Integrated electrode and high density feedthrough system for chip-scale implantable devices

at SciVerse ScienceDirect

Biomaterials xxx (2013) 1e10

Contents lists available

Biomaterials

journal homepage: www.elsevier .com/locate/biomater ia ls

Integrated electrode and high density feedthrough system forchip-scale implantable devices

Rylie A. Green a,*,1, Thomas Guenther a,1, Christoph Jeschke a, Amandine Jaillon a,b,Jin F. Yu a, Wolfram F. Dueck a,c, William W. Lim a, William C. Henderson a,Anne Vanhoestenberghe d, Nigel H. Lovell a, Gregg J. Suaning a

aGraduate School of Biomedical Engineering, Sydney, NSW 2052, Australiab Engineering, École Centrale de Lyon, 69130 Écully, Francec Electrodes and Interfaces Technology Cluster, Cochlear Ltd., Sydney, NSW, Australiad Implanted Devices Group, University College London, London WC1E 6BT, UK

a r t i c l e i n f o

Article history:Received 9 April 2013Accepted 29 April 2013Available online xxx

Keywords:High-density arrayPlatinum electrodesAluminaFeedthroughsHermetic

* Corresponding author.E-mail address: [email protected] (R.A. Green)

1 Joint lead authors.

0142-9612/$ e see front matter Crown Copyright � 2http://dx.doi.org/10.1016/j.biomaterials.2013.04.054

Please cite this article in press as: Green RA,Biomaterials (2013), http://dx.doi.org/10.101

a b s t r a c t

High density feedthroughs have been developed which allow for the integration of chip-scale featuresand electrode arrays with up to 1141 stimulating sites to be located on a single implantable package. Thislayered technology displays hermetic properties and can be produced from as little as two laminated200 mm thick alumina sheets. It can also be expanded to a greater number of layers to allow flexiblerouting to integrated electronics. The microelectrodes, which are produced from sintered platinum (Pt)particulate, have high charge injection capacity as a result of a porous surface morphology. Despite theirinherent porosity the electrodes are electrically stable following more than 1.8 billion stimulation pulsesdelivered at clinically relevant levels. The ceramic-Pt constructs are also shown to have acceptablebiological properties, causing no cell growth inhibition using standard leachant assays and supportneural cell survival and differentiation under both passive conditions and active electrical stimulation.

Crown Copyright � 2013 Published by Elsevier Ltd. All rights reserved.

1. Introduction

There is a growing need for implant technologies that enablehigh-density electrode arrays to be placed within organs, providinghigh resolution control of physiological processes. This is likely toinclude devices such as the bionic eye and cochlear implant, wherea greater number of electrodes are expected to improve the qualityof vision or sound perception, respectively, experienced by thepatient. A major limitation to realising such devices is the need formanufacturing processes which enable stimulating electrodes to bedeveloped with very small features in close proximity to the targetneurons, while preserving the need for discreet, addressablechannels and device hermeticity. Key to addressing this challenge isthe design of feedthroughs (vias) from the electrode surface orarray connection point to the internal, encapsulated electronics.This integral pathway must be produced in a reliable and robustmanner which preserves device efficacy and safety across chronicimplantation time frames (decades). This includes not only her-metic, conductive and isolated channels, but also an insulative

.

013 Published by Elsevier Ltd. All

et al., Integrated electrode and6/j.biomaterials.2013.04.054

material which is compatible with the biological environment andtissue in which it is placed.

Polycrystalline ceramics have been investigated for severaldecades as potential materials for implant encapsulation packagesand feedthrough substrates [1e4]. They are an advantageous choicedue to their inherent low water permeability and high stability incorrosive environments [5,6]. As such, ceramics based on highpurity Al2O3 have been extensively investigated and incorporatedinto regulatory approved Class III implantable devices for hermeticsealing [7,8]. Other ceramics such as aluminium nitride (AlN), zir-conia (ZrO2), silicon carbide (SiC), and silicon nitride (Si3N4) havebeen investigated for ceramic-to-metal assemblies in these devices[1]. In fact ceramic-to-metal feedthroughs are considered one ofthe most promising options, being robust, durable, and havingtighter hermeticity and better electrical insulation properties thanthe alternative glass-to-metal feedthroughs or polymer encapsu-lation technologies. However, ceramics have significant processingchallenges, requiring high temperature firing that leads to materialshrinkage and the introduction of internal stresses, exacerbated bythe dissimilar materials at the ceramic-metal interface. As a result,these connections have only been used for relatively large feed-throughs at modest density, with the smallest, most advancedmethods producing pitches in the range of 200e600 mm [2].

rights reserved.

high density feedthrough system for chip-scale implantable devices,

R.A. Green et al. / Biomaterials xxx (2013) 1e102

Ceramic-to-metal feedthroughs are typically produced by thecreation of vias through the insulating ceramic and then brazing orco-firing metal pins or screen printing metal paste through theseholes. Sintering of the ceramic is also required, and if the process isnot adequately controlled, voids form throughout the metal filledvias, reducing hermeticity and potentially creating open circuits.Current fabrication processes have resulted in limited high densityelectrical feedthroughs which are time intensive to produce, sufferfrom low repeatability, lack scalability and have significant processfailure modes [1,2]. Of particular importance Auciello and Shi notethat there is no hermetic microchip-size coating that provides areliable, long-term performance as an encapsulating coating [8].

One of the major obstacles to the design of micron-scale her-metic feedthroughs is the difficulty in dealing with the thermalexpansion coefficients of dissimilar materials. A common mode offailure for feedthroughs is cracking at the interface of the conduc-tive and insulative components as the internal material stressexceeds the strength of the material with varied temperature. Inprevious studies Suaning et al. demonstrated that the use of hori-zontal tracks through bulk alumina can produce feedthroughs withreduced complications from interfacial stress cracking [9]. How-ever, to enable the development of miniaturised, high density ar-rays, there is a need to increase the packing efficiency of thesetracks. As a result of this work, a multilayer system has beendeveloped to enable substantial increases in feedthrough packingby utilising the bulk surface of the implantable electronics package.

