M

Pa

b

c

d

a

ARR1AA

KDMPL

1

gbolbmsppr

ooaammga

U

0d

Sensors and Actuators B 153 (2011) 145–151

Contents lists available at ScienceDirect

Sensors and Actuators B: Chemical

journa l homepage: www.e lsev ier .com/ locate /snb

icromachined dissolved oxygen sensor based on solid polymer electrolyte

eng Wanga,∗, Yi Liub, Héctor D. Abrunab, Jason A. Spectorc, William L. Olbrichta,d

School of Chemical and Biomolecular Engineering, Cornell University, Ithaca, NY 14853, USADepartment of Chemistry and Chemical Biology, Cornell University, Ithaca, NY 14853, USALaboratory for Bioregenerative Medicine & Surgery, Cornell-Weill Medical College, New York, NY 10065, USADepartment of Biomedical Engineering, Cornell University, Ithaca, NY 14853, USA

r t i c l e i n f o

rticle history:eceived 27 July 2010eceived in revised form2 September 2010ccepted 30 September 2010

a b s t r a c t

A silicon microprobe to measure dissolved oxygen levels is described. The sensors are prepared by over-laying platinum thin film electrodes with a solid state proton conductive matrix (PCM) coating. Theplatinum thin film electrodes are fabricated on silicon substrates by standard photolithographic tech-niques while the PCM coating is achieved by drop-casting methods. The size and materials of the device

vailable online 21 October 2010

eywords:issolved oxygen sensoricrofabrication

roton conductive matrixinear sweep voltammetry

make it potentially suitable for medical implantation. The devices are tested in deionized water (DI water),phosphate buffered saline (PBS), and bovine blood serum (BBS). Through linear sweep voltammetry (LSV),single devices are shown to have decent performances in terms of long term stability, reliability, hys-teresis, linearity, and sensitivity. Variations among different devices are characterized and correlated.The simplicity and cost effectiveness of the fabrication and packaging procedures and the decent in vitroperformances of these devices make them good candidates as miniaturized, disposable, and implantabledissolved oxygen sensors for biological and biomedical use.

. Introduction

Dissolved oxygen sensors are important for measuring the oxy-en content in body fluids and tissues [1,2] and can be used asasic devices for integrated biosensor applications [3–5]. Dissolvedxygen sensors presently available on the market are relativelyarge devices that are difficult to use in some biological andiomedical applications. The devices often require labor-intensiveanufacturing and assembly owing to the complicated and delicate

tructures embedded in them. Microfabrication technology such ashotolithography provides the possibility of easy and cost effectiveroduction of smaller devices while improving the usability andobustness of these devices.

Miniaturization has been an important trend in the fabricationf sensors for the past four decades. For the specific case of dissolvedxygen sensors based on electrochemical methods, the immedi-te benefits from miniaturization are little oxygen consumptionnd enhanced spatial resolution. One of the earliest versions of

iniaturized dissolved oxygen sensors was fabricated from glassicropipettes. For example, in the work by Whalen et al. [6], a

lass capillary tube was drawn out in a pipette puller and filledlmost to the tip with molten metal. The metal in the recess was

∗ Corresponding author at: 120 Olin Hall, Cornell University, Ithaca, NY 14853,SA. Tel.: +1 607 342 8522.

E-mail addresses: pw96@cornell.edu, pw96@live.com (P. Wang).

925-4005/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.snb.2010.09.075

© 2010 Elsevier B.V. All rights reserved.

subsequently plated with gold and the recess filled with collodion.The sharp microelectrode had a long, tapering point, with a tip of1–2 �m. The sharp tip could be utilized to impale cells in intracel-lular oxygen level measurements. These types of microsensors hadto be manually fabricated and assembled one by one. In the pastthree decades, owing to the progress in semiconductor and micro-machining techniques, various types of miniature oxygen sensorshave been proposed [7–20]. The majority of these sensors wereClark-type. The Clark-type sensor operates by electrochemicallyreducing the dissolved oxygen in solution and measuring the elec-trical current to determine the oxygen level. The main advantageof the Clark-type design is that all the electrodes required for oper-ation are contained within an inert gas permeable membrane. Thismembrane allows the device to be used in delicate biological mediasuch as blood. Isolating the electrodes from the media improvesthe accuracy of the measurement, since it eliminates other elec-troactive species from giving additional reduction current signals.In the three-electrode configuration, the construction of the work-ing, reference, and counter electrodes could be easily integratedby micromachining techniques. It was also possible to incorpo-rate on-chip microstructures, on-chip electronic circuits, and otherchemical microsensors (e.g. glucose sensors) into the fabrication

