corrected due toredistribution of the current in the wire (see Appendix 6A). To give a rough estimate,when we incorporate the skin-effect into eq. 6.3.3a we get:

Chapter 6

123

( )RS Cu

Cu

p hb h b

p hb h b

[] [ ]Ω = ⋅−

⋅ ++

= ⋅−

⋅ ++

ρα

δ

ρα ωµ

ρ

D

D

2

2

1 1 1

2

1 1

8

1

2

2(6.3.3c)

In our telemetric system, we are working with copper wires (µr ~ 0.999 and ρ ~2⋅10-8 Ωm) of typically 10 µm in diameter at a frequency of 3 MHz. Therefore, the skindepth δ ~ 0.07/√f ~ 40 µm, which is larger than the diameter of the wire so that theinfluence of the skin-effect on the total series resistance is negligible. Nevertheless, intelemetric systems working at high frequencies, the Ohmic “DC” losses can beminimised by increasing the conducting area of the wire, but this gain is limited by theskin-effect. Moreover, increasing the width of a wire will decrease the maximumnumber of turns possible per unit area. So, a careful choice of the dimensions of thecoil and working frequency of the circuitry is necessary.

6.3.3Parallel resistance

The parallel resistance is caused by the finite resistance of the insulating layer onwhich the coil is placed. The power loss in this layer becomes important when thisresistance approaches the admittance of the capacitor, XC = 1/ωC. The parallelresistance is:

RPOx Oxx

b bD

p[] [ ]Ω = ⋅

−ρ ρ

α2

2 21=

x2 (6.3.5)

In most practical situations, the influence of the parallel resistance is far less than thatof the series resistance. We will disregard it too.

6.3.4 Parasitic capacitance

The parasitic capacitance consists of three different capacities (figure 6.3.4a,b);the capacitance between the coil’s turns Ctt, the capacitance between turns andsubstrate Cst, and the capacitance between contact pads and substrate Ccs. Theelectrical equivalent is shown in figure 6.3.4c. These capacities might be approximatedby that of parallel plates and, in case of the unfilled circular coil and using eq. 6.3.2, theexpression becomes:

( )( )C F

h D

Nx

hD

x

p

Dtt [ ] =⋅ ⋅ −

= ⋅ + −−

ε 4 21

4

11 1

εα

α (6.3.6a)

( )C [F]b

x

bD

x pst =⋅ ⋅

= ⋅ −ε ε

α

2

2

2

21 (6.3.6b)

C FA

sc

cp[ ] =

⋅2

2

εx

(6.3.6c)

where ε = ε0εr = 8.84·10-12⋅εr [F/m] is the permitivity of the material between turns orbetween turn and substrate (εr = 1 for air and εr = 3.8 for SiO2), x1 and x2 are thedistances between two adjacent turns and between turns and substrate respectively,and Asc is the contact pad area. The factor N in the expression for Ctt is caused by the


124

linearly decreasing voltage on going from the outside of the spiral to its centre. Thetotal parasitic capacitance Cpar of the receiver coil is then given by:

C F C C Cpar tt st sc[] [ ] = + +14

1

2(6.3.6d)

In our microcoil, the height of the wire h, and the distance between two adjacent turnsx1, are almost the same, but x2 is much smaller. Therefore, Ctt can be disregarded.Moreover, the contact pad is typically much smaller than the coil area thus Csc is alsonegligible so that we only have to consider Cst. In the case of a completely filled coilseparated from a conducting substrate by only a relatively small distance, equations6.3.6b and 6.3.6d then give:

C FbD

x ppar [ ] =ε 2

24(6.3.6e)

Sometimes, the parasitic capacitance caused by the conducting ground plane,such as the silicon wafer, is much too high. In such cases, the parasitic capacitance canbe lowered drastically by replacing the oxidised silicon wafer by a glass substrate.Then, the capacitances to the substrate will be neglected with respect to thecapacitance between the turns and only Ctt has to be calculated. Equation 6.3.6a is asimple parallel plate approximation to estimate this capacitance. However, because thedistance between the turns is typically of the same order as the diameter of the turns,eq. 6.3.6a cannot be used since electrical fringing, i.e. parasitic fields, start to play arole. A better estimate may be that of a capacitance between two long, parallel circularconducting wires of diameter d [Pe88]:

(c)

Figure 6.3.4. Cross-section of a planar coil showing the parasitic capacitances (a)between the turns Ctt, between turns and substrate Cst, (b) between contact pads andsubstrate Ccp; (c) electrical equivalent of the parasitic capacitances.

( )C F

D

Np

d

p

d

tt [ ] =⋅ ⋅

⋅ +

−

π ε -

ln

4

12

(6.3.6f)

(a) (b)

Chapter 6

125

where, p is the pitch (the distance between the axes of two adjacent turns).

6.3.5 Quality-factor

The quality-factor or efficiency of the receiver coil of the telemetric system cannow be calculated and optimised with respect to geometrical dimensions using theexpressions just found (equations 6.3.1i, 6.3.3a, and 6.3.6b,d). After all, a coil withturns to the centre maximises not only the self-inductance but also the series resistance.Since the self-inductance and the series resistance do not depend on the number ofturns in the same way, there is a configuration which leads to an optimum ratio of self-inductance to resistance. This configuration is that of a coil with a small opening in thecentre, i.e. the turns are not wound right up to the centre, which is expressed by thefactor α. For an insulator with a negligible resistance Rp, this intrinsic Q-factor is (eq.6.2.1c):

( )( ) ( )

( )( )( )

QR

L

C

hbp

D

Dx

bpn

is par

[]

. .

=

=−

⋅− +

−+

−+

+

1

1

1 1

10 2235

1

10 726

2 2

2

ρ αµ α

πεαα

αα

(6.3.7a)

This expression can be optimised with respect to the factor α, i.e. the ratiobetween the inner diameter and outer diameter of the planar microcoil. If we do so, wefind that this maximum is reached for α = 1, i.e. only one turn. This makes sensebecause the parasitic capacitance decreases linearly with the total length of the spiral tozero. Of course, α = 1 is not practical because it means that D = Di, i.e. there is nospiral at all.

As already mentioned, in our application we would like to tune both the receiverand transmitter units for the same frequency to couple energy as much as possible.Therefore, extra capacitors are placed in parallel with both coils. Such an extra tuningcapacitor is possible as long as the intrinsic resonance frequency is lower than thedesired operating frequency. Because we have set the operating frequency of thereceiving unit, the main parameter which determines the quality-factor in this case isthe ratio of the reactance of the self-inductance to the series resistance as found by eq.6.2.1b.

( ) ( )( )

( )( )Q

L

R

Dhb

pno

o

S

o= =− +

−+

−+

+

ω µω απρ

αα

αα

1

4

1

10 2235

1

10 726 . . (6.3.7b)

This ratio has its maximum at δ(L/R)/δα = 0, for α ~ 1/4.

6.3.6 Mutual inductance

The mutual-inductance between two coils, M [H], (see fig. 6.3.2) depends on theself-inductance of the transmitter coil Lt and receiver coil Lr and a parameter k whichis a measure of the coupling and depends on the relative position of the coils. When therelative position is such that lines of flux from one inductance link with turns of the


126

other, the two inductances are said to be inductively coupled and mutual-inductanceexists between them. Mutual-inductance may be defined in terms of the number of fluxlinkages in the second coil per unit current in the first coil, or vice versa. However, amore practical definition of the mutual-inductance is the voltage induced in the secondcircuit when the current in the first circuit changes at a unit rate. If the current flowingin the first circuit is sinusoidal, then the voltage induced in the second circuit is:

( ) ( )V M

dI t

dtM

d I j t

dtj MI j tr

t= = =sin( )

cos( )ω

ω ω (6.3.8)

Another quick method to find the mutual-inductance between two circuits is tomeasure first the self-inductance of a coil alone. Now, when a second coil is placed inthe neighbourhood of this coil, the inductance of the first coil will change with a valueidentical to the mutual-inductance. The maximum value of mutual inductance that canexist between two coils is √(LrLt), which occurs when all the flux of one coil links withall the turns of the other. The ratio of the mutual inductance actually present to themaximum possible value that can occur is called the coefficient of coupling and iswritten as:

kM

L Lt r

= (6.3.9a)

In general, the coefficient of coupling between two coils will only be close to oneif the coils are quite close to each other. Close in this respect means that the distancebetween the two coils is smaller than their size. When both coils are tuned at the sameresonance frequency, the coupling is at its maximum. For this so-called criticalcoupling kc, eq. 6.3.9a turns into:

kQ Qc

t r

=1

(6.3.9b)

When k > kc, the resonance curve has two peaks and when k < kc, the resonancecurve is lower and smaller than in case of k = kc. In the case of two inductively coupledLC-circuits, eq. 6.2.1b should be modified. For such a “band-filter” the next expressionholds:

Qoo= ⋅

ωω

2

2 (6.2.1d)

The factor √2 is caused by the extra LC-circuit with respect to eq. 6.2.1b.

