Implantable MEMS Intraocular Pressure (IOP) Regulator Massachusetts Institute of Technology 2.372J/6.777J May 15, 2008

Group D: Nadia Cheng Asiri Ediriwickrema Joy Johnson Kristofor Payer Advisor: Prof. Carol Livermore

Table of Contents 1

Introduction......................................................................................................................................... 2

2

Microfluidics....................................................................................................................................... 3

3

Piezoresistive pressure sensor............................................................................................................. 4 3.1

The Wheatstone Bridge............................................................................................................... 4

3.2

Determining ratio of output-to-input signals .............................................................................. 5

3.3

Designing resistor dimensions .................................................................................................... 7

4

3.3.1

Noise ................................................................................................................................... 7

3.3.2

Self-heating ......................................................................................................................... 8

3.3.3

Uniform-stress assumption ................................................................................................. 8

Electrochemical Micro-Actuator ........................................................................................................ 9 4.1

Electrochemical cell design and composition............................................................................. 9

4.1.1 4.2

Electrochemical Cell Composition ..................................................................................... 9

Micro-Actuation & Valve Design............................................................................................. 10

4.2.1

IOP vs. deflection: Valve and Fluid Channel Design ....................................................... 10

4.2.2

Actuator Pressure vs. Deflection: Corrugated Membrane Design.................................... 12

4.2.3

Cavity channel area vs. O2 loss rate: Cavity Design ........................................................ 13

4.3

Power consumption of electrochemical cell ............................................................................. 15

5

Fabrication ........................................................................................................................................ 15

6

Packaging.......................................................................................................................................... 21

7

Feedback Controller.......................................................................................................................... 22

8

Device Power .................................................................................................................................... 23

9

Calibration......................................................................................................................................... 24

10

Telemetry ...................................................................................................................................... 24

11

Simulated Performance................................................................................................................. 25

12

Conclusions................................................................................................................................... 26

13

References..................................................................................................................................... 27

1

1 Introduction Glaucoma is a common eye disease associated with elevated intraocular pressure (IOP). According to the World Health Organization, glaucoma is the global leading cause of blindness and is the leading cause of irreversible blindness in the US. There is a variety of possible devices that can be surgically implanted into the eye to treat elevated IOP. The increase in pressure is attributed to the build up of ocular fluid, the aqueous humor, in the anterior chamber of the eye. Aqueous humor is produced in the posterior chamber of the eye and enters the anterior chamber through the pupillary opening. There are two channels where the ocular fluid can drain out, the more common route being the orbicular meshwork, Schlemm's canal and episcleral veins. The remaining fluid drains through the veosceleral route.1 Normal IOP should be within 10 - 21 mmHg.2 Usually, elevated IOP is a result of lower outflow than inflow of aqueous humor, and the increase in pressure can cause damage to the optic nerve tissue, loss of peripheral vision, and ultimately blindness.3 Conventional IOP monitoring devices have difficulty monitoring pressure inside the eye and lack the capacity to continually monitor pressure. Continuous monitoring of IOP is important for disease management. Since individual IOP is dynamic, discrete IOP measurements may indicate healthy pressure levels in patients where blindness continues to progress. Additionally, a significant complication in glaucoma treatments is postoperative hypotony, where eye pressure becomes to low causing irreparable damage. Microelectromechanical systems (MEMS) pressure sensors and actuators have been designed previously to sense and treat elevated IOP; we are proposing the fabrication of an integrated, closed-loop system to sense and maintain an ideal IOP level. A piezoresistive pressure sensor will work upstream and detect ocular fluid pressure, which will consequently engage an electro-chemical valve actuator to modify the fluidic resistance in the microfluidic channel. Proportional control will be used to keep IOP at the desired level. Figures 1 and 2 show schematics of our MEMS device from the top and cross-sectional views, respectively.

Figure 1—Schematic of top view of device

2

Figure 2—Schematic of cross-section of device

2 Microfluidics The microfluidic network in this device consist of a series of channels: a silicone rubber tube which connects the eye to the device, a channel over the sensor, over the actuator, after the actuator and finally a silicone rubber tube which drains from the device back to the body. Each of these channels acts as a resistor with an associated pressure drop7. The key to designing a device that is well suited to controlling fluidic resistance is to make all other resistors negligible in comparison. Additional resistances come into play at micro-channel transitions: changing geometries or cross-sectional areas13. For the geometries looked at for the proposed device the pressure drop across transition regions was negligible, on the order of 1 Pa. Finally, gravity can also add resistance and cause a pressure drop on the order of 200 Pa. This is certainly not negligible compared to the pressures being controlled (with a resolution of 130 Pa), however for simplification and because it varies depending on the orientation of the device, these effects will be ignored13. The aqueous humour (eye fluid) is produced at a rate of 1-3 μL/min13; the Reynolds Number for these flow rates in channel geometries of interest are on the order of 1-5. Laminar, pressure-driven flow can be reduced to Poiseuille flow7. This simplifies the equations and allows fluidic resistances to be calculated which relate the pressure across the channels to the flow rate. The fluidic resistance in Poiseuille flow is a function of the channel geometry (length and cross-sectional area), wetted perimeter and fluid viscosity. A general equation, Equation 1, can be written for the fluidic resistance in some arbitrary geometry13: 2 ⋅ k shape ⋅ L ⋅η (Equation 1). R= 2 Dh ⋅ A Where A is the cross-sectional area of the channel, Dh is the hydraulic diameter given by Equation 2 below and kshape is a shape factor which is given in Table 1. 4A Dh = (Equation 2) P P is the wetted perimeter of the channel. Channels of interest in the proposed device include a circular cross-section (for the silicone-rubber tubes) and rectangular channels where the width is much greater than the height. For such channels, the various constants can be look up in tables13 and plugged into this equation. The pressure drop across the channels can be calculated for a known resistance and a given flow rate. Table 1 below describes the channel geometries chosen for the proposed device along with 3

the pressure drops for an eye fluid production rate of 2.5 μL/min. It is clear to see that the pressure drop along the actuator channel dominates; flow rates of 1 μL/min and 3 μL/min were also investigated and for all cases the actuator resistance dominates. Region Geometry Shape Factor Dimension 1 Dimension 2 Length Pressure Drop