Vertical access vias facilitate connection to the horizontal her-metic tracks from either side of the implant package, as shown inFig. 1. This “kink” design is also conducive to the application ofcompressive forces during fabrication which are required to formstable Pt-ceramic bonds described by Allen and Borbige [10]. Thevertical section of this feedthrough design is achieved by creatingplatinum (Pt) paths through both the upper and lower ceramiclayers. They are not required for preservation of hermeticity e

instead, hermeticity is achieved solely by the horizontal bridgingtrack. They do however facilitate flexible routing from the outputpads of a neurostimulation integrated circuit to an array of micro-electrodes. As a result, this design creates a technology platform forsingle assembly implantable packages, where the electronics aresafely encapsulated within the hermetic package, and the outersurface can directly stimulate neural tissue through a high densityarray of microelectrodes.

An important step in realising such a design is the functionalcharacterisation of materials and methods used to create the as-sembly. To ensure robustness in the design and fabrication tech-niques, preservation of hermeticity will be confirmed and theelectrode arrays characterised in comparison to state-of-the-artmicromachined Pt foil electrodes. The tolerance of the biologicalenvironment to the materials and potential fabrication by-productsneeds to be established including the efficacy of electrode arrays as

Fig. 1. Design for hermetic single assembly neurostimulation packages interfacin

Please cite this article in press as: Green RA, et al., Integrated electrode andBiomaterials (2013), http://dx.doi.org/10.1016/j.biomaterials.2013.04.054

neural interfaces in direct contact with target tissue. To this end, amethod for fabricating thin, high density ceramic-to-metal feed-throughs with integrated electrodes is presented using aluminaand Pt particulate to produce a composite material where theresulting electroactive sites are Pt bound with trace alumina. Thefabrication of this multilayer system is detailed and baselineproperties including feature size and hermeticity are characterised.The charge injection limit of these electrodes and stability understimulation is explored and cell compatibility is investigatedthrough cell growth inhibition studies and neural cell culture underimplant-specific electrical stimulation.

2. Materials and methods

2.1. Fabrication of multi-layer feedthroughs

In its most simplistic form, a typical feedthrough with chip bonding and elec-trode sites can consist of a two layer structure. Alumina (Al2O3) particulate, sus-pended in wax binder was formed into a tape with a thickness of 200 mm for each ofthe layers. An Nd:YAG laser (DPL Genesis Marker, ACI, Nohra, Germany) with ascanner head was used for micromachining processes with micro-scale definition.The upper layer carries the electrode sites and the base layer carries the chip bondpads. In between these layers, horizontal tracks are formed to provide hermeticbridging tracks, connecting the chip to the electrode sites. This system can beexpanded to a multilayer assembly to achieve flexible routing through addition ofone or more middle layers carrying directionally-patterned, horizontal, hermetictracks. A two ceramic layer system is depicted in Fig. 2 and detailed below.

The alumina tape was mounted on a microscope slide to aid alignment and lasermicromachining was used to produce grooves and vias. Pt particles, suspended in athinner (Terpeniol), were screen printed to fill these features. Subsequently, thepaste was dried in a convection oven (S.E.M, 25G, Tech Trader, Aust.) for 15 min at130 �C to extract the volatile solvents from the paste. The base layer was flipped andthe entire area was screen printed with Pt paste using a 50 mm thick stainless steelfoil as a stencil. After another drying step, the paste was patterned with the aid of alaser to form the horizontal tracks.

The alumina sheets were then aligned and laminated by placement in a uniaxialhot press (model 4368, Carver, Wabash, IN, USA) for 10 min at 80 �C and 50 MPa tomelt the wax binders and integrate the surfaces of the ceramic layers. The entireassembly was placed in a furnace (STT-1700-6, Sentro Tech, Strongsvill, OH, USA)and sintered at 1500 �C for 180 min. Sinter setters (Keralpor, Kerafol, Eschenbach,Germany) were used to ensure flatness of the substrates. After firing, the assemblywas polished on both sides with diamond polishing pads (MD1200, Struers, Willich,Germany) to remove excess Pt paste and debris from the surfaces, leaving embeddedPt pads in alumina behind.

2.2. Feedthrough performance

2.2.1. Feature sizeHigh density electrodes with feedthroughs were fabricated to produce a 1141

channel electrode array using a two layer system. To achieve this, the inner lami-nated side of the base layer was patterned with the hermetic tracks, using themethodology described above. Images were taken via a stereoscope (Leica M80 withIC80HD camera, Wetzlar, Germany) and features measured via the Leica Suitesoftware to yield average feature width and pitch. Film thickness was also measuredby creating samples that were cross-sectioned prior to analysis. To enable sectioningof the thin ceramic assembly, samples were embedded in epoxy resin, cut with adiamond tip microtome and polished using a diamond polishing pad and diamondslurries (MD1200, DiaPro DAC and DiaPro NAP, Struers, Willich, Germany).

g with neural tissue. Shown as visual prosthesis with epiretinal placement.

high density feedthrough system for chip-scale implantable devices,

Fig. 2. Fabrication process for producing two layer hermetic assemblies with chip bonding pads and electrode sites.

R.A. Green et al. / Biomaterials xxx (2013) 1e10 3

2.2.2. HermeticityPreliminary hermeticity testing was conducted using a helium (He) leak de-

tector (HLT 560 Smart Test, Pfeiffer Vacuum GmbH, Asslar, Germany) with a sensi-tivity limit of 1.0 � 10�12 atm�cm3/s for He spray testing and standard military andaerospace protocols for microelectronics as outlined in MIL-STD-883.1014.A4,Environmental tests for sealing [11]. A custom designed adaptor was fabricated tohold the ceramic discs on the leak detector using a 2�10�6 atmvacuum beneath theceramic/feedthrough disc. He was sprayed above the sample to serve as a sense gasto be detected in the event of any leaks. Noisewas reduced through use of a dynamicfilter function which enabled background He to be subtracted from the measuredsignal.

