218 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ... · 218 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 3, NO. 3, JULY 2006 Development of an Automated Microspotting

218 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 3, NO. 3, JULY 2006

Development of an Automated Microspotting Systemfor Rapid Dielectrophoretic Fabrication of Bundled

Carbon Nanotube SensorsKing Wai Chiu Lai, Student Member, IEEE, Carmen Kar Man Fung, Student Member, IEEE,

Victor Tak Sing Wong, Student Member, IEEE, Mandy Lai Yi Sin, Wen Jung Li, Member, IEEE, andChung-Ping Kwong, Senior Member, IEEE

Abstract—An automated carbon nanotube (CNT) microspot-ting system was developed for rapid and batch assembly ofbulk multiwalled carbon nanotubes (MWNTs)-based microelec-tromechanical system sensors. By using the dielectrophoreticand microspotting technique, MWNT bundles were successfullyand repeatedly manipulated between an array of microfabri-cated electrodes. Preliminary experimental results showed thatmore than 75% of CNT functional devices can be assembledsuccessfully using our technique, which is considered to be a goodyield for nanodevice manufacturing. Besides, the devices weredemonstrated to potentially serve as novel thermal sensors fortemperature and fluid-flow measurements. This feasible batchmanufacturable method will dramatically reduce production costsand production time of nanosensing devices and potentially enablefully automated assembly of CNT-based devices.

Note to Practitioners—This paper was motivated by the problemof manipulate carbon naotube (CNT) across gold microelec-trodes effectively and precisely. The purposed system potentiallyapplies to other nano-sized particles that are neutral and withhigh polarizability. Existing methods of CNT assembly includeguided CNT growth, external forces, polar molecular patterning,and atomic force microscopy (AFM) manipulation, which istime-consuming and unrealistic when considering batch produc-tion of CNT-based sensors. This paper reported a novel methodto build CNT-based sensors across the microelectrodes by usingour automated microspotting system. This system is integratedwith a dielectrophoretic and microspotting technique. We firstexplained the dielectrophorectic effect on CNT—this methodis very effective to manipulate CNT. Then, the microspottingtechnique was developed to spot a micron-sized CNT dilutiondroplet on the desired positions of a microchip substrate. Finally,dielectropheretic manipulation can be used to position CNT bun-dles across the microelectrodes. In this paper, we experimentallyshowed the difficulties to spot a micron-sized droplet, and theproblems can be overcome by sharpening the spotting probechemically and using the special spotting method. We then show

Manuscript received November 26, 2004; revised February 7, 2006. Thiswork was supported in part by the Hong Kong Research Grants Council(CUHK4206/00E and 4381/02E), in part by the Chinese Academy of Sciences’Distinguished Overseas Scholar Grant, and in part by a grant from the ShunHing Institute of Advanced Engineering of CUHK. This paper was recom-mended by Associate Editor N. Xi and Editor M. Wang upon evaluation of thereviewers’ comments.

K. W. C. Lai is with Michigan State University, East Lansing, MI 48824 USA(e-mail: [email protected]).

C. K. M. Fung, M. L. Y. Sin, W. J. Li, and C.-P. Kwong are with the Centre forMicro and Nano Systems, The Chinese University of Hong Kong, Hong Kong,China.

V. T. S. Wong is with the University of California, Los Angeles, CA USA90095.

Digital Object Identifier 10.1109/TASE.2006.877395

the yield of CNT-based sensors fabricated by using this systemis very promising. We also reported that the CNT-based sensorshave low power consumption, and CNT can be used as the sensingelement of the thermal sensor. The experimental results indicatedthis approach is feasible to develop batch manufacturing of nanodevices.

Index Terms—Carbon nanotube (CNT), CNT sensors, di-electrophoretic manipulation, nano batch fabrication, nanomanufacturing.

I. INTRODUCTION

CARBON nanotubes (CNTs), have been widely studieddue to their electrical (e.g., see [1]), mechanical (e.g., see

[2]), and chemical properties since Sumio Iijima discoveredthem in 1991 [3]. Due to the CNTs’ miniature size and tendencyto cling together in nature, connecting, aligning, and isolatingprocesses of CNTs are difficult. Atomic force microscopy(AFM) is typically used to manipulate each nano-sized tubeone-by-one [4]. However, this is time-consuming and unre-alistic when considering batch production processes. Othermethods of CNT assembly include guided CNT growth [5],[6]; external forces [7], [8]; and polar molecular patterning [9].The first method grows organized CNT structures (directedassembly) by chemical-vapor deposition (CVD), whereas thelatter two methods are for pregrown nanotubes. When con-sidering these methods, the electric force field is still betterin isolating, aligning, and connecting CNTs for nanoscalecircuits applications [10]. Hence, electric-field-based CNT ma-nipulation is widely used today in laboratories worldwide. K.Yamamoto et al., pioneered the work in electric-field-assistedmanipulation of CNT bundles [11], [12]. However, in order toperform a precise and efficient batch manipulation process forCNTs, we developed an automated CNT microspotting systembased on the technique of electric-field assist assembly (i.e.,dielectrophoresis). We have shown that many CNT devicescan be successfully fabricated in a fast and automated mannerby this method, which enables high yield and high precisionalignment of CNT bundles.