Table 1—Selected channel geometries Silicone Tube 1 Sensor Actuator Post-Actuator Circle Rectangle Rectangle Rectangle (w>>h) (w>>h) (w>>h) 16 24 15.5 24 300 μm (dia) 1 mm (w) 150 μm (w) 1 mm (w) 20 μm (h) 20 μm (h) 20 μm (h) 2 cm 2 mm 1 mm 1 mm 4 Pa 130 Pa 0.5-4.5 kPa 65 Pa

Silicone Tube 2 Circle 16 300 μm (dia) 2 cm 4 Pa

3 Piezoresistive pressure sensor 3.1 The Wheatstone Bridge The IOP is measured using a pressure sensor located upstream relative to the actuator. The sensor consists of a thin square silicon membrane and four diffused p-type doped piezoresistors; silicon is chosen based on its high variation in resistance when it deforms mechanically.4 The four resistors create a Wheatstone bridge: each resistor is placed in the center near an edge of a membrane, where the strain is greatest, thus maximizing the output voltage.5 A Wheatstone bridge is favorable because it eliminates common-mode resistance changes in the differential signal (e.g., temperature effects are balanced).6 As depicted in Figure 3, all the resistors are positioned with their lengths along the [110]. Resistors R1 and R3 are “longitudinal” resistors, while R2 and R4 are “transverse” resistors; the longitudinal stress on R1 and R3 is the transverse stress at R2 and R4 , and vice versa.7 The resistors are relatively long in one direction and narrow in another so that the current density and electric field are both along the long axis of the resistor; this simplifies the following analysis.7

Figure 3—Schematic of placement of piezoresistors on the pressure sensor’s diaphragm membrane

4

All the resistors are designed to have the same dimensions to have the same unstrained resistance R . When pressure is applied, resistors R1 and R3 each result in an increased resistance R + ΔR , while resistors R2 and R4 each result in a decreased resistance R − ΔR .8 The Wheatstone bridge circuit is illustrated in Figure 4, where VS is the input signal of the circuit and VO is the output signal.

Figure 4—Wheatstone bridge circuit constructed from resistors in Figure 3

3.2 Determining ratio of output-to-input signals In simplifying the signal analysis, it is assumed that the stress across each resistor is uniform, thus small resistors must be used (the selection of resistor dimensions is discussed later). Two parameters, π l and π t , are defined as the longitudinal and transverse piezoresistance coefficients, respectively. These piezoresistance coefficients are functions of material and doping type but are weak functions of doping level for doping below about 1019 cm-3.7 For p-type doped silicon along a [110] direction, π l = 71.8 and π t = -66.3.7 Therefore, if σ l and υ are the longitudinal stress of the longitudinal resistors and Poisson’s ratio, respectively, then the output-to-input signal ratio is:7

VO R1 R3 − R2 R4 (π l + υπt )σ l + (π t + υπl )σ l ≈ = VS (R1 + R2 )(R3 + R4 ) 2[1 + (π l + υπt )σ l − (π t + υπl )σ l ]

(Equation 3).

Finite-difference methods have been used to solve for the maximum longitudinal stress of a square diaphragm as a function of plate dimensions—edge length, L , and plate thickness, H —and the uniform load, P , applied to the plate:7 2

⎛L⎞ ⎟ P (Equation 4) ⎝H⎠ Equation 3 becomes a function of the diaphragm dimensions and the applied load only:

σ l = 0.204⎜

5

2 2 ⎡ ⎡ ⎛L⎞ ⎤ ⎛L⎞ ⎤ (π l + υπ t ) ⎢0.294⎜ ⎟ P ⎥ + (π t + υπ l ) ⎢0.294⎜ ⎟ P ⎥ ⎝ H ⎠ ⎥⎦ ⎝ H ⎠ ⎥⎦ VO ⎢⎣ ⎢⎣ = (Equation 5). 2 2 VS ⎧⎪ ⎫ ⎡ ⎤ ⎡ ⎤ L L ⎪ ⎛ ⎞ ⎛ ⎞ 2⎨1 + (π l + υπ t ) ⎢0.294⎜ ⎟ P ⎥ + (π t + υπ l ) ⎢0.294⎜ ⎟ P ⎥ ⎬ ⎝ H ⎠ ⎥⎦ ⎝ H ⎠ ⎥⎦ ⎪⎭ ⎪⎩ ⎢⎣ ⎢⎣ V Because it is desirable to maximize O , L should be optimized while complying with structural VS H constraints. It has been advised to keep the maximum stress experienced by the membrane to be less than one-fifth of its fractural stress, which is considered to be about 0.3 GPa for silicon.9

The length of the membrane has been selected to be 1000 microns (maximized for the specific device constraints). After running several simulations and considering the thickness of the piezoresistors (discussed later), the membrane thickness has been chosen to be 10 microns. By utilizing Equation 2, these dimensions result in a maximum stress of 14.7 MPa, which is less than one-fifth of the fractural stress. V Figure 5 shows O plotted against IOP ranging from 500—5000 Pa, which is what the device is VS expected to measure: Vo/Vs vs. Pressure 0.01 0.009 0.008 0.007

Vo/Vs

0.006 0.005 0.004 0.003 0.002 0.001 0 500

1000

1500

2000

2500 3000 3500 Pressure [Pa]

4000

4500

5000

Figure 5—VO/VS for a pressure range of 500—5000 Pa IOP range

The plot shows favorable results when considering that VS = 1 Volt, giving output voltages ranging from 1 mV—10 mV. A 1 V input signal is reasonable for the device in terms of selecting an appropriate implantable power source as well as considering the effects self-heating, the latter of which will be discussed later.

6

3.3 Designing resistor dimensions Several areas are considered simultaneously in selecting resistor dimensions: noise, self-heating, and the uniform-stress assumption. After using MATLAB to run through various combinations of values to arrive at favorable results, the resistor dimensions are chosen to be length l = 100 microns, width w = 10 microns, and thickness t = 0.5 microns. The following analysis will assume these dimensions to show that they are reasonable for the device. Several of the following equations utilize the resistance, R , of the piezoresistor:9 ρ ×l (Equation 6), R= t×w where ρ is the resistivity of the piezoresistor. The resistivity for a p-type doped piezoresistor is computed as10 1 (Equation 7), ρ≈ qμ p p where q = 1.6e-19 C is the charge of an electron, μ p = 500 cm2/V-s is the hole mobility for p-type doping, and p = 1017 cm-3 is the doping concentration (this value is considered high but is widely accepted for p-type piezoresistors). Therefore, our piezoresistors each have a resistance of 25 kΩ, which appears to be reasonable after surveying similar devices discussed in various literatures.