With the knowledge of the self- and mutual-inductances of the receiver andtransmitter units, we are now able to calculate the energy transfer between bothcircuits. When the power in the transmitter circuit Pt is transferred by way of mutual-inductive coupling into the receiver unit, then the power received is:

P [W] PM

LMIr t

tt= ⋅ =

1

22 (6.3.10)

6.4 Design and fabrication of planar microcoils

In a telemetric system the major requirement is a high power transfer, so that asmuch power as possible is available for the implant. This means that the receiver coilshould have a high Q, i.e. a high inductance value, a low series resistance and a lowcapacitance. An important problem with a microcoil is the decrease of the inductance,due to a smaller area. This decrease can be compensated for by a large number of

Chapter 6

127

turns; however, the relatively long length of a spiral conductor required for a largenumber of turns produces a higher series resistance. For this reason it is desirable touse a low-resistivity coil material. In our case we decided to use copper, which has aresistivity of 2.0 µΩ⋅cm (for thin films).

For an experimental evaluation we fabricated two types of coil, each 4.5 mm indiameter and 112 turns which results in a pitch of 20 µm. The first planar coil (I) of 10 µm wire width, is 1 µm thick on 1.9 µm thick silicon oxide. It is wound to the centreto maximise the self-inductance. The second coil (II) differs from the first in height ofthe conductor (11 µm), width of the wire (14 µm), and thickness of the oxide layer (1µm). All three influence the quality of the coil at resonance, because an increase inconductor height and width decreases the series resistance and the decrease in insulatorthickness increases the parasitic capacitance.

Two different fabrication techniques have been used to make the two types ofcoil: sputtering with lift-off (type I) and sputtering + electroplating (type II). The mainadvantage of electroplating is the possibility of high structures. Fabrication of coil typeII is schematically shown in figure 6.4.1. The process scheme was as follows: Anadhesion layer of 20 nm Cr was sputtered on 1 µm silicon oxide, followed by 0.5 µmof Cu (step 1). Two layers of 7.5 µm thick resist (Ma-P275, Micro allresist) were spunon the wafer with a 1 min. 90°C baking period in-between (step 2).

Figure 6.4.1. Fabrication process for electroplated planar coil (II):An electroplating set-up was made using a commercial Cu electrode and electrolyte(2.25 M H2SO4, 0.28 M CuSO4.5H2O with surface active agents). A layer of 11 µmof Cu was grown by electroplating for 30 min. (step 3) using the approximate formulaof the growth rate, δh [µm/min], at room temperature:

δh j= ⋅0 016. (6.4.1)

After the deposition of thecopper seed layer and

spinning of the photoresist,this pattern can be used asa mold for

the electroplating of acopper coil.

Finally, the seed layer is locallyremoved by ion beam etching.


128

where a current density j = 20 mA/cm2 was used. After removal of the resist layer, theseed layer of sputtered Cr + Cu was removed at unwanted locations by ion beametching (step 4). In figure 6.4.2, SEM photos of the electroplated coil are shown. Thesurface of the electroplated copper is rough due to the quite high current density.

(a) (b)

Figure 6.4.2. SEM photographs of (a) top-view of the centre, and (b) cross-section ofthe electroplated coil (II).

6.5 Results and Discussion

The geometrical parameters of the coils and the calculated self-inductance (eq.6.3.1j), DC resistance (eq. 6.3.3b) and parasitic capacitances (eq. 6.3.6e) aresummarised in table 6.5-1. Figure 6.5.1 shows how the characteristics change with thefilling of the coil; expressed with the variable α = Di/D. In all cases, the resistance andcapacitance decreases as α increases. However, the self-inductance shows a maximumat α ~ 0.14. The values at α = 0 are the ones given in table 6.5-1. Figure 6.5.2 showsthat the intrinsic resonance frequency and quality factor increase with increasing α.When the frequency of the transmitter is set to a constant value, for example 3 MHz,the quality factor no longer rises with α. Instead it has a maximum at α = 1/4independent of the real work frequency. Therefore, this factor is now called theoperational or work quality factor. In general, the higher the intrinsic resonancefrequency, the higher the frequency of the transmitter we can use and the better theperformance with respect to the collected energy. Tuning of the receiver unit with anextra parallel capacitance will decrease the performance under all circumstances. Theparasitic capacitance of the planar receiver coil is mainly determined by the capacitancebetween the turns and the substrate. Therefore, increasing the thickness of theinsulating oxide will also increase the intrinsic quality factor. When a very high intrinsicquality factor is desired, then a glass substrate is useful.

Design Parameter Coil I Coil IID Outer diameter 4.50 mm 4.50 mmDi Inner diameter 0 mm 0 mmα Ratio inner-outer diameter (= Di /D) 0 0N Number of turns 112 112l Length of conductor (= 2ND) 1.008 m 1.008 mp Pitch between adjacent turns (= D/2N) 20 µm 20 µm

Chapter 6

129

b Width of turns 10 µm 14 µmh Height of turns 1 µm 11 µmx2 Thickness oxide insulation 1.9 µm 1.0 µmAcp Contact pad 1x0.5 mm2 50x50 µm2

Calculated parameters Coil I Coil IILr Self-inductance (eq. 6.3.1j) 23.7 µH 21.6 µHCtt Capacitance between turns 0.008 pF 0.15 pFCst Capacitance turns to substrate 179 pF 477 pFCsc Capacitance contact pads to substrate 17.5 pF 0.045 pFCpar Total parasitic capacitance (eq. 6.3.6e) 54 pF 119 pFρ Resistivity of thin film Cu 2.0 µΩcm 2.0 µΩcmRdc DC Resistance (eq. 6.3.3b) 2025 Ω 131.5 Ωf0 Resonance Frequency (eq. 6.2.2a) xxx 2.98 MHzQ Quality factor (eq. 6.2.1c) xxx 3.08Zo Characteristic impedance (eq. 6.2.3c) xxx 426 Ω

Measured parameters Coil I Coil IIRdc DC Resistance 2060 Ω 129 Ωf0 Resonance Frequency xxx 3 MHzQ Quality factor: eq.6.2.1d and fig.6.5.4b xxx 1.8Zo Characteristic impedance: fig. 6.5.4b 405 Ω

Table 6.5-1. Parameters of the two coil configurations

Figure 6.5.1. The dependency of the coil parameters, L, R, C on the ration of theinner- and outer diameter of the coil, α.


130

Figure 6.5.2. The relationship between the operational quality factor at a constantfrequency of 3MHz and the intrinsic resonance frequency with its intrinsic quality-factor as a function of the ratio of the inner- and outer diameter of the coil, α.

The total impedance of the two configurations was calculated and thecorresponding Bode-diagrams are shown in figure 6.5.3. Coil I does not show aresonance peak. In contrast, coil II exhibits a peak at 2.98 MHz, mainly due to themuch lower series resistance.

Impedance measurements were performed using an HP 4194A Impedance / GainPhase Analyser. The results are shown in figure 6.5.4. The 1 µm high coil does notshow inductance, due to the relatively high series resistance 2060 Ω (fig. 6.5.4a). Theelectroplated coil II clearly shows inductance properties (fig. 6.5.4b). The resonancefrequency is 3 MHz and the DC resistance is 129 Ω. The measured maximumimpedance (405 Ω) is close to the theoretical value of 426 Ω, calculated for a DCresistance of 129 Ω. At 10 MHz there seems to be a second maximum in impedance.This might be caused by a difference in resonance frequency between the receiver andtransmitter coil.

