A

tapppsftrgh©

K

1

flRhssbrOal

0d

Available online at www.sciencedirect.com

Biosensors and Bioelectronics 23 (2008) 1637–1644

Carbon post-microarrays for glucose sensors

Han Xu a, Kartikeya Malladi b, Chunlei Wang c,∗, Lawrence Kulinsky a,Mingje Song b, Marc Madou a

a Department of Mechanical and Aerospace Engineering, University of California, Irvine, CA 92697, USAb Department of Material Science and Engineering, University of California, Irvine, CA 92697, USA

c Department of Mechanical and Materials Engineering, Florida International University, Miami, FL 33174, USA

Received 12 September 2007; received in revised form 7 January 2008; accepted 29 January 2008Available online 8 February 2008

bstract

A novel design and fabrication method of glucose sensors based on high aspect ratio carbon post-microarrays is reported in this paper. Apart fromhe fact that carbon has a wide electrochemical stability window, a major advantage of using carbon post-microarrays as working electrodes for anmperometric glucose sensor is the large reactive surface per unit footprint substrate area, improving sensitivity of the glucose sensor. The carbonost-microarrays were fabricated by carbon-microelectromechanical systems (C-MEMS) technology. Immobilization of enzyme onto the carbonost-electrodes was carried out through co-deposition of glucose oxidase (GOx) and electrochemically polymerized polypyrrole (PPy). Sensingerformance of the glucose sensors with different post-heights and various post-densities was tested and compared. The carbon post-glucose sensorshow a linear range from 0.5 mM to 20 mM and a response time of about 20 s, which are comparable to the simulation result. Sensitivity per unitootprint substrate area as large as 2.02 mA/(mM cm2) is achieved with the 140 �m high (aspect ratio around 5:1) carbon post-samples, which iswo times the sensitivity per unit footprint substrate area of the flat carbon films. This result is consistent with the hypothesis that the number of

eaction sites scales with the reactive surface area of the sensor. Numerical simulation based on enzymatic reaction and glucose diffusion kineticsives the optimum geometric design rules for the carbon post-glucose sensor. Glucose sensors with even higher sensitivity can be achieved utilizingigher carbon post-microarrays when technology evolution will permit it.

2008 Elsevier B.V. All rights reserved.

S); M

dgtg2

tofios

eywords: Glucose sensor; Carbon-microelectromechanical systems (C-MEM

. Introduction

The growing interest in the development of analytical devicesor detection and monitoring of physiological substances hased to the emergence of the modern development of biosensors.eliable glucose sensing is one of the critical technologies in theealth care field for monitoring of blood glucose level in patientsuffering from diabetes mellitus. It is well known that the mosterious complications of diabetes mellitus such as renal failure,lindness, neuropathy, and peripheral vascular disease, can beeduced by a tight control of blood glucose levels (Group, 1993).

n the other hand, diabetes medication overdose may lead to

n increased risk of hypoglycemia, i.e. too low blood glucoseevel, which can also have serious consequences including sud-

∗ Corresponding author. Tel.: +1 305 348 1217.E-mail address: wangc@fiu.edu (C. Wang).

aitfBam

956-5663/$ – see front matter © 2008 Elsevier B.V. All rights reserved.oi:10.1016/j.bios.2008.01.031

icroarrays; Polypyrrole

en death (Wickramasinghe et al., 2004). A measure of bloodlucose control is normally achieved through daily insulin injec-ions. Intermittent or continuous blood glucose monitoring bylucose sensors is used to assess the level of control (Heinemann,006).

The stable immobilization of enzymes that allow for con-rolled sensing is one of the major issues in the developmentf amperometric biosensors. Recently, new developments in theeld of conducting polymers have led to their application notnly for the immobilization of biological enzymes in biosen-ors (Trojanowicz et al., 1995; Ross and Cammann, 1994), butlso for microactuators (Xu et al., 2006). The electrochemicalmmobilization of enzymes within conducting polymers has cer-ain advantages over conventional chemical or physical methods

or the manufacture of amperometric biosensor (Cosnier, 1999;artlett, 1993). For example, it is possible to control the amountnd spatial distribution of the enzyme in the polymer matrix byodification of electrochemical deposition conditions such as a

