Low-Cost Flexible Printed Circuit Technology Based Microelectrode

Sensors and Actuators A 169 (2011) 89–97

Contents lists available at ScienceDirect

Sensors and Actuators A: Physical

jo urnal homepage: www.elsevier.com/locate/sna

Low-cost ﬂexible printed circuit technology based microelectrode array for

extracellular stimulation of the invertebrate locomotory system

a,∗,1 b

Alper Bozkurt , Amit Lal

North Carolina State University, Department of Electrical and Computer Engineering, Raleigh, NC 27695-7911, USA

Cornell University, School of Electrical and Computer Engineering, Ithaca, NY 14853, USA

a r t i c l e i n f o

a b s t r a c t

Article history: The biobotic control of invertebrates through functional electrical stimulation of neural and neuromus-

Received 6 October 2010

cular tissue is under active exploration. Implantable microelectrodes are often designed to be used in

Received in revised form 30 March 2011

chronic long term applications in vertebrates and subjected to strict endurance and resolution require-

Accepted 12 May 2011

ments. However, these constraints can be relaxed in invertebrate-related applications to allow low cost

Available online 20 May 2011

production for high-volume markets. In this study, we propose ﬂexible printed circuit board (ﬂex-PCB)

based electrodes for implantable neuromuscular stimulation, address related shortcomings, and suggest

Keywords: 2

modiﬁcations in the fabrication process. We were able to obtain a charge storage capacity of 3.18 mC/cm

Neural stimulation

␮ × ␮

Electrodes and 1 kHz impedance of 52 k with gold electroplated 100 m 100 m electrode sites on the ﬂex-PCB

Gold electrodes. The electrodeposition of iridium oxide and electrochemical polymerization of PEDOT with

Iridium oxide dopant PSS on microelectrodes enhanced the charge storage capacity to 38.9 and 124.3 mC/cm where

Conductive polymers the 1 kHz impedance magnitude was 16 k and 3 k , respectively. This improvement in electrochemical

PEDOT performance was also corroborated by current pulsed voltage excursion studies. The long term dip test

Flexible electronics

in saline solution supports the potential of ﬂex-PCB electrodes for neural electrostimulation of insects,

Invertebrate physiology

while revealing potential instability in PEDOT-PSS coatings with continuous high current density pulsing.

1. Introduction the cost of fabrication. The vertebrate implantable microelectrodes

are often designed to be used in chronic long term applications.

The functional electrical stimulation of the neuromuscular sys- Hence, the endurance of these devices is an important design cri-

tem has been increasingly used as a clinical treatment option for terion during the implantation process and against the long term

restoration, rehabilitation, and control of movement [1]. Recently, reactions from the biochemical agents that exist in biological tis-

this technique has been under investigation as a means to con- sue. In addition, higher sensitivity and speciﬁcity are required

trol and direct the locomotion of invertebrate organisms, insects to obtain a successful outcome from the sophisticated vertebrate

in particular, for applications ranging from ecological monitor- motor control system. Therefore, implantable stimulating micro-

ing to search-and-rescue missions in a disaster [2,3]. Such a electrodes are often produced following high-cost conventional

“biobotic” control paradigm may beneﬁt extensively from meta- thin-ﬁlm processing technologies derived from semiconductor fab-

morphic development to couple electrically active microelectrodes rication techniques [5]. On the other hand, immunoreaction is

into the electrically responsive tissue of the insect to enable an less an issue for the invertebrates and the lifespan of the elec-

automated mass production line [4]. The cost of microfabricating trodes in tissue is much less due to the shorter life-time of insects.

the metamorphosis-implanted electrodes on this production line Moreover, depending on the electrical actuation and movement

is an important concern, especially for the biobotic applications in control scheme, lower spatial resolution for site speciﬁcity can be

which a larger number of insects would be employed. tolerated due to the relatively less intricate insect motor control

Nevertheless, the design constraints of these invertebrate system of the insect. This difference allows the use of larger pads

microelectrodes can be compromised to a greater extent with with more pitch sizes. All these features provide the opportunity

respect to vertebrate microelectrodes, thus enabling a reduction in to fabricate implantable electrodes through highly standardized

and commonly available ﬂexible printed circuit board (ﬂex-PCB)

technologies where the production cost is optimized for high

∗ volume markets. Polyimide, the most common substrate for ﬂex-

Corresponding author. Tel.: +1 919 515 7349.

PCB production, is a biocompatible material [6]. These substrates

E-mail address: [email protected] (A. Bozkurt).

This work was performed while A.B. was at Cornell University. are often covered with copper as a conductive layer, which can

90 A. Bozkurt, A. Lal / Sensors and Actuators A 169 (2011) 89–97

Fig. 1. Cross-section diagram of the one-layer ﬂexible circuit board electrode (*immersion gold, ** electroplated gold). Nickel–phosphorus layer is only for immersion plating

(ENIG).

easily be coated with gold for biocompatibility as a standard part The tissue growing through these oriﬁces provides mechanical

of the ﬂex-PCB production line. In addition to these, the elec- anchoring in in vivo setups, but that is not in the scope of this

␮ ␮

trode size (50–100 m on each side) and density (50–100 m paper.