A scripting programme utilising Matlab Software (Mathworks, Australia) facil-itated the recording of the leak rate data via the RS-232 interface of the leak detector.Differential measurement of He leakage was determined for five, two layer samples,each with 1141 feedthroughs.

2.3. Electrode performance

To simplify electrode performance testing, multilayer feedthrough devices weredesigned and fabricated to produce an array with two hexagonally (hex) arrangedelectrode configurations, shown in Fig. 3. These arrays were produced to enableelectrical and biological characterisation using a configuration which reflectedprevious implant electrode arrays developed for visual prostheses [12e14]. Themethod used to create these devices is identical to that described above in 2.1, withmodification of the uppermost layer to allow routing to 14 electrodes, each of200 mm diameter. This diameter of the electrodes was chosen for comparison tostate-of-the-art laser micromachined electrodes, which have been previouslydescribed in Dodds et al. [12]. The electrodeswere initially characterised by scanningelectron microscopy (JEOL Neoscope, Japan) with images taken at 10 kV and opticalprofilometry (Contour GT-K, Bruker, Singapore) at 40� magnification. Profilometryscans were used to calculate the surface index (real surface area/geometric area).X-ray photoelectron spectroscopy (Kratos XSAM800, Kratos Analytical Ltd., UK) was

Please cite this article in press as: Green RA, et al., Integrated electrode andBiomaterials (2013), http://dx.doi.org/10.1016/j.biomaterials.2013.04.054

also used to chemically analyse the electrodes and determine the purity of the Ptsurface.

2.3.1. Charge injection limitCharge injection limit of electrodes formed from Pt particulate was determined

using established methods [15e18]. Briefly, electrodes were immersed in phosphatebuffered saline (PBS) which was sparged with N2 gas for 20 min to eliminate un-bound oxygen. A three electrode cell was formed with a large, low impedance Ptcounter electrode, an isolated Ag/AgCl reference electrode (3M KCl) and the Pt testelectrode. A cyclic voltammetry (CV) sweep was performed to confirm the voltagelimits of the water window were at �600 mV and 800 mV versus the Ag/AgClreference for ceramic-Pt electrodes. Subsequently, the three electrode cell wasconnected to a biphasic stimulator (custom made, constant current device withinter-stimulus shorting) and cathodic-first pulses were delivered with a fixed phaselength and amplitude. The resulting voltage transient was captured on an oscillo-scope (TDS2004B, Tektronix, USA) and the residual voltage remaining on thecathodic electrode, Emc was recorded directly following reversal of the charge in-jection. A schematic of this waveform has been previously published in Ref. [15]. Thecurrent was incrementally increased until Emc reached �600 mV, defining the limitof charge injection for this phase length. This process was repeated for a range ofphase lengths from 0.1 to 0.8 ms. Comparisons of injection limits were made byparallel assessment of Pt foil electrodes fabricated using established techniques[12,15] where the electrodes were laser micromachined and embedded in a poly(-dimethyl siloxane) (PDMS) insulation. Both the ceramic-Pt and PDMS-Pt arrays hadthe same twin-hex design with 200 mm diameter electrodes.

2.3.2. Biphasic stimulation stabilityTo ensure electrodes produced from sintered Pt particulates were electro-

chemically stable, a stimulation stability study was conducted. The ceramic sub-strate electrode arrays were clamped in a sandwich assembly, which produced acontained well overlying each pair of hex electrode arrays. Nitrogen gas sparged PBSwas placed in each well, covering the electrodes and completing the circuit.

high density feedthrough system for chip-scale implantable devices,

Fig. 3. Schematic representation of electrode constructs on outer layer using multilayer ceramic-Pt paste feedthrough technology. Two hexagonal (hex) arrays are connected to fourbond pads, representing two central electrodes, each with a hex guard.

R.A. Green et al. / Biomaterials xxx (2013) 1e104

Constant current, charge balanced (symmetric), biphasic stimulations were appliedbetween the centre electrode and surrounding guard electrodes of one hex. Thisstimulation configuration has been proposed to contain the electric field, preventingundesirable current spread beyond the target tissue [19]. The adjacent hex on eachceramic substrate was used as a passive control with no stimulation, except duringmeasurement. A cathodic-first stimulus with amplitude of 220 mA and phase lengthof 0.4 ms (67.22 mC/cm2 per phase) was applied with an inter-stimulus interval of0.1 ms during which all electrodes were shorted together. Stimulation was carriedout for 21 days at 1000 Hz, equivalent to 1.8 billion stimulations. Prior tocommencing and at the conclusion of the study, the voltage transient of the stim-ulation was recorded from an oscilloscope and the amplitude of the waveform wasnoted for both active and passive hex arrays.

CV and electrochemical impedance spectroscopy (EIS) were used as additionalmetrics to assess electrode performance. To create the three electrode cell, a large,low impedance Pt counter electrode was introduced to the wells and an isolated Ag/AgCl reference electrode was used to record the interfacial voltage of the ceramic-Ptelectrodes under test. An eDaq eCorder and potentiostat (eDaq, Australia) were usedto perform both studies. CV was conducted by cycling the voltage between�600mVandþ800mV vs Ag/AgCl at 150mV/s for 50 cycles. The charge storage capacity (CSC)was calculated from the final cycle for each electrode. The EIS was conducted byapplication of 50 mV sinusoids at 1e10,000 Hz. A Bode plot was used to comparechanges in Pt performance before and after 1.8 billion biphasic stimulations.