1 ioelec

ttcsiiotpBpifiet

tTiWacaieFfrroia2tcfp

rEirmmahTbppasbctip

eecatscr(t(ae

foas(ddtaesTrpbmarg

2

2

itFaiihi5tSaLf

638 H. Xu et al. / Biosensors and B

ype of the polymer, as well as monomer and enzyme concen-rations in the electrochemical solution. Conducting polymersan be electrodeposited on a wide variety of electrode shapeso enzymes can be immobilized on 3D surfaces. Furthermore,nterference from other electro-oxidizable analytes such as urates expected to be suppressed due to the size exclusion propertyf conducting polymers (Shin and Kim, 1996). Immobiliza-ion of glucose oxidase in conducting polymers, particularly inolypyrrole (PPy), is a well-established technology (Fortier andelanger, 1991; Ramanathan et al., 1995, 1996). Specifically,olymerization of pyrrole, the monomer, forms an electroactiveon exchange PPy film on the surface of the substrate. The PPylm helps to create a number of electroactive centers, thus short-ning each electron transfer path, and speeding up the electronransfer rate from reaction sites to electrode surface.

The development and the performance of biosensors dependo a large extent on the materials employed for their construction.he most common materials for electrochemical transducers

nclude either inert metals, such as platinum or gold (Khan andernet, 1997; De la Guardia, 1995), or carbonaceous materi-

ls (Gorton, 1995; Csoregi, 1994). Recently, microfabrication ofarbonaceous material has received a lot of attention due to manypplications that can be envisioned such as microelectrodesn electrochemical sensors (Hu et al., 2007) and miniaturizednergy storage/energy conversion devices (Yang et al., 2007).or example, carbon paste employed as the supporting matrixor the construction of biosensors exhibited low background cur-ent and relative facility of fabrication and was used by manyesearch groups (Wang et al., 2002; Cui et al., 2005). Carbon nan-tubes have also been explored for a wide variety of applicationsncluding biosensors, where they can alter the catalytic activitynd affect the specificity of biological systems (Yemini et al.,005; Reches and Gazit, 2004). Other carbonaceous materialshat may be used as electrochemical transducers are graphite,arbon fibers, porous and glassy carbons. These materials allowor easy enzyme immobilization while at the same time theyossess a wide electrochemical stability window.

Microfabrication technology based on carbonaceous mate-ials can greatly extend the practical application of MEMS.merging carbon microfabrication technology includes focused

on beam milling (Miura et al., 1998; Irie et al., 2003) andeactive ion etching (Tay et al., 2003). Recently, an alternative,ore economical C-MEMS microfabrication technique for 3Dicrostructures, involving the pyrolysis of patterned photoresist

t different temperatures and in different ambient atmospheres,as been developed (Wang et al., 2005; Wang and Madou, 2006).his C-MEMS technology offers several advantages in MEMS-ased electrochemical and biological sensor applications com-ared to conventional carbon microfabrication processes: (1)hotoresist can be finely patterned by lithography techniquesnd hence a wide variety of shapes and 3D topologies are pos-ible; (2) high aspect ratio 3D C-MEMS enables the design ofiosensors with improved sensitivity; (3) carbon has good bio-

ompatibility for life-science related applications; (4) differentemperature treatments result in different electrical and mechan-cal properties; (5) the method leads to consistency of materialroperties within the same temperature range; (6) the sensing

qeO1

tronics 23 (2008) 1637–1644

lectrodes and contact pads can all be made of carbon with lin-ar dimensions ranging from millimeters to sub-microns; (7)ompared to metal interconnects, the whole carbon system isble to provide negligible contact resistance, which is believedo be beneficial to further increase the signal-to-noise ratio of theystem; (8) carbon has a wider electrochemical stability windowompared to that of Pt or Au electrodes and is suitable for a wideange of applications including electrochemistry and biosensing;9) batch processing of C-MEMS allows high volume produc-ion and low-cost fabrication; (10) addition of nanocomponentssuch as nanopowders, nanofibers) is possible to link the micro-nd nano-scales and to obtain new mechanical, chemical andlectrical properties.