pitch size) can be determined with conventional ﬂex-PCB technolo- Laser drilling and milling have been widely used to deﬁne minia-

gies, which can be successfully accommodated for insect-biobot ture structures on ﬂex PCB substrates. Carbon rich debris has been

applications. Using this manufacturing technology, the circuit known to occur as a result of photothermal and photochemical

components for control and data handling can also be directly mechanisms leading to the ablation of the polymer [9]. To remove

assembled on the microelectrode substrate, which is a beneﬁcial the laser ablation debris, we cleaned the received probes ultrason-

property in terms of the size, power, and noise performance of the ically by soaking them in acetone and then isopropyl alcohol (IPA)

electronics. baths, each for 10 min. Acetone and IPA are known to affect the

So far, the use of ﬂex-PCB technology in neural engineering molecular orientation of the polyimide; meanwhile, the ultrasonic

research has been limited to manufacturing ﬂexible interconnects pulsing has the potential to irritate the deposited gold surface, so

between silicon-based microelectrode arrays and control micro- we performed an electrochemical analysis to characterize the effect

electronics [7]. This technology also has been incorporated in of the probe cleaning procedure.

in vitro culture dishes to “record” the electrical activity of cultured

cardiac-cells [8]. The fabrication and use of implantable in vivo ﬂex-

2.2. Electrochemical deposition

PCB “stimulation” electrodes, however, require further analysis and

modiﬁcation.

To enhance the electrochemical properties of the ﬂex-PCB

In this study, we report on the improvement of electrical and

electrodes, we electrodeposited iridium oxide and electropolymer-

electrochemical behavior of ﬂex-PCB probes that will be used as

ized poly(3,4-ethylenedioxythiophene) poly(styrene-sulfonate)

implantable insect tissue stimulation microelectrodes. We present

(PEDOT-PSS) over the gold coated electrodes to facilitate extra

the in vitro characterization of morphological, electrical, and elec-

charge transfer across the interface. For the deposition of iridium

trochemical properties of these surface modiﬁcations through

oxide, we followed a recipe similar to [10]. Seventy-ﬁve milligrams

scanning electron microscopy, cyclic voltammetry, electrochemi-

of iridium (IV) chloride hydrate was dissolved in 50 ml of deionized

cal impedance spectroscopy, and current pulsed voltage excursions

water by stirring for 30 min at room temperature, which formed

in phosphate buffered saline solutions.

a black colored solution. Then, 0.5 ml of 30% hydrogen peroxide

solution were added to the solution and stirred for 10 min, which

2. Methods and materials

turned the color of the solution to yellow. Adding 250 mg of oxalic

acid dihydrate turned the color of the solution blue, and that was

2.1. Flexible printed circuit board fabrication

stirred for another 10 min. Small amounts of anhydrous potassium

carbonate were added to the solution slowly, until a pH of 10.9 was

The layouts of the stimulation electrodes were prepared using

obtained. The solution was kept at room temperature for 2 days to

Target3001 software (Ing.-Buero FRIEDRICH, Germany) and the

reach equilibrium before the deposition.

generated Gerber ﬁles were submitted to Hughes Circuits, Inc. (San

PEDOT-PSS monomer solution was prepared by stirring

Marcos, CA, USA) for production. For the substrate, 100 ␮m (∼4 mil)

35 mg of 3,4-ethylenedioxythiophene (EDOT) with 250 mg

® TM

thick Kapton polyimide ﬁlm (AP8545, DuPont , DE, USA) was

poly(styrenesulfonic acid sodium salt) in 25 ml of deionized water

used, which was laminated with 17 ␮m (0.5 oz/ft ) thick copper

for 2 h until all of the globules of EDOT that had formed were

on one side only. The thinnest available material was 50 ␮m (2 mil)

dissolved [11]. All the chemicals for both solutions were purchased

TM 2

thick polyimide (AP7156, DuPont ) with 0.25 oz/ft of copper. The

from Sigma Aldrich, Inc.

patterned traces on the copper layer were coated with 20 ␮m of liq-

In both cases, ﬂex-PCB electrodes were immersed in the solution

uid photo imageable (LPI) soldermask (PSR 900 FXT, Taiyo, Inc., NV,

to act as anodic working electrodes. The cathodic counter electrode

USA) for insulation. The minimum allowable width of the traces

was a 2.5 cm × 2.5 cm platinum sheet. Galvonastatic charges of 40

was 75 ␮m (∼3 mil) with a line-to-line spacing of 75 ␮m. Next, ␮ 2

and 100 A/mm were applied for 60 and 120 s for iridium oxide

openings were made in the insulation through laser etching where

and PEDOT-PSS respectively. The electrodes were rinsed with DI

the stimulation pads are desired. The laser etching of LPI enables

water and nitrogen-dried.

insulation openings in the order of 75 ␮m (3 mil) on each side. The

outer shape of the electrodes was deﬁned using laser milling to

2.3. Surface morphology characterization

fit into a FFC (flat-flex-connector). The copper pads were coated

␮

using both electroplated (1.5 m thick, ASTM B 488 type III Grade

The coatings on the electrodes were visually analyzed through

A) and electroless nickel – immersion gold (ENIG) (125 nm and

␮ optical microscopy (Olympus BHMLJ) and scanning electron

2.5 m thickness respectively) for biocompability. We had actu-

microscopy (SEM, Zeiss Supra). The roughness parameters (rough-

ation pads of two sizes: 75 ␮m × 150 ␮m and 100 ␮m × 100 ␮m.

ness average and surface curtosis) were obtained using a Wyko

The fabrication layers can be seen in Fig. 1. In some probes, we

NT9100 Optical Proﬁling System.

used oriﬁces that had been opened at the tip with laser drilling.