2.4. Compatibility with biological systems

As the materials and methods used to fabricate these ceramic-Pt implants arenot yet tested in chronic preparations, there is a need to assess their impact onbiological systems. These properties were investigated through two studies, firstly acell growth inhibition study based on the ISO standard 10993-5 [20], followed by astudy intended tomore closely represent the implant environmentwhere the neuralmodel cell, PC12, was cultured directly on electrode arrays which were electricallystimulated.

2.4.1. Cell growth inhibitionThe cell growth inhibition studywas carried out with extracts of the ceramic and

Pt assembly, at each step of manufacture. The extracts were applied to non-confluent monolayers of fibroblastic (L929) cells and compared to known controls.Following the 10993-5 standard [20], 15 cm2 of exposed surface was produced for 5test samples: the unpolished ceramic, polished ceramic, lasered ceramic, unpolishedsintered Pt paste and the final high-density feedthrough construct. Followingmanufacture and steam sterilisation, samples were extracted in 2.5 mL of saline for24 h at 70 �C.

L929s were seeded in 35mm Petri dishes, at 20,000 cells/cm2 in Eagles modifiedessential medium (EMEM, Sigma Aldrich, Aust.). After 24 h, the mediawere removedand replaced with extract solution. The extract solutionwas comprised of 25% salinefrom the sample extract diluted in 75% EMEM. Cells were counted following 48 h ofexposure using a cell counter (ViCell, Beckman Coulter, USA). A latex positive controland ethanol gradients were run in parallel for experimental validation. Negativecontrols consisted of both unmodified EMEM and a saline diluted blank extractsolution. The cell counts were normalised to the saline control to obtain a relativeindication of cell growth inhibition resulting from the material extract.

2.4.2. Neural cell cultureThe twin hex ceramic-Pt electrodes and samples of the 1141 high density

feedthrough arrays were autoclaved (Getinge, Australia) and placed in well plates.For active stimulation of the twin-hex arrays only, well plates were modified to

Please cite this article in press as: Green RA, et al., Integrated electrode andBiomaterials (2013), http://dx.doi.org/10.1016/j.biomaterials.2013.04.054

enable leads to pass out of the chamber and connect to the biphasic stimulator.Wellswere coated with collagen I (SigmaeAldrich, Aust.) at a concentration of 5 mg/mL topromote cell adherence. PC12 cells were plated at a concentration of 20,000 cells/cm2 in 99% roswell park memorial institute (RPMI) medium and 1% horse serum(HS) (RPMI and HS from SigmaeAldrich, Aust) supplemented with nerve growthfactor (NGF) (Jomar, Aust.) at a concentration of 100 ng/mL. Each study consisted ofthree ceramic-Pt arrays which were actively stimulated, three passive controls andthree passive high density arrays. Cells were left for 24 h post plating to enable timeto adhere at which time stimulation commenced on the active samples andcontinued for 72 h.

A 6-channel stimulation unit (custom-made, UNSW) was used to stimulateactive ceramic-Pt samples. Stimulation of PC12s was conducted based on past in vivoexperiments by Green et al. [14] and Wong et al. [21]. These stimulation parameterswere found to be sufficient to evoke a cortical response in the feline retinal model,and therefore considered a good basis of comparison for in vitro cellular interactions.The stimulation delivered cathodic-first, charge balanced, symmetric stimulus at220 mA and 400 ms (67.22 mC/cm2 per phase, based upon geometric surface area)through the centre electrode of each hex array. The ceramic-Pt constructs consistedof two arrays each having a centre electrode (cathodic) and a surrounding ring of sixelectrodes to serve as guard (anodic) electrodes. The guard electrodes wereelectrically-connected to one another.

At 96 h post plating, fluorescent micrographs of the cell response were taken ona fluorescent microscope with an attached camera (Axioskop, Carl Zeiss, Germany).Cell density (cells/cm2) and neurite outgrowth density (mm/cm2) were determinedwith Image J (v1.44) software package and Neuron J (v1.4.2) plugin (NIH freeware,USA). Neurites weremeasuredwhen the length of projection exceeded a single bodylength.

Each experiment was conducted with three duplicate wells of each substrate.For each well, five fluorescent micrographs were taken by scanning the film toobtain fluorescent micrographs from different sectors of the well area. Cell densityand neurite outgrowth supported by each substrate were presented as mean � SE.

3. Results

3.1. Feedthrough performance

Measurements made from light microscopy and profilometryreveal that features can be reliably produced with a minimum linewidth of 20 mm and a centre to centre distance (pitch) of 40 mm. Pthorizontal track thickness, shown in Fig. 4A, was kept below10 mm tofacilitate stability in lamination and improve isolation of individualtracks during the lamination process. High density feedthrough andelectrode arrays were produced at up to 1000 channels/cm2,see Fig. 4B. The feedthroughs were found to have a resistance of0.05 U/cm2.

For hermeticity testing it was found that the precise measure-ment of the He leak rate was limited by the detector sensitivity andO-ring permeability. When samples were produced with hermeticproperties the leak rate was found to be below 1�10�12 atm∙cm3/sand consequently recorded as a zero leak on the detector. Addi-tionally, due to permeation of the He through the rubber O-ring

high density feedthrough system for chip-scale implantable devices,

Fig. 4. Layered constructs produced from ceramic and Pt co-firing method show: A. cross-section of hermetic feedthroughs produced with four layers, B. high density (1141)electrode array and C. prototype package assembly.

R.A. Green et al. / Biomaterials xxx (2013) 1e10 5

material, quantification of leak over extended time frames was notpossible, being restricted to approximately 20 s before He perme-ated into the detector via the O-ring. Therefore, He spray testingwas conducted within these limitations as a production control.Positive control samples, namely ceramic blanks without feed-throughs, showed a leak rate <1 � 10�12 atm∙cm3/s (He) for therestricted testing timeframe. Negative controls with known leaks atspecific failure points were also fabricated to benchmark againsttest samples. The negative controls were shown to have signifi-cantly higher leak rates of up to 1�10�9 atm∙cm3/s, which could bedetected from the onset of spraying with He. This leak behaviourwas easily discernible from the O-ring permeation, due to the timecourse of the leak detection being almost instantaneous for nega-tive controls. All five high density feedthrough samples exhibitedleak behaviour which followed the same rates of He detectionrecorded for the positive hermetic controls. No leak was detectablefor any of the test samples within the 20 s testing period.