This work reports on utilization of C-MEMS technologyor biosensing applications. Design, fabrication, and testingf enzymatic glucose sensors based on high density and highspect ratio 3D carbon post-microarrays are described in thistudy. Glucose sensing performance for different post-heights20 �m, 85 �m, 140 �m, and flat carbon film) and various post-ensities (100 �m, 80 �m, 60 �m, and 40 �m in center-to-centeristance) is tested and analyzed. The results demonstrate thathe sensitivity of the carbon posts with a height of 140 �m isround two times that of the flat carbon films, which could bexplained by the fact that larger reactive surface area of the sen-or contains correspondingly larger number of reaction sites.he result of numerical simulation based on both enzymatic

eaction and diffusion shows that the sensitivity of the carbonost-glucose sensor depends not only on the total surface area,ut also on the glucose diffusion characteristics. A set of opti-um geometric rules encompassing the well width, post-height

nd post-diameter have been summarized from the simulationesults to guide the design of high sensitivity C-MEMS-basedlucose sensors.

. Experimental

.1. Fabrication of carbon post-microarrays

A typical process flow of fabricating carbon post-microarrayss shown in Fig. 1. SU-8, a negative tone photoresist, was used ashe source of carbon in this C-MEMS microfabrication process.irst, a Si(1 0 0) wafer used as the substrate was cleaned withcetone and isopropyl alcohol and pre-baked at 95 ◦C for 30 minn a convection oven to remove possible moisture and contam-nants. Various process recipes were used to produce differenteights (25 �m, 100 �m, 150 �m, and flat carbon film) and var-ous densities (20 �m/40 �m, 30 �m/60 �m, 40 �m/80 �m, and0 �m/100 �m representing post-diameter/center-to-center dis-ance) cylindrical SU-8 posts that were created from NANOTM

U-8 25 and 100 photoresists. A typical process for spinning on100 �m thick SU-8 photoresist film was carried out using aaurell photoresist spinner at 500 rpm for 12 s then at 1400 rpm

or 30 s, followed by a 10 min soft bake at 65 ◦C and subse-

uent soft bake for 30 min at 95 ◦C. Next, the photoresist wasxposed with a patterning mask using a mercury lamp on anriel exposure tool (light intensity of 500 mJ/cm2) for about25 s. The SU-8 posts with different heights have the same post-

H. Xu et al. / Biosensors and Bioelectronics 23 (2008) 1637–1644 1639

F tratep (d) ph(

dS1t1drpswftfwaa2um

2

iospm7G

s

tat(dtms

2

tcir(tpsspTg

3

ig. 1. Typical process flow for fabricating carbon post-microarrays: (a) Si subshotoresist by UV exposure; (c) photoresist development to produce SU-8 posts;e) 3D schematic of the carbon post-microarrays.

iameter of 50 �m and center-to-center distance of 100 �m. TheU-8 posts with different densities have the same post-height of00 �m. The exposed photoresist samples were then subjectedo a post-exposure bake (PEB) for 1 min at 65 ◦C followed by0 min at 95 ◦C. Development was carried out using an SU-8eveloper from MicroChem (NANOTM SU-8 Developer) to getid of the unwanted SU-8 coverage. Photoresist-derived carbonost-microarrays were finally obtained in a two-step pyroly-is process in an open ended quartz-tube diffusion furnace, inhich samples were post-baked in a N2 atmosphere at 300 ◦C

or about 30 min, then heated in a N2 atmosphere (2000 sccm)o 900 ◦C. The samples were then held at 900 ◦C for 1 h in aorming gas (5% H2 + 95% N2) environment (2000 sccm) afterhich the samples were allowed to cool down to room temper-

ture in N2 atmosphere. The heating rate was about 10 ◦C/min,nd the total cooling time was about 8 h. A Hitachi S-4700-field-emission scanning electron microscope (FESEM) was

sed to characterize the morphology of the resulting C-MEMSicrostructures.