This paper reports the dielectrophoretic technique forforming bundled multiwalled carbon nanotube (MWNT) resis-tive elements between gold (Au) microelectrodes which canbe fabricated by a standard microlithographic technique. Inaddition, experimental results showing the validity of a batchassembly of CNT-based sensor arrays using the automated CNT

1545-5955/$20.00 © 2006 IEEE

LAI et al.: DEVELOPMENT OF AN AUTOMATED MICROSPOTTING SYSTEM 219

Fig. 1. Dielectric spherical particle undergoes dielectrophoretic motion. Theunbalanced force impart on the sphere causes it to move toward the smallerelectrode. The figure above shows the charge distribution in a particular instantwhere the small electrode is at a higher potential than the large electrode. Forac bias voltage, the potential will be interchanged periodically on the electrodesand the resulting force on the particle is still acting on the same direction due touneven electric-field distribution.

microspotting system are also presented. We have proven thatCNT devices built by this method are capable of performingvery low power thermal sensing (i.e., 1000 times lower thanconventional MEMS polysilicon-based thermal sensors) andcan be fabricated in a fast and efficient manner. The architectureof the automated CNT microspotting system will be described.The development of this architecture requires overcoming someproblems, such as eliminating the adhesion force between CNTdilution and spotting probe tip, and aligning the probe tip to themicroelectrodes, etc.

II. DIELECTROPHORETIC MANIPULATION OF

CARBON NANOTUBE

A. Theoretical Background and Modeling

Dielectrophoresis is a phenomenon where neutral particlesundergo mechanical motion inside a nonuniform ac electric field[13]. As shown in Fig. 1, the force applied to a particle couldbe nonuniform at a particular location and time between a pairof electrodes. In our CNT manipulation experiment, MWNTswere dispersed inside a liquid medium (e.g., ethanol) and, there-fore, other forces (e.g., viscous force) besides dielectrophoreticforces were imparted on CNTs during the manipulation process.During the CNT manipulation process, the total force acting onan MWNT is the sum of a number of independent forces [14],[15]. With negligible gravitational force, three dominate forceswere the dielectrophoretic (DEP) force, viscous force, and elec-trothermal force. For other small particles with volumes com-parable to MWNTs, thermal effects can dominate, but the highpolarizability of MWNTs makes the DEP force large enough toproduce deterministic movements.

Dielectrophoresis refers to the force exerted on a polarizedparticle in the nonuniform electric field [13], [16]. It can bewritten as

(1)

Fig. 2. Complex polarization factor K when the longest axis of the MWNT isparallel to the electric field applied.

where is the electric field and is the induced dipole momenton an MWNT. It is assumed that the MWNT is a long prolatespheroid with the longest axis aligned with the electric field;therefore, the induced dipole moment is

(2)

where , , and are the absolute permittivity of the medium,the half length of MWNT, and the radius of MWNT, respec-tively. In (2), the complex polarization factor is given by

(3)

where and are the absolute complex permittivitiesof MWNT [17] and the medium (ethanol) [18], respec-tively. The depolarization factor can be approximated by

[16]. Since the electric-field componentsare in phase, the time-averaged DEP force deduced from theabove equations is

(4)

where is the gradient of the square of the root-mean-square (rms) of the electric field. The gradient is affected byboth the geometry of the electrodes and the applied voltage.The former factor controls the overall position of the MWNTbranches between the electrodes because it determines the di-rection of the average DEP force. As the potential differenceapplied is proportional to the magnitude of the electric field, alarger potential can increase the magnitude of the DEP force andoffer a larger torque to align the MWNT with the electric-fielddirection [16]. If the voltage is increased to a sufficiently highlevel (in our case, 16 V), the DEP force is high enough todraw the CNT to the maximum electric-field region.

The frequency of the applied voltage also affects the mag-nitude of DEP force since the complex permittivity spectra ofMWNT and ethanol are functions of the frequency. Due to thehigh conductivity of the assumed purified MWNT

[17] and the small depolarization factor resultingfrom its long dipole , the real part ofthe complex polarization factor is positive as shown in Fig. 2.The corresponding positive dielectrophoretic force draws the


Fig. 3. (a) Photograph of array of Au microelectrodes on a substrate.(b) Optical image showing a pair of Au microelectrodes before CNT.

MWNT to the electric field maxima. At low frequencies ( 10MHz), the CNT is always more polarizable than the mediumand undergoes positive DEP, the force is toward points of highelectrical field strength, and the particles collect initially at theelectrode edges. The orientation of CNT is dependent on theelectrode alignment and the CNTs are always aligned parallelto the electric field. In order to understand the physical phenom-enon during dielectrophoretic manipulation, we have previouslyconducted simulations on the effect of different microelectrodegeometries on carbon nanotube alignments and the results canbe found in [19].

B. Experimental Details

Based on the physical phenomenon of dielectrophoresis, wesuccessfully manipulated bundled CNTs on fabricated micro-electrodes. The experimental process flow for the CNT’s ma-nipulation process can be divided into three parts: 1) fabrica-tion of microelectrodes, 2) sample preparation, and 3) DEP ma-nipulation. An array of Au microelectrodes which were electri-cally connected together to have same voltage and were fabri-cated on glass or substrates by a standard microfab-rication technique, were previously described in [20] and areshown in Fig. 3. Prior to the MWNT manipulation, bundledMWNTs were prepared by a sonication treatment. The bundledMWNTs used in the experiments were ordered commerciallyfrom [21], which were prepared by the manufacturer using thechemical vapor deposition method. The axial dimension and thediameter of the MWNTs were 1 m–10 m and 10 nm–30 nm,respectively. In order to form a stable colloidal suspension ofMWNTs and minimize the degree of aggregation, 50 mg of thesample was ultrasonically dispersed in a 500-mL ethanol solu-tion and the resulting solution was diluted to 0.01 mg/mL forlater usage. During the CNT manipulation process, the substratewith fabricated Au microelectrodes was placed on the vacuumpump-based stage of a micromanipulator station, which allowedprobing of the microelectrodes by microprobes. The vacuumpump-based stage was one part of the micromanipulator sta-tion, which was used to hold the substrate in position whenprobing and applying the ac voltage to the electrodes on the sub-strate. Au microelectrodes were then excited by an ac voltageof typically 16 V peak-to-peak with a frequency of 1 MHz. TheCNT/ethanol solution was then transferred to the substrate bya microspotting probe that could be moved by a three-axis mi-crorobotic system. The detailed experimental flow will be pre-sented in the next section.