A. Bozkurt, A. Lal / Sensors and Actuators A 169 (2011) 89–97 91

2.4. Electrochemical impedance spectroscopy and cyclic

voltammetry

The body of the electrodes was shaped to ﬁt into an FFC

connector, enabling external electrical connection for both elec-

trochemical deposition and characterization. Gamry Femtostat

(FAS2) was used for all the electrochemical characterizations

where a 0.1 M PBS buffer was used as the electrolyte. We used

a 2.5 cm× 2.5 cm platinum sheet as the counter electrode and an

Ag AgCl electrode as the reference. The complex impedance was

recorded through electrochemical impedance spectroscopy (EIS)

to characterize the charge transport mechanism. An alternating

sinusoidal current with 25 mV amplitude (typical EIS small-signal

value) and zero bias voltage was used as the input signal. The

impedance was recorded between 10 Hz and 100 kHz at 10 dis-

crete frequencies per decade. The obtained graphs were ﬁt to the

equivalent circuit parameters for further analysis in which the

circuit parameters were initially estimated using graphical cues.

Then, EIS300 software (Gamry Instruments) was used to ﬁt the

impedance values iteratively to the equivalent circuit models to

reﬁne the circuit parameter estimates until a 1–10% difference was

obtained between the measured and calculated impedances.

␮

Fig. 2. (a) Ablation debris caused by laser milling and drilling, scale bar: 100 m

The cyclic voltammetry (CV) curve reveals the ability of the

(b) SEM of contaminated gold pads (c and d) removal of the debris with ultrasonic

interface to store the charge. A scan rate of 500 mV/s was used

cleaning. The stored charge and impedance at 1 kHz is given at the bottom for pre-

− |

to sweep the range between 0.6 and 0.7 V vs. Ag AgCl to remain and post cleaning.

within the water window. The water window is deﬁned through

the safe potential limits associated with the hydrogen and oxy-

n = 20 sites for PEDOT-PSS and iridium oxide coated sites. The sam-

gen evolution in electrolysis of water. The area under the CV curve

ple size for the long term dip test was n = 20 for both PEDOT-PSS and

was calculated as an indication of the amount of stored/injectable

gold. The presented values in this study represent average values

charge in the water window without any gas evolution.

with standard deviations provided when available. The error-bar

regions on the graphs indicate the standard error of the mean.

2.5. Voltage excursion studies

2.7. Long term dip test

To be considered biocompatible, the charge needs to be trans-

ferred to the tissue either through reversible faradic reactions or

The long term stability of the electrodes against water perme-

capacitive discharge where no new substance that may harm the

ability was evaluated under both charge injected and non-injected

tissue is produced. This situation limits the amount of voltage that

conditions. The electrodes were dipped into a saline solution for 40

can be created across the interface. Otherwise, electrical failure of

days while the impedance at 1 kHz and injectable charge amount

the implant occurs either because of electrochemical damage to

(through current pulsed voltage excursion analysis) were recorded

the electrode itself or because of the biological damage to the sur-

on a daily basis. The electrodes were monitored in the absence

rounding tissue caused by these electrochemical reaction products.

and presence of charge transfer between the metal pad and the

Typically, a balanced charge biphasic waveform is used to prevent

electrolyte solution. The charge was applied as cathodic ﬁrst,

tissue damage at the electrode/tissue interface induced by irre-

␮ 2

charge balanced biphasic current pulses (150 C/cm for gold and

versible electrochemical reactions. In this paradigm, the current

2 2

2 mC/cm and 10 mC/cm for PEDOT-PSS) with 2 ms of pulse width

is ﬁrst passed in one direction, followed immediately by a second

and 2% duty cycle.

pulse which passes the current in the opposite direction using an

equal amount of charge. It is difﬁcult to extrapolate the results stud-

ied with slow potential sweep methods of cyclic voltammetry to the 3. Results and discussions

time scale of these pulsed potentials. On the other hand, studying

the voltage transients through the pulse excursions allows one, in 3.1. Cleaning of the laser ablation

principle, to address the electrically safe charge injection limita-

tions to keep the operation in a range where no harmful reactant is After the ﬂex-PCB was manufactured, the electrodes had an

released to the tissue. extensive amount of surface debris, especially at the laser patterned

To generate charge balanced, cathodic ﬁrst, biphasic pulse cur- tips as a result of carbonization and charring of the ablated organic

rents, a programmable stimulator was used (Multichannel Systems material (Fig. 2). It is very likely that such contamination adversely

STG2008). Pulses with 2 ms pulse width, 2% duty cycle and charges affects the biocompatibility of the substrate and the electrochem-

between 0.2 and 20 mC/cm were sent to the electrodes in the ical quality of the formed tissue-electrode coupling; therefore, it

saline solution. The resulting waveforms were recorded using an needs to be removed before tissue implantation. We were only

oscilloscope and then post-processed to ﬁt to the equivalent circuit able to “partially” remove the debris by just wiping the electrodes

parameters. with acetone/alcohol or leaving the electrodes in acetone and alco-

hol overnight without applying any ultrasound, both of which are

2.6. Statistical analysis the most common cleaning techniques when implanting micro-

machined probes into tissue. It has been reported in the literature

We used a student t-test to assess the statistical signiﬁcance that it is possible to dry-clean the debris through applying relatively

(p < 0.05) for all the variables. The data were collected from n = 25 weak laser pulses or plasma-ashing [12]. Because it is costly to have

sites for sputtered, electroplated and immersion plated gold and access to the required laser and plasma etcher, and both methods