process. Clark-type oxygen sensors require a compartment filledby aqueous electrolyte between the membrane and the electrode.This requirement has dramatically complicated the fabrication pro-cedure, decreased the robustness of the device, and lead to the needfor rehydration in long term storage. Hydrogel and solid polymer

1 Actuators B 153 (2011) 145–151

eeabtisrdpcttctsnmsattf

2

2

mpiwtsuaLrt

Fig. 1. Summary of photolithographic processes. (a) Deposition of silicon dioxide

46 P. Wang et al. / Sensors and

lectrolyte have been shown to be able to replace the aqueous-lectrolyte-filled compartment. Among these methods, the PCMpproach is especially attractive due to the simplicity and scala-ility of the fabrication and packaging procedures [21]. However,here are some common shortcomings from an industrial and med-cal point of view in most of the previous work on miniature oxygenensors constructed by micromachining techniques: First, the fab-ication procedures are lengthy and time-consuming, making theseevices inamicable for large-volume production and thus not dis-osable. Second, the materials used for the working, reference, andounter electrodes in these devices are different. This means thathe thin films used as the working, reference, and counter elec-rodes have to be deposited and patterned one by one, which greatlyomplicates the processing procedures. Third, none of them is wellransferable for clinical applications. The functional parts of theseensors are small, but the whole sensors are big and thereforeon-implantable. In this paper, we describe an extremely simpleonolithic procedure for fabricating and packaging disposable, dis-

olved, PCM-based oxygen sensors that are suitable for medicalpplications. We test these devices in PBS and BBS and evaluateheir performances in terms of long term stability, reliability, hys-eresis, linearity, sensitivity, and device-to-device variation, with aocus on medical usefulness.

. Experimental methods

.1. Preparation of microelectrode probes

Microelectrode probes were fabricated using standard micro-achining techniques. A schematic of the electrode fabrication

rocess is presented in Fig. 1. To start the fabrication, a 500 nmnsulation layer of silicon dioxide was deposited on a 4-in. silicon

afer (500 �m thick, single side polished, <1 0 0> crystal orienta-ion) by plasma enhanced chemical vapor deposition (PECVD) (GSIingle wafer PECVD, GSI Group Sciences) The lift-off technique was

sed to define the electrode geometry (Fig. 2). The geometry andrrangement of the three electrodes are similar to those in Ref. [15].ayers of 30 nm of Ti and 130 nm of Pt were electron beam evapo-ated (CHA Mark 50 e-Beam Evaporator) on the substrate to formhe electrodes. Wafers were then coated with a 400 nm capping

Fig. 2. (a) Schematic illustration of the device planar structure (not in scale).

insulation layer, (b) deposition and patterning of platinum metallization layer, and(c) deposition and patterning of silicon nitride capping layer.

layer of low stress silicon nitride by PECVD. The recording areas andthe bonding pads were defined by selectively etching the siliconnitride layer using reactive ions (Oxford 81 Etcher, Oxford Plas-maLab 80+ RIE System). Wafers were diced into 1.1 mm × 7.3 mmchips (Kulicke & Soffa 7100 Dicing Saw). For each chip, bonding padswere connected to insulated fine wires by silver conductive epoxy(MG Chemicals) and then insulated from the environment with reg-ular epoxy (Epoxy 907, Miller-Stephenson). The exposed area of the

working electrode was designed to be ca. 0.43 mm2. The areas ofthe working, reference, and counter electrodes are maintained inthe ratios 1:0.2:0.44. The gap between the working electrode andthe reference electrode is 15 �m and the gap between the referenceelectrode and the counter electrode is 60 �m.

(b) Schematic illustration of the electrode configuration (not in scale).