Figure 6.5.3. Calculated Bode-diagrams for two coil configurations consideringparallel resistance negligible, width of wire is 10µm for coil (I), 14µm for coil (II),height of coil (I) is 1µm, and of coil (II) is 11µm.

Chapter 6

131

Figure 6.5.4. Impedance measurements of (a) 1 µm high coil (I), and (b) 11 µm highcoil (II).

The dependence of the energy transfer on the resistive load of the receiver circuitwas measured using the transmitter-receiver set-up shown in figure 6.5.5a. For firsttests, a germanium diode (type AAZ15, UD = 0.3 V) was used for single-sidedrectification, in order to maximise the output voltage which drives the electronics. Ofcourse, when the voltage is high enough, double-sided rectification is desirable todouble the energy transfer to the load. Values of the several components are alsodisplayed in this figure. The circuit was tuned to the resonance frequency of the samplecoil. The output voltage and output power of the receiver coil Vr, as a function of theload can be seen in figures 6.5.5b,c. Two different distances between the transmittercoil and the receiver coil, 1 mm and 3.5 mm were used. The energy transfer can be ashigh as 2 mW. At a closer distance, the coupled flux is higher and, therefore, the

(a)

(b)


132

voltage and power transfer is increased. The output voltage increases with increasingload resistance because the current through the load decreases. The power transfer tothe load increases when the load is decreased from 100 kΩ down to 400 Ω. The powertransfer will be maximum when the load is adjusted to the characteristic impedance ofthe receiver LC-circuit, i.e. 426 Ω.

(a)

(b)

(c)

Figure 6.5.5. (a) Transmission circuitry; (b) Output voltage as a function of the load;(c) Power dissipated by the load.

Chapter 6

133

Using formula 6.3.8 and the measurements shown in fig. 6.5.3, the mutualinductance M and the coupling factor k can be approximated. These values arepresented in table 6.5-2. The values of the coupling factor seem quite low but areconsidered normal for this type of weak coupling.

Coil Separation M k1.0 mm 0.6 µH 0.0363.5 mm 0.3 µH 0.018

Table 6.5-2. Approximation of mutual induction M and coupling factor k using coilII.

A transmitter-receiver set-up was made using the receiver circuitry fromfig.6.5.6. For first tests only one germanium diode (type AAZ15) was used forrectification. The transmitter was placed at a distance of 1 mm. Using the electroplatedcoil, it was possible to have at least 1 mW and 2V available on the output terminals. Apressure regulator microactuator, based on electrochemical actuation [Ne9601], wasconnected and a constant current of 28 µA was driven through the actuator. Theactuator dissipated 50 µW. Although the efficiency is low, the power transmission ishigh enough to power the electronics.

Figure 6.5.6. Transmission set-up for actuation of the pressure regulator.


134

6.6 Feedback controller

To summarise once again: The electrochemical actuator is used to control theeye pressure by deflecting a membrane which changes the eye fluid resistance of adrainage tube. The drainage tube is replacing the natural intraocular eye fluid outflow.The eye pressure is monitored with the help of a pressure sensor upstream the valve.By balancing the removal of eye fluid with the generation of this fluid, a constant eyepressure is obtained. This can be achieved by combining a feedback controller, with thesensor and actuator to form a regulating system. The energy needed for the regulatingsystem is delivered by two inductively coupled coils as was presented above. To testthis regulating system, an electrochemical macrocell was constructed, and the pressuregenerated in the cell was controlled. In the following paragraphs, the regulating systemis considered.

6.6.1 Experimental

To start, it is necessary to have a hermetically closed macrocell. In principle thisis not difficult; however, we would like to have a cell which can be easily modifiedwith respect to the electrode materials and electrolytes, and it also should be suitable tomonitor small amounts of produced gases. The final version incorporated some basichigh vacuum components from the vacuum technology, as shown in fig. 6.6.1. Astainless steel blank flange was machined (a hole was drilled) to mount the pressuresensor. The pressure sensor was connected by a Viton ‘O’ ring and Teflon tape toensure a leak-free construction. The electrodes were inserted inside a standard K40“window-glass” that has a stainless steel support ring. Holes, 1mm in diameter, weredrilled in the glass through which the electrodes were inserted. The residual gaps werefilled with vacuum glue (Torrseal) to prevent leakage. After filling the cell with theelectrolyte, all the parts were assembled and closed tightly by a flange clamp.

The gas evolution from the electrolysis can be measured using a pressure sensor,sensing at constant volume. The pressure sensor is a miniature vented gauge transducerof Data Instruments (MM10013). The sensor in stainless steel housing is

Figure 6.6.1. Basic set-up of the closed macrocell with standard high vacuumcomponents.

Chapter 6

135

temperature compensated and able to measure pressures up to 7 bar (100 PSI). It usespiezoresistive elements driven by a stabilised 5 V supply from an external voltagesource (for example, Voltcraft TNG35). To minimise the ripple from this source, avoltage regulator is utilised (LM7805). The voltage monitor of the pressure sensor is aprogrammable unit from Hewlett Packard (HP34401a) with an IEEE488 (GPIB) portto enable data exchange. Data from the pressure sensor is recorded with HP-VEEsoftware.

The electrodes tested for the electrolysis of water (gold, platinum) were wires of1 mm diameter, area ca. 0.3 cm2, purity 99.99% from Goodfellow.

A constant cell pressure can be achieved by combining a feedback controller,with the sensor and actuator to form a regulating system. In fig. 6.6.2a, such a totalregulating system is shown and in fig. 6.6.2b the electronic circuit of the feedbackcontroller is given in detail. It consists of three sections: a 100 fold amplifier for thesensor signal (OP27), an integrator built around a comparator to pre-set the cellpressure (CA3140), and a current amplifier to feed the electrolysis system (BD139).After amplification, the signal is fed to the CA3140 where it is compared to therequired pressure. This is achieved by a trim-potentiometer “Pressure Set” which isable to maintain a pressure inside the cell of up to 7 bar. The signal is integrated tofilter out any noise. The integration time is controlled by a replaceable capacitor, in thiscase 10nF. The capacitor across the base emitter of the current amplifier BD139 ismeant to prevent parasitic oscillations.

(a) (b)

Figure 6.6.2. (a) Schematic of the feedback controller to stabilise the cell pressure.(b) Electronic circuit of the feedback controller to settle the cell pressure.


An electrochemical experiment with the macrocell was performed using two goldelectrodes; by applying a constant current, oxygen and hydrogen gases are created.The result is given in figure 6.6.3. The plot shows three increases; the first is caused bya current through the cell of 500 µA for 200 s (3.8 V), the second 100 µA for 1000 s(3.0 V), and the third 20 µA for 5000 s (2.5 V). The total charge passed through thecell was the same for all cases. Clearly, the increase in pressure is identical for all threecases. The drop in pressure after switching the electrolysis current off is very slow witha response time more than 100 days. This means that the leakage from the cell is slight.Although this pressure, in open circuit condition, decreases very slow as it is desired,


136

unfortunately gold electrodes cannot be used because the actuator will not bereversible: in the pressure reduction state the oxygen and hydrogen gases will not bereduce to water but just the electrodes from where they are generated will be changed.

Au / Au / water

1.641.6451.65

1.6551.66

1.6651.67

1.6751.68

1.685

0 200 400 600 800 1000 1200 1400 1600time [min]

Pre

ssur

e [1

05 P

a]

500 µA, 200 s

100 µA, 1000 s

τ > 100 days

20 µA, 5000 s

Figure 6.6.3. Behaviour of the EC cell with gold electrodes. The open circuit decay isvery long, over 100 days.

The influence of the catalytic action of the platinum [At9801, Wh8301] wastested by using two gold electrodes and a piece of Pt inserted in the macrocell (it isfixed by glue on the glass). The result is shown in fig. 6.6.4. Comparing this fig. withfig. 6.6.3, the open circuit decay decreases significantly from more than 100 days to ca.12 days. This drop is thought to be caused by the catalytic action of the platinum. Bycomparing the voltage applied to the cell for the case of Pt-Pt (not shown) and Au-Auelectrodes, respectively, it is observed that in the case of the Au-Au electrodes thepotential is ca 0.5 V bigger in order to get the same current as for Pt-Pt electrodes.This extra voltage is related to the activation energy.