3.2. Electrode performance

The fabrication process described above was used to producesamples where the upper layer was configured to represent thesame hexagonal electrode array as used in flexible electrode arraysproduced by the authors from Pt foil and PDMS [22]. These arrayswere produced using identical geometries for comparative studiesof electrode performance. Stereoscopic and SEM images of eachelectrode type are shown in Fig. 5. The diameter of each electrodepad is 200 mm resulting in a geometric surface area of 0.0314 mm2.The surface index of ceramic-Pt electrodes was found to be 1.7� 0.1,indicating a 70% increase in geometric surface area resulting fromelectrode roughness. In comparison the PDMS-Pt electrodes werefound to have a surface index of 1.2 � 0.2.

The XPS data revealed that the ceramic-Pt electrodes wereproduced with high purity of Pt at the contact sites with only a

Please cite this article in press as: Green RA, et al., Integrated electrode andBiomaterials (2013), http://dx.doi.org/10.1016/j.biomaterials.2013.04.054

small percentage of alumina present at the surface, detailed inTable 1. Control values obtained from 99.95% purity Pt are used todetermine background contamination from silicon (Si), carbon (C)and oxygen (O) elements.

The charge injection limit of electrodes produced from Pt boundin a ceramic matrix was found to be higher than those producedfrom laser micromachined Pt foil in PDMS, see Fig. 6. Both electrodetypes demonstrated phase dependent behaviour with themaximum injection of ceramic-Pt electrodes at 0.24 mC/cm2 for0.8 ms phases compared to 0.14 mC/cm2 with 0.8 ms phases forPDMS-Pt electrodes.

Biphasic stimulation of electrodes demonstrated that ceramic-Pt electrodes are stable following stimulation for 1.8 billion pul-ses, as shown in Fig. 7. There was no significant change in either thevoltage of the biphasic waveform or the charge storage capacity foractively stimulated electrodes when a paired t-test was used tocompare data (p ¼ 0.05). Additionally, no statistically significantchange was found for the passive controls. The Bode plot compar-ison was found to be indistinguishable following stimulation andhence only pre-stimulus data has been included for completenessof baseline electrode characterisation in Fig. 8.

3.3. Compatibility with biological systems

The cell growth inhibition studies demonstrated that leachantsfrom the materials used to produce the ceramic-Pt unit did notimpact on the cell growth and survival properties, shown in Fig. 9.In fact, all material samples which represented both the materialsand the fabrication process at each step yielded viable cell numberswithin 5% of the blank control.

Neural cells were cultured on the functional electrode arraysand cell response was analysed under both passive growth andelectrically stimulated conditions. Cells were also grown on thehigh density electrodes in the passive condition. Neural cells

high density feedthrough system for chip-scale implantable devices,

Fig. 5. Stereoscopic (top) and SEM (bottom) images of ceramic-Pt electrodes (left) compared to laser micromachined PDMS-Pt electrodes (right).

0.15

0.20

0.25

0.30

0.35

ctio

n L

im

it (

mC

/cm

2)

PDMS-Pt Array Ceramic-Pt Array

R.A. Green et al. / Biomaterials xxx (2013) 1e106

adhered well to the Pt electrode sites in all test conditions; seeFig. 10, showing preferential attachment to the Pt over the alumina.

The cell density, plotted in Fig. 11, for passive electrodes was16,300 cells/cm2, but with electrical stimulation this was reducedto 9100 cells/cm2. Similarly, the neurite outgrowth was higher onpassive arrays than the electrically stimulated arrays, but in bothcases the cells did extend neurites and form networks.

4. Discussion

Ceramic and Pt feedthrough technologies present a flexible androbust system for producing single assembly neuroprosthetic de-vices. With careful control of fabrication parameters this systemcan be used to create hermetic technology platforms with anin-built high density array of functional electrodes. The electrodes

Table 1Elemental composition of Pt electrode sites formed from particulate in an aluminasubstrate.

Element/spin Peak (eV) Ceramic-Pt(%composition)

Pt foil(%composition)

O1s 532.2 35.9 32.3C1s C 289.1 2.1 2.8C1s B 288.0 1.4 3.6C1s A 285.0 35.4 41.8Al2s A 119.3 2.4 e

Si2p A 102.6 3.4 5.2Pt4f7 B 72.6 5.1 3.8Pt4f7 A 71.0 14.2 10.5

0.00

0.05

0.10

0.00 0.20 0.40 0.60 0.80

Ch

arg

e In

je

Phase Width (ms)

Fig. 6. Charge injection limit of Pt electrodes produced from Pt paste embedded inceramic compared to laser micromachined Pt foil embedded in PDMS. Error barsrepresent 1 SD for n ¼ 8.

Please cite this article in press as: Green RA, et al., Integrated electrode and high density feedthrough system for chip-scale implantable devices,Biomaterials (2013), http://dx.doi.org/10.1016/j.biomaterials.2013.04.054

0

100

200

300

400

500

600

700

800

Active Passive

To

tal v

oltag

e a

cro

ss

ele

ctro

des

(m

V) Before After

0

10

20

30

40

50

60

70

80

90

100

Active Passive

Ch

arg

es

to

rag

ecap

acity

(m

C/c

m2)

Before After

Fig. 7. Electrical properties of ceramic-Pt electrodes following 1.8 billion biphasic stimulation pulses at clinically relevant levels. The voltage drop across electrode pairs is shown onthe left, and the charge storage capacity of electrodes is represented on the right (n ¼ 4, Error bars are 1SD).