92 A. Bozkurt, A. Lal / Sensors and Actuators A 169 (2011) 89–97

Fig. 3. SEM and proﬁlometer images of the sputtered (left), electroplated (middle) and immersion plated (right) gold surfaces; arrow indicating the partial plating at the

corner of immersion plating. The scale bars on the SEM indicate 20 ␮m and bottom edge of the each proﬁlometer image is 30 ␮m. The bottom color bar for the proﬁlometer

shows the range of 0–3 ␮m for right and middle images and 0–300 nm for left image. The roughness average and surface kurtosis, respectively, are provided for each image.

can deform the patterned shapes, wet-cleaning was used to clean ods provide simpler layouts and production because the voltage

the debris, thus removing all the contamination. The contaminated application is not necessary during deposition.

and cleaned probe tips after ultrasonic acetone and IPA cleaning can On the other hand, one of the foremost concerns in electri-

be seen in Fig. 2. The EIS and CV measurements revealed that the cally stimulating the neuro-muscular system is the injection of

ultrasonic cleaning process had an insigniﬁcant amount of degrada- charges from electronically conductive metal electrode into ion-

tion in the electrochemical performance of the gold surface (Fig. 2). ically conductive tissue. Therefore, a comparison is required of

Therefore, it is safe and necessary to incorporate a ﬁnal ultrasonic the electrochemical properties of these surfaces in order to eval-

cleaning step in the production of the ﬂex-PCB microelectrodes in uate charge injection efﬁciency. Electroplated gold was found to

order to use them as neuromuscular stimulation probes. have more surface roughness and thus a larger electroactive sur-

It may also be possible to use the same laser for ablation to face area with respect to the immersion plated gold (Fig. 3) due to

remove the debris. However, this requires adjusting the intensity of the electrical ﬁeld effects during the deposition. We also compared

the laser pulses to be reapplied for cleaning after the ablation pro- these surfaces with the gold directly sputtered on the polyimide

cess, for it is often difﬁcult to have access to and be able to modify surface to compare the ﬂex-PCB fabrication with conventional

the tools in the commercial foundries for research purposes; this microfabrication methods. The directly sputtered gold had negligi-

option was not considered in the scope of this paper. ble roughness when compared to the ﬂex-PCB based gold surfaces

(Fig. 3). We obtained the electrochemical impedance and cyclic

voltammetry curves to characterize the effect of these surface

3.2. Comparison of gold plating techniques

topologies on the interfacial charge transfer characteristics for neu-

ral stimulation applications.

Gold is the most common ﬁnish-coating for ﬂex-PCBs for cov-

One important point that needs to be mentioned is the poorness

ering the underlying copper pads to improve soldering or wire

of the step coverage for the immersion plating method. Although

bonding performance. However, in this study, gold plating was

the sidewalls of the copper pads were successfully covered with

required for biocompatibility in covering the copper, which is

gold through electroplating, the side surfaces partially remained

highly poisonous for tissue. During ﬂex-PCB production, the depo-

uncoated in the case of immersion plating (see the arrow in Fig. 3).

sition of gold is generally achieved either through electrochemical

To solve this issue, the metal traces were designed to be larger than

(electroplating) or chemical (electroless or immersion plating)

the 100 ␮m × 100 ␮m insulation openings covering them (Fig. 4).

means. The electroplating of gold requires extra temporary traces

By using this method, the corners were kept under the insulation

on the substrate to raise the potential of the pads during the depo-

and partial deposition was avoided. However, larger traces were

sition. These traces should be removed during the release of the

required in this case, which increased the dimensions of the elec-

microelectrodes. Therefore, electroless or immersion plating meth-

Fig. 4. Immersion gold plated electrodes with insulation openings larger (left) and smaller (right) than the underlying traces to avoid partial coverage of the corners. Scale

bar shows 25 ␮m.

A. Bozkurt, A. Lal / Sensors and Actuators A 169 (2011) 89–97 93

Fig. 5. Averaged amplitude and phase plots obtained through EIS (left) and averaged CV plots (right) for 100 ␮m × 100 ␮m gold pads deposited through sputtering,

electroplating and immersion plating. Inner table shows the area under each CV curve indicating the stored charge for each material.

Table 1

capacitance this time, representing the non-ideality of the surface

The averaged values of the model parameters extracted through the curve ﬁtting.

caused by increased roughness due to the underlying copper and

Material Rs ( ) Cdl (nF) Rct (k ) Cins (nF) Yo n the effects of the electrical ﬁelds during electroplating. In the case

of immersion plating, covering the corners with an insulating layer

Sputtered 4196 0.025 – – – –

Electroplated 4246 2.26 – – 4.89E-08 0.54 introduced a capacitance (Cins) in series to the CPE, despite the fact