Fig. 4. Microspotting system with syringe pump and micromanipulator.

Fig. 5. (a) Microscopic images showing different capillary probe tips weresharpened by our chemical etching process. (b) SEM image showing a sharp tipwith the inner diameter.

III. AUTOMATED CNT MICROSPOTTING SYSTEM

A computer-controlled CNT microspotting system wasdeveloped to perform the dielectrophoretic CNT manipulationautomatically. This system allows the CNT/ethanol solutionto be “spotted” between each pair of microelectrodes pre-cisely. The volume of CNT/ethanol droplets can also be wellcontrolled, resulting in a high yield and high precision rapidassembly method for CNT-based devices. The system devel-opment process included dielectrophoretic CNT manipulation,fluid transport through fabricated microprobes, and positioncontrol of the micromanipulator placement between the mi-croelectrodes on the substrate. The developed microspottingsystem is shown in Fig. 4. In order to spot the CNT/ethanolsolution droplets precisely between the microelectrodes, severalspotting experiments were conducted and will be presented inthe following parts.

A. Fabrication of Nanometric Probe

The tip profile of the microspotting probes directly affects thesize of the micro droplets. Smaller probe tip diameters allow thesystem to eject smaller size droplets. Capillary probes were usedin the microspotting experiments and were fabricated by ournovel chemical etching process, which employed glass tubingas a sacrificial barrier to control and sharpen the probe intodifferent tip profiles as shown in Fig. 5. Detailed fabricationprocesses can be found in [22]. Capillary tubing (TSP002 150with an inner diameter of 2 m and outer diameter of 126 m,which was the smallest size provided by Polymicro Technolo-gies) could be sharpened into capillary probes (with the sameinner diameter and smallest outer diameter of 6.5 m) for ourexperiments.


Fig. 6. Microscopic images (side view) showing the comparison of dropletssize ejected from different outer diameter probe tips.

Fig. 7. Microscopic images showing a droplet cannot be ejected and placedon the substrate because of the large adhesion force between the fluid and theprobe outer surface.

B. Microinjection Experiment

To understand the micro fluidic spotting processes, anexperimental system was built as shown in Fig. 4. This mi-crospotting system consists of five major components: capillaryprobes, a syringe pump, a programmable micromanipulator, acharge-coupled-device-based (CCD-based) video microscopesystem, and a computer. The whole system was installed ona vibration isolation table. Before the injection experiment,a silicon substrate was placed on the stage of the microspot-ting system. A fabricated capillary probe was mounted ona programmable X-Y-Z micromanipulator (MP285, ShutterInstrument Company), which was used to move the probe to thedesired position of the substrate. A syringe pump (V6 syringedrive modules, Kloehn Ltd.) was connected to the capillaryprobe and was used to drive the probe to eject the fluid dropletsto the substrate. Note that the ejection of fluid using a syringepump is not always possible (i.e., for a 2- m diameter capillaryprobe, the hydraulic pressure from this pump would not beenough to eject fluid to the substrate. However, it was success-fully demonstrated that the ultrasmall fluid droplet m)could be ejected by using our fabricated probe ( m) outerdiameter; 10- m inner diameter) as shown in Fig. 6.

It was found that the adhesion force between the outlet ofthe capillary probe and the fluid droplet was very large duringthe experiment (i.e., a droplet cannot be ejected directly ontothe substrate as shown in Fig. 7). The fluid droplet typicallycomes out from the outlet of the capillary probe, but it stickson the capillary probe side wall until the droplet becomes verylarge. The effect of adhesion force can be eliminated by different

Fig. 8. Microscopic images showing a droplet can be placed on the substrateby reducing the initial gap distance between the probe tip and substrate. Theimages in each column represent different trials of an initial gap distance. Theimages in the first row show the initial position of the capillary probe. Theimages in the second row show fluid that was ejected from the probe. The imagesin the third row show the fluid droplet could be ejected when D = 20 �m andD = 10 �m.

Fig. 9. Microscopic images showing a typical single spotting process,especially (d) stable interfaces formed (e) breakpoint of fluid.

methods such as reducing the gap distance between the probe tipand the substrate [i.e., making the probe tip touch the substrate(droplet contact method)], or increasing injection pressure. Theeffect of moving the probe tip closer to the substrate to placedroplets on the substrate is as shown in Fig. 8.

The detailed analysis of a typical single droplet contact in-jection process (droplet contact method) is shown in Fig. 9. Theprobe tip was moved to 30 m above the substrate [Fig. 9(a)].Pressure was then applied to eject fluid. The strong surface en-ergy/force caused the concave shape fluid at the tip [Fig. 9(b)].


Fig. 10. Contact angle at a liquid-solid interface of Young’s equation.

Fig. 11. Microscopic image showing the definition of contact angles at thefluid breakpoint.