Au / Au / water

1.655

1.6575

1.66

1.6625

1.665

1.6675

1.67

500 600 700 800 900 1000time [min]

Pre

ssur

e [1

05 P

a] t > 100 days

Au / Au / water with a Pt piece

2.6

2.7

2.8

2.9

3

3.1

3.2

1200 1400 1600 1800 2000 2200 2400 2600time [min]

Pre

ssur

e [1

05 P

a]

τ ~ 12 days

Figure 6.6.4. The EC cell with Au-Au electrodes and a Pt piece. (a) The open circuitdecay decrease significantly from more than 100 days to (b) ca. 12 days, when thecatalytic Pt is in the system.

To maintain a constant cell pressure, the feedback controller was used. Theresult of the feedback controller upon the pressure inside the cell is shown in fig. 6.6.5.A step of 16.8 mbar was set to the comparator and the regulator actuated theelectrochemical cell in such a way that the desired pressure was obtained and keptconstant. Clearly, the cell pressure remains constant. After eight steps made with thecontroller the feedback is switched off and the pressure begins to drop. By setting avalue of the pressure the feedback circuit was controlling automatically the

τ > 100 days

(a) (b)

Chapter 6

137

electrochemical cell pressure: the circuit is stabilising the cell pressure much better thanwould be possible without the controller.

Pt / Pt / waterwith regulator circuit

2.8

3

3.2

3.4

3.6

3.8

4

0 10 20 30 40 50 60 70 80time [min]

Pre

ssur

e [1

05 Pa]

steps of 0.17*105 Pa

Figure 6.6.5. Output of the cell pressure with feedback controller.

6.6.3 Pt-O2-H2 reversible electrochemical actuator

As it was above shown, platinum acts as a catalyst for hydrogen and oxygen toreact back to water. This effect can be used for a reversible actuator. In fig. 6.6.6 a cellis sketched that has two rooms, each room has an electrode, and between the roomsthere is an opening. The purpose of this opening is to connect the two electrodes viathe electrolyte (i.e. allows a high ionic flux), and to minimise intermixing of the twotypes of gases. The diameter of the opening is in such a way chosen that the cloggingof the opening by gas bubbles, due to supersaturation of the electrolyte solution, isminimal [Co9401]; otherwise the electrical resistance of the cell increases in anuncontrolled manner. One room has a flexible membrane that deflects when a gaspressure is generated. By applying a current to the platinum electrodes hydrogen andoxygen gas will evolve. Thus in each room a different type of gas exists.

Fig. 6.6.6. Sketch of a reversible electrochemical actuator using two platinumelectrodes, oxygen and hydrogen gases. The reduction of the gas pressure is based onthe catalytic action of platinum on hydrogen and oxygen to react back to water.

Due to the gas pressure the membrane will deflect. When the current is reversed,oxygen and hydrogen are again generated but this time in the opposite room: hydrogenin the room where in a previous electrolysis state oxygen was created. Since there is a

Pt Pt

flexible membrane

Cross-section

Feedback ON

Top view

Feedback OFF


138

Pt electrode, oxygen and hydrogen will react together forming water, so the pressurewill decrease. When the cell is in the open circuit state, the pressure will be maintainedas well as the membrane deflection because each type of gases are confined in separaterooms. Hence, this system forms a chemically reversible electrochemical cell. Bychanging the dimensions of the channel-connection between the rooms it is possible tocontrol the leakage rate of oxygen and hydrogen gases between both rooms.Moreover, by changing the concentration (molarity) of the electrolyte we are able tocontrol the electrical conductance of the channel-connection without changing theleakage of the generated gases. In this way we are able to control the performance ofthe cell.

Qualitative experiments done with a macrocell by measuring the volume of theoxygen and hydrogen gases show that this principle works and the gases’ volume isreduced by reversing the current.

Chapter 6

139

6.7 Conclusions

This chapter reports fabrication, testing and characterisation of an electroplatedplanar microcoil, which is used as the receiver coil in a telemetry system consisting oftwo inductively coupled coils. Calculations on the electrical parameters of the receivercoil have shown that for coils with a diameter of 4.5 mm and wire of 14 µm indiameter, a few mW can be transmitted. A gain in quality factor can be reached; (i) bymaking high structures, resulting in a decrease of the series resistance of the coil; (ii)by increasing the thickness of the insulator, or better, replacing the silicon substrate bya glass substrate; or (iii) by increasing the number of turns since this increases the self-inductance enormously. When the frequency of the transmitter is fixed, then a little isgained by choosing a ratio between the inner- to outer diameter of 0.25. When a glasssubstrate is used, the parasitic capacitance between the turns has to be taken intoaccount. The model described for the electrical equivalent of the receiver coil predictsvalues of the intrinsic resonance frequency, intrinsic quality factor, series resistance ofthe coil, and the power transfer from the coil into the load, which are in goodagreement with measurements.

The results of an electrochemical macrocell have been also presented. It consistsof a hermetically sealed cavity filled with water, two electrodes, and a pressure sensor.When a current is applied to the two electrodes, oxygen and hydrogen gas wereformed. The increase of pressure is directly related to the charge driven through thecell. It was seen that the pressure drops relatively fast (a few days) to its initial valuewhen platinum electrodes are used. The reason is that platinum has a low activationenergy for the reaction of hydrogen with oxygen gas to water. It is shownexperimentally that the decay of the pressure after switching the current off (thepassive state of the cell) indeed improves by the use of gold electrodes. However, adisadvantage of this higher activation energy for gold is that a higher potential isnecessary to drive the cell.

The feedback circuit automatically controlled the cell pressure: by setting a valueof the pressure, the regulator actuated the electrochemical cell in such a way to givethe desired pressure.

When the efficiency of the electrolysis process is high (i.e. low ohmic losses,overpotentials), it can be said that the EC actuator is able to create and maintain a veryhigh pressure inside a closed cell. Pressures beyond 7 bar have been reached withoutany difficulties. Due to this characteristic, the electrochemical cell can openpossibilities far beyond the reach of other types of actuators, such as the thermalactuator, piezo actuator, or magnetic actuator.


140

Appendix 6A

At high frequencies, when the skin-effect becomes important, i.e. when thepenetration depth, δ, is not negligible compared to the half width of the wire, b/2, theresistance of a straight wire has to be corrected [Bu26]:

R Rsac dc[ ]Ω = +

1

1

482

4b/

δ(6A.1)

For two or more adjacent wires, the current distribution in one wire is affectedby the magnetic field produced by the adjacent wire as well as by the magnetic fluxproduced by the current in the wire itself. This phenomenon called the proximity effectcauses the resistance to be greater than in case of simple skin-effect. A formula whichholds for a coil with only two turns is [Lo9201]:

R Rp p p pprox dc= +

+

+

+

+

+

11

481 12

1

6

1

18

1

402 2 2 2 2

2 2 4 6 8b b b b b/ / / / /

δ.....

(6A.2)

where p is the pitch, the distance between the axes of two adjacent turns. If b/2

d << 1,

only the first term of the infinite series is relevant, where d is the diameter of the wire.For a single layer disc coil with many spaced turns (N > 35), an approximate formulafor the total resistance is [Bu26]:

R R utotal dc[ ]Ω = +

+ ⋅

1

1

481 122 2

2 2b b/ /

δ δ(6A.3)

The variable u depends on rwD

(see table 6A-1), where D is the diameter and rw is the

winding depth of the coil. For a planar coil wound to the centre, the constant u = 8.64.The proximity losses can be neglected if d > 2b [Lo9201].

An exact calculation of the frequency-dependent resistance is difficult. For moreaccurate expressions, numerical methods are used [Wa8701].

rDw u r

Dw u

0.000 3.290 0.250 4.7490.025 3.315 0.275 5.0410.050 3.373 0.300 5.3640.075 3.459 0.325 5.7180.100 3.567 0.350 6.1040.125 3.702 0.375 6.5230.150 3.859 0.400 6.9680.175 4.042 0.425 7.4360.200 4.251 0.450 7.9110.225 4.486 0.475 8.6380.250 4.749 0.500 8.638

Table 6A-1. Correction parameter u for calculation of the effective resistance usingeq. 6A.3. It depends on the winding depth rw and diameter D of the coil. Values arefrom [Bu2601].

Chapter 6

141

References

[At9801] P.W. Atkins, Physical chemistry, Oxford University Press, 1998.[Bu2601] S. Butterworth, Effective resistance of inductance coils at radio frequency, Parts

I&II, Experimental wireless and the wireless engineer, April 1926, p. 203-210 &309-316.