.2. Enzyme immobilization

The GOx enzyme for amperometric glucose sensors wasmmobilized onto the carbon post-microarrays in the processf electrochemical polymerization of PPy. Dodecylbenzene-ulfonate (DBS−) was used as dopant species during theolymerization process. The solution for electrochemical poly-

erization consisted of 0.1 M potassium phosphate buffer (pH

.0 at 25 ◦C), 0.1 M pyrrole, 0.1 M NaDBS, and 100 U ml−1

Ox. Polymer film of 5 �m1 in thickness was deposited on

1 Profilometer was used to measure the thickness of PPy layer on the flatubstrate electropolymerized for the same period of time as PPy layer elec-

db

t

(1 0 0) with SU-8 film of desired thickness spin-coated; (b) patterning the SU-8otoresist pyrolysis converts SU-8 post-microarrays to carbon post-microarrays.

he carbon post-working electrode (1 cm2 in footprint substraterea) in potentiostatic mode (0.65 V vs. Ag/AgCl reference elec-rode, gold plate as counter electrode) of a PC4/750 potentiostatGamry Instruments). After electrochemical polymerization, theerived glucose sensor was washed several times with dis-illed water to remove any loosely bound enzyme and pyrrole

onomers, and then stored in 0.1 M potassium phosphate bufferolution (pH 7.0 at 25 ◦C).

.3. Amperometric measurement

Glucose sensing test was implemented using the same poten-iostat equipment (Ag/AgCl as reference electrode, gold plate asounter electrode). The glucose stock solution (1 M) was storedn the refrigerator at 4 ◦C and diluted when necessary to give theequired standard concentration with a phosphate buffer solutionpH 7.0 at 25 ◦C). The response of the PPy/GOx glucose senoro glucose of various concentrations was examined in 25 ml ofhosphate buffer solution at a constant potential of 0.7 V ver-us Ag/AgCl. After the background current became constant,pecific amounts of 0.1 M glucose were added to the solution toroduce glucose concentrations ranging from 0.5 mM to 20 mM.he sequential current changes corresponding to the increasinglucose concentrations were recorded by the same potentiostat.

. Numerical simulation of sensor performance

A series of simulations for performance of sensor array withifferent post-heights, post-diameters and well widths (distanceetween two posts, excluding the polymer film thickness) have

ropolymerized on corresponding C-posts.

1640 H. Xu et al. / Biosensors and Bioelectronics 23 (2008) 1637–1644

F postsp

bpf

4

4

cmGtc

G

r

H

bmttc

fd

smcaFwiFc8ptdatintabptatpd

ig. 2. SEM pictures of carbon post-microarrays. (a) and (b) 20 �m high carbonosts before and after electrodeposition of PPy/GOx.

een implemented for the design optimization of the carbonost-electrode geometry. Please refer to the Supplementary Dataor details of the sensor models used for simulation.

. Results and discussion

.1. Electrode geometry

Presently, enzymatic glucose sensing based on electrochemi-al reaction is the most widely used clinical approach for glucoseeasurement (Renard, 2002). In enzymatic glucose sensors,Ox is typically used as the biological enzyme to form the elec-

rochemical transducer. With the presence of oxygen, GOx canatalyze the electro-oxidation of glucose:

lucose + O2GOx−→gluconic acid + H2O2 (1)

When H2O2 comes further to anode surface, the followingeaction takes place:

2O2+0.7 V−→ O2 + 2H+ + 2e− (2)

When dipped into glucose solution, the electrode immo-ilized with GOx can be used as an amperometric glucoseonitoring system. The redox reaction of H2O2 near the elec-

rode surface results in a current increase that is proportionalo the glucose concentration in the solution. Thus the glucose

oncentration can be known by reading the anodic current signal.

With proper control of process parameters, as well as care-ul handling of each process step, carbon post-microarrays withifferent post-heights and densities were fabricated. Typical

v83t

before and after electrodeposition of PPy/GOx. (c) and (d) 140 �m high carbon

canning electron microscope photos of the carbon post-icroarrays are shown in Fig. 2. Fig. 2(a) and (b) represents

arbon post-microarrays with post-height of 20 �m before andfter the electrochemical deposition of PPy/GOx, respectively;ig. 2(c) and (d) demonstrates the carbon post-microarraysith post-height of 140 �m before and after the electrochem-