3.3.1 Noise The resistor dimensions should be designed so that the noise signal is negligible relative to the resolution of the output signal. The noise seen by each resistor is primarily composed of Johnson noise, V J , and flicker noise, V f .9 The total noise for a single resistor, Vnoise , can be expressed as a superposition of the components:9 2 Vnoise = V J2 + V f2

(Equation 8).

The Johnson noise is a thermal voltage noise, and is expressed as9 V J2 = 4k B TR( f max − f min ) (Equation 9), where k B is the Boltzmann constant, T is the absolute temperature, and ( f max − f min ) defines the working bandwidth.11

The flicker noise, which is caused by fluctuation in electrical conductivity and is only noticeable for low frequencies, is defined as αVS2 ⎛ f max ⎞ ⎟ V f2 = ln⎜ (Equation 10), plwt ⎜⎝ f min ⎟⎠ where α = 1.5e-6 is a dimensionless parameter for high doping concentrations.11 The working bandwidth used in this analysis is defined by f max = 1 kHz and f min = 1 Hz, which are values found in literature for a similar device. 11 Using these equations for Johnson noise and flicker noise, the total noise voltage for a single resistor in our device is 7.9675e-7 V. At worst, the device will see a total noise voltage twice that of a single resistor’s (for a Wheatstone bridge configuration, the output signal is the voltage across two resistors in 7

series). Considering that the desired pressure resolution of the device is 100 Pa, the “resolution output voltage”, from Equation 3, is 1.8994e-4 V. Therefore, the noise voltage of the device is significantly less than the output signal and can be neglected.

3.3.2 Self-heating The resistor dimensions must also be designed so that the change in resistance due to self-heating is significantly less than the change in resistance due to strain. A maximum input voltage, limited by a user-defined upper limit for the amount of self-heating allowed, can be calculated. In the case for which there is no self-heating and the change in resistance is only due to strain,8 ΔR VO = (Equation 11). R VS The change in resistance due to self-heating, for a voltage-controlled device, is12 V2 ΔRmeas = α R RT S (Equation 12), R where α R can be approximated as 2500e-6 K-1 for silicon7, and the thermal resistance is 1 l RT = (Equation 13). κ wt The thermal conductivity is κ = 148 W/(K-m).7 By forcing ΔRmeas to be less than 1% of ΔR ( ΔRmeas < 0.01ΔR ), rearranging Equation 11 and Equation 12, and assuming a minimum output voltage of 1 mV (Figure 5), the maximum input voltage is computed to be 2.6448 V. Because we are forcing self-heating effects to be negligible for input signals below this value, an earlier mentioning of setting the input voltage to be 1 V is confirmed to be reasonable.

3.3.3 Uniform-stress assumption In order to simplify the output-to-input signals analysis, we assume uniform stress across each resistor. In order to validate this assumption, the stresses at the center and at the end of each resistor are compared. Note that the longitudinal resistors can be folded in half in a very narrow “U” shape so that both halves effectively experience the greatest amount of strain near the edge of the membrane:

Figure 6—The longitudinal resistors can be folded in half so that both halves are effectively experiencing the maximum strain near the edge of the membrane.

For small deflection w( x, y ) , the stresses of a plate in pure bending are7

σx = −

Ez 1 −ν 2

⎛ ∂2w ∂2w ⎞ ⎜⎜ 2 + ν 2 ⎟⎟ ∂y ⎠ ⎝ ∂x

(Equation 14)

8

Ez ⎛ ∂ 2 w ∂2w ⎞ ⎜ (Equation 15), σy =− + ν 2 ⎟⎟ 1 − ν 2 ⎜⎝ ∂y 2 ∂y ⎠ where E = 150 GPa is Young’s modulus, and z is the vertical distance from the center of the plate. A trial solution for wˆ ( x, y ) has been made:7

c1 ⎡ ⎛ 2πy ⎞⎤ ⎛ 2πx ⎞⎤ ⎡ 1 + cos⎜ ⎟⎥ (Equation 16), ⎟⎥ ⎢1 + cos⎜ ⎢ 4⎣ ⎝ L ⎠⎦ ⎝ L ⎠⎦ ⎣ where the coefficient c1 is approximated as7 wˆ ( x, y ) =

6 P(1 − υ 2 ) L4 c1 = (Equation 17). π 4 EH 3 Using Equation 14—Equation 17, the maximum change in stress across the length of each piezoresistor, from its midpoint to its end, is 2.45%, and the change in stress across its height is 10%. Therefore, the selected resistor dimensions allow us to assume that each piezoresistor is experiencing uniform stress in our analysis.

4 Electrochemical Micro-Actuator 4.1 Electrochemical cell design and composition In order to control the flow of aqueous humour, or eye fluid, and therefore the intraocular pressure (IOP) in the eye, we propose the use of an electrochemical micro-actuator whose gas pressure build-up will deflect a membrane that changes the flow resistance of the channel it obstructs.

4.1.1 Electrochemical Cell Composition The micro-actuator is an electrochemical cell consisting of a platinum working electrode (anode) and a platinum counter electrode (cathode), each in separate cavities, connected by a narrow channel in order to separate the electrochemical reactions and subsequent gases as shown in Figure 7. Platinum thin film electrodes are used at the anode and cathode because platinum will allow hydrogen gas bubbles in the cathode cavity and oxygen in the anode cavity without reacting with the metal eliminating the need for a semi- permeable passivation membrane over the electrode. Additionally, Platinum has faster redox reaction rates than that of most other commonly used metals and the fabrication of same metal electrodes is significantly less complicated. The entire cavity is filled with an aqueous electrolyte connecting the two electrodes in the system via a small channel between the two cavities. The cavities are covered with a thin corrugated silicon nitride membrane which when deflected will act as a valve obstructing the flow of eye fluid, subsequently changing fluid resistance and intraocular pressure. The membrane over the anode is significantly thicker, than that of the membrane over the cathode in order to ensure that it does not deflect, as the deflection in our system is based on oxygen gas production. A positive current is applied to the cell drives electrolysis at the anode, allowing oxidation to occur at the working electrode while hydrogen is formed at the counter electrode. (If the reverse current is applied, hydrogen is oxidized at the counter electrode and reduction occurs at the working electrode). The electrolysis of the aqueous electrolyte causes the advent of oxygen bubbles at the working electrode, which will cause a gas pressure build-up and, subsequently, a deflection in the thin membrane covering its cavity. The deflecting membrane can then obstruct the channel through which ocular fluid flows in our device.