When the fluid touched the substrate, a tip-fluid interface anda fluid-substrate interface were formed [Fig. 9(c)]). When theprobe tip remained stationary and the volume of the droplet wasconstant, the interfaces became stable because the fluid surfaceand interface forces were balanced [Fig. 9(d)]. Afterward, whenthe probe tip was moved upward, the fluid interface was brokenat the middle because surface energy of the fluid was not strongenough to maintain the stable configuration. In other words, thefluid surface force was weaker than the fluid-substrate interfa-cial force [Fig. 9(e)]. Finally, a fluid droplet was left on the sub-strate [Fig. 9(f)]. The fluid solid surface force interaction can becharacterized by the following equation [23]:

(5)

where is the radius of the droplet, is the liquid-air inter-facial tension, is the contact angle, and is the separationdistance between the two surfaces. The contact angle, whichis determined from the Young’s equation [23], is the resultof liquid-air, solid-air, and liquid-solid interfacial tensions asshown in Fig. 10. To approximate the liquid surface force at thebreakpoint [Fig. 9(e)], the microspotting experiments of ethanolwere conducted for various initial tip-substrate separations (10

m–30 m). The contact angles and separations were measuredfrom the microscopic images at the breakpoint (Fig. 11). Thecontact angles were defined as shown in Fig. 11 and were from23.2 to 38 as shown in Fig. 12. The separations betweenthe tip and substrate were from 68 m–78 m. With ethanol’ssurface tension of 2.23 mN/m at 20 C, based on (5), thesurface force was estimated consistently between 0.50 to 0.63

from the measured contact angles. Hence, the adhesionforce between the CNT dilution and a given substrate must begreater than for this droplet contact method to workproperly.

Fig. 12. Contact angles measured from microscopic images.

Fig. 13. Microscopic images (top view) showing the probe tip moving to a pairof microelectrodes, and a droplet was ejected from the probe to the substratesuccessfully.

Fig. 14. Illustration of CNT position assignment.

C. CNT/Ethanol Solution Spotting Process on Microelectrodes

After understanding the basic dynamics of the microspottingprocess, further experiments were carried out using droplet con-tact method to spot the CNT/ethanol solution onto a substratewith an array of microelectrodes. The experiments were con-ducted by using our fabricated probe (10- m inner diameter,40- m outer diameter). The CNT/ethanol solution droplet sizesspotted by these probes were approximately 50 m as shown inFig. 13.

Since the microchips with arrays of microelectrodes are fab-ricated by a standard microlithographic technique, a CIF-maskcomputer file is required and predesigned in order to producea lithographic mask. Hence, a custom “CNT layer,” which canbe defined in the CIF-mask computer file, could be added to themask file to locate different positions of the microelectrodes.The CIF-mask computer file could then be imported to our au-tomated CNT microspotting system, where all positions of mi-croelectrodes are obtained and stored in a computer to controlthe movement of a x-y-z microrobotic manipulator. A schematicof this procedure is illustrated in Fig. 14.

A probe was attached to the programmable X-Y-Z micro-manipulator (MP285, Shutter Instrument Company), and a


Fig. 15. Probe tip must align to the initial microelectrode.

Fig. 16. Probe tip moved and dropped CNT/ethanol solution on differentmicroelectrodes.

CCD video camera was connected to the microscope (Micro-manipulator P7000) to locate the initial position of the arrayof microelectrodes through the real-time video image observedon a computer screen. The probe tip was then required to bealigned to the initial (upper-left) microelectrode. A samplemicroscope image of probe tip-electrode alignment was shownin Fig. 15. After this initial probe tip-electrode alignment wascalibrated, the spotting process began. The probe tip was movedsequentially to positions above each microelectrodes, then itwas moved downwards to the substrate to a specified distance,and the syringe pump was commanded to provide a sufficientpressure to eject the CNT dilution to exit the probe tip. Thus,the dilution droplet was sufficiently close to the substrateand, hence, would also form a liquid-solid interface with thesubstrate as previously discussed. Note that the DEP voltagewas applied between each pair of microelectrodes to form theCNT sensing elements during the whole process. After theCNT dilution droplet spotted on the desired microelectrode,the probe tip was moved to the next pair of microelectrodesto repeat the process again. The microspotting process wascompleted until all pairs of microelectrodes were spotted withthe CNT dilution. Sample images of the sequential spotting ofmicroelectrode pairs are shown in Fig. 16.

D. Experimental Results of CNT Manipulation

As mentioned earlier, the ac electric field was applied to themicroelectrodes during the spotting of the CNT dilution. Afterspotting the CNT/ethanol dilution on the microelectrodes, theethanol evaporated away (after a few seconds) and left the CNTs

Fig. 17. (a) Drawing of the design for the chip with arrays of Aumicroelectodes. (b) Scanning electron microscopic (SEM) images showing theformations of MWNTs between different pairs of Au microelectrodes.

to reside between the gaps of the microelectrodes. From exper-imental results, we observed that bundled CNTs were attractedtoward the Au microelectrodes under the nonuniform electricfield and connected across the microelectrodes after the processwas described in pact C. The connections of bundled CNTs forsample pairs of microelectrodes on the substrate after one spot-ting cycle of CNT dilution are shown in Fig. 17. In order to con-firm the linkage of bundled CNTs between each pair of micro-electrodes, room temperature resistance corresponding to eachpair of microelectrodes was measured. The CNT connectionprocess was deemed successful between two microelectrodeswhen the room temperature resistance was measured from sev-eral kilo-ohms to several hundred kilo-ohms. Since the conduc-tivity of CNTs depends on their lattice geometries during theirgrowth process, the conductivities of individual CNTs cannotbe well controlled, which results in the variation of conductivi-ties in individual CNTs. During the dielectrophoresis process toform CNT bundles across microelectrodes, CNT bundles wererandomly connected between microelectrodes. Therefore, it islogical that different CNT samples exhibited different conduc-tivities. Several chips were checked using a scanning electronmicroscope (SEM) to validate the CNT connections betweenthe microelectrodes.