Actuators B 153 (2011) 145–151 147

2

iagoamwSmci

1ditwdw

2

sAttwTtn

2

i1BrtnwIS

3

stnimaPicb[tcti

P. Wang et al. / Sensors and

.2. PCM coating

After fabrication and packaging, the microprobes weremmersed sequentially in isopropyl alcohol, dilute sulfuric acid,nd deionized water, each for 10 min, and then dried by a nitro-en gun. At this point, the microprobe was ready for depositionf the PCM. The PCM is a mixture of 5 wt.% solution of NafionTM,perfluorinated ion-exchange resin (hydrogen ion form), in aixture of lower aliphatic alcohols and water (Aldrich, Mil-aukee, WI, USA) and polyvinylpyrrolidone (PVP-360) (Sigma,

t. Louis, MO, USA) with 3 wt.% 2,6-bis(azidobenzylidene)-4-ethylcyclohexanone (Aldrich) in a ratio of 4:1. The NafionTM

omponent provides ionic (proton) conductivity while the PVP-360mproves adhesion of the PCM to the substrate.

For better adhesion, the device was dipped in a mixture of Silane74A (Sigma) and ethanol in a ratio of 1:50 for 5 min and allowed tory at 60 ◦C for 5 min. Next, the PCM was applied by either drop cast-

ng or dipping coating. In dipping coating, the device was dipped inhe PCM solution for 5 min. In drop casting, 0.4 �L of PCM solutionas applied to the surface of the device (area uncovered by epoxy,imension 1.1 mm × 1.5 mm) via a micropipette. Finally, the deviceas allowed to dry at 60 ◦C for at least 24 h.

.3. PCM surface morphology and film thickness characterization

Atomic force microscopy (AFM) was utilized to characterize theurface of the PCM. AFM was conducted on a Veeco Dimension 3100mbient AFM system employing an Olympus cantilever. Poten-

ial step chronoamperometry (PSC) was performed to estimate thehickness of the PCM film. All electrochemical experiments in thisork were carried out on a CHI600D electrochemical work station.

he settings of the electrochemical recording system were suchhat cathodic currents were defined to be positive and anodic to beegative.

.4. Evaluation of sensor performances

We investigated the characteristics of the oxygen electrodes bymmersing the device in either PBS (137 mM NaCl, 2.7 mM KCl,0 mM Na2HPO4, 10 mM KH2PO4, pH 7.4, EMD Chemicals, Inc.) orBS (Rockland), allowing them to equilibrate, and monitoring theiresponse by CV or LSV techniques. We reduced the oxygen concen-ration in the solution by means of purging the solution with pureitrogen (high purity, Airgas). The output of the oxygen electrodeas calibrated with a commercial dissolved oxygen meter (InO2,

nnovative Instruments, Inc.). Sodium lactate was purchased fromigma–Aldrich.

. Results and discussion

The physical dimensions of the device have been designed to bemall enough for medical implantation purposes but large enougho minimize internal electrical resistance. Micromachining tech-iques offer a unique opportunity for creating large amounts of

dentical devices from a single processing batch in a cost effectiveanner. In our design, a total of 287 devices can be obtained fromsingle 4-in. silicon wafer. The monolithic procedure in thin filmt electrode fabrication has greatly reduced the processing stepsnvolved. Pt is a common material of choice for the working and theounter electrode in oxygen reduction. Pt has also been shown toe able to serve as the reference electrode under certain conditions

22]. For oxygen reduction, we have found that Pt can be used ashe reference electrode for sensing purposes. A single layer of PCMoating has also dramatically reduced processing steps and time. Inhis work, the PCM coating serves the dual purposes of acting as theonic conductive electrolyte and the interfering molecule blocking

Fig. 3. AFM image of the surface of the PCM. The dimensions of the image are 12 �mby 12 �m.

membrane. The drop-casting or dip-coating approach makes theprocessing very inexpensive.

Only the PCM film prepared by drop-casting was investigatedin detail. Although dip-coating procedures are easier and faster tocarry out and the devices do, to some extent, work well in terms ofoxygen sensing, we have found that dip-coating processed devicesare not as robust as drop-casting processed ones and this methodmay lead to early device failure in repetitive operations. When thechips with Silane 174A adhesion layers are dipped in the PCM coat-ing solution, Silane 174A may be partially stripped, which will leadto inferior adhesion and earlier device failure. On the other hand,due to the extreme simplicity of dipping-coating procedure, it isdesirable to pursue adhesion layer enhancement treatment in thefuture to make such devices more reliable.

