ieee transactions on nanotechnology, vol. 4, no. 5 ... · the speciﬁc cantilever used in this...

IEEE TRANSACTIONS ON NANOTECHNOLOGY, VOL. 4, NO. 5, SEPTEMBER 2005 605

“Videolized” Atomic Force Microscopy forInteractive Nanomanipulation and Nanoassembly

Guangyong Li, Student Member, IEEE, Ning Xi, Member, IEEE, Heping Chen, Student Member, IEEE,Craig Pomeroy, Student Member, IEEE, and Mathew Prokos

Abstract—The main problem in nanomanipulation andnanoassembly using atomic force microscopy (AFM) is its lack ofreal-time visual feedback during manipulation. Fortunately, thisproblem has been solved by our recently developed augmentedreality system, which includes real-time force feedback andreal-time “videolized” visual feedback. Through the augmentedreality interface, the operator can monitor real-time changes of thenanoenvironment during nanomanipulation through a movie-likeAFM image. In this paper, the behavior of some nanowires underpushing is theoretically analyzed and the interaction among thetip, substrate, and nanowires has been modeled. Based on thesemodels, the real-time interactive forces can be used to locally up-date the AFM image in order to obtain movie-like visual feedbackin video frame rate. This augmented reality enhanced systemcapable of manipulation of nanoparticles and nanowires helps theoperator to perform several operations without the need of a newimage scan. AFM-based nanoassembly becomes feasible throughthis newly developed system.

Index Terms—Atomic force microscopy (AFM), augmentedreality, haptic feedback, nanomanipulation.

I. INTRODUCTION

THE techniques for nanomanufacturing can be classifiedinto “bottom-up” and “top-down” methods. Self-assembly

in nanoscale is the main “bottom-up” technique, which is ap-plied to make regular symmetric structures of nanoobjects [1].However, many potential nanostructures and nanodevices areasymmetric patterns, which cannot be manufactured using self-assembly. The semiconductor fabrication technique is a ma-tured “top-down” method, which has been used in the fabrica-tion of microelectromechanical systems (MEMS). However, itis hard to build nanostructures using this method due to limita-tions of the lithography in which the smallest feature that canbe made must be larger than half the wavelength of the lightused in the lithography. Atomic force microscopy (AFM) [2]has been proven to be a powerful technique to study samplesurfaces down to the nanometer scale. Not only can it charac-terize sample surfaces, it can also modify the sample surfacethrough manipulation [3]–[6], which is a promising nanofab-rication technique that combines the advantages of the “top-down” and “bottom-up” methods. In recent years, many kindsof nanomanipulation schemes have been developed to manip-ulate nanoobjects [7], [8]. The main problem with the existing

Manuscript received July 6, 2004; revised February 25, 2005. This workwas supported in part by the National Science Foundation under GrantIIS-9796300, Grant IIS-9796287, and Grant EIA-9911077, and by the Officeof Naval Research under Grant N00014-04-1-0799.

The authors are with the Department of Electrical and ComputerEngineering, Michigan State University, East Lansing, MI 48824 USA(e-mail: [email protected]).

Digital Object Identifier 10.1109/TNANO.2005.851430

manipulation schemes is the lack of real-time visual feedbackduring manipulation. Each operation has to be verified by an-other new image scan before the next operation. Obviously, thisscan–design–manipulation–scan cycle is time consuming andmakes mass production impossible.

Recently, some researchers have been trying to combine theAFM with both the haptic technique and a virtual reality in-terface to facilitate nanomanipulation [9], [10]. Although a vir-tual reality has been constructed, which can display a static vir-tual environment and a dynamic tip position, it does not displayany real-time changes in the environment during manipulation.Therefore, the operator is still blind because he/she cannot seethe real-time changes in the environment. Thus, any methodsthat can update the AFM image as close as possible to the ac-tual environment in real time will help the operator to performseveral operations without the need of a new image scan. A validmodel that can calculate real-time changes of the real environ-ment is very important in order to provide the operator with areal-time visual display during manipulation. In [11], a modelof tip–substrate–particle interaction has been presented in a spe-cial case in which the tip does not contact the substrate surfaceduring manipulation. However, the tip is usually contacting thesubstrate surface in the general case in order to guarantee suffi-cient pushing force during manipulation.