To validate the consistency of the rapid assembled batch as-sembly of CNT devices, repeated experiments on different chipswere performed. A plot of statistical data for different exper-iments was generated (Fig. 18), which showed the maximum,minimum, average, and standard deviation (S. D.) among themeasured resistances on each sample. From the experimental re-sults, we observed that the range of the room temperature resis-tances is from several kilo-ohms to several hundred kilo-ohmsas stated earlier. Besides, we experimentally found that the suc-cess rate for different sensor chips is consistent with the overallsuccess rate, which is equal to or greater than 75% (Fig. 19).


Fig. 18. Plots of statistical data of measured resistances between the Aumicroelectrodes on different samples.

Fig. 19. Plots of overall yield for different sensors chips.

The success rate is defined as the ratio of the number of suc-cessful CNT-connected microelectrodes to the total number ofmicroelectrodes on a substrate.

IV. CARBON NANOTUBE AS SENSING ELEMENTS FOR

THERMAL SENSING

In this section, we will present an example of an applicationfor CNT elements linked across microelectrodes. Some of theresults presented in this section were previously reported in [19].However, we have verified that the experimental results reportedhere can be applied in general to CNT sensors fabricated usingthe method reported in [19] and also the automated method re-ported in this paper.

A. Thermal Sensitivity

In order to investigate the temperature dependence of the bun-dled CNT-based sensors, several fabricated sensor chips werepackaged on a printed-circuit board (PCB) for data acquisi-tion and were placed at the hotplate (Dataplate Digital Hot-plate PMC 732 series, Barnstead International). The hotplatetemperature was monitored by a Fluke type-K thermocouple at-tached on the surface of the PCB. Then, the resistance change

Fig. 20. TCR variations of a typical MWNT device in five consecutivemeasurements. The experimental data are linearized for the approximation ofTCRs.

Fig. 21. Room temperature resistance drift of a CNT sensor during sevenrepeated thermal cycling.

of the CNT-based sensors was measured against the tempera-ture. The temperature-resistance relationship of the CNT-basedsensors was measured and a representative data set is shown inFig. 20. Experiments showed that, in most cases, a few thermalannealing cycles (applying different currents to heat up the sen-sors for a few cycles) were necessary before the measured tem-perature coefficient of resistances (TCRs) of each sensor can bestabilized (i.e., considerable room temperature resistances driftwere observed in repeated measurements during the thermal an-nealing cycles). The statistical data of the room temperature driftof a typical CNT sensor for each cycle of measurement wasmeasured and plotted in Fig. 21. It was found that the magni-tude of the change in resistance during annealing is 40% aftertwo cycles. The resistance is stabilized afterwards. One possiblereason to account for this phenomenon is that during the thermalannealing cycling processes, some CNTs with weak adhesion tothe Au microelectrodes may detach due to the mismatch in tem-perature coefficient of expansion, causing the drift of the roomtemperature resistance after each thermal cycling. The bulk re-sistance due to the detachment of some CNTs after annealingcan be either increased or decreased because the CNT bundlesare randomly oriented and formed between the microelectrodes(i.e., the change should be different for different samples). How-ever, we have observed that the resistance is stabilized after thethermal annealing cycles process. Therefore, our current workis to use polymer and conventional silicon-based thin films toprotect the CNTs and provide a better fixture of CNTs with themicroelectrodes.


Fig. 22. Typical I-V characteristics of the MWNT bundles. Three repeatedmeasurements were performed to validate the repeatability. The straight lineis the theoretical expectation using Ohm’s Law and the room temperatureresistance of bundled MWNTs in our testing sample was about 87.34 k.

The TCR of the CNT-based sensors was measured to deter-mine their thermal sensitivity. From the experimental results,the bundled CNTs were sensitive to the change of temperatureand their resistance dropped with increasing temperature. Thenegative TCR of MWNTs has been reported previously in [24],though the measurements were based on individual MWNTs.The temperature-resistance dependency of bundled MWNTsimplied their thermal sensing capability. Based on experimentalresults, the TCR of the CNT-based sensor was obtained bymeasuring the resistance change of the bundled CNTs with thecorresponding temperature and was calculated based on

(6)

where is the resistance at room temperature , andis the temperature coefficient of resistance. The range of theTCR for the CNT-based sensors was found to be from 0.1 to

C.

B. Power Consumption

Electrical properties of individual CNTs have been studiedby many research groups [24], [25] in the past few years. Inthis work, we have determined I-V characteristics of the bun-dled CNTs by a source meter (Model 2400 General-PurposeSourceMeter, Keithley, Instruments, Inc.) under constant cur-rent configuration. Similar to those of individual CNTs reportedby other researchers, the bundled CNTs also showed powercharacteristics (Fig. 22), that is, bundled CNTs can be regardedas a complex network of individual CNTs. From the results ofexperiments conducted on the CNT-based sensors, the currentrequired to induce self-heating of the CNT devices was in the

A range at several volts, which suggests that bundled CNTscould be used as resistive elements for low power consumptiondevices. This self-heating effect is evident in Fig. 22, where theI-V relationship becomes nonlinear when the current across theCNT resistive element is above a certain level. That is, whenthe temperature (induced by current heating) across the CNTsis high enough to change the sensing element’s resistance, itsI-V characteristic becomes nonlinear. Typically, the two-probe