9

SiNi membrane

Aqueous electrolyte Cavity for electrolyte

cathode (Pt)

reduction: electrolysis, production of 02 bubble.

anode (Pt) counter

working

oxidation: production of H2

Figure 7— Electrochemical Setup

Previous work has shown the feasibility of such an electrochemical cell for the purposes of actuation13. However, significant novel changes are made to optimize the actuation of the electrochemical cell to act as a valve that controls fluid resistance in our device.

4.2 Micro-Actuation & Valve Design 4.2.1 IOP vs. deflection: Valve and Fluid Channel design The fluid flow resistance is adjusted in our device by controlling and varying the inner cross-section of the flow channel, which is referred in this paper as a change in the gap. Initially, we use the following equation to relate the pressure drop across the channel, essentially the IOP across the membrane as it deflects due to the pressure build up of our electrochemical cell.1 2k shape lμ Δp = φ (Equation 18) Dh2 A In Equation 18, A(m2) is the flow area, Dh(m) is the hydraulic diameter which we define as Dh = 4*A/P, P(m) is the wetted perimeter of the flow channel cross-section, kshape is a channel shape-dependent constant that we model as a rectangular structure due to the aspect ratio of our semicircular crosssection. A, is the flow area thus it is calculated as the A = gap*length. Φ is the typical fluid flow rate for eye fluid, μ is the viscosity of the eye fluid, and l is length of the valve. Our device’s fluid channel has a height of 20 μm. All of the aforementioned variable values used in are shown in Table 2. Table 2-Valve and Fluid Channel Parameters gap 1*10-6 - 20*10-6 m length 150*10-6 m A gap*length P 2*(length + gap) Dh (4*A)/P μ 1*10-3 Pa*s Φ 1.67*10^-11 - 5*10^-11 m^3/mi kshape 15.5 l

1*10-3 m

10

Implementing the design values in Table 2 into the Equation 18 and comparing them to device estimates used in the literature1 we generate Figure 8 to compare the change in gap due to the deflection of the silicon nitride membrane to the pressure across the membrane, or in the case of this device the IOP in the device fluid channel. From the figure, we can easily see that to restore the system to a normal IOP of 2250 Pa, the actuator does not have to be fully deflected, thus efficiently utilizing the membrane as an obtrusive valve to control pressure. (In Figure 8, our derived design from the previous calculations is represented in the purple stars, other variations are represented as follows: black circles, l=2mm and length=300μm; blue diamonds, l=1mm and length=300μm; black circles, l=1mm and length=50μm.)

Figure 8—Gap spacing vs. pressure across channel

In the previous simulation, we assumed an eye fluid flow rate of 2μ/ min to 3.3*10-11m3/s. However, looking at Figure 9, which shows the change in fluid flow rate to pressure over the membrane for a gap of 11microns we see that the pressure is dependent on the fluid flow rate (i.e., fluid resistance). Normally, the eye fluid flow rate is 1-3ul/min so using 2ul/min is an accurate approximation for the rate at which the fluid is flowing in our device (however it is a sensitive value and at any rate in the range the pressure can be controlled to the normal base pressure).

11

Figure 9—Fluid flow vs. pressure across membrane

Contingent upon the fact that the fluid flow rate can be modified within the aforementioned typical range, it is now known that the smaller the gap distance, or the higher membrane/valve deflection the pressure drop across the valve can be adjusted to equal normal eye pressure of 2250Pa. Additionally we see that the optimal IOP can be attained without full membrane deflection.

4.2.2 Actuator Pressure vs. Deflection: Corrugated Membrane Design In order to reduce stress on the membrane, increase its flexibility, and thus increase deflection with low actuator pressure, sinusoidal corrugations are added to the SiNi membrane. Using Equation 19, the dimensionless actuator pressure (Pa4/Eh4) is correlated to the deflection, y, of the corrugated membrane.1 Pa 4 y ⎛ y⎞ = Ap + B p ⎜ ⎟ 4 h Eh ⎝h⎠

3

(Equation 19)

where,

where, , 12

, . In Equation 19, h is the membrane thickness, H is the corrugation depth, l is the corrugation frequency, N is the number of corrugations, v is the Poisson’s ratio, and E is the Young’s modulus of SiNi. The parameters q, s, and R are as shown above. The design parameters that we optimized for our membrane are shown in Table 3. Table 3—Membrane design parameters

v E H h l N

0.3 300*103 Pa 9*10-6m 1*10-6m 6*H 13

Implementing Equation 19 with the design parameters in Table 3, the plot in Figure 10 shows that the actuator pressure to fully actuate the valve (20μm) is less than 1bar. Thus, our corrugated membrane design is effective in requiring a low pressure from the electrochemical cell build up of oxygen gas.

Figure 10—Membrane deflection vs. actuator pressure

4.2.3 Cavity channel area vs. O2 loss rate: Cavity Design In order for the electrochemical cell to be complete, the two electrodes must be connected by the electrolyte. The dual cavity design for the electrochemical actuator requires that a narrow channel is fabricated not only to connect the two cavities, and corresponding electrodes, but to also block the 13

diffusion of oxygen to the cathode. Theoretically, the oxygen production rate at the anode can be calculated based on an applied current, i, using Equations 20 and 21. Equation 20 gives the moles of O2 produced for a certain applied current. Equation 21 is simplified for 4 electrons which are required to obtain 1 molecule of O2.1 Q it = (Equation 20) N= nF nF i ⎛ dN ⎞ (Equation 21) ⎟ = ⎜ ⎝ dt ⎠ th 4 F Thus, for example using the theoretical rate of oxygen production, for a 5μA applied current over a typical pressure buildup time of 10 minutes (600s) is 1.2*10-11 mole/s.