Fig. 3 shows the AFM image taken on a 12 �m × 12 �m sitearea of the PCM surface. Most area of the surface appears to besmooth and pin-hole free. According to the image statistics, themean roughness of the PCM surface is ca. 4 nm. The surface ten-sion of the coating solution leads to a smooth surface morphologyupon drying. The needle or bump-like surface structures can beattributed to defects (e.g. tiny dust particles).

The PCM coating should be thick enough to prevent interferingspecies (e.g. proteins) from reaching the Pt surface but also thinenough to minimize oxygen diffusion resistance. The thickness ofthe PCM film is difficult to be measured accurately. We choose toestimate this parameter in two different ways. Based on the non-volatile content of the PCM coating solution and the measurementof the solution-covered geometric area, we estimate that the thick-ness of the PCM film is 6.7 ± 0.7 �m (the variance in film thicknessis estimated based on the variance of the solution dispensing appa-ratus). A second way to estimate the film thickness is based onPSC technique. Prior to measurement, the electrodes were elec-trochemically stabilized in air-equilibrated DI water by CV from0 V to −0.6 V at 2 V/s for 5 min and then at 0.06 V/s for 1 h. LSVmeasurements were subsequently carried out from 0 V to −0.6 V at0.06 V/s for air-equilibrated DI water and oxygen-depleted DI water(Fig. 4a). The difference between the two curves is attributed tooxygen reduction. In PSC measurements, the potential was stepped

from 0 V to −0.6 V instantaneously and the current was recordedwith time (Fig. 4b). In order to single out the faradaic current cor-responding to oxygen reduction, we subtract the current recordedin oxygen-depleted DI water from air-equilibrated DI water. The

148 P. Wang et al. / Sensors and Actuators B 153 (2011) 145–151

F oxygea

bdriw(tbacmOaPt

ctsgtatctllbbidR

sacTs

TC

ig. 4. (a) Linear sweep voltammograms obtained in air-equilibrated DI water andnd oxygen-depleted DI water.

ackground current recorded in the absence of oxygen representsouble-layer charging effects and other faradaic processes occur-ing on the electrode surface. The background-subtracted currents fitted according to equation (1.4) and (1.10) in Ref. [21]. In this

ay, the thickness of the PCM film is estimated to be 5.0 ± 1.4 �mn = 10). Since the film thickness determines the mass transfer resis-ance to oxygen diffusion, the dynamical response of the device cane affected by variation in film thickness. From the film thicknessnd the oxygen diffusion coefficient in the film, the diffusion timeonstant of the device is estimated to be ca. 0.15 s. According to PSCeasurements, there is a 28% relative variation in film thickness.ne possible method to reduce device-to-device variation is to usecontrolled stencil placed on top of the electrode to carry out theCM coating step. Alternatively, thin film coating techniques otherhan drop-casting can be explored as replacements.

The device can be operated in two distinct ways to detect oxygenoncentrations in the solution [23]. In the first way, LSV is utilizedo quantify the oxygen level: for a given sweep rate, the currentampled at a given potential should be proportional to the oxy-en concentration. In the second way, PSC is utilized to quantifyhe oxygen level: for a given potential step, the current sampledt a given time should be proportional to the oxygen concentra-ion. These theoretical relations are valid in both the mass transferontrolled regime and the kinetics controlled regime. Operation inhe mass transfer controlled regime has the advantage of yieldingarger signals. In principle, LSV and PSC techniques are equiva-ent to each other. Many commercial dissolved oxygen sensors areased on PSC technique. In practice, we have found that, for PCM-ased dissolved oxygen sensors, LSV technique is a better choice

n terms of enhancing single-device repeatability and reducingevice-to-device variations (similar observations were reported inef. [21]).

Prior to measurements, the electrodes were electrochemicallytabilized in air-equilibrated PBS or BBS by CV from 0 V to −0.6 V

t 2 V/s for 5 min and then at 0.06 V/s for 1 h. In Fig. 5a and b, someycles in the cyclic voltammograms at 0.06 V/s for 1 h are shown.he electrochemical activity of the electrode can reach the steadytate faster and diminishes less in PBS compared to in BBS. The

able 1urrent values at −0.6 V in LSV measurements.