room temperature resistance of the bundled MWNTs rangedfrom several kilo-ohms to several hundred kilo-ohms as reportedpreviously. These variations may be a result of the random con-nections between MWNT bundles during the dielectrophoresisprocess. However, there are other possible factors as it is gener-ally recognized that the CNT resistance among many reportedexperiments varied (e.g., see [26] and [27]). It was reported in[26] that several factors, such as quality of the nanotubes andthe electrical contacts, may account for the variations of themeasured resistance across CNTs. The existence of metal–nan-otube contact resistance was explained theoretically in [27] (i.e.,in an idealized model, conduction between the CNT and metalmay be forbidden by the block symmetry of the electrodes). Onthe other hand, it was studied experimentally that contact resis-tance in the order of hundreds of kilo-ohms or megaohms wasmeasured when positioning a single CNT on a metal electrode[28] or depositing metal electrodes across a selected single CNT[29]. Besides, it was indicated in [30] that contact resistanceexists between the nanotube bundles and the metal electrodesdue to observed discrepancy of three orders of magnitude whendifferent methods were used to measure the resistance. There-fore, the contact resistance should have an influence on the vari-ation of resistance. The magnitude of contact resistance of oursensors has to be further investigated in the future. Neverthe-less, several groups have reported that the contact resistancecan be decreased by orders of magnitudes by post-treatments,such as local electron beam exposure [26], rapid thermal an-nealing process [31], electroplated Au over the contacted elec-trodes [32], and focused ion beam to deposit metal on CNT,which showed that the contact resistances were decreased below100 or in the range of a few tens of s [28]. Hence, futurework for our group will involve applying one of these techniquesto reduce the contact resistance of the CNT-based sensors.

Despite the random variations in room temperature resistanceof different MWNT bundles, it was found that all of the sampleshave similar I-V characteristics. In comparison, the operationpower of many MEMS polysilicon-based thermal sensors is inthe order of milliwatts [33]. In other words, the power consump-tion of CNT-based sensors is three orders of magnitude lowerthan that of many MEMS polysilicon-based thermal sensors.

C. Frequency Response of the Electronic Circuit With BundledMWNT as Feedback Resistor

The frequency response of a sensor measures its efficiencyto pick up time-dependent input changes from the surroundingenvironment. In order to pick up time-dependent fluctuations ofthe sensing environment, sensors with fast frequency responseare highly desired. For turbulent boundary layer research in thefield of aerodynamics, shear stress sensors with fine resolution

m), fast frequency response ( 5 kHz), and high sen-sitivity are highly sought [34].

A typical bundled MWNT sensing element was connectedin a constant current mode configuration in the frequency-re-sponse experiment (Fig. 23). To test the frequency response ofthe driving circuitry with bundled MWNT as a feedback re-sistor, an input square wave of 2 V peak-to-peak at 10 kHzwas fed into the negative input terminal of the circuit shown inFig. 23 and the output response was determined (Fig. 24). From


Fig. 23. Schematic diagram showing the constant current-mode circuitconfiguration.

Fig. 24. Frequency response of the constant current circuit with bundledMWNTs as feedback resistors. The cutoff frequency was estimated to be 177kHz according to [33].

our experimental measurements, the electronic circuit with bun-dled MWNT functioning as a feedback resistor exhibited veryfast frequency response to the input electronic signal. Using theapproximation between the time constant and cutoff frequency[33]

(7)

where is the cutoff frequency, is the time constant of theresponse, and the estimated cutoff frequency of the system wasabout 177 kHz. As a comparison, the typical frequency responseof the constant current-mode circuitry with MEMS polysiliconsensors, as feedback resistors without frequency compensation,is around several hundred hertz to several hundred kilohertz[35].

V. CONCLUSION

An automated microspotting system to rapidly fabricateCNT nanosensors based on dielectrophoretic manipulationwas presented. We have proven that the success rate for batchassembling bundled CNTs on arrays of microelectrodes is ap-proximately 75% and the time required to form one CNT sensoris in a few seconds. This demonstrated that the dielectrophoreticmanipulation process performed by the automated microspot-ting system is a feasible technology to rapidly assemble CNT

functional devices in a precise manner. With the ability to ma-nipulate bundled MWNTs between microfabricated electrodes,bundled MWNTs were investigated as sensing elements formicrothermal sensing applications. We have successfully inte-grated some CNT-based sensors into a constant current-modeconfiguration for dynamic characterizations, such as air flowand frequency response. We have also performed I-V and TCRmeasurements on several CNT-based sensors. Due to the lowpower consumption (in the range of ) and fast frequencyresponse ( 177 kHz) of bundled MWNT sensing elements,CNT-based sensors were shown to be promising in thermalsensing applications.

ACKNOWLEDGMENT

The authors would like to thank Dr. W. Y. Cheung of the De-partment of Electronic Engineering of CUHK for his help withthis project.

REFERENCES

[1] S. Frank, P. Poncharal, Z. L. Wang, and W. A. de Heer, “Carbon nanotubequantum resistors,” Science, vol. 280, pp. 1744–1746, 1998.

[2] E. W. Wong, P. E. Sheehan, and C. M. Lieber, “Nanobeam mechanics:elasticity, strength, and toughness of nanorods and nanotubes,” Science,vol. 277, pp. 1971–1975, 1997.

[3] S. Iijima, “Helical microtubules of graphitic carbon,” Nature, vol. 354,pp. 56–58, 1991.

[4] T. Shiokawa, K. Tsukagoshi, K. Ishibashi, and Y. Aoyagi, “Nanostruc-ture construction in single-walled carbon nanotubes by AFM manipu-lation,” in Proc. Microprocesses and Nanotech. Conf., Shimane, Japan,2001, pp. 164–165.

[5] B. Q. Wei, R. Vajtai, Y. Jung, J. Ward, R. Zhang, G. Ramanath, andP. M. Ajayan, “Organized assembly of carbon nanotubes,” Nature, vol.416, pp. 495–496, 2002.

[6] J. Hu, M. Ouyang, P. Yang, and C. M. Lieber, “Controlled growth andelectrical properties of heterojunctions of carbon nanotubes and siliconnanowires,” Nature, vol. 399, pp. 48–51, 1999.