R.A. Green et al. / Biomaterials xxx (2013) 1e10 7

formed have properties comparable to conventional micro-machined Pt electrodes and are stable under high rates of biphasicstimulation. The naturally porous surface also provides an increasein area across which charge transfer can occur. The materials usedto fabricate these devices are shown to be compatible with cellsystems and can interface with neural tissue under both passiveand electrically active conditions.

Combining existing material technologies with new micro-machining processes provides new opportunities to produce min-iaturised implant technologies on the chip-scale level. Theproposed difficulties associated with high stress generation at theinterface of dissimilar materials [23] have been shown to becontrolled with careful manipulation of materials and processingmethods. Key to this process is the embedding of the Pt trackswithin the bulk ceramic such that thermal stress can be dissipatedwithout delamination. The resulting system allows fabrication ofdevices with track widths of 20 mm and pitch of 40 mm. This is asubstantial advance on the 200 mm pitch reported by Tooker et al.[2] and enables functional paths to be created for more than

Fig. 8. Bode plot of frequency dependent electrochemical impeda

Please cite this article in press as: Green RA, et al., Integrated electrode andBiomaterials (2013), http://dx.doi.org/10.1016/j.biomaterials.2013.04.054

1000 channels/cm2 and improves on previous developments whichdemonstrated 360 functional hermetic channels [24].

While hermeticity was assessed using standard techniques andequipment [11], it was found that the dominant source of leakingwas at the interface of the leak detector, and not associatedwith thefeedthrough construct. As a result, comparative studies weredevised to enable preliminary hermeticity testing to be performed.On this basis it was found that samples containing 1141 feed-throughs were hermetic with leak rates being below1 � 10�12 atm∙cm3/s for the duration of the study before O-ringleaking dominated the signal. This concurs with previous tests byGuenther et al., which demonstrated similarly low leak rates(below the sensitivity limit) with 100 feedthrough channels [25].This initial level of hermeticity exceeds the resolution limits of mostcommercially available leak detectors, but also highlights the ne-cessity of improving the sensitivity of leak detectors for this criticalapplication. However, it is also important to note that these leakrates were equivalent to those recorded from known hermeticmaterial controls and hence a good method for component

nce spectroscopy of 200 mm diameter ceramic-Pt electrodes.

high density feedthrough system for chip-scale implantable devices,

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Alumina (unpolished) Alumina (polished) Alumina (lasered) Sintered Pt in aluminamatrix

High densityfeedthroughs

Ra

tio

o

f v

ia

ble

c

ell n

um

be

rs

(Te

st s

am

ple

/ N

ull)

Fig. 9. Material leachant studies show negligible cell growth inhibition from any material component of the ceramic-Pt feedthrough electrode system compared to blank controls.Error bars represent 1 SD for n ¼ 3.

R.A. Green et al. / Biomaterials xxx (2013) 1e108

selection and quality assurance for feedthrough manufacture.Studies by Schuettler et al. have used thesemethods successfully onsealed hermetic packages [24], which do not require an O-ringinterface. It is expected that as this single assembly design isadvanced to a complete sealed package, these methods will findsimilar application.

Having established a flexible and robust technology platform,the next step in developing a single assembly device is in ascer-taining the safety and efficacy of the implant electrode sites. The Ptelectrodes formed within the alumina substrate were assessed incomparison to lasermicromachined Pt foil embeddedwithin PDMS.The charge injection limit of ceramic-Pt electrodes was found to behigher than that of the PMDS-Pt electrodes and values published forPt [16,26]. Rose and Robblee determined the injection limit for a Ptelectrode with 0.2 ms phases to be 0.05e0.15 mC/cm2 [26]. At thesame phase length the cathodic charge injection limit of ceramic-Ptelectrodes was 0.21 mC/cm2 compared to 0.1 mC/cm2 for thePDMS-Pt electrodes. There are higher reported values for Pt chargeinjection limit but these are based on long stimulation pulses whichutilise reversible Faradaic chemical reactions to carry charge [27]. Itis generally thought that the safest stimulation mechanism for Ptelectrodes will use double layer charging as the dominant methodof cell activation to minimise chemical changes in the local tissueenvironment [26,28]. Although this may prove idealistic as highcharge injection is often required to activate injured and diseasedtissue. The higher charge injection limit found for ceramic-Ptelectrodes is likely a result of the porous surface produced fromthe sintering process (pictured in Fig. 12) increasing the real surface

Fig. 10. PC12 cells cultured for 96 h on ceramic-Pt electrode arrays unde

Please cite this article in press as: Green RA, et al., Integrated electrode andBiomaterials (2013), http://dx.doi.org/10.1016/j.biomaterials.2013.04.054

area relative to the geometric surface area. This is supported by theoptical profilometry data which indicated that the ceramic-Ptelectrodes had a 41.7% higher real surface area than the PDMS-Ptelectrodes. The electrode chemistry is not expected to have a sig-nificant impact on charge transfer. Only traces of the insulatingalumina were detected by XPS yielding a ratio of 1.2:14.2 (Al2:Pt) or7.8% aluminium bound as alumina. Additionally, the Pt content ofthe electrode is comparable to 99.95% purity Pt. This sintered Ptparticulate approach may provide an attractive alternative toelectrode coating technologies which have been previously pro-posed for improving safe charge injection from Pt electrodes [15,17].

Pt electrodes produced through ceramic substrates were alsocharacterised following high frequency biphasic stimulation. Thismethod of accelerated electrical ageing has been used previously toelicit failure mechanisms in electrode coatings [15]. In this study,more than 1.8 billion pulses were delivered to the electrode surface.It was shown that there was no change in the electroactivitydetermined by three metrics: the voltage transient recorded from abiphasic pulse, the charge storage capacity and the frequencydependant impedance. Additionally, inspection of the electrodesunder SEM indicated no discernible change in surface morphology.It is expected that electrodes created by this process are very stable,having comparable properties to those previously reported formicromachined Pt [14,16,29,30].