Current values (nA) 0 3 min 6 min 9 m

PCM coated electrode in PBS 419 426 434 433PCM coated electrode in BBS 294 291 285 282Bare electrode in PBS 624 538 544 590Bare electrode in BBS 364 357 351 347

n-depleted DI water. (b) Current–time curves obtained in air-equilibrated DI water

cathodic current includes oxygen reduction at the Pt surface, doublelayer charging, and other parasitic reactions.

After the electrochemical stabilization procedure, the electrodeswere left in quiescence in the solution for 3 min. Then LSV curveswere recorded from 0 V to −0.6 V at 0.06 V/s (Fig. 5c). Notice thatthese curves are not identical to those in the last cycle of CV in thecorresponding medium. This is due to the mass transfer complica-tions involved in CV techniques. The initial potential was chosenwhere there is almost no oxygen reduction. The final potential waschosen where oxygen reduction is supposed to happen at its masstransfer limiting rate. Below −0.6 V (especially −0.7 V and below),hydrogen evolution will occur and the gas bubbles will destroythe PCM coating. The scan rate was chosen such that the dataacquisition time is 10 s, neither too short for double-layer chargingcurrent to dominate nor too long as unnecessary for engineeringpurposes. In general, the LSV response of the device depends onthe medium, i.e. the LSV curves obtained in DI water, PBS, and BBSare different. One of the well-adopted structural models for Nafiondescribes its microscopic structure as clusters interconnected bynarrow channels (about 1 nm in diameter) which determine thetransport properties of ions and water [24]. The chemical speciespresent in the aqueous medium (e.g. ions, proteins) can affect themolecular structures of the film by either diffusing into the chan-nels or adsorbing onto the surface. These processes are associatedwith the changes in the CV behaviors during the electrochemicalstabilization. The electrochemical stabilization process settles thefilm down to equilibrium or quasi-equilibrium with the mediumand makes the device become more consistent and show less longterm drifting.

After the first LSV curve was recorded, more LSV curves weretaken subsequently at the time points of 3 min, 6 min, 9 min, 30 min,1 h, and 12 h. The current values at −0.6 V are used to quantifythe electrochemical activity of the electrodes. For comparison, barePt electrodes were also examined following identical experimen-

tal procedures (including electrochemical stabilization). Data areshown in Table 1. Both the short term and long term performancesof the PCM coated electrodes are better than those of the bare Ptelectrodes. It is unexpected that results obtained for PCM coated

in 30 min 1 h 12 h Coefficient of variation

519 547 532 0.12308 297 325 0.05865 1098 686 0.29413 366 764 0.36

P. Wang et al. / Sensors and Actuators B 153 (2011) 145–151 149

F tion. (L

ebp

itac(3Pvabtdi2c

tvisrias2o

ig. 5. (a) Cyclic voltammograms obtained in PBS during electrochemical stabilizainear sweep voltammograms obtained in PBS and BBS.

lectrodes in BBS are better than those obtained in PBS, which coulde observed irrespective of the devices used, the batch of the PBSrepared, the batch of the BBS purchased, etc.

From now on, we will focus on device performance evaluationn BBS which is more clinically relevant in long term implanta-ions. Device-to-device variation was examined by LSV techniquesnd results are quantified by the current values at −0.6 V. For PCMoated electrodes, the current values at −0.6 V are 247 ± 65 nAn = 10). For bare Pt electrodes, the current values at −0.6 V are46 ± 78 nA (n = 10). Compared to PCM coated electrodes, the baret electrodes have a slightly smaller coefficient of variation (0.23ersus 0.26). However, when long term device reliability is placeds the most important consideration, PCM coated electrodes shoulde used. Therefore, from now on we will focus on PCM coated elec-rodes. Notice that if the current at −0.6 V is assumed to be theiffusion limited cathodic current, then a relative variation of 26%

n this current will roughly translate into a relative variation of6% in the PCM thickness, which is consistent with film thicknessharacterization results in the previous discussion.