[7] J. E. Fischer, W. Zhou, J. Vavro, M. C. Llaguno, C. Guthy, R. Haggen-mueller, M. J. Casavant, D. E. Walters, and R. E. Smalley, “Magneticallyaligned single wall carbon nanotube films: preferred orientation andanisotropic transport properties,” J. Appl. Phys., vol. 93, pp. 2157–2163,2003.

[8] Y. Huang, X. Duan, Q. Wei, and C. M. Lieber, “Directed assembly ofone-dimensional nanostructures into functional networks,” Science, vol.291, pp. 630–633, 2001.

[9] S. G. Rao, L. Huang, W. Setyawan, and S. Hong, “Large-scale assemblyof carbon nanotubes,” Nature, vol. 425, pp. 36–37, 2003.

[10] R. Krupke, F. Hennrich, H. Löhneysen, and M. M. Kappes, “Separationof metallic from semi-conducting single-walled carbon nanotubes,” Sci-ence, vol. 301, pp. 344–347, 2003.

[11] K. Yamamoto, S. Akita, and Y. Nakayama, “Orientation of carbonnanotubes using electrophoresis,” Japanese J. Appl. Phys., vol. 35, pp.L917–L918, 1996.

[12] , “Orientation and purification of carbon nanotubes using AC elec-trophoresis,” J. Phys. D: Appl. Phys., vol. 31, pp. L34–L36, 1998.

[13] H. A. Pohl, Dielectrophoresis: The Behavior of Neutral Matter inNonuniform Electric Fields. Cambridge, MA: Cambridge Univ.Press, 1978.

[14] A. Ramos, H. Morgan, N. G. Green, and A. Castellanos, “AC elec-trokinectics: a review of forces in microelectrode structures,” J. Phys.D: Appl. Phys., vol. 31, pp. 2338–2353, 1998.

[15] T. Heida, W. L. C. Rutten, and E. Marani, “Understanding di-electrophoretic trapping of neuronal cells: modeling electric field,electrode-liquid interface and fluid flow,” J. Phys. D: Appl. Phys., vol.35, pp. 1592–1602, 2002.

[16] T. B. Jones, Electromechanics of Particles. Cambridge, MA: Cam-bridge Univ. Press, 1995.

[17] C. A. Grimes, E. C. Dickey, C. Mungle, K. G. Ong, and D. Qian, “Ef-fect of purification of the electrical conductivity and complex permit-tivity of multiwall carbon nanotubes,” J. Appl. Phys., vol. 90, no. 8, pp.4134–4137, 1995.


[18] P. Petong, R. Pottel, and U. Kaatze, “Water-ethanol mixtures at differentcompositions and temperatures. A dielectric relaxation study,” J. Phys.Chem. A, vol. 104, pp. 7420–7428, 2000.

[19] R. H. M. Chan, C. K. M. Fung, and W. J. Li, “Rapid assembly of carbonnanotubes for nanosensing by dielectrophoretic force,” Nanotechnology,vol. 15, pp. S672–S677, 2004.

[20] V. T. S. Wong and W. J. Li, “Dependence of AC electrophoresis carbonnanotube manipulation on microelectrode geometry,” Int. J. NonlinearSci. Numer. Simulation., vol. 3, no. 3–4, pp. 769–774, 2002.

[21] Sun Nanotech Co. Ltd., , Beijing, China.[22] K. W. C. Lai and W. J. Li, “KL probes for robotic-based cellular nano

surgery,” in Proc. IEEE Nano, vol. 1, 2003, pp. 152–155.[23] C. H. Mastrangelo and C. H. Hsu, “Mechanical stability and adhesion of

microstructures under capillary forces. I. Basic theory,” J. MEMS, vol.2, no. 1, pp. 33–43, 1993.

[24] T. W. Ebbesen, H. J. Lezec, H. Hiura, J. W. Bennett, H. F. Ghaemi, and T.Thio, “Electrical conductivity of individual carbon nanotubes,” Nature,vol. 382, pp. 54–56, 1996.

[25] P. G. Collins, M. S. Arnold, and P. Avouris, “Engineering carbon nan-otubes and nanotube circuits using electrical breakdown,” Science, vol.292, pp. 706–709, 2001.

[26] A. Bachtold, M. Henny, C. Terrier, C. Strunk, and C. Schonenberger,“Contacting carbon nanotubes selectively with low-ohmic contacts forfour-probe electric measurements,” Appl. Phys. Lett., vol. 73, no. 2, pp.274–276, 1998.

[27] J. Tersoff, “Contact resistance of carbon naotubes,” Appl. Phys. Lett.,vol. 74, no. 15, pp. 2122–2124, 1999.

[28] B. Q. Wei, R. Vajtai, and P. M. Ajayan, “Reliability and current car-rying capacity of carbon nanotubes,” Appl. Phys. Lett., vol. 79, no. 8,pp. 1172–1174, 2001.

[29] F. Wakaya, K. Katayama, and K. Gamo, “Contact resistance of multiwallcarbon nanotubes,” Microelectron. Eng., vol. 67–68, pp. 853–57, 2003.

[30] L. Marty, V. Bouchiat, A. M. Bonnot, M. Chaumont, T. Fournier, S. De-cossas, and S. Roche, “Batch processing of nanometer-scale electricalcircuitry based on in-situ grown single-walled carbon nanotubes,” Mi-croelectron. Eng., vol. 61–62, pp. 485–489, 2002.

[31] J. O. Lee, C. Park, J. J. Kim, J. H. Kim, K. W. Park, and K. H. Yoo,“Formation of low-resistance ohmic contacts between carbon nanotubeand metal electrodes by a rapid thermal annealing method,” J. Phys. D:Appl. Phys., vol. 33, pp. 1953–1956, 2000.