cal deposition of PPy/GOx, respectively. It can be seen fromig. 2(b) and (d) that the PPy/GOx films produce uniformoating after the electrochemical deposition process. As SU-photoresist posts are converted into carbon post-via pyrolysishysical dimensions of the posts are shrinking. Reduction ofhe height was dependent on the initial (pre-pyrolysis) post-imensions. For the SU-8 posts with heights of 25 �m, 100 �m,nd 150 �m, after pyrolysis, the post-heights were reducedo 20 �m, 85 �m, and 140 �m, respectively. From the SEMmages it can be seen that the C-MEMS posts shrink lessear the base (where posts join the substrate) than at the loca-ions higher up (likely due to constraints imposed by the gooddhesion of SU-8 to the substrate). The diameter of the car-on posts half-way up the post-decreases from 50 �m beforeyrolysis to 30 �m after pyrolysis. SEM microscopy investiga-ion was carried out on the samples of the same process batchnd consistent shrinkage values were observed from sampleo sample. The C-MEMS microfabrication technique based onhotolithography and pyrolysis is proved to be highly repro-ucible and controllable. Similarly, for the sample batch of

arious post-densities, the post-height decreases from 100 �m to5 �m, while the post-diameters decrease from 50 �m, 40 �m,0 �m, and 20 �m to 30 �m, 25 �m, 20 �m, and 15 �m, respec-ively.

H. Xu et al. / Biosensors and Bioelec

Fbd

4

csfacrgalSppetisg

4

btmSu

i

ps9tetS

imsFsttsfrcowtcdotwsialw(

eTiabwotpdoSohft

ig. 3. (a) A typical sensing curve of the PPy/GOx covered carbon post-iosensors. (b) Response of PPy/GOx covered carbon post-biosensors toifferent glucose solution of concentrations up to 20 mM.

.2. Amperometric characterization

Fig. 3(a) shows the amperometric response signal of a typi-al PPy/GOx film coated carbon post-glucose sensor in glucoseolutions of various concentrations. The response time2 of theabricated glucose sensors is estimated from Fig. 3(a) to beround 20 s, which is in line with the response of commer-ial glucose sensors (Guo et al., 1992). The least-square linearegression analysis of the calibration characteristics in Fig. 3(b)ives a correlation coefficient of 0.9895, indicating that themperometric response signal of the glucose sensors is in theinear range for glucose concentrations from 0.5 mM to 20 mM.ame glucose sensing experiments were repeated for five sam-les of each carbon post-height, including flat carbon film (zeroost-height). Resulting narrow spread of the measurements isvidenced in small error bars (representing one standard devia-ion of the current measurement per given glucose concentration)n Fig. 3(b). The PPy/GOx film coated carbon post-glucose sen-ors demonstrate good reproducibility for the determination oflucose concentration within the linear range.

.3. Simulation results

Transient responses of the glucose sensor to different glucoseulk concentrations have been simulated in COMSOL Mul-iphysics FEMLAB® environment using the one-dimensional

odel (refer to the sensor model section and Fig. S3 in theupplementary Data). The polymer film thickness of 5 �m wassed in the simulation (corresponding to the average measured

2 Response time is defined as the time for the sensor signal to reach 95% ofts final value.

t

aTear

tronics 23 (2008) 1637–1644 1641

olypyrrole film thickness of the manufactured glucose biosen-ors). It is shown from the transient response result that the5% response time, depending on the bulk glucose concentra-ion, ranges from 20 s to 100 s, which are comparable to thexperimental result. Steady state response of the glucose sensoro various bulk glucose concentrations (refer to Fig. S4 in theupplementary Data) demonstrates a linear sensitivity curve.

Two-dimensional model (refer to the sensor model sectionn the Supplementary Data) has been used to optimize the geo-

etric design of the carbon post-electrode microarrays with theame set of parameter values as for the one-dimensional case.or a typical example of 140 �m high carbon post, it can beeen that in the steady state the bulk glucose concentration inhe well between posts are not uniform along the vertical direc-ion. The glucose concentration at the polymer film and bulkolution interface along the vertical direction inside the wellor the 140 �m high carbon post-electrodes is simulated and theesult is shown in Fig. 4(a). This non-uniformity of the bulk con-entration results from the consumption of the glucose at the topf the well before it has a chance to diffuse to the bottom of theell bottom and it leads to non-uniform current density along

he electrode surface in the vertical direction. Integration of theurrent density along the electrode surface gives the line currentensity which qualitatively represents the total current collectedn each post-electrode of constant post-diameter. The change ofhe line current density with respect to carbon post-height andell width for carbon post-diameter of 30 �m is obtained by

imulation and shown in Fig. 4(b). For each well width, with thencrease of post-height the line current density reaches a peaknd then decreases. A conclusion can be drawn from the simu-ation results that the maximum line current density is obtainedhen the post-height is roughly equal to two times the well width