One requirement for an implanted dissolved oxygen sensor ishat it should be insensitive to agitation caused by body fluid con-ection or patient movement. In our experiments, the device wasmmersed in air-equilibrated BBS contained in a beaker. A magnetictirrer was used to generate forced convection in the solution. Theotation speed of the stir bar was varied from 0 to the possible max-

mum. LSV measurements were carried out and the current valuest −0.6 V are used to quantify the results. The statistics on mea-urements with all different rotation speeds gives a mean value of71 nA, a standard deviation of 12 nA, and a coefficient of variationf 0.04. This variation is acceptable if we take into account that the

b) Cyclic voltammograms obtained in BBS during electrochemical stabilization. (c)

readings from a single device in long term tests in air-equilibratedBBS have a coefficient of variation of 0.05 (see Table 1). The insensi-tiveness of the device operation to stirring implies that the oxygenmass transfer resistance is dominated by diffusion through the PCMfilm.

These micromachined dissolved oxygen sensors were designedand intended for long term implantation use. One important tar-get application is post-operative tissue microenvironment (e.g.oxygen level) monitoring in free tissue transfers. The oxygen con-centrations to be monitored in this and most of other medicalsituations range from 0 % atm to 21 % atm (x % atm correspondsto the oxygen concentration in a solution in equilibrium with agas phase of x % atm pure oxygen). For example, in the aforemen-tioned application, 21 % atm oxygen level indicates the normalcondition with well-established blood flow. Near 0 % atm oxygenlevel indicates the occurrence of ischemia. For any reliable sensor,any measurement should be independent of the measurement his-tory of the sensor. In the case of a dissolved oxygen sensor, thereading should depend on the oxygen level in the solution beingmonitored, but not on the oxygen level in the solution previouslymonitored. We prepared oxygen-depleted BBS by purging nitro-gen into the solution. The device was alternatively immersed inair-equilibrated BBS and oxygen-depleted BBS. LSV measurementswere carried out and the current values at −0.6 V are used to quan-tify the results. The statistics (Fig. 6a) has shown that the values

obtained in air-equilibrated BBS are consistent with each otherand the values obtained oxygen-depleted BBS are consistent witheach other, regardless of the measurement history. The offset cur-rents at the zero oxygen concentration level are primarily causedby device capacitance and faradaic side-reactions. The currents

150 P. Wang et al. / Sensors and Actuators B 153 (2011) 145–151

F BS. (bc h thren

otiodhc

toFtctvppoAmfcra

ebt

ig. 6. (a) Current values at −0.6 V obtained in air-equilibrated/oxygen-depleted Bentration levels. (c) Current values from 3 devices at −0.6 V obtained in BBS witormal/lactated BBS.

btained in air-equilibrated BBS are significantly different fromhose obtained in oxygen-depleted BBS (p < 0.0001). These resultsndicate that the device has the potential to be used in low-levelxygen monitoring. We have also observed that, for PCM-basedissolved oxygen sensors, measurements based on PSC techniqueave much larger hysteresis in terms of response to oxygen con-entration change.

The next measurement is to evaluate device linearity and sensi-ivity. Measurements were carried out on 10 devices at 3 differentxygen levels (Fig. 6b). Fig. 6c shows the data for 3 of the devices.or each device, the current/oxygen-concentration has a linear rela-ionship as expected. The Pearson product-moment correlationoefficient is calculated for each device and the average is foundo be 0.996. One problem reflected in Fig. 6c is device-to-deviceariation. Since single device has shown good linearity, we pro-ose that a two-point calibration will be sufficient for each devicerior to use. The sensitivity is determined to be 4.7 ± 1.8 nA/(% atmxygen). Another important parameter of a sensor is its resolution.s we have discussed, the readings of a single device have approxi-ately 5% fluctuations in long term tests. The average reading taken

rom 10 devices in air-equilibrated BBS is 247 nA (see previous dis-ussion). The average sensitivity is 4.7 nA/(% atm oxygen). Thus, theesolution of the device is roughly equal to 5% × (247 nA)/(4.7 nA/(%

tm oxygen)) = 2.6 % atm oxygen.