[Ca8301] D. Cahana, A new transmission line approach for designing spiral microstripsinductors for microwave integrated circuits, in IEEE MTT-S Int. MicrowaveSymp. Dig. (Boston), 1983, p. 245-247.

[Ch8901] D.K. Cheng, Field and wave electromagnetics, Addison-Wesley, USA, 1989.[Co9401] B.E. Conway, J.O’M. Bockris, R.E. White (eds), Modern Aspects of

Electrochemistry, No. 26, Plenum Press, 1994, p.105-163.[Du6101] J.M.C. Dukes, Printed Circuits, their design and applications, MacDonald,

London, 1961, p. 120-135.[Gr7301] F.W. Grover, Inductance calculations: working formulas and tables, Dover

publications, 1973.[Gr7401] H.M. Greenhouse, Design of planar rectangular microelectronic inductors, IEEE

Transactions on parts, hybrids, and packaging, PHP-10(2), 1974, p. 101-109.[Ka6901] R. Kadefors, E. Kaiser and I. Petersen, Energizing Implantable Transmitters by

Means of Coupled Inductance Coils, IEEE Transactions on Bio-medical Eng. 16,1969, p.177-183.

[Ka8401] K. Kawabe, H. Koyoma, K. Shirae, Planar inductors, IEEE Transactions onMagnetics, MAG-20 (5), 1984, p. 1804-1806.

[La8801] D. Lang, Broadband model predicts S-parameters of spiral inductors, Microwaves& RF, January 1988, p. 107-110.

[Le9301] J.F. Lehmann, Therapeutic Heat and Cold, Williams & Wilkins, 1982.[Lo9201] A.W. Lotfi, F.C. Lee, Proximity losses in short coils of circular cylindrical

windings, 23rd IEEE Power Electronics Specialists Conference, Toledo, Spain,June 29-July 3, p. 1253-1260, 1992.

[Na9501 M. Nardin and K. Najafi, A multichannel neuromuscular microstimulator with bi-directional telemetry, Tech. Digest Int. Conf. Solid-State Sensors and Actuators,Transducers’95, 25-29 June, Stockholm, Sweden, p. 59-62, 1995.

[Pe8801] E. Pettenpaul, H. Kapusta, A. Weisgerber, H. Mampe, J. Luginsland, I. Wolff,CAD models of lumped elements on GaAs up to 18 GHz, IEEE Transactions onMicrowave Theory and Techniques, 36 (2), p. 294-306, 1988.

[Pu9501 R. Puers, Linking sensors with telemetry: impact on the system design, Techn.Digest Int. Conf. Solid-State Sensors and Actuators, Transducers’95, 25-29June, Stockholm, Sweden, p. 47-50, 1995.

[Re9301] J.R. Reitz, F.J. Milford and R.W. Christy, Foundations of ElectromagneticTheory, Addison-Weshley Pub. Company, USA, 1993.

[Te4301] F.E. Terman, Radio Engineers Handbook, McGraw-Hill, London, p46-78, 1943.[Wa8701] P. Waldow, I. Wolff, Dual bounds variational formulation of skin effect problems,

in IEEE MTT-S Int. Microwave Symp. Dig. (Las Vegas), p. 333-336, 1987.[Wh2801] H.A. Wheeler, Simple inductance formulas for radio coils, Proc. IRE, 16 p. 1398-

1400, 1928.[Wh8301] R.E. White, J.O’M. Bockris, B.E. Conway, Modern Aspects of Electrochemistry,

no. 15, Plenum Press, 1983.

Chapter 7

143

The main results of the research presented in this thesis aresummarised in this chapter. Based on the experiments, thefeasibility of the electrochemical microactuator as animplantable eye pressure regulator is discussed. Somesuggestions for further research are given.

This thesis describes the design, fabrication and testing of a new type ofmicroactuator for medical applications. Observations and conclusions for eachcomponent of a total system are presented.

A total system for the eye pressure will consist of an implantable microsystemand an external remote unit. The microsystem will contain an active microvalve, amicromechanical intraocular pressure sensor, and an electronical system. The activemicrovalve (that adjusts the eye pressure) is composed of an electrochemicalmicroactuator and a flow channel. The electronical system will consist of an electronicfeedback controller (that will stabilise the eye-pressure to the requested value byactuating the electrochemical actuator), a microprocessor (for data exchange) andwireless energy supply (that transmits energy to the actuator by inductive coupling of apair of coils), that includes a microcoil and a receiver system.

7.1 The active microvalve

The electrochemical microactuator, based on reversible electrochemicalreactions, is used to adjust a fluid pressure by changing the fluid resistance of a flowchannel by means of a deflecting membrane. The required pressure difference acrossthe active microvalve is within a range of 1000 to 4000 Pa over atmospheric pressurefor low liquid flows, 1 to 3 µl/min. The actuator consists of an electrochemical cell anda flexible membrane that deflects because of the pressure of oxygen gas generated byelectrolysis (chapter 3 and 5).

GENERAL CONCLUSIONS

7

General discussions and conclusions

144

7.1.1Electrochemical cell

The design, fabrication and operation of an electrochemically actuatedmicrovalve were described in chapter 5. In the prototypes a Cu/aq.1M CuSO4/Ptelectrodes-electrolyte system was used in an electrochemical cell with dimensions2x2x1 mm3, made by standard silicon micromachining.

The time to build-up the gas pressure is mainly determined by the charge(current) passed through the cell. With gas generation, the response is not immediate(for currents up to 50 µA) due to the solubility of oxygen in the solution and the timerequired to develop pressure in an approximately constant volume. The decrease of thegas pressure follows the diffusion-controlled reaction of the electrochemical reductionof oxygen, which is relatively slow.

Under open circuit conditions the gas pressure built up in a previous electrolysisstep should, in principle, remain constant. In practice, this ideal situation was verydifficult to achieve; the gas produced at one electrode (e.g. O2 at Pt) may react at theother electrode (e.g. Cu). A way to prevent this is to protect the second electrode witha semi-permeable membrane like Nafion®. This approach was investigated, asdescribed in chapter 5. A pressure decay time of ca. 6 min was observed for a firstprototype. This decrease in pressure may be explained by the poor protection of thecopper electrode against oxygen due to a combination of the following reasons: (i) theNafion membrane is thin so oxygen diffuses rapidly through it. However, a very thickNafion will have a higher electrical resistance increasing the total cell resistance so ahigher voltage will be necessary to drive the actuator. Thus, a compromise has to bemade; (ii) a poor adhesion of Nafion to the substrate. With the use of the polyimidemesh Nafion adhesion becomes better; the area of Nafion is decreased, while itsthickness is increased leading to an increase of the decay time in open circuitconditions from 6 min (in the first prototype) to 93 min (for the second type).

When the microcell was driven at 1.6 V and currents below 50 µA, pressures of2 bar could be easily obtained. This implies that relatively large pressures (up to tens ofbars) can be reached with only low power consumption (in the µW range).

The results obtained with the microcell are still not sufficiently satisfactory as animplantable eye pressure regulator due to the irreversible loss of the oxygen gaspressure. However, the research presented here indicates that micromachinedelectrochemical actuators can be fabricated, and have attractive properties which areuseful for applications with less stringent requirements.

7.1.2Deflecting membrane

Two types of deflecting membrane were processed (chapter 4.2): 1.2x1.2 mm2, 1µm thick flat and corrugated low-stress LPCVD silicon nitride and 7 µm flatpolyimide. Corrugated membranes showed larger deflections (as high as 70 µm) thanflat membranes, for the same pressure load and size. Moreover, the corrugationsextend the linear range of the pressure-deflection behaviour. The performance of thecorrugated membranes depends, among other factors, on the shape of the corrugation,i.e. sharp corners will make the membrane more brittle, as was the case of ourcorrugated silicon nitride membranes. From the results presented here, the polyimidemembrane is the best candidate to be used in the valve system due to its high

Chapter 7

145

deflection, 50 µm. This deflection meets the requirements as given in chapter 2.However, with better shape corrugations, corrugated silicon nitride membranes may bealso used.