Theoretically, the oxygen loss rate through the narrow channel separating the two cavities can also be calculated using Equation 22.14 ⎛ dN ⎞ N N AN C O2 (Equation 22) ⎟ = I O2 = DO2 ⎜ DN ⎝ dt ⎠ loss D02 = kb*T/ (6*pi*η) 2 is the diffusion coefficient of oxygen through the narrow channel connecting the cavities (calculated from the literature), A is the area of that channel, d is the channel thickness, and CO2 is the concentration of oxygen on the outer edge of the channel. In order to optimize loss, Equation 22 is modeled with different channel areas in Figure 11, the thickness must be at least 25*10-6 m to be feasibly fabricated.

Figure 11—Cavity channel area vs. O2 loss

As a result of this analysis the following dimensions describe the channel connecting the cavities,

14

channel thickness of 5*10-6 m, channel length and width of 100*10-6 m for a loss of 1.3e-10 mol/s which is an order of magnitude lower than that of the theoretical oxygen production rate.

4.3 Power consumption of electrochemical cell In order to ensure that the power consumption of the electrochemical cell does not exceed what is possible to supply to the device in addition to that needed by the pressure sensor, Equation 23 is applied. 4 FV gas Δp E cell (Equation 23) P= RT t In Equation 23, Δp is 1 bar, F is the Faraday constant, Ecell is the voltage generated from the electrochemical reactions (~1.1V), t is the pressure build-up time (~600s), R is the gas constant, Vgas is the cell volume (~1mm3) and T is body temperature (310K). If the previous design values are used, we see a power consumption of ~27uW, and if the cell voltage is kept between 1-2V the power will not exceed 50uW.

5 Fabrication The fabrication process that is used to build this integrated device follows the bulk micromachining philosophy. The starting materials consist of a Pyrex wafer which is etched to form the microfluidic network; an n-type SOI wafer with 10 μm device layer and 2 μm buried oxide (BOX) is the platform for the working parts of the sensor and actuator; a double-side polished silicon wafer with pre-patterned copper-filled vias15 will provide contact to the platinum electrodes and will encapsulate the sensor and actuator cavities. These three wafers bond together in a stack after processing to form the completed devices which are then separated and packaged. The fabrication process flow is listed in Table 4. The photolithography processes that are listed will generally consist of HMDS treatment to promote resist adhesion to the substrate, coating the wafer with some type of photoresist, a pre-bake to drive out solvents in the resist and an exposure through the appropriate photomask. After exposure the pattern is developed and a final post-bake is performed. There are several variations to this photolithography process including the use of image-reversal resist for lift-off and the use of dry-film resist. Dry-film resist16 is laminated on the wafer using a specialized tool; this type of resist has the advantage of being a thick, uniform film which can be applied over deeply etched features. Some specialty dry-film resists can span large cavities like the roof of a tent; this allows deep cavities to be protected during etch steps with a film that can be patterned photolithographically. It is important to note that the implanted ions will diffuse during any high-temperature processing. In the proposed process, there are only two steps which could potentially qualify as high-temperature: the anneal and the LPCVD nitride deposition. The nitride deposition, however, is still relatively lowtemperature compared to the anneal: around 750oC compared to 1100oC. Therefore, the majority of ion diffusion takes place during the anneal step and while the subsequent diffusion cannot be ignored, its effect is small. Because there is metal on the wafers prior to fusion bonding, it is desired to anneal the fusion bond at a low temperature on the order of 200-300oC. There is a bonding technology called plasma-assisted-bonding (PAB) which uses an oxygen plasma to pre-activate the surface of the wafers prior to bonding. When PAB is combined with a low-temperature anneal (300oC) it has been demonstrated that similar bond strengths to traditional fusion bonding can be achieved17. A commercial bonder made by Suss MicroTec18 implements this technology. 15

The shadow mask technique is used to deposit aluminum bond-pads for contacting the implanted resistors. Generally, shadow masking is used as a last-resort method. Because the fragile membrane is already released when the aluminum is deposited, this method provides the safest method for depositing the metal without damaging the membrane. The resulting wafer and individual chip are shown in Figure 18. Step Starting Material 1. Photolithography (pos) 10 2. Etch 3. Strip Resist 4. Clean 5. Oxidation 6. Photolithography (pos) 10 7. Ion Implant 8. Strip Resist 9. Photolithography (pos) 10 10. Ion Implant 11. Strip Resist 12. Clean 13. Anneal 14. Photolithography (pos) 15. Etch 16. Strip Resist 17. Photolithography (pos) 18. Etch 19. Strip Resist 20. Clean 21. Deposit Oxide 22. Deposit Nitride 23. Photolithography (pos) 10 24. Etch 25. Strip Resist 26. Photolithography (pos) 10 27. Etch 28. Strip Resist 29. Photolithography (pos) 30. Etch19 31. Strip Resist 19 32. Etch 33. Clean 34. Strip oxide 16

35. Dry film resist coat 36. Photolithography (neg) 37. Etch 38. Strip Resist

Table 4—Fabrication Description SOI Mask 0: Alignment Marks. Standard 1 μm resist. Dry etch into silicon. Cl2 etch chemistry. Ash (O2 plasma) and piranha RCA Grow 20 nm thermal oxide Mask 1: Piezoresistor. Standard 2 μm resist. p+ implant: NA = 1017 cm-3, final junction depth 0.5 μm Ash and piranha Mask 2: Trace Implant. Standard 2 μm resist. p++ implant: NA = 1022 cm-3, final junction depth 2 μm Ash and piranha RCA High temp anneal to drive in implants Mask 3: Corrugations. Standard 6 μm resist. Dry etch 8 μm into silicon. SF6 etch chemistry for a fairly isotropic etch profile. Ash Mask 4: Membrane. Thick, 10 μm resist. Dry etch 1 μm into silicon. SF6 etch chemistry; smooth sharp edges on corrugations and recess 1 μm. Ash and piranha RCA 3 μm PECVD oxide on backside of wafer 1 μm LPCVD low-stress nitride Mask 5: Sensor. Thick, 10 μm resist Dry etch 1 μm silicon nitride, etch stop on silicon. CHF3/O2 etch chemistry; sensor membrane and metal contact areas. Ash and piranha Backside of wafer. Mask 6: Nested Cavity. Standard 2 μm resist. Dry etch 1 μm nitride, 3 μm oxide, etch stop on silicon. CF4/H2 etch chemistry Ash and piranha Backside of wafer. Mask 7: Cavity. Standard 10 μm resist. DRIE 100 μm into silicon (timed etch). SF6/C4F8 etch chemistry. Ash DRIE 400 μm into silicon, etch stop on buried oxide. SF6/C4F8 etch chemistry Piranha (no ash) BOE or HF vapor to remove remaining PECVD oxide and buried oxide. Laminate both sides of wafer with dry film resist Backside of wafer. Mask 8: Membrane release XeF2 etch approx 10 μm to release corrugated membrane Ash