Finally, both cell studies demonstrated that the biologicalproperties of ceramic-Pt electrodes are suitable for implant devices.While historically alumina has been employed in dental implants[31] and has some history in device packaging [32,33], it has not

r A. passive conditions and B. implant specific electrical stimulation.

high density feedthrough system for chip-scale implantable devices,

0

2

4

6

8

10

12

14

16

18

0

2000

4000

6000

8000

10000

12000

14000

16000

18000

20000

Passive Active 1141 Feedthrough

Ne

urite

Ou

tg

ro

wth

(m

m/c

m2)

Ce

ll D

en

sity

(ce

lls

/cm

2)

Cell Density

Neurite Outgrowth

Fig. 11. Cell density and neurite outgrowth from PC12s grown on ceramic-Pt electrodes for 96 h, both in the passive electrical state and under active implant biphasic stimulation.Additionally, neural cell growth was shown to be unaffected on passive high density electrode arrays. Error bars represent standard error for n ¼ 3.

R.A. Green et al. / Biomaterials xxx (2013) 1e10 9

been used to directly interfacewith neural tissue. Alumina has beenshown to be cytocompatible with osteoblasts and minimallyinhibitory to fibroblasts [34]. Similarly, Pt has been utilised formany implantable electrode applications, but most applicationsuse micromachined Pt foil or sputtered Pt on adherent base layers[35e38]. This study indicates that Pt produced from sintered par-ticulates has a negligible impact on cell growth, but also has adesirable morphology for neural cell interactions. It was shown byPC12 studies that the Pt electrode sites had preferential interactionwith the cells at the border region with the ceramic insulation.Alumina is hydrophobic and known to have minimal cell interac-tion when produced with a smooth surface [39], but the Pt pro-duced from particulates has greater wettability than typical Pt foil.It is thought that this combination of surface morphology andchemistry improves cell interactions, encouraging cell attachmentand supporting growth. This material response may be beneficial toencouraging neural responses which can be localised to the elec-trode site, improving stimulation selectivity.

Fig. 12. High magnification (�5000) SEM of Pt electrode produced from particulate inan alumina matrix, showing highly porous surface resulting from the sintering process.

Please cite this article in press as: Green RA, et al., Integrated electrode andBiomaterials (2013), http://dx.doi.org/10.1016/j.biomaterials.2013.04.054

5. Conclusions

High density feedthroughs can be successfully producedthrough alumina and Pt particulate sintering processes. Thisceramic-Pt technology platform can be expanded through addi-tional layering to produce chip-scale implantable packages withintegrated electronics and an outer layer of microelectrodes whichcan directly interface with neural tissue. These devices are stableunder stimulation with more than 1.8 billion pulses and the elec-trodes have improved charge transfer properties when comparedto machined Pt microelectrodes. The materials used have beenfound to be cytocompatible with desirable neural cell interactions.

Acknowledgements

The authors acknowledge partial funding from University ofNew South Wales, Silver Star grant scheme, PS24596 and BionicVision Australia, a special research initiative (SRI) of the AustralianResearch Council (ARC).

References

[1] Jiang G, Zhou D. Technology advances and challenges in hermetic packagingfor implantable medical devices. In: Zhou D, Greenbaum E, editors. Implant-able neural prostheses 2. Biological and medical physics, biomedical engi-neering. New York: Springer; 2010. p. 27e61.

[2] High-density, bio-compatible, and hermetic electrical feedthroughs usingextruded metal vias. In: Tooker A, Shah K, Tolosa V, Sheth H, Felix S, Delima T,et al., editors. Hilton head Workshop: Solid-state Sensors, Actuators, andMicrosystems 2012.

[3] Stieglitz T. Manufacturing, assembling and packaging of miniaturized neuralimplants. Microsystem Technologies 2010;16(5):723e34.

[4] Vanhoestenberghe A, Donaldson N, Lovell N, Suaniqng G. Hermetic encapsu-lation of an implantable vision prosthesis e combining implant fabricationphilosophies. Biomed Tech 2008;53(1):224e6.

[5] Jung E. Packaging options for MEMS devices. MRS Bull 2003;28:51e4.[6] Park JB, Lakes RS. Ceramic implant materials. Biomaterials. New York:

Springer; 2007. p. 139e71.[7] Sweitzer R, Scholz C, Montezuma S, Rizzo JF. Evaluation of subretinal implants

coated with amorphous aluminum oxide and diamond-like carbon. J BioactCompat Polym 2006;21(1):5e22.

[8] Auciello O, Shi B. Science and technology of bio-inert thin films as hermetic-encapsulating coatings for implantable biomedical devices: application toimplantable microchip in the eye for the artificial retina. In: Zhou D,Greenbaum E, editors. Implantable neural prostheses 2. Biological and medicalphysics, biomedical engineering. New York: Springer; 2010. p. 63e84.

high density feedthrough system for chip-scale implantable devices,

R.A. Green et al. / Biomaterials xxx (2013) 1e1010

[9] Suaning GJ, Lavoie P, Forrester J, Armitage T, Lovell NH. Microelectronic retinalprosthesis: III. A new method for fabrication of high-density hermetic feed-throughs. Conf Proc IEEE Eng Med Biol Soc 2006;1:1638e41.

[10] Allen RV, Borbidge WE. Solid state metal-ceramic bonding of platinum toalumina. J Mater Sci 1983;18(9):2835e43.

[11] Department of defense. Microcircuits, test standards e environmental testsfor sealing. USA 2006.

[12] Dodds CWD, Wong YT, Byrnes-Preston PJ, Rendl M, Lovell NH, Suaning GJ.Performance of laser fabricated stimulating electrode arrays for a retinalprosthesis in saline. Conf Proc IEEE Eng Med Biol Soc 2009:88e91.