7.1.3Valve

The valve consists of the deflecting membrane and a transition flow channelwhich makes the connection between the silicone rubber tube implant and themicroactuator. The operation of the active microvalve was simulated (chapter 4.3) andtested (chapter 5.4). The fluid flow was analysed with the use of analytical equations,for a simple channel geometry, while the pressure drop across the deflecting membranewas estimated using a numerical model. With the numerical simulation, the influence ofthe membrane deflection on the flow resistance, as well as on the pressure across theactive valve could be seen: The volume flow rates are small, so the membrane is notdeformed due to the fluid pressure. From the experimental results it can be concludedthat for a microactuator pressure up to 1⋅105Pa (1 bar), the fluid pressure across theactive valve can be easily adjusted in the requested range, 1000 - 4000 Pa. So far, thetransition flow channel was made with fine mechanics in perspex. The final channel willbe made in silicon or molded in a biocompatible material.

7.2 Intraocular pressure sensor

A complete eye pressure regulator system will have to include a micromachinedintraocular pressure sensor. A comparison between two measuring principles,piezoresistive and capacitive was made in chapter 2. A definitive selection has not yetbeen made.

7.3 Electronical system

Inductive powering of implantable devices is a widely accepted solution toreplace implanted batteries. Therefore, the transmission of energy by inductive coupledcoils was used to power the electrochemical actuator. The electronical system willconsist of an feedback controller (that will stabilise the eye-pressure to the requestedvalue by actuating the electrochemical actuator), a microprocessor (for data exchange)and wireless energy supply (that transmits energy to the actuator by inductive couplingof a pair of coils), that is made up of a microcoil and a receiver system.

7.3.1Microcoil

The modelling, design, fabrication and characterisation of an electroplated planarmicrocoil receiver was discussed in chapter 7. Calculations on the electrical parametersof the receiver coil have shown that for coils with a diameter of 4.5 mm and wire of 14 µm in diameter, a few mW can be transmitted. A gain in quality factor can be reached(i) by making high structures, resulting in a decrease of the series resistance of the coil;(ii) by increasing the thickness of the insulator between coil and silicon substrate, orbetter, by replacing the silicon substrate by a glass substrate; or (iii) by increasing thenumber of turns since this increases the self-inductance enormously. When thefrequency of the transmitter is fixed, then little is gained by choosing a ratio betweenthe inner- to outer diameter of 0.25. When a glass substrate is used, the parasitic

General discussions and conclusions

146

capacitance between the turns has to be taken into account. The model described forthe electrical equivalent of the receiver coil predicts values of the intrinsic resonancefrequency, intrinsic quality factor, series resistance of the coil, and the power transferfrom the coil to the load, which are in good agreement with measurements. Themicrocoil realised here meets the system requirements in terms of size andperformance.

7.3.2Receiver, Transmitter

A transmitter-receiver set-up was made on PCB. The transmitter was placed at adistance of 1 mm. Using the electroplated coil, it was possible to have at least 1 mWand 2V available on the output terminals. The electrochemical microactuator wasconnected and a constant current of 28 µA was driven through the actuator. Theactuator dissipated 50 µW. Although the efficiency is low, the power transmission ishigh enough to power electronics.

7.3.3Feedback controller

A stable eye-pressure will be achieved by incorporating a feedback controllerbetween the eye pressure sensor and the active microvalve. The electronic circuit of afeedback controller was made on a printed circuit board (PCB) and tested with amacro electrochemical cell and pressure sensor. The feedback circuit was controllingautomatically the electrochemical cell pressure: by setting a value of the pressure, theregulator actuated the electrochemical cell in such a way that the desired pressure wasobtained (as described in chapter 6).

7.3.4Microprocessor

The microprocessor will handle the exchange of data between the feedbackcontroller and the receiver: the information coming from the intraocular pressuresensor will be send to the external remote unit. This was not yet constructed.

7.4 Biocompatibility

The whole valve system is in contact with the eye fluid, thus biocompatibility is avery important issue that has to be taken into consideration. Materials in direct contactwith body tissue used so far are silicon, silicon nitride and polyimide that are notbiocompatible. Therefore, a cover with a thin layer of Teflon®, silicone rubber ortitanium might be sufficient. Besides this requirement, gases and liquids should notdiffuse through the membrane in order to obtain a tight electrochemical cell. Siliconand silicon nitride are known to have a quite low diffusion coefficient for gases andliquids.

7.5 Overall conclusion

From the present results it may be concluded that the design of theelectrochemical microcell as suggested in this thesis is not sufficient to meet therequirements needed for an implantable glaucoma eye pressure regulator. The maindrawback is the irreversible loss of oxygen gas at one of the electrodes (Cu) used inthe electrochemical cell. This implies a more frequent use of the power supply, than is

Chapter 7

147

desirable. The prevention of this loss is strongly dependent on whether or not it ispossible to find an electrode material that will not react with the generated gas.

Substantial work must be done on the actuator (to increase the time constant andto find a satisfying way for assembly and sealing without the use of adhesives) and onthe pressure sensor. Having energy available, we hope that the pressure sensor will notbe the main bottle neck for the system. The other subsytems can be designed andrealised using available technology.

Nevertheless, for applications where the requirements are not so strict as in ourcase, an active microvalve actuated using an electrochemical driving principle couldfind a broad market, such as pumps or active valves, due to its possibility to produce ahigh pressure for large deflections having a low power consumption.

7.6 Suggestions for electrochemical actuator

Suggestions for further research that could improve the performance of theactuator include:a1. another design of the Nafion or a better polymer, with a lower oxygen diffusion and

a better adhesion to the substrate. At present, such a polymer is, to our knowledge,not available.

a2. It will be necessary to replace the Cu/Cu2+ electrode. A possible alternative wouldbe to use a metal electrode passivated by a coherent oxide layer, which does notreact with oxygen gas. Some possibilities have been investigated electrochemically(chapter 3.6), but experiments to test the long-term stability of such systems in amicrocell are necessary.

b1. the two substrates were bonded by an epoxy glue, because high temperaturebonding (> 250°C) could not be used due to the Nafion membrane. Furtherresearch is necessary for a gas-tight bonding at low temperature.

b2. the electrochemical cavity was sealed with the electrolyte in it by using the sameepoxy. Electrochemical sealing requires more research since the electroplatingsealing was not gas-tight.

Summary

149

Glaucoma is a disease causing damage to the optic nerve head due to a too higheye pressure. This damage will lead to visual field loss, and finally to blindness. Thecurrent surgery treatment improves the drainage of the eye fluid by introducing adraining device. The problem is that one cannot predict in advance the value of the eyepressure after surgery and it cannot be adjusted to the optimum eye pressure of thepatient, after surgery. A continuous adjustment of the eye pressure would simplify andimprove the present treatment.

The goal of this research is to develop a micromachined actuator that could becombined with existing glaucoma implants. This microactuator, that acts as an activevalve would allow the eye pressure to be adjusted continuously around a desired value.The actuator can be used to adjust the eye fluid pressure, for patients suffering fromglaucoma, by changing the fluid resistance of an implanted flow channel due todeflection of an integrated membrane.

To obtain a low energy consumption and to have the possibility of discontinuoussupply of power, it was opted for electrochemical actuation, which is based on theelectrolysis of an aqueous electrolyte solution. The reversible electrochemicalreactions, which are driven by an external current source, lead to gas evolution or gasreduction (depending on the direction of the current). In a closed system thecorresponding gas pressure rise or drop is used to change the deflection of a flexiblemembrane, which in turn can close or open a liquid channel. If such an electrolytic cellis operated under open-circuit conditions, the pressure and thus the deflection state ofthe diaphragm will, ideally, be maintained. This means that no energy is required tomaintain the state of the valve. It has been proven that relatively large pressures (up totens of bar) and large deflections can be reached in this electrochemical actuator with alow energy consumption.

The complete microactuator system should have a maximum size of 5x5x2 mm3

and will be powered by wireless energy supply. The energy is transmitted by using apair of coils which are inductively coupled; one coil is implanted and the other one isoutside the body.