The potential values of these devices for in vivo use were furtherxamined by looking at their in vitro performances in simulatedody environment. One important shortcoming of experimenta-ion in out-of-body BBS is the lack of physiological responses. For

) Linear sweep voltammograms obtained in BBS with three different oxygen con-e different oxygen concentration levels. (d) Current values at −0.6 V obtained in

example, when ischemia occurs (near zero oxygen concentration),the contents of the local body fluid change slightly and unknownsmall electroactive molecules might be produced and accumulatetransiently in the body fluid. Despite the fact that these moleculesshould only be present at very low concentrations, it is possible thatthey can interfere with the oxygen sensing process by one wayor another. It is well-known that lactate is the most abundantlyproduced molecules during ischemia. As a preliminary study, wecompared the responses of the device in air-equilibrated BBS andlactated air-equilibrated BBS. Specifically, BBS containing 10 mMsodium lactate was prepared by adding sodium lactate to themedium. The device was alternatively immersed in normal BBS andlactated BBS and was thoroughly rinsed by DI water when trans-ferred from lactated BBS to normal BBS. LSV measurements werecarried out and the current values at −0.6 V are used to quantifythe results. The statistics reveal that there is no significant differ-ence between measurements obtained in the two media (p > 0.05)(Fig. 6d).

4. Conclusions

Solid-state micro dissolved oxygen sensors have been fabri-cated. The small size of the whole device and the simplicity of the

fabrication process make these devices promising for medical uses.Single devices are shown to have decent performances in termsof long term stability, reliability, hysteresis, linearity, and sensi-tivity. Variations among different devices are characterized andcorrelated and methods for reducing such variations are proposed.

Actua

A

Ct

A

t

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

P. Wang et al. / Sensors and

cknowledgements

We would like to thank CNF, CCMR and a Cornell-Weill Medicalollege grant. Peng Wang would like to thank Prof. A. Stroock forhe use of an oxygen meter for calibration purposes.

ppendix A. Supplementary data

Supplementary data associated with this article can be found, inhe online version, at doi:10.1016/j.snb.2010.09.075.

eferences

[1] H. Baumgartl, D.W. Lubbers, Microcoaxial needle sensor for polarographicmeasurement of local O2 pressure in the cellular range of living tissue. Itsconstruction and properties, in: E. Gnaiger, H. Forstner (Eds.), PolarographicOxygen Sensors, Springer Verlag, Berlin, 1983, pp. 37–65 (Chapter 1.4).

[2] M. Althoff, H. Acker, The influence of carotid body stimulation on oxygen ten-sion and microcirculation of various organs of the cat, Int. J. Microcirc. Clin. Exp.4 (1985) 379.

[3] B.D. McKean, D.A. Gough, A telemetryinstallation system for chronicallyimplanted glucose and oxygen sensors, IEEE Trans. Biomed. Eng. 35 (7) (1988)526.

[4] M. Nanjo, G.G. Giulbault, Enzyme electrodes for l-amino acids and glucose,Anal. Chim. Acta 73 (1974) 367.

[5] M. Mascini, A. Memoli, Comparison of microbial sensors based on amper-ometric and potentiometric electrodes, Anal. Chim. Acta 182 (1986)113.

[6] W.J. Whalen, J. Riley, P. Nair, A microelectrode for measuring intracellular pO2,J. Appl. Physiol. 23 (1967) 798.

[7] M. Koudelka, Performance characteristics of a planar Clark-type oxygen elec-trode, Sens. Actuators 9 (1986) 249.

[8] H. Suzuki, E. Tamiya, I. Karube, Fabrication of an oxygen electrode using semi-conductor technology, Anal. Chem. 60 (1988) 1078.

[9] H. Suzuki, A. Sugama, N. Kojima, Miniature Clark-type oxygen electrode with athree-electrode configuration, Sens. Actuators B 2 (1990) 297.

10] H. Suzuki, N. Kojima, A. Sugama, F. Takei, K. Ikegami, Disposable oxygenelectrodes fabricated by semiconductor techniques and their application tobioelectrodes, Sens. Actuators B 1 (1990) 528.

11] H. Suzuki, N. Kojima, A. Sugama, F. Takei, Development of a miniature Clark-type oxygen electrode using semiconductor techniques and its improvementfor practical applications, Sens. Actuators B 2 (1990) 185.