16

Starting Material 1. Photolithography (Lift-Off) 2. Deposit Metal 3. Liftoff 4. Photolithography (pos) 5. Etch 6. Strip Resist 7. Photolithography (Lift-Off) 8. Deposit Metal 9. Liftoff 10. Photolithography (pos) 11. Etch 12. Strip Resist 13. Clean 17 14. Bond 15. Anneal 16. Metal Deposition

Cu-Via Wafer Backside of Wafer, Mask 9: Solder bump pads. Image-reversal resist. eBeam 20 nm titanium (adhesion layer), 500 nm copper Acetone, then solvent rinse Mask 10: Electrolyte Channel. Standard 2 μm resist. Dry etch 250 nm into silicon (timed etch). Cl2/HBr etch chemistry. Ash and nanostrip Mask 11: Pt Electrodes. Image-reversal resist. eBeam 20 nm titanium (adhesion layer), 500 nm platinum Acetone, then solvent rinse Mask 12: Fill Holes. Double thick 20 μm resist (2 coats, 10 μm each) DRIE through wafer. SF6/C4F8 etch chemistry Ash and nanostrip Nanostrip Plasma-assisted fusion bond Low-temp (200-300oC) anneal. Align shadow mask (Mask 13), deposit aluminum for metal contacts to implants

Starting Material 1. Clean 2. Deposit Metal 3. Photolithography (pos) 4. Etch 5. Etch 6. Strip Resist 7. Strip Metal 8. Drill Holes 9. Clean 16 10. Dry film resist coat 11. Diesaw 12. Clean 13. Bond 16 14. Dry film resist coat 15. Photolithography (neg) 16. Deposit Metal 17. Liftoff 18. Clean

Pyrex Wafer Piranha Sputter 50 nm chromium (adhesion layer), 500 nm gold Mask 14: Glass Channel. Standard 2 μm resist. Etch gold (KI:I2 gold etchant) and chromium (CR-7 chrome etchant) Wet etch glass (HF:HNO3:H2O mixture) 20 μm deep (timed etch) Ash or nanostrip Gold etchant and chromium etchant Mask 15: Port-Holes. Use ultrasonic drilling. Piranha Laminate both sides of Pyrex wafer Partially diesaw etched side of Pyrex to raise the lanes for bond-pads Piranha Anodic bond Pyrex to top side of SOI stack. Laminate both sides of stack with thick dry film resist Topside of wafer. Mask 16: CCW Layer eBeam 50 nm chromium (adhesion layer), 2 μm gold for compression cold welding (CCW) Acetone and solvent rinse Extended oxygen plasma ash

Step Starting Material 1. Clean 2. Oxidation 3. Metal Deposition 5. Photolithography (pos) 6. Electroplate 7. Strip resist 8. Remove unwanted seed layer 9. Diesaw 10. Clean 11. Attach wires 12. Encapsulate in PDMS

Table 5—Coil Fabrication Description Silicon RCA Grow 1 μm of wet thermal oxide Sputter 20 nm chromium, 500 nm copper Mask 17: Coils. Double thick 20 μm resist. Plate 11 μm of copper. Acetone/solvent clean plus nanostrip Ion Beam Etching to remove Cr/Cu between coils Cut into individual coil segments Solvent clean plus plasma ash Solder-bond wires for connection to package Silicone rubber protects device, also gets coated with parylene later.

17

Mask Mask 1: Piezoresistor Mask 2: Trace Implant Mask 3: Corrugations Mask 4: Membrane

Figure Figure 12 Figure 12, Figure 13 Figure 13 Figure 13

Mask 5: Sensor

Figure 13

Mask 6: Nested Cavity

Figure 14

Mask 7: Cavity

Figure 14

Mask 8: Membrane Release Mask 9: Solder Bump Pads Mask 10: Electrolyte Channel Mask 11: Pt Electrodes

Figure 14

Mask 12: Fill Holes Mask 13: Al Contacts

Figure 16 Figure 13

Mask 14: Glass Channel Mask 15: Port Holes Mask 16: CCW Layer Mask 17: Coils

Figure 17 Figure 17 Figure 17 Mask not depicted

Figure 15 Figure 16 Figure 16

Table 6—Mask Key Description Small, green boxes are exposed (overall dark field mask) Larger, blue regions are exposed (rest dark field) Co-centric “ring” pattern is exposed (rest dark) Box spans all corrugations is exposed (rest dark); will also expose same areas as previous mask (point is to round corners) Square box spanning sensor membrane and small squares at the end of “trace implants” are exposed (rest dark) Three large squares are exposed (rest dark), this is a mirror image mask because it is on the backside of wafer Two large squares and elongated rectangle are exposed (rest dark), again on back of wafer Negative resist: Bright-field mask, large, center box (under membrane) is dark. Again on backside of wafer. Image Reversal Resist: Bright-field mask, large brown regions are dark. Mirror image; backside of Cu-via wafer. Single square is exposed (rest dark) Image Reversal Resist: Bright-field mask, bigger white squares are dark. Two, small squares are exposed (rest dark) Shadow mask, the small square features are holes in the shadow wafer, rest is solid Inside channel (where numbers are located) is exposed (rest dark) Two circular areas are drilled away Negative resist: Bright-field mask, yellow ring is dark The coils themselves would be exposed and the rest of the mask would be dark. The dimensions are for those listed in Table 6.