SUMMARY

Samenvatting

150

Glaucoma is een ziekte die schade aan het optische zenuwcentrum veroorzaakt dooreen te hoge oogdruk. Deze schade kan tot een beperking van het gezichtsvermogenleiden en uiteindelijk tot blindheid. De huidige medische behandeling verbetert deafvoer van de oogvloeistof door een afvoersysteem te implementeren in het oog. Hetprobleem is dat het niet mogelijk is het niveau van de oogdruk te voorspellen na deingreep. Bovendien kan de oogdruk niet bijgesteld worden tot de optimale oogdrukvan de patiënt. Een doorlopende bijregeling van de oogdruk zou de huidigebehandeling van glaucoma aanzienlijk verbeteren en vereenvoudigen.

Het doel van dit onderzoek is het ontwikkelen van een micromechanischeactuator die gecombineerd kan worden met de bestaande glaucoma implantaten. Dezemicroactuator, die werkt als een soort actieve klep, zorgt ervoor dat de oogdrukcontinu geregeld wordt rond een gewenste waarde. De actuator doet dit door deuitwijking van een geïntegreerde membraan te regelen, waardoor de doorstroom vaneen vloeistofkanaal wordt bepaald.

Voor het verkrijgen van een laag energieverbruik en de mogelijkheid totgedoseerde toevoer, is er uiteindelijk gekozen voor een electrochemische actuatie vande actieve klep. Deze actuator is gebaseerd op de elektrolyse van een electroliet opwaterbasis. De omkeerbare electrochemische reacties, die gestuurd worden door eenexterne stroombron, leiden tot de ontwikkeling van gas dan wel tot het verdwijnenervan. De richting van dit proces (produktie of reductie) wordt bepaald door destroomrichting. In een gesloten systeem zal de corresponderende gasdruktoename ofafname gebruikt worden om de uitwijking van een buigzaam membraan te regelen endaardoor de doorstroom van een kanaal te bepalen. Als er geen stroom naar/van deelectrochemische cel stroomt zal de druk en dus de membraan uitwijking enkanaalweerstand, ideaal gezien, behouden blijven. Dit betekent dat er geen energienodig is om de toestand van de doorstroomsnelheid vast te houden. Het is bewezen datrelatief hoge drukken (tot vele tientallen atmosfeer) en grote uitwijkingen mogelijk zijnmet behulp van de beschreven electrochemische actuator terwijl het energieverbruiktoch relatief laag is ten opzichte van andere actuatieprincipes.

De microactuator mag compleet niet groter dan 5 x 5 x 2 mm3 zijn en zaldraadloos van energie voorzien worden. Dit gebeurt door middel van twee spoelen dieinductief gekoppeld zijn; de ene zit in het oog en de andere bevindt zich er buiten.

SAMENVATTING

Rezumat

151

Glaucoma este o boala de ochi ce cauzeaza deteriorarea nervului optic datoritaunei presiunii oculare prea mari. Aceasta deteriorare va avea ca efect pierdereacampului vizual si, in final, va conduce la orbire. Tratamentul chirurgical actualincearca sa micsoreze presiunea oculara prin introducerea unui sistem de drenare ceimbunatateste drenarea lichidului ocular. Dificultati apar din faptul ca nu se poateprezice ce presiune oculara va fi dupa operatie, iar dupa operatie ajustarea la presiuneaoculara optima a pacientului nu este posibila. O ajustare continua a presiuni oculare arsimplifica si imbunatati tratamentul prezent.

Obiectivul prezentului proiect de doctorat este producerea unui microactuator cepoate fi folosit impreuna cu implantul existent pentru glaucoma. Acest microactuator,ce actioneaza ca o valva activa, va permite ca presiunea oculara sa fie ajustata continuula valoarea dorita. Actuatorul ajusteaza presiunea lichidului ocular prin modificarearezistentei la curgere, printr-un canal implantat, a lichidului ocular, cu ajutorul uneimembrane flexibile.

Pentru a obtine un consum de energie cat mai redus si pentru a avea posibilitateaunei utilizari discontinue a sursei de energie, s-a optat pentru folosirea reactiilorelectrochimice ca principiu de actuare. Electroliza unei solutii apoase de electrolitpoate duce la formarea de gaze sau la reducerea volumului lor, in functie de directiacurentului electric. Intr-un sistem inchis, cresterea sau scaderea presiuni de gaz estefolosita la producerea deflexiei unei membrane flexibile care, la randul ei, poate inchidesau deschide curgerea lichidului ocular prin canalul implantat. Daca o astfel de celulaelectrochimica este utilizata in conditii de circuit electric deschis, presiunea gazului sideci deflexia membranei va fi, in caz ideal, mentinuta. Aceasta inseamna ca nu estenecesara nici un fel de energie electrica pentru a mentine nivelul de deflexie almembranei, deci starea valvei. S-a dovedit experimental, cu acest actuatorelectrochimic, ca pot fi obtinute presiuni relativ mari (pana la zeci de bar) si deflexiimari, folosind un consum redus de energie.

Dimensiunea maxima permisa pentru intregul sistem ce va continemicroactuatorul este de 5x5x2 mm3. Microactuatorul este alimentat electric folosindtransmisia de energie electrica de la distanta, utilizand cuplajul inductiv intre 2 bobine:o bobina va fi implantata iar cealalta in afara corpului.

REZUMAT

Bibliography

152

C. Neagu, M. Hamberg, J.G.E. Gardeniers, D.J. Yntema and M. Elwenspoek, Anelectrochemical microactuator, Proc. IEEE Workshop on MEMS, Amsterdam, the Netherlands,p. 106-110, 1995.

C.R. Neagu, J.G.E. Gardeniers, M. Elwenspoek and Prof. J.J. Kelly, An ElectrochemicalMicroactuator: Principle and First Results, J. Microelectromech. Syst. 5(1), p. 2-9, 1996.

C.R. Neagu, J.G.E. Gardeniers, M. Elwenspoek and J.J. Kelly, An electrochemical active valve,Proc. Electrochemical Microsystems Technologies, Grevenbroich, Germany, August 28-30,1996.

C.R. Neagu, J.G.E. Gardeniers, M. Elwenspoek, J.H.J. Fluitman, An electrochemically actuatedmicrovalve, Proc. Int. Conf. on New Actuators, Actuator’96, Bremen, Germany, June 26-28,1996,

C.R. Neagu, A. Smith, J.G.E. Gardeniers, M. Elwenspoek and Prof. J.H.J. Fluitman,Characterisation of a planar microcoil for implantable microsystems, (poster) Proc.Eurosensors X, Sept. 8-11, Leuven, Belgium, p.343-346, 1996.

C.R. Neagu, J.G.E. Gardeniers, M. Elwenspoek and J.J. Kelly, An electrochemical active valve,Electrochimica Acta 42(20-22), p.3367, 1997.

C.R. Neagu, H.V. Jansen, A. Smith, J.G.E. Gardeniers, M. Elwenspoek, Characterisation of aplanar micro coil for implantable microsystems, Sensors & Actuators A 62(1-3), p.599-611,1997.

H.V. Jansen, M.J. de Boer, R.J. Wiegerink, N.R. Tas, E.J. Smulders, C.R. Neagu, and M.C.Elwenspoek, The black silicon method VII: RIE lag in high aspect ratio trench etching ofsilicon, Microelectronic engineering 35, p. 45-50, 1997.

C.R. Neagu, H.V. Jansen, J.G.E. Gardeniers, M. Elwenspoek, The electrolysis of water: anactuation principle for MEMS with a big opportunity:”, (poster) Proc. Int. Conf. on NewActuators, Actuator’98, Bremen, Germany, June, 1998.

W. Olthuis, S. Böhm, C. Neagu, P. Bergveld, Various current regimes applicable for sensing inµTAS, to be presented at MicroTAS '98, Banff, Canada, October, 1998.

BIBLIOGRAPHY

Epilogue

153

“What do you like doing the best in the world, Pooh?Well, said Pooh, What I like the best- and then he had tostop and think. Because although Eating Honey was avery good thing to do, there was a moment just beforeyou began to eat which was better than when you were,but he didn’t know what it was called”

(from The house at Pooh Corner by A.A. Milne)

Yippee! With satisfaction and some hesitation I can say that the 'Book' is ready.Relaxed I look back to the last four years: It was nice, exciting, full of adventure, andvery instructive. It would have been almost impossible to do all the research for thisthesis by myself. I would, therefore, like to thank everybody who participated inputting together the 'Book'.