H = 2 × W; H: post-height, W: well width).Simulation has also been implemented to investigate the

ffect of post-diameter on the total current per unit footprint area.he current per unit footprint area is obtained by calculate the

ntegration of the current density over the total electrode surfacerea. The current density at the bottom of the substrate is negligi-le. The main governing principle is a trade-off (for a given wellidth and post-height) between the increase in the surface areaf each post-with the increase in post-diameter counteracted byhe corresponding decrease in the number of posts per unit foot-rint area as post-diameter is increased. Simulation results thatemonstrate this trade-off for the sensor array with well widthf 40 �m and post-height of 80 �m are presented in Fig. 4(c).ame simulation algorithm has been implemented for a seriesf well width values and their corresponding optimum post-eights. It has been found that the maximum current per unitootprint area is obtained when the post-diameter is equal tohe sum of well width together with two times of polymer filmhickness. (D = W + 2 × T; D: post-diameter, T: film thickness).

Simulation results for the optimum current per unit footprintrea with respect to well width are demonstrated in Fig. 4(d).

he value of the optimum current per unit footprint area forach well width is obtained by selecting the optimum post-heightnd the corresponding optimal post-diameter. The optimum cur-ent per unit footprint area is increasing with well width. The

1642 H. Xu et al. / Biosensors and Bioelectronics 23 (2008) 1637–1644

Fig. 4. (a) Current density at the PPy/bulk interface along the vertical direction inside the well for 140 �m high carbon post-electrode. (b) Change of line currentd t perp with rp t per

mlcftdcpoapD

4

dtbpftcs(t

ttFoToft1a0a0tse(mptCompared to nanofabrication of biosensors, microfabricationoffers a more mature, easier and cheaper way to fabricate highsensitivity biosensors.

ensity with respect to carbon post-height and well width. (c) Change of currenost-height is 80 �m. (d) Change of the optimum current per unit footprint areaicked for each well width value in the simulation to obtain the optimum curren

aximum rate of increase is reached for posts with well widthess than 50 �m; further increase in well width will result inurrent that will increase although at the slower rate. Ease ofabrication is an important consideration and since it is difficulto manufacture carbon posts higher than 200 �m there exists aesign constraint imposed by limitations of present microfabri-ation technology. Therefore, to design the GOx-based carbonost-electrode biosensor microarray with optimal sensitivity, theptimum range for carbon post-well width is between 50 �mnd 100 �m, and the optimum post-height and the optimumost-diameter are selected based on the rules H = 2 × W and= W + 2 × T, respectively.

.4. Experimental results

In experiments, sensitivity of the glucose sensors can beerived from the slope of the linear regression line in the calibra-ion curve. The calculated reactive surface area that is coveredy the PPy/GOx film, as well as sensitivities of the carbonost-glucose sensors of different post-heights with the sameootprint substrate area (1 cm2) are summarized in table (refer

o Table S1 in the Supplementary Data). The sensitivities of thearbon post-glucose sensors of different well widths with theame footprint substrate area (1 cm2) are summarized as wellrefer to Table S2 in the Supplementary Data). In the calcula-ion of the reactive surface area, the post-diameter was taken at

t

unit footprint area with respect to carbon post-diameter. Well width is 40 �m,espect to well width. The optimum post-height and optimum post-diameter areunit footprint area.