Electroless plated 5115 0.79 1.41 18.2 9.46E-07 0.26 that it solved the step coverage related problems. We also needed to

add a parallel charge transfer resistance (Rct) to Cins to have a satis-

factory model ﬁt. This ohmic representation, modeled as resistance,

can be interpreted as the leakage from the copper to the electrolyte

trodes. Moreover, copper traces can be patterned more precisely

solution as a result of the poorly coated parts of the copper surface

(down to 1 ␮m resolution) with respect to the insulation openings

␮ with a relatively thinner gold layer. The similarities of recorded Rs

(10–30 m resolution). Therefore, the variation in the pad size is

values for all the surfaces validate the models. The higher roughness

more controllable when the insulation holes are wider than the

of electroplated gold causes a higher capacitance, and a smoother

metal traces, as in the case of electroplated gold (Fig. 4). Another

sputtered gold surface results in lower capacitance. The CPE has

important point is that any desired thickness for the gold can be

n −1

an impedance deﬁned as (Yo(jω) ) where lowness of Yo means

obtained in the electroplating process, allowing for several microns

more non-ideality. As n gets closer to 1, the non-ideality begins to

of thickness for a safe coverage of the underlying copper. Nev-

demonstrate a more capacitive behavior. The increased roughness

ertheless, immersion plating results in a difﬁcult-to-control ﬁnal

of the electroplated gold also matches the lower Yo value, where the

thickness in the range of 100–200 nm.

non-ideality is more capacitive with respect to the more resistive

The analysis of the CV curves reveals that the plated gold, in both

leaky immersion plated gold surface.

the immersion and electroplating processes, had a much higher

The above analysis shows that the maximum amount of charge

number of stored charges available for injection than the sput-

that can be capacitively discharged from the gold coated cop-

tered gold (Fig. 5). In fact, the electroplated gold had 1.3 times more

per pads of ﬂex-PCB probes are in the order of 3–4 mC/cm in

charge with respect to the immersion plated gold. The EIS curves

the water window, which is the borderline to induce a biome-

can be used to interpret these differences. For this purpose, the

chanical response through neuromuscular stimulation of the insect

obtained EIS curves were satisfactorily ﬁt to the equivalent circuit

ﬂight muscles [4]. To obtain a higher amount of charge for a more

models [13,14] in Fig. 6 (Table 1). In the case of sputtered gold, the

robust neural stimulation, one needs to exceed the water window,

impedance was formed by the resistance of the electrolyte solu-

which causes oxygen reduction, where the free radical products

tion (Rs) and a series double layer capacitor (Cdl) formed at the

of oxygen are known to be harmful to the cellular function in the

relatively smooth gold–saline solution interface. The electroplat-

long term [15]. Higher voltages also potentially expose poisonous

ing of gold over copper resulted in a similar equivalent circuit, but

copper to the tissue by dissolving/delaminating the gold coating.

with a constant phase element (CPE) in parallel to the double layer

Fig. 6. Equivalent circuit models for the electrolyte–metal interface on gold pads deposited through sputtering, electroplating and immersion plating.

94 A. Bozkurt, A. Lal / Sensors and Actuators A 169 (2011) 89–97

Fig. 7. Electrodeposited iridium oxide (middle column) and electropolymerized PEDOT-PSS (right column) over the electroplated gold pad (left column) at three different

␮

scales. The top row shows the optical image of the pad (scale-bar: 25 m) with a magniﬁed view in the middle squares (each side 25 ␮m). The higher resolution image was

obtained through SEM as presented at the bottom (scale-bar: 2 ␮m).

Therefore, direct use of ﬂex-PCB as a stable and long-term stimula- The central feature of the improvement with PEDOT-PSS depo-

tion electrode is not feasible for applications requiring more than sition was the introduction of ionic dopants with extra available

3–4 mC/cm to initiate a biomechanical response. mobile charge carriers, causing an enhanced charge transport into

the electrolyte solution [11]. In the case of the iridium oxide coat-

ing, the electrochemical reaction involved a reversible conversion

3.3. Improvement with iridium oxide and PEDOT coating +3 +4

between redox states of Ir and Ir where the produced or dissi-

pated charge was balanced by the counterions in the electrolyte

Because the ﬂex-PCB electrodes were manufactured using com-

through a faradic charge transfer mechanism [16]. The charge

mercial foundries and received as ﬁnished products, we were only

applied to the electronically conductive electrode developed a volt-

able to pursue the use of wet deposition techniques that require

age at the interface and this forced iridium oxide to transfer to a

no masks to coat the surface. For the coating material, we studied

higher valance by ejecting protons into the ionically conductive

iridium oxide and PEDOT which are currently considered as the

solution, which was reversed when the charge was removed from

best performing conductive thin ﬁlms to improve the charge trans-

the surface towards the electrode. Also, for either material, addi-

port efﬁciency from metal stimulating electrodes into the aqueous

tional surface irregularities are formed, increasing the roughness

media [11]. Coating the electrodes with either material required a

or porosity to alter the impedance properties [11,14].

gold coated surface, rather than bare copper. The immersion gold

Although the electrochemical properties were improved,

was problematic during the coating of either material when there

exceeding the water window still causes irreversible reactions,

were poorly coated regions; therefore, electroplated gold was used

where gas formation and soluble dissolution products occur. These

for the starting surface.

products diffuse away to the tissue, promoting tissue damage, and

The optical and SEM images of the bare and coated surfaces can

cannot be converted back to the corresponding reactants. The volt-

be found in Fig. 7. The PEDOT-PSS facilitated a signiﬁcant decrease

age excursion studies of the sent biphasic pulses revealed that the

in the impedance curve, much greater than the decrease achieved

amount of charge that can be safely injected to the tissue was in

with iridium oxide (Fig. 8). The decrease in the impedance indicates 2

the order of 10 mC/cm for PEDOT-PSS, whereas this charge was

a lower charge transfer resistance and thus a more efﬁcient elec- 2 2

approximately 1 mC/cm for iridium oxide and around 0.1 mC/cm

trical signal transduction from metal to the electrolyte. This result

for the gold surface (Fig. 9). More than these values, the charges

is also matched with a much higher number of injectable charges

caused a polarization voltage (Vp) higher than −0.6 V, pushing

for PEDOT-PSS with respect to the iridium oxide as characterized

the operation to unsafe regions outside the water window. The

by the cyclic voltammetry (Fig. 8).