[32] D. W. Austin, A. A. Puretzky, D. B. Geohegan, P. F. Britt, M. A. Guillorn,and M. L. Simpson, “The electrodeposition of metal at metal/carbonnanotube junctions,” Chem. Phys. Lett., vol. 361, pp. 525–529, 2002.

[33] C. Liu, J. B. Huang, Z. Zhu, F. Jiang, S. Tung, Y. C. Tai, and C. M. Ho,“A micromachined flow shear-stress sensor based on thermal transferprinciple,” J. MEMS, vol. 8, no. 1, pp. 90–99, 1999.

[34] J. B. Huang, C. Liu, F. Jiang, S. Tung, Y. C. Tai, and C. M. Ho, “Flu-idic shear stress measurement using surface-micromachined sensors,”in Proc. IEEE Region 10 Int. Conf. Microelectronics and VLSI, HongKong, China, 1995, pp. 16–19.

[35] J. B. Huang, F. K. Jiang, Y. C. Tai, and C. M. Ho, “MEMS-basedthermal shear-stress sensor with self-frequency compensation,” Meas.Sci. Technol., vol. 10, no. 8, pp. 687–696, 1999.

King Wai Chiu Lai (S’03) received the B.Eng.and M.Phil. degrees in mechanical and automationengineering from The Chinese University of HongKong (CUHK), in 2000 and 2002, respectively, andthe Ph.D. degree in automation and computer-aidedengineering from CUHK in 2005.

Currently, he has focused on microassembly andfabrication of bio-MEMS devices. His researchoutput has been published in various journals andpresented in several international conferences. Hehas collaborated in the research of teleoperation in

microassembly of MEMS devices with researchers from the Robotics andAutomation Laboratory of Michigan State University, East Lansing. He alsoworked with researchers of the Photonic Packaging Laboratory of CUHK inthe design and packaging of microoptoelectromechanical systems (MOEMS).

Dr. Lai has received several awards such as the Best Conference Paper fromthe International Conference on Intelligent Mechatronics and Automation(ICIMA) and Best Student Paper from the 2000 International Conference onInformation Society in the 21st Century (IS2000).

Carmen Kar Man Fung (S’03) received the B.Eng.degree in mechanical and automation engineeringin 2000 and the Ph.D. degree in 2005 from theDepartment of Automation and Computer-AidedEngineering, The Chinese University of Hong Kong.

Dr. Fung received the Best Student Paper Awardfrom the IS2000 in 2000; the Academic CreativityAward from the Chung Chi College, CUHK, in 2000;and the Best Poster Paper Award of the IEEE Inter-national Conference on Nanotechnology 2003, SanFrancisco, CA.

Victor Tak Sing Wong (S’00) received the B.Eng de-gree in automation and computer-aided engineeringfrom the Chinese University of Hong Kong (CUHK),Hong Kong, in 2003.

Currently, he is a Student Researcher with theCentre for Micro and Nano Systems of CUHKfrom 2001 to 2003, and is currently pursuing thePh.D. degree in mechanical engineering from theUniversity of California, Los Angeles. His currentresearch interests involve the study of micro andnanoscale sciences and the development of micro and

nanofabrication technologies for material science and biological applications.

Mandy Lai Yi Sin received the B.Sc. degree inphysics from The University of Hong Kong, HongKong, China, in 2000, the M.Phil. degree in physicsfrom The Chinese University of Hong Kong (CUHK)in 2002, and is currently pursuing the Ph.D. degreein automation and computer-aided engineering atCUHK.

Her current research interest is the development ofthe micro/nanosensors for chemicals based on multi-walled carbon nanotube.

Wen Jung Li (S’97–A’97–M’00) received the B.S.and M.S. degrees in aerospace engineering fromthe University of Southern California Los Angeles(UCLA) Los Angeles, CA, in 1987 and 1989,respectively, and the Ph.D. degree in microelec-tromechanical systems (MEMS) from UCLA in1997.

His industrial experience includes The AerospaceCorporation, El Segundo, CA; Silicon Microstruc-tures, Fremont, CA; and the NASA/CalTech JetPropulsion Laboratory, Pasadena, CA. He received

the Aerospace Corporate Fellowship, Silicon Microstructures Inc. EmployeeAward, and a NASA technical innovation award for his contributions to thoseorganizations. He joined the Department of Automation and Computer-AidedEngineering of The Chinese University of Hong Kong (CUHK) in 1997. In thepast eight years, he has published many papers in international journals andconference proceedings on MEMS and nanotechnology-related work. He wasthe Guest Editor of the Focused Section on Micro and Nano Manipulations ofthe IEEE/ASME TRANSACTIONS OF MECHATRONICS, and is the Director of theCentre for Micro and Nano Systems at CUHK.

Dr. Li is a member of the Technical Committee on Nanorobotics andNanomanufacturing of the IEEE Nanotechnology Council, and a DistinguishedOverseas Scholar of the Chinese Academy of Sciences. His research interest isto develop MEMS devices for micro/nanosensing and manipulation.

Chung-Ping Kwong (SM’89) received the Ph.D. de-gree in electronics from The Chinese University ofHong Kong (CUHK) and the Ph.D. degree in mathe-matics from the City University of Hong Kong.

Currently, he is Professor in the Departmentof Automation and Computer-Aided Engineering,CUHK. He has published widely in the areas ofcontrol theory, adaptive filtering, neural networks,and fuzzy systems. His recent research interest isapplied algebraic topology.

Documents

218 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ... · 218 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 3, NO. 3, JULY 2006 Development of an Automated Microspotting