he base of the carbon post-structure.3 The results demonstratehat glucose sensors with higher posts have larger sensitivity.or example, sensitivity of the 140 �m high posts (aspect ratiof around 5:1) is about two times that of the flat carbon film.his is mainly because there are more enzymatic reaction sitesn the surface of higher posts due to their larger reactive sur-ace area. The sensitivities per unit footprint substrate area ofhe flat carbon film, 20 �m, 85 �m, and 140 �m high posts are.02 mA/(mM cm2), 1.28 mA/(mM cm2), 1.74 mA/(mM cm2),nd 2.02 mA/(mM cm2), respectively, which are larger than.26 mA/(mM cm2), the sensitivity per unit footprint substraterea of a Au/PPy/GOx glucose sensor (Zhu et al., 2005), and.75 mA/(mM cm2) of a Pt/PPy/GOx glucose sensor reported inhe literature (Piechotta et al., 2005). Even though the glucoseensors with carbon nanotubes serving as enzyme-embeddedlectrodes were reported to have high sensitivity (2.33 nA/mM)4

Wang and Musameh, 2005), it is significantly more difficult toanufacture reliable nanotube biosensor because of the com-

lexity of deposition and positioning of nanotubes of the sameype, same length and same properties at the desired location.

3 The small diameter variation along the post does not have major effect onhe surface area.

4 For 2.33 nA/mM value reported, no surface area was given.

H. Xu et al. / Biosensors and Bioelec

Fig. 5. Variation of sensitivity and total surface area with carbon post-height.

ripehgdtFeri

gtsfhc

5

mtics8p2

cRaaiswtebtocioastottssao(s

A

A

i

R

BCC

C

DFGD

GHHI

Khan, G.F., Wernet, W., 1997. Analytical Chemistry 69, 2682–2687.

Fig. 6. Variation of sensitivity with well width.

The values of the reactive surface area and sensitivity withespect to the post-heights (for well width of 60 �m) are depictedn Fig. 5. While the reactive surface area varies linearly with theost-height, the trend of the sensitivity change deviates from lin-ar relationship with the post-height, especially for sufficientlyigh posts. This is the result of the non-uniformity of the bulklucose concentration along the electrode surface due to theepletion of glucose at the top of the electrode before it diffuseso the bottom of the well. The nonlinear trend of sensitivity inig. 5 matches well with the simulation result of the well widthqual to 60 �m in Fig. 4(b). The variation of the sensitivity withespect to the well width is depicted in Fig. 6 and is summarizedn Table S2 (Supplementary Data).

To further enhance the performance of the carbon post-lucose sensors, electrochemical deposition parameters (such ashickness, density, and surface roughness) of the PPy/GOx filmhould be optimized. Forthcoming improvements of C-MEMSabrication technology will make it possible to manufactureigher carbon posts thereby further increasing sensitivity ofarbon post-based biosensors.

. Conclusion

Design of a novel glucose sensor based on carbon post-icroarrays is reported in this work. Fabrication process of

he high aspect ratio and density carbon post-microarrays andmmobilization of enzyme onto the microarrays through electro-hemical polymerization of pyrrole are also described. Glucose

ensing characteristics for different post-heights (140 �m,5 �m, 20 �m, and flat carbon films with diameter of 30 �m afteryrolysis) and various densities (15 �m/40 �m, 20 �m/60 �m,5 �m/80 �m, and 30 �m/100 �m of post-diameter/center-to-

M

P

tronics 23 (2008) 1637–1644 1643

enter distance after pyrolysis) are tested and compared.esponse of the PPy/GOx carbon post-glucose sensors showslinear range from 0.5 mM to 20 mM and a response time of

round 20 s, which offers possibility of its practical applicationn monitoring blood glucose level of diabetic patients. Under-tanding of the chemistry and physical kinetics of the processas confirmed by the close correspondence between experimen-

al measurements and results of the numerical simulation. Thexperimental results demonstrate that the sensitivity of the car-on posts with an aspect ratio 5:1 is around two times that ofhe flat carbon films, which can be explained by the presencef more reaction sites on the larger reactive surface area of thearbon post-based glucose biosensors. Therefore, it is feasible toncrease the sensitivity of the miniaturized glucose sensors with-ut losing signal quality by optimizing carbon post-geometrynd synthesis condition of the conducting polymer films. Theensitivity of the glucose sensor does not grow linearly withhe total surface area, it depends on the diffusion characteristicsf the glucose onto the electrodes. Geometric optimization ofhe carbon post-microarrays considering both enzymatic reac-ion and diffusion kinetics has been implemented in numericalimulation. To obtain the maximum sensitivity of the glucoseensor, the optimum range of the well width is between 50 �mnd 100 �m for present microfabrication capabilities, while theptimum post-height is supposed to be two times the well widthH = 2 × W), and the selection of the optimum post-diameterhould follow the design rule: D = W + 2 × T.