12] C.C. Wu, T. Yasukawa, H. Shiku, T. Matsue, Fabrication of miniature Clarkoxygen sensor integrated with microstructure, Sens. Actuators B 110 (2005)

342.

13] H. Suzuki, A. Sugama, N. Kojima, Micromachined Clark oxygen electrode, Sens.Actuators B 10 (1993) 91.

14] H. Suzuki, A. Sugama, N. Kojima, F. Takei, K. Ikegami, A miniature Clark-typeoxygen electrode using a polyelectrolyte and its application as a glucose sensor,Biosens. Bioelectron. 6 (1991) 395.

tors B 153 (2011) 145–151 151

15] G. Jobst, G. Urban, A. Jachimowicz, F. Kohl, O. Tilado, Thin-film Clark-typeoxygen sensor based on novel polymer membrane systems for in vivo andbiosensor applications, Biosens. Bioelectron. 8 (1993) 123.

16] Z. Yang, S. Sasaki, I. Karube, H. Suzuki, Fabrication of oxygen electrode arraysand their incorporation into electrodes for measuring biochemical oxygendemand, Anal. Chim. Acta 357 (1997) 41.

17] H. Suzuki, H. Ozawa, S. Sasaki, I. Karube, A novel thin-film Ag/AgCl anode struc-ture for microfabricated Clark-type oxygen electrodes, Sens. Actuators B 54(1998) 140.

18] H. Suzuki, T. Hirakawa, S. Sasaki, I. Karube, An integrated module for sensingpO2, pCO2, and pH, Anal. Chim. Acta 405 (2000) 57.

19] H. Suzuki, H. Arakawa, I. Karube, Fabrication of a sensing module using micro-machined bioelectrodes, Biosens. Bioelectron. 16 (2001) 725.

20] H. Suzuki, T. Hirakawa, I. Watanabe, Y. Kikuchi, Determination of blood pO2

using a micromachined Clark-type oxygen electrode, Anal. Chim. Acta 431(2001) 249.

21] G.W. McLaughlin, K. Braden, B. Franc, G.T.A. Kovacs, Microfabricated solid-statedissolved oxygen sensor, Sens. Actuators B 83 (2002) 138.

22] K.K. Kasem, S. Jones, Platinum as a reference electrode in electrochemical mea-surements, Platinum Met. Rev. 52 (2) (2008) 100.

23] A.J. Bard, L.R. Faulkner, Electrochemical Methods, John Wiley & Sons, Inc., NJ,2001.

24] T.D. Gierke, W.S. Hsu, The cluster-network model of ion clustering in perfluoro-sulfonated membranes. Perflnorinated lonomer membranes, in: A. Eisenberg,H.L. Yeager (Eds.), ACS Symp. Ser. 180, American Chemical Society, Washington,DC, 1982, p. 283.

Biographies

Peng Wang obtained his Bachelor’s degree in 2006 from Tsinghua University, China.He then proceeded to pursue graduate study at Cornell University. His main aca-demic interests are microfabrication and electrochemistry.

Yi Liu obtained his Ph.D. degree in 2009 from Cornell University. He is currently apostdoctoral associate in the Department of Chemistry at Massachusetts Instituteof Technology and is working on electrocatalysts for oxygen production from wateroxidation for solar fuel energy storage.

Héctor D. Abruna obtained his B.S. at Rensselaer Polytechnic Institute in 1975, andM.S. subsequently there in 1976. He completed his Ph.D. at University of NorthCarolina, Chapel Hill in 1980. He is now the Emile M. Chamot Professor in the Depart-ment of Chemistry and Chemical Biology at Cornell University and interested in allresearch fields related to electrochemistry.

Jason A. Spector obtained a M.D. in 1996 at New York University Medical center.He is assistant professor of surgery at Weill-Cornell Medical College and interestedin reconstructive plastic surgery.

William L. Olbricht obtained his B.S. at Stanford University in 1973 and Ph.D. atCalifornia Institute of Technology in 1980. He is current professor at Cornell Univer-sity in both chemical and biomolecular engineering and biomedical engineering. Hisresearch involves the application of fluid mechanics, mass transfer, microfabrication,and electrochemistry to problems of biomedical interest.