[13] Green RA, Devillaine F, Dodds C, Matteucci P, Chen S, Byrnes-Preston P, et al.Conducting polymer electrodes for visual prostheses. Conf Proc IEEE Eng MedBiol Soc 2010:6769e72.

[14] Green RA, Matteucci PB, Hassarati RT, Giraud B, Dodds CWD, Chen S, et al.Performance of conducting polymer electrodes for stimulating neuro-prosthetics. J Neural Eng 2013;10(1):016009.

[15] Green RA, Hassarati RT, Bouchinet L, Lee CS, Cheong GLM, Yu JF, et al. Sub-strate dependent stability of conducting polymer coatings on medical elec-trodes. Biomaterials 2012;33(25):5875e86.

[16] Green RA, Toor H, Dodds C, Lovell NH. Variation in performance of platinumelectrodes with size and surface roughness. Sensors Mater 2012;24(4):165e80.

[17] Cogan SF, Troyk PR, Ehrlich J, Plante TD. In vitro comparison of the charge-injection limits of activated iridium oxide (AIROF) and platinum-iridium mi-croelectrodes. IEEE Trans Biomed Eng 2005;52(9):1612e4.

[18] Cogan SF. Neural stimulation and recording electrodes. Ann Rev Biomed Eng2008;10:275e309.

[19] Abramian M, Lovell NH, Morley JW, Suaning GJ, Dokos S. Activation of retinalganglion cells following epiretinal electrical stimulation with hexagonallyarranged bipolar electrodes. J Neural Eng 2011;8(3):035004.

[20] ISO 10993-5. Biological evaluation of medical devices e part 5: tests forin vitro cytotoxicity. International Organisation for Standardization; 2008.

[21] Wong YT, Chen SC, Seo JM, Morley JW, Lovell NH, Suaning GJ. Focal activationof the feline retina via a suprachoroidal electrode array. Vision Res2009;49(8):825e33.

[22] Suaning GJ, Schuettler M, Lovell NH. Fabrication of multi-layer, high-densitymicro-electrode arrays for neural stimulation and bio-signal recording. In:IEEE EMBS Conf Neural Eng 2007.

[23] Ordonez J, Jeschke C, Guenther T, Stieglitz T, Suaning GJ. Laser-structuring ofthick-film conductors and its application on an optical stimulator. In: Proc14th Ann Conf IFESS 2009.

Please cite this article in press as: Green RA, et al., Integrated electrode andBiomaterials (2013), http://dx.doi.org/10.1016/j.biomaterials.2013.04.054

[24] Schuettler M, Ordonez JS, Santisteban TS, Schatz A, Wilde J, Stieglitz T.Fabrication and test of a hermetic miniature implant package with 360electrical feedthroughs. Conf Proc IEEE Eng Med Biol Soc 2010:1585e8.

[25] Chip-scale hermetic feedthroughs for implantable bionics. In: Guenther T,Dodds CWD, Lovell NH, Suaning GJ, editors. Conf Proc IEEE Eng Med Biol Soc2011:6717e20.

[26] Rose TL, Robblee LS. Electrical stimulation with Pt electrodes; VIII. Electro-chemically safe charge injection limits with 0.2 ms pulses. IEEE Trans BiomedEng 1990;37(11):1118e20.

[27] Brummer SB, Turner MJ. Electrical stimulation with Pt electrodes: II-estimation of maximum surface redox (theoretical non-gassing) limits. IEEETrans Biomed Eng 1977;BME-24(5):440e3.

[28] Merrill DR, BiksonM, Jefferys JGR. Electrical stimulation of excitable tissue:designof efficacious and safe protocols. J Neurosci Methods 2005;141(2):171e98.

[29] Grill WM, Mortimer JT. Stability of the input-output properties of chronicallyimplanted multiple contact nerve cuff stimulating electrodes. IEEE TransRehabil Eng 1998;6(4):364e73.

[30] Lago N, Yoshida K, Koch KP, Navarro X. Assessment of biocompatibility ofchronically implanted polyimide and platinum intrafascicular electrodes. IEEETrans Biomed Eng 2007;54(2):281e90.

[31] Kawahara H, Hirabayashi M, Shikita T. Single crystal alumina for dental im-plants and bone screws. J Biomed Mater Res 1980;14(5):597e605.

[32] Rihui H, Chang-Jin K. Low-temperature monolithic encapsulation usingporous-alumina shell anodized on chip. J Microelectromech Syst 2009;18(3):588e96.

[33] Jeon G-J, Kim WY, Lee HC. New selective two-step anodization of porous anodicalumina for thin-film encapsulation. Microelectron Eng 2013;103(0):99e105.

[34] Naji A, Harmand M-F. Cytocompatibility of two coating materials, amorphousalumina and silicon carbide, using human differentiated cell cultures. Bio-materials 1991;12(7):690e4.

[35] Stieglitz T, Beutel Hr, Schuettler M, Meyer JU. Micromachined, polyimide-based devices for flexible neural interfaces. Biomed Microdevices 2000;2(4):283e94.

[36] Naples GG, Mortimer JT, Scheiner A, Sweeney JD. A spiral nerve cuff electrodefor peripheral nerve stimulation. IEEE Trans Biomed Eng 1988;35(11):905e16.

[37] Clark GM, Tong YC, Black R, Forster IC, Patrick JF, Dewhurst DJ. A multipleelectrode cochlear implant. The J Laryngol Otology 1977;91(11):935e45.

[38] Bhatti PT, Wise KD. A 32-Site 4-Channel high-density electrode array for acochlear prosthesis. IEEE J Solid-state Circuits 2006;41(12):2965e73.

[39] WilliamsDF.Biocompatibilityofclinical implantmaterials. BocaRaton:CRC;1981.

high density feedthrough system for chip-scale implantable devices,