First of all I would like to thank my first promoter Jan Fluitman for giving me theopportunity to start this PhD, and to Miko Elwenspoek and John Kelly for continuingthis special role. Although Miko has always such a busy schedule, his door was alwaysopen for questions. I appreciate everything he did for me, not only as a supervisor butalso at the human level. I am grateful to John Kelly for his understanding and patientsupervision into the 'world of electrochemistry'. If it wasn't for John Kelly to help outwith the structuring of this thesis, in particular with chapter 3, it would have been aheavy 'stone' to swallow.

I would also like to thank Han Gardeniers, my daily supervisor for all thevaluable discussions, competent comments, and for tolerating my foolish/sillyquestions.

Highly appreciated are Henri's always critical comments: "No compromise inscience ... Life is something else". He was the one who read the ‘very’ first draft of mythesis!

Without all of you, I would have never succeeded in delivering the thesis in this form,at least not in this readable form!

EPILOGUE

Epilogue

154

My first roommates Vincent Spiering and Gert-Jan Burger, thank you for yourguidance and support in helping with the transition to the Dutch mentality. Especiallyyou Vincent, you are a great and indispensable friend! Thank you very much for yourmoral support. Sandra Burger, thank you for the joyful and useful evenings, whilehelping me understand some aspects of the Dutch life. Unfortunately, she was not theonly one who was 'a little bit stressed' by my 'helping-offer' behaviour: Ineke and Keesvan der Werf had to deal with me much more often. I am grateful for their moralsupport and patience during all these years. I can't imagine living in Enschede withoutall the weekends spent at their house, where, I can proudly say I have my 'own room'.There, I also had my first Dutch lessons given by their daughter, Hanneke. By the way,the basic Dutch words to survive the 'micmec society' are "patat en mayonaise" alongwith the essential word: BIER, NU!! Therefore, I thank the Micmec group for theirgood guidance into the real Micmec culture and habits! Besides this, I enjoyed going toconcerts and shows with Bert Otter, Roger Bruis, Johnny Sanderink, attendingArgentinean tango with Elke Duncker, or listening to choir music, where Ineke van derWerf, Rob Kooyman or Walter Olthuis were performing. I also enjoyed very muchgoing with Marianna Sijtsema for a swim at 7:00 o'clock in the morning during thewinter!

I appreciate the remarkable assistance of the MESA 'blue(s) boys and girls' of theclean room: Bert Otter, Johnny Sanderink, Huib van Vossen, Gerard Roelofs, StanKruger, Nicole Closset, Riel Weeinink, Kees Eijkel and the others who made thefabrication of my device possible. Thanks for your help and kindness and for putting upwith my 'demands'!. However, the 'blue boys' will not survive without Sharron Kochand Hermine Knol, who make the communication easier and possible, I mean the'heads' can hear what the 'neck' says.

I am also obliged to Michiel Hamberg for working with me in the clean roomduring the first few month and for pushing me to finalise the first electrochemical cellbefore his leaving for work to Stockholm, Sweden. I would also like to thank HenkKok for his first tests on low temperature bonding and electroplating.

An important contribution to the research was done by 'my' students, withoutthem this thesis would have been much thinner. Eric Visser; Erik Leussink with who Iwas trying 'to break the personality' of all those electrochemical reactions which didn'twant to listen to our wishes; Tino Ebbers and the 'toreador' Ansgar Smith making the‘Taurus - magnetic field' fit quite nicely in a in 5x5 mm2 coil!

Erwin Berenschot and Meint de Boer, without you our group would have had arecord in 'broken wafers' or in other words, we would have had 'two left hands' in theclean room .But, with respect to the beer … ha ha … we are an octopus!

Furthermore, I would like to thank STW for supporting this project financiallyand for funding a three months practical training at Case Western Reserve University,Cleveland, Ohio, USA. The work and life experienced there made me appreciate evenmore what I have here, in the Netherlands. Michael Huff, A. Heuer, Hal Kahn, WilliamBenard, Jeff Melzak, Russ deAnna were very nice and did their best for me to have agood time. Special thanks to Beth Fuller and Dave Smith. Also the Romanians: Irina

Epilogue

155

and her family, Marian, Monica, Sorin, Titus whom I met there, took good care of meand made me feel at home, what a nice feeling!

While attending different conferences I had the chance to meet special peopleand have interesting talks, which made me look forward to the next conference: YlvaBäcklund, Peter Enoksson, Roland Zengerle, Michael Reed, Tony Ricco, Chuck Sloan,P. Hesketh, Bill Trimmer. A.J. Bard and B.E. Conway thank you for your competentremarks. I appreciate the fruitful discussions with Milena Koudelka-Hep and Peter vander Waal.

I would like to mention John Slot, Alex Volanschi and Albert Prak for being notonly my best neighbours but also friends. Thanks to Elke Duncker and Easy for beingpatient with me all these years and to Roger Bruis for heaving a good time together.Also, I thank to Ben Kloek and his family, Takako and Marie for making me toremember Neuchatel with pleasure.

I am grateful and proud to have Michael Reed, Jan Greve, Piet Bergveld, JohnKelly, Miko Elwenspoek, Han Gardeniers, Albert van der Berg and Mircea Rusu in thedefence committee judging my thesis.

Of course the spirit in the Micmec played an important role. I thank all mycolleagues for providing high knowledge, for the pleasant ambience and theirenthusiasm. The 'cooking club', Edwin Smulders, Vincent Spiering, Joost van Kijk,Henri Jansen, Albert Prak, made it easier for me to integrate in the group. It was niceto watch Edwin doing his best to cook a delicious '5 stars dinner' and the others askingwhen they will get something … quantitative ... to eat. Jasper Nijdam, even if weshared only for a short time the same room, you were a pleasant roommate. DickEkkelkamp for the tee, orders and for preventing a 'tragedy', which would havehappened if we would have gone to Norway by boat; instead we went by plane, boat,bus and train! Henk van Wolferen for helping me in obtaining ‘good’ sound andphoto’s. Thanks to Judith Beld, Jose Nijhuis, Simone Moekotte I didn't feel aloneamong so many 'jochies', and of course to Judith for taking care that Robert and Hanget the most points for the World Cup football in France!!

Furthermore, I would like to thank my whole family for their encouragement andsupport. I thank my father Mircea Rusu for being were I am now due to hisenthusiastic stories about physics, and both him and my mother Cornelia for theirinspiring comments about his thesis. I am grateful to Mircea Neagu for being a greatand supportive friend, even at the distance. I manifest my gratitude to Knuffie for hissupport and for having put up with me for the last one and a half years.

Biography

156

Cristina Neagu was born on September 4, 1966 in Bucharest, Romania. Shegraduated from Faculty of Physics, University of Bucharest in 1990 with M.Sc. degreein Applied Physics. From 1990 to 1993 she worked at the Institute of Atomic PhysicsBucharest, Laser Active Media Dept. There she studied the laser crystals for theestimation of their growth technology by cathodoluminescence imaging and SEM. In1993, winning an European fellowship for micro-characterisation, she went for 8months at MESA Institute of the University of Twente, the Netherlands in theBiophysical Techniques Group, Prof. J. Greve. The research subject was on theinvestigation of the applicability of the (silver-enhanced) immunogold labelling method(with FITC and nano-gold particles) for the high-resolution human lymphocytesantigen detection by combined atomic force and fluorescence microscopy and flowcytometry. Since March 1994 she is employed as Ph.D. by STW (NetherlandsTechnology Foundation) at MESA Research Institute, in the MicromechanicalTransducers Group, Prof. Miko Elwenspoek. The work was on the development of anmicromechanical actuator that could be used, in combination with a glaucoma implant,to the adjustment of the eye fluid pressure, for the patients suffering from glaucoma,which is described in this thesis. This was interchanged by a 3 months stay in theMicromechanical group (Prof. M. Huff and Prof. A.H. Heuer) at Case WesternReserve University, Cleveland, Ohio, U.S.A. There she worked on a RF magnetrondeposition of thin films, high temperature shape memory alloys (TiNiPd). In August1998 she will start working at MESA Institute, the Biophysical Techniques Group, onsingle molecule DNA analysis.

Besides these activities she likes to go on a horse-back ridding, swimming,volleyball and to travel.

BIOGRAPHY

Date post:	23-Jun-2020
Category:	Documents
Upload:	others
View:	1 times
Download:	0 times

A Medical Microactuator based on an Electrochemical Principle · A Medical Microactuator based on...

Documents