Figure 12

18

Figure 13

Figure 14

Figure 15

19

Figure 16

Figure 17

Figure 18

20

6 Packaging Throughout the packaging process Table 7, it is desirable to keep the devices as clean as possible. During die saw steps a UV-release tape can be used to prevent unwanted water and particles from entering the various open ports on the device. The electrolyte filling step is relatively straightforward: by immersing the device in the solution and pulling vacuum any bubbles that are trapped in the cavities get pulled out and electrolyte fills all cavities. After the fill-holes are sealed the device can be placed back into a vacuum chamber to evacuate the fluid out of the microfluidic channels. The compression-cold-welding20 (CCW) process is key in the packaging of the integrated device. Because there is electrolyte and metal present in the device, the packaging should be done at a low temperature. CCW provides a hermetic seal between a glass or silicon chip and lid and between the lid and base of a package; this allows for the glass face of the device to be exposed in the body and greatly simplifies the interface between the silicone-rubber tubes and the chip. The CCW process entails complimentary tongue-and-groove features made of some sort of metal: gold, platinum, aluminum, titanium and combinations have been demonstrated. The tongue-groove should be 1-100 μm high. They are aligned and compressed together using a pressure on the order of 0.5-1 GPa. Care must be taken to avoid fracturing the device, but most importantly is that this process takes place at or below room temperature. The parylene deposition is done in two steps. During the first deposition, the interior of the fluidic network is exposed. This step conformally coats the surface of the device that interacts with the aqueous humour. During the second, thicker parylene deposition the microfluidic section is clamped off so that large build-up of parylene does not clog the channels. The final device is completely biocompatible: the silicone rubber tubes, glass surface of the chip, PDMS encased coils Table 5, and hermetically sealed titanium package are all biocompatible. The full device is additionally coated in 5 μm of parylene to enhance the biocompatibility and provide further protection. A simulated version of the entire device package is shown in Figure 19 and 20.

Packaging Step 1. Tape 2. Diesaw 1 3. Diesaw 2 4. Remove Tape 5. Clean 13 6. Fill Electrolyte 7. Cap fill-holes 8. Mount chips to bench 9. Wirebond 10. CCW 11. Attach Inductor Coils 12. Parylene Deposition 13. Attach silicone rubber tubing

Table 7—Packaging Description Laminate both sides of wafers with diesaw tape Diesaw partway through Pyrex to release strips over bond-pads Cut wafer into individual devices UV release tape Ash Immerse device in electrolyte solution and place in vacuum chamber (pulls bubbles out of cavities). Seal holes with epoxy. Solder bumps on backside of device connect device chip to bench and electrically connect Cu-vias for Pt electrodes. Batteries and circuitry mounted in separate “chamber” in the package with solder or epoxy. Make appropriate electrical connections. Connections between “chambers” in the package and to external electrical components are integrated into the package design. Compression cold welding to hermetically seal the package (patent from MicroCHIPS21). Wire connection to coils is encased in silicone polymer Deposit 0.5 μm of Parylene in deposition chamber. Will conformally coat the outside of package, wires and inductor coils as well as the inside of the microfluidic channel. Insert into drilled holes and epoxy into place

21

14. Parylene Deposition 15. Test

Clamp silicone rubber tubing and deposit 4 μm more parylene as extra protection. Remove clamps. Packaging finished, test devices.

Figure 19

Figure 20

7 Feedback Controller In the device, the electrochemical actuator acts as a valve, deflecting a membrane, which is used to control the IOP which changes the eye fluid resistance in the channel. The IOP is measured and monitored using the piezoresistive sensor placed prior to the actuator membrane in the device channel. When the eye fluid generation and removal are balanced a constant base pressure is achieved, eliminating the onset of glaucoma-related disease13. In order for this to occur, the pressure sensor and actuator are used to form a regulating system via a feedback controller circuit as shown in Figure 21.

22

Figure 21—Feedback controller circuit

Represented in detailed circuit form, the following is an explanation of the feedback controller circuit. The filtered voltage from the sensor (mV), which corresponds to the pressure sensed by the piezoresistors, is the input of the circuit. This input voltage is then fed into a 100 fold amplifier (Opamp). This signal is then sent into another op-amp(Op-Amp1) which is a comparator with an integrator built around it to pre-set the cell pressure which is represented by a DC voltage source for modeling purposes (this DC source is actually a trim-potentiometer which is able to maintain a stable pressure inside the electrochemical cell). The output signal of the comparator is then integrated to filter out any noise and then fed into the transistor (BD139), or current amplifier whose output current will be sent across the resistor R3 (representing the current supplied to the electrochemical system to drive the electrolysis and subsequent actuation of the membrane). The values for the circuit elements are as follows: R1 = 100kΩ, R2 = 1kΩ, R3 = 100Ω, C1 = 100nF, C2= 10nF. It must be noted that the system dynamics are very slow, the IOP pressure changes on the order of minutes to hours and the eye fluid cycles every 100 hours. However the actuator reacts on the order of a couple of minutes (~10) making the proportional control of the system relatively simple.

8 Device Power In order to power the device with relatively low and wireless power on a MEMS scale, a thin film lithium-ion battery is employed. The battery is to be manufactured by Oak Ridge Micro-Energy, Inc.22 As shown in Figure 22 the capacity of the battery can be 50μAh at 37°C (body temperature) and supply a potential from 2-3.8 volts which is more than needed for the device. The battery is less than 1mm thick and the area is determined by the method of packaging, however it is also adequate to fit into our package.

23

Figure 22—Capacity vs. voltage potential for battery

To cycle (sense the pressure and adjust the micro-actuator) every half hour, at this capacity 210 cycles are performed before the battery would need to be recharged, thus the battery would only need to be recharged approximately every four days.

9 Calibration Our device needs to be calibrated in two areas: 1) the pressure sensor needs to be calibrated to relate fluid pressure to output voltage, and 2) the actuator needs to be calibrated to relate desired fluid resistance to input current. The pressure sensor can be calibrated by pumping fluid through a tube that is connected to the inlet port of our device (e.g., by using a syringe pump). A commercial pressure sensor can be integrated into the tube so that the measured pressure can be related to the signal read by the actuator, resulting in a calibrated pressure-voltage curve. Similarly, the actuator can be calibrated by applying a known pressure at the inlet and measuring the outflow rate with a commercial MEMS flow meter. Both forward and reverse current should be applied to measure the membrane deflection and its response time.

10 Telemetry Device operation can be monitored using inductively coupled telemetry. An inductor microcoil package (Figure 23) is included with the device to allow for remote communication with an outside source. The inductor coil will be implanted closer to the skin and connected to the device using insulated copper wires. Studies show that wireless energy and data transfer is this microcoil technology23. The inductive coil can communicate to the Doctor, and recharge the battery which will occur approximately once a week. The remote communication unit will be attached to glasses and the patient will wear them when recharging the thin film battery.