Fig. 8. Average improvements in the impedance and charge injection capacity with the deposition of iridium oxide and PEDOT-PSS over electroplated gold as characterized

through EIS (left) and CV (right). Inner table shows the area under each CV curve indicating the stored charge for each material.

A. Bozkurt, A. Lal / Sensors and Actuators A 169 (2011) 89–97 95

Fig. 11. Polarization, drive and residual voltages obtained for different biphasic

current amplitudes applied to gold, iridium oxide and PEDOT-PSS surfaces.

Fig. 9. Voltage excursions obtained by passing biphasic current pulses (top) to the

electrode–electrolyte junction. The bottom graph demonstrates the drive voltage

with the obtained −0.6 V polarization voltage across each material. The required

For PEDOT-PSS coating, the introduced ionic dopants with extra

currents to obtain this voltage are given next to each material.

mobile charge carriers, in addition to the improved surface topol-

ogy obtained during electropolymerization, provided more charge

storage (higher Cdl) and an easier and more efﬁcient charge trans-

port (lower Rct) with respect to the extra reversible faradic reactions

introduced with the deposition of iridium oxide. For the given elec-

trode size, the lower impedance obtained with the PEDOT-PSS at

the metal electrolyte requires a lower stimulus voltage that will

induce enough current to initiate a biomechanical response dur-

ing neuromuscular stimulation applications. The lowness of the

stimulus voltage in turn inﬂuences the size, supply voltage and

power consumption of the electronic circuitry providing the stim-

ulus pulses.

Fig. 10. (a) Equivalent circuit model used for ﬁtting the voltage excursions (b) deﬁ-

nitions of the polarization, drive and residual voltage on the voltage transient curve.

polarization voltage was calculated by subtracting the voltage drop

due to the solution resistance (Id Rs where Id is induced drive cur-

rent and Rs is the solution resistance) from the overall drive voltage

(Vd) recorded at the electrode (Fig. 10). PEDOT-PSS also caused a

smaller residual voltage, which was an indication of the irreversible

reactions occurring due to the non-ideality of the system. Ideally,

the residual voltage should return to 0 V at the end of the second

pulse of the biphasic sequence. The residual, drive and polarization

voltages (Fig. 10) that were recorded for different injected charges

are presented in Fig. 11.

The voltage transients for these three surfaces can be thoroughly

characterized and compared by ﬁtting them into the equivalent cir-

cuit models (Fig. 9), thus allowing formulation of general insights

into the behavior of electrodes under pulsed conditions (Table 2).

Table 2

The averaged values of the model parameters extracted from the voltage excursion curves.

Material Qinj (mC/cm ) Rs (k) Rct (k) Cdl (nF)

Gold 0.2 3.8 ± 2.3 – 40.4 ± 0.9

2 4.2 ± 2.1 – 43.9 ± 0.6

4 5.1 ± 1.7 – 49.7 ± 0.5

Iridium oxide 0.2 3.2 ± 2.6 32.2 ± 5.2 96.9 ± 22.2

2 5.2 ± 2.3 28.9 ± 4.1 99.6 ± 17.1

4 6.1 ± 2.0 26.1 ± 3.8 100.4 ± 14.5

PEDOT-PSS 0.2 4.1 ± 2.7 2.7 ± 1.6 705.6 ± 253

Fig. 12. Average impedance amplitude at 1 kHz and injectable amount of charge

2 4.7 ± 2.1 2.5 ± 1.2 773.9 ± 282

recorded over 40 days in saline solution. The bottom graph shows voltage pulsed

4 5.2 ± 1.8 2.0 ± 1.1 892.1 ± 232

gold–electrolyte interface and nonpulsed interface presented at the top.

96 A. Bozkurt, A. Lal / Sensors and Actuators A 169 (2011) 89–97

Fig. 13. Average impedance amplitude at 1 kHz (left) and injectable amount of charge (right) for high (bottom) and low (middle) current density pulsed and nonpulsed (top)

PEDOT-PSS surfaces in saline solution recorded over 40 days. The inner graphs are the magniﬁed and linear version of the impedance data for the ﬁrst 26 and 17 days for

nonpulsed and pulsed interfaces respectively. Arrows emphasize the data on the last two days.

In addition to this advantage, the aqueous monomer solution for lized afterwards (Fig. 12). The gold pads continuously pulsed with

electropolymerization of PEDOT-PSS is much easier to prepare, can 150 ␮C/cm biphasic pulses and followed a similar trend with non-

be stored longer and requires less expensive chemicals with respect pulsed pads but with a higher standard deviation. We also had

to the electrodeposition solution of iridium oxide. Iridium oxide some of the pads completely covered with polyimide as the insula-

also required certain amount of current to be applied during the tion layer, thus without any insulation openings. The impedance of

electrodeposition and was less tolerant to the deviations in these these pads was “open” during the course of the measurement and

values. The deviations from the required value caused poor adher- did not change at all. Therefore, the penetration of water through

ence to the underlying substrate where delamination was observed the polyimide insulation was not an issue for water dipped ﬂex-PCB

during later voltage or current pulsing, which was also reported in electrodes during the duration of the test. This may indicate that

the literature [16]. the improvement in the impedance and injectable charge by time

was caused by the penetration of water into the exposed metal pads

which potentially facilitated the charge transport mechanism.