cknowledgement

This work is partially supported by NSF DMI-0428958.

ppendix A. Supplementary data

Supplementary data associated with this article can be found,n the online version, at doi:10.1016/j.bios.2008.01.031.

eferences

artlett, P.N.a.C.J.M., 1993. Electroanalytical Chemistry 362, 1–12.osnier, S., 1999. Biosensors & Bioelectronics 14, 443–456.soregi, E., 1994. Construction and miniaturization of peroxide based ampero-

metric biosensors, Lund University.ui, X., Liu, G., Lin, Y., 2005. Nanomedicine: Nanotechnology, Biology and

Medicine 1, 130–135.e la Guardia, M., 1995. Microchimica Acta 120, 243–255.ortier, G., Belanger, D., 1991. Biotechnology and Bioengineering 37, 854–858.orton, L., 1995. Electroanalysis 7, 23–45.iabetes Control and Complications Research Group, 1993. New England Jour-

nal of Medicine 329, 977–986.uo, J.L.L., et al., 1992. Chinese Journal of Analytical Chemistry 20, 828–830.einemann, L., 2006. Diabetes Research and Clinical Practice 74, S82–S92.u, C., Yang, C., Hu, S., 2007. Electrochemistry Communications 9, 128–134.

rie, M., Endo, S., Wang, C.L., Ito, T., 2003. Diamond and Related Materials12, 1563–1568.

iura, N., Numaguchi, T., Yamada, A., Konagai, M., Shirakashi, J., 1998.Japanese Journal of Applied Physics Part 2—Letters 37, L423–L425.

iechotta, G., Albers, J., Hintsche, R., 2005. Biosensors & Bioelectronics 21,802–808.

1 ioelec

R

R

RRRST

T

W

W

W

WW

X

644 H. Xu et al. / Biosensors and B

amanathan, K., Ram, M.K., Verghese, M.M., Malhotra, B.D., 1996. Journalof Applied Polymer Science 60, 2309–2316.

amanathan, K., Sundaresan, N.S., Malhotra, B.D., 1995. Electroanalysis 7,579–582.

eches, M., Gazit, E., 2004. Nano Letters 4, 581–585.enard, E., 2002. Current Opinion in Pharmacology 2, 708–716.oss, B., Cammann, K., 1994. Talanta 41, 977–983.hin, M.C., Kim, H.S., 1996. Biosensors & Bioelectronics 11, 161–169.ay, B.K., Sheeja, D., Yu, L.J., 2003. Diamond and Related Materials 12,

185–194.rojanowicz, M., Geschke, O., Krawczyk, T.K.V., Cammann, K., 1995. Sensors

and Actuators B—Chemical 28, 191–199.ang, C., Madou, M., 2006. McGRAW-HILL Yearbook of Science & Technol-

ogy, 44–46.

Y

YZ

tronics 23 (2008) 1637–1644

ang, C.L., Jia, G.Y., Taherabadi, L.H., Madou, M.J., 2005. Journal of Micro-electromechanical Systems 14, 348–358.

ang, J., Li, S., Mo, J.-W., Porter, J., Musameh, M.M., Dasgupta, P.K., 2002.Biosensors and Bioelectronics 17, 999–1003.

ang, J., Musameh, M., 2005. Analytica Chimica Acta 539, 209–213.ickramasinghe, Y., Yang, Y., Spencer, S.A., 2004. Journal of Fluorescence 14,

513–520.u, H., Wang, C., Wang, C., Zoval, J., Madou, M., 2006. Biosensors and Bio-

electronics 21, 2094–2099.

ang, S., Song, H., Chen, X., Okotrub, A.V., Bulusheva, L.G., 2007. Elec-

trochimica Acta 52, 5286–5293.emini, M., Reches, M., Rishpon, J., Gazit, E., 2005. Nano Letters 5, 183–186.hu, M.Z., Jiang, Z.D., Jing, W.X., 2005. Sensors and Actuators B—Chemical

110, 382–389.