24

Figure 23—Inductor microcoil package

11 Simulated Performance The overall device performance was modeled using Simulink (Figure 24), and the resulting output pressure is plotted alongside disturbance pressure in Figure 25. The simulation is modeled by including the controller system discussed earlier and the actuator as the plant. A disturbance pressure is introduced afterwards which models the fluctuating IOP, and the loop is closed through the piezoresistor and voltage amplifier.

Figure 24—Simulink diagram for overall device performance.

The results of the simulation support the functionality of the device. We see that a 2000 Pa disturbance from a baseline IOP of 2250 Pa is controlled by using the device. There was some error in the control loop; however, the resulting pressure is within acceptable range. Limitations are apparent once the sensed IOP increases above 4000 Pa, yet the controlled pressure is still within the acceptable range. This limitation is due to the size of the valve which has an upper limit to the pressure drop it can provide.

25

Figure 25—IOP pressure disturbance simulation.

12 Conclusions Glaucoma is a common disease, and several different techniques are being explored for treating the disease. Integrating MEMS technologies are being explored to monitor and treat increasing IOP levels. A piezoresistive pressure sensor was designed to detect IOP ranging from 500-5000 Pa with a resolution of 130 Pa, and the 1-10 mV signal output was greater than voltage output due to noise. The low powered electrochemical valve actuator incorporated a corrugated membrane to control IOP using a controlled drainage mechanism. The valve actuator would have to be around 50 percent deflection to maintain a normal IOP of 2250 Pa, and its dual cavity design allows for reversible reaction. The primary power source for the device is a thin film battery which is ideal for addressing its low power requirements. Inductor coils are used as a secondary power source for weekly charging of the device. Additionally, these coils will allow for communication with a Physician as well and monitoring the IOP set point. Slow system dynamics supports the use of proportional control, which results in a reasonable response. The overall simulation including this control logic and actuator and sensor dynamics confirmed the expected function of this device. This device can be fabricated using bulk micromachining and wafer bonding. Device packaging utilizes biocompatible polymers and results in a hermetically sealed device. Our analysis and stimulation results support the utility of this device and the size and implantation techniques are comparable to current devices. 26

13 References 1

Titcomb, Lucy C., Treatment of glaucoma: part 2, online < http://www.pjonline.com/Editorial/19990828/education/glaucoma.html> 2 Titcomb, Lucy C., Treatment of glaucoma: part 1, online < http://www.pjonline.com/editorial/19991002/education/glaucoma.html> 3 National Eye Institute, online < http://www.nei.nih.gov> 4 Creemer. J. Fredrick, F. Fruett, G. Meijer, P. French, The Piezojunction Effect in Silicon Sensors and Circuits and its Relation to Piezoresistance, IEEE Sensors Journal, Vol 1, No.2, August 2001 5 Kanda, Yozo, A. Yasukawa, Optimum design considerations for silicon piezoresistive pressure sensors, Sensors and Actuators A 62, 2007 6 Szentpáli, Béla, M. Ádám, T. Mohácsy, Noise in piezoresistive Si pressure sensors, Proceedings of SPIE: Noise and information in nanoelectronics, sensors, and standards III, Austin, TX, 24-26 May 2005. 7 Senturia, Stephen D., Microsystem Design, pgs. 472—477, Spring Science+Media, Inc., New York, NY, 2001. 8 Bae, Byunghoon , K.Park, M.Shannon, MEMS Application of Actuators and Sensors for Glaucoma Treatment (from MEMS/NEMS Handbook, Volume 5), Spring Science+Media, Inc., New York, NY, 2006. 9 Pramanik, C., H. Saha, U. Gangopadhyay, Design optimization of a high performance silicon MEMS piezoresistive pressure sensor for biomedical applications, Journal of Micromechanics and Microengineering, 2006. 10 Plummer, James D., M. Deal, P. Griffin, Silicon VLSI Technology, pg. 18, Prentice Hall, Upper Salle River, NJ, 2000. 11 Mohammed, Ahmed A.S., W. Moussa, E. Lou, High Sensitivity MEMS Strain Sensor: Design and Simulation, Sensors, 2008. 12 Lecture notes for MIT course 6.777—Design and Fabrication of MEMS, Lecture 13: The Thermal Domain II, Spring, 2008. 13 Neagu, C. R. (1996). A Medical Microactuator based on an Electrochemical Principle, University of Twente, The Netherlands. PhD Dissertation: 162. 14 Chin, F.G.a.D.T. (2004). “Determination of diffusivity and solution of oxygen in phosphoric acid using a transit time on a rotating ring-disc electrode.” Journal of Applied Electrochemistry, 23(4): 452-455. 15 IceMos Technology, “SOI Products.” [Online]. Available: http://www.icemostech.com/ice/interconnect.aspx. [Accessed: 13 May 2008]. 16 DuPont, “Dry Film Photoresists,” www2.dupont.com. [Online]. Available: http://www2.dupont.com/Imaging_Materials/en_US/products/dryfilmPhotoresist/index.html. [Accessed: 13 May 13, 2008]. 17 M. Visser et al., “Strength and leak testing of plasma activated bond interfaces,” in Sensors and Actuators A, 1997-1998, pp. 434-440. 18 Suss MicroTec, “nP200 / nP12 Surface Activation System,” www.suss.com. [Online]. Available: http://www.suss.com/products/wafer_bonder/automated_bond_cluster/np200_np12. [Accessed: 13 May 2008]. 19 Madou, Marc J., Fundamentals of Microfabrication, CRC Press LLC, Boca Raton, FL, 2000. 20 J. Coppeta, et al., “Compression and cold weld sealing methods and devices,” U.S. Patent 20060115323, 1 June, 2006. 21 MicroCHIPS. [Online]. Available: http://mchips.com. [Accessed: 13 May 2008]. 22 Oak Ridge Micro-Energy, I. (2008, 4/18/08). "Thin Film Lithium-Ion Batteries." from http://www.oakridgemicro.com/tech/specs.pdf. 23 Kim, Sohee, and Oliver Scholz. "Implantable Active Telemetry System using Microcoils." Proceedings of the 2005 IEEE 27 (2005).

27