3.4. Testing water permeability

The average impedance of the nonpulsed PEDOT-PSS coated

electrodes also followed a decreasing trend (from 1.6–1.1 k)

While PEDOT-PSS coating results in a much improved electro-

which was stabilized around the 10th day. In the meantime, the

chemical performance, the lifetime of this inherently conductive

injectable charge rose to 14.2 from 10 mC/cm in 4 day period

polymer for neural recording and stimulation applications is still

(Fig. 13). The continuously pulsed electrodes (with pulse ampli-

under investigation. Absorption of water by polyimide is also

tude of 2 mC/cm ) followed a similar trend. We also observed this

known to cause delamination between metal and polyimide as

trend with the electrodes pulsed with 10 mC/cm pulses, however

a result of the associated stresses [17]. The potential of PEDOT-

just until the 17th day. Then, the number of failing pads started to

PSS coated ﬂex-PCB electrodes for neuromuscular stimulation was

increase day by day. When we observed the pads around the 40th

assessed with a 40-day saline solution dip test. When immersed

day, all the pads (both pulsed and non-pulsed) were delaminated

in the saline solution, the average 1 kHz impedance of the elec-

spots on the PEDOT-PSS coating where the average impedance

trodeposited gold pads decreased from 21 k to 5 k in 5 days and

was around 80 and 85 k and the average injectable charge was

were stabilized afterwards. The average injected charge to cause

2 0.5 and 0.3 mC/cm for nonpulsed and pulsed pads, respectively

a polarization voltage of −0.6 V was 150 ␮C/cm on the ﬁrst day

2 (Fig. 13).

and increased to 510 ␮C/cm during the ﬁrst 15 days and stabi-

A. Bozkurt, A. Lal / Sensors and Actuators A 169 (2011) 89–97 97

During these tests, we observed that the change in the electro- References

chemical properties of PEDOT-PSS was much lower with respect

[1] P.H. Peckham, J.S. Knutson, Functional electrical stimulation for neuromuscular

to the gold; therefore, PEDOT PSS provided a more stable inter-

applications, Ann. Rev. Biomed. Eng. 7 (2005) 327–360.

face during the ﬁrst 4 weeks after having been dipped in the saline

[2] A. Bozkurt, R. Gilmour, A. Lal, Balloon assisted ﬂight of radio controlled insect

solution. The pulsing of the gold did not change the average elec- biobots, IEEE Trans. Biomed. Eng. 56 (9) (2009) 2304–2307.

[3] H. Sato, C.W. Berry, Y. Peeri, E. Baghoomian, B.E. Casey, G. Lavella, J.M. Van-

trochemical performance of the stimulation electrodes but caused

denBrooks, J.F. Harrison, M.M. Maharbiz, Remote radio control of insect ﬂight,

a higher variation due to the possible surface degradation with con-

Front. Integr. Neurosci. 3 (24) (2009) 1–11.

tinuous charge transfer. In the case of PEDOT-PSS, a similar result [4] A. Bozkurt, R. Gilmour, A. Sinha, D. Stern, A. Lal, Insect machine interface based

neuro cybernetics, IEEE Trans. Biomed. Eng. 56 (6) (2009) 1727–1733.

was observed during the ﬁrst 17 days of the operation. Then, the

[5] K.D. Wise, Silicon microsystems for use in neuroscience and neural prostheses:

high density current pulsing caused an earlier delamination of the

interfacing with the central nervous system at the cellular level, IEEE Eng. Med.

PEDOT-PSS. While the nonpulsed and low density current pulsed Biol. Mag. 24 (2005) 22–29.

PEDOT-PSS electrodes survived the ﬁrst 4 weeks, one needs to be [6] P.J. Rousche, et al., Flexible’ polyimide-based intracortical electrode arrays with

bioactive capability, IEEE Trans. Biomed. Eng. 48 (3) (2001) 361–371.

cautious about the applications that require continuous pulsing of

[7] P. Norlin, M. Kindlundh, A. Mouroux, K. Yoshida, U.G. Hofmann, A 32-site neural

the pads with higher current densities.

recording probe fabrication by DRIE of SOI substrates, J. Micromech. Microeng.

The four-week lifetime of these electrodes ensures a stable 12 (2002) 414–419.

[8] L. Giovangrandi, K.H. Gilchrist, R.H. Whittington, G.T.A. Kovacs, Low-cost

operation during the aforementioned neuromuscular stimulation

microelectrode array with integrated heater for extracellular recording of

applications with insects, for an insect’s lifespan is usually less than

cardiomyocyte cultures using commercial ﬂexible printed circuit technology,

a month. The neurostimulation and locomotion control capability Sens. Actuators B: Chem. 113 (2006) 545–554.

[9] K. Coupland, Laser cleaning of ablation debris from CO2-laser-etched vias in

of the electrodes were demonstrated in an earlier study [4].

polyimide, Appl. Surf. Sci. 127–129 (1998) 731–737.

This study also presents a good starting point for further exten-

[10] K. Yamanaka, Anodically electrodeposited iridium oxide ﬁlms (AEIROF) from

sions to the deposition of other emerging conductors such as carbon alkaline solutions for electrochromic display devices, Jpn. J. Appl. Phys. 28 (4)

(1989) 632–637.

nanotubes, graphene or other inherently conductive polymers to

[11] X.T. Cui, D.D. Zhou, Poly (3, 4-ethylenedioxythiophene) for chronic neural stim-

improve the electrochemical properties of the ﬂex-PCB electrodes

ulation, IEEE Trans. Neural Syst. Rehabil. Eng. 15 (2007) 502–508.

to be used for neural stimulation applications. [12] G. Koren, J.J. Donelon, CO2 laser cleaning of black deposits formed during the

excimer laser etching of polyimide in air, Appl. Phys. 45 (1988) 45.

[13] G.T.A. Kovacs, Introduction to the theory, design, and modeling of thin-

4. Conclusion

ﬁlm microelectrodes for neural interfaces, in: D.A. Stenger, T.M. McKenna

(Eds.), Enabling Technologies for Cultured Neural Networks, Academic, London,

1994.

In this paper, we reported on the potential of ﬂexible PCB tech-

[14] T. Stieglitz, Electrode materials for recording and stimulation, in: K.W Horch,

nology to manufacture electrodes for electrostimulation of insect

G.S. Dhillon (Eds.), Neuroprosthetics: Theory and Practice, World Scientiﬁc,

neuromuscular tissue and address the improvements in the electri- Singapore, 2004, pp. 475–516.

cal and electrochemical behavior of these electrodes. We conducted [15] S.L. Morton, M.L. Daroux, J.T. Mortimer, The role of oxygen reduction

in electrical stimulation of neural tissue, J. Electrochem. Soc. 141 (1994)

in vitro experiments in phosphate buffered saline solutions to

122–130.

characterize morphological, electrical and electrochemical proper-

[16] R.D. Meyer, S.F. Cogan, T.H. Nguyen, R.D. Rauh, Electrodeposited iridium oxide

ties of various coatings over copper-on-polyimide traces through for neural stimulation and recording electrodes, IEEE Trans. Neural Syst. Reha-

bil. Eng. 9 (2001) 2–11.

scanning electron microscopy, cyclic voltammetry, electrochemi-

[17] T. Xu, R.J. Farris, Stresses associated with diffusion in Kapton and Polyacrylics

cal impedance spectroscopy and current pulsed voltage excursions.

ﬁlms, J. Appl. Polym. Sci. 99 (5) (2006) 2661–2670.

Electrodeposition of copper traces with gold outperforms the gold

coating with immersion deposition in terms of charge storage Biographies

and impedance properties. The electrochemical properties were

even further enhanced by electrodeposition of iridium oxide and

Alper Bozkurt is an Assistant Professor with the Department of Electrical and Com-

electrochemical polymerization of PEDOT with dopant PSS on

puter Engineering in North Carolina State University. He received the B.Sc. degree in

microelectrode sites where extra reversible faradic reactions and electrical and electronics engineering from Bogazici University in 2001, the M.Sc.

ionic dopants with extra mobile charge carriers were introduced degree in biomedical engineering from Drexel University in 2004 and the Ph.D.

degree in electrical and computer engineering from Cornell University in 2010. He

to the system, respectively. PEDOT-PSS provided electrochemically

performed research at Cornell to interface microtechnologies with metamorphic

stable interfaces that can inject more charges when compared to

development of insects for building remotely controlled biobotic organisms (insect

iridium oxide in the safe stimulation limits without any hydrolysis cyborgs). His previous research at Drexel included the development of functional

near infrared spectroscopy systems for brain–machine interfaces to augment cog-

or gas evolution. Both gold and PEDOT-PSS coated surfaces survived

nition and for clinical monitoring of the newborn brain in neonatal intensive care

for more than four weeks, which is the lifespan of most of the insects

units. Bozkurt is a recipient of the Calhoun Fellowship from Drexel University and

in the long term water immersion test. Both charge injection capac- Donald Kerr Award at Cornell University.

ity and impedance magnitude at 1 kHz were enhanced throughout

Amit Lal received the B.S. degree in electrical engineering from California Insti-

this time period. Pulsing of the electrodeposited gold electrodes tute of Technology in 1990, and the Ph.D. degree in electrical engineering from

the University of California, Berkeley, in 1996. He is now an associate professor

with high density current pulses did not change the average elec-

at Cornell University, in the School of Electrical and Computer Engineering and his

trochemical properties; instead, a higher variation resulted in these

current research interests include the use of silicon-based high-intensity ultrasonic

parameters. For PEDOT-PSS, high density current pulsing caused actuators for sensing and actuation, radioactive thin ﬁlms for self-powered micro-

instabilities and delamination after a two-week period. Therefore, electromechanical systems (MEMS), and microfabrication for realizing integrated

systems. He holds 17 patents and has published more than 145 research papers in

these ﬁndings support the potential of ﬂex-PCB electrodes to be

the area of microsystem engineering. Most recently, he served as a Program Man-

used in short-term electrostimulation of neural and neuromuscular

ager at DARPA in the Microsystems Technology Ofﬁce, from 2005 to 2009. At DARPA

tissue. he managed ten and started six new programs in the area of navigation, low-energy

computation, bio-robotics, and atomic microsystems. He is the recipient of the NSF

CAREER award, and several best paper awards at the IEEE Ultrasonics and Frequency

Acknowledgment

Control Symposium, and IEEE NEMS conferences. He is also a recipient of the Depart-

ment of Defense Exceptional Service Award, and a Best Program Manager Award for

his work at DARPA.

This work was fully supported by DARPA HI-MEMS Program.