In this paper, an AFM-based nanomanipulation systemassisted by augmented reality has been developed. An operatorcan feel the real-time three-dimensional (3-D) interactionforces through the haptic system and simultaneously observethe real-time changes of the nanoenvironment through the“videolized” AFM image updated in video frame rate. In orderto display the real-time changes of the nanoenvironment, amodel of tip–substrate–object has been developed for the casesin which the tip is contacting the substrate surface duringmanipulation. It has been shown that the nanoparticles can beeasily manipulated under assistance of this augmented realityenhanced system [12]. However, assembly of nanostructuresusually involves manipulation of nanoparticles, nanowires,nanotubes, and many others. Modeling the behavior of ananowire or a nanotube pushed by an AFM tip is much morecomplex than that of a nanoparticle because, in the case ofthe nanoparticle, usually only translation occurs, while forthe nanowire and nanotube, both translational and rotationalmotion or even deformation occur during manipulation. Thebehavior of nanowires with a small aspect ratio under pushingand the interaction among tip, substrate, and nanowires havebeen modeled in this paper. Using the models to update theAFM image based on the measured real-time force information,real-time visual feedback in video frame rate is achieved. The

1536-125X/$20.00 © 2005 IEEE

606 IEEE TRANSACTIONS ON NANOTECHNOLOGY, VOL. 4, NO. 5, SEPTEMBER 2005

Fig. 1. AFM-based augmented reality system. (left) AFM system includingother accessories for imaging functions. (right) Augmented reality interfaceproviding an interface for an operator to simultaneously control the tip motionthrough a haptic joystick, view the real-time AFM image, and feel the real-timeforce during manipulation.

ability of the augmented reality system has been extendedfrom manipulating nanoparticles to nanowires with a smallaspect ratio. The experimental study not only validates themodels, but also proves the effectiveness of the AFM-basednanomanipulation system for assembly of nanostructures.

II. DEVELOPMENT OF THE AFM-BASED

AUGMENTED REALITY SYSTEM

The augmented reality system aims to provide an operatorwith real-time “videolized” visual display and force feedbackduring nanomanipulation. It includes two subsystems: the AFMsystem and the augmented reality interface, which are shown inFig. 1.

The AFM system called Bioscope (Veeco Instruments,Woodbury, NY) is equipped with a scanner with a maximum

scan range of 90 m 90 m and a range of 5 m.Peripheral devices include an optical microscope, a charge-cou-pled device (CCD) camera, and a signal access module, whichcan access most real-time signals inside the AFM system.The inverted optical microscope and the CCD camera helpthe operator to locate the tip, adjust the laser, and search forthe interesting areas on the substrate. The augmented realityinterface is a computer equipped with a haptic device (Phantom,Sensable Company, Woburn, MA). Through the signal accessmodule, the deflection signal can go directly into the A/Dconvertor card inside the computer. The augmented realityinterface provides enhanced media for the operator to view thereal-time “videolized” AFM image and feel the force feedbackduring nanomanipulation. The “videolized” real-time visualdisplay is a dynamic AFM image of the operating environment,which is locally updated based on the environment model,tip-subject interaction model, real-time force information, anda local scanning mechanism, as shown in Fig. 2. The twosubsystems are connected through Ethernet.

The positioning precision is the one of the most critical issuesin AFM-based nanomanipulation. There are many factors thatcause position errors, but the following three are the most signif-icant ones. The first cause of position errors is due to the thermaldrift. If the manipulation is not performed in a controlled en-vironment, the thermal drift is random and sometimes can ap-proach several hundred nanometers from our observation. The

Fig. 2. Detailed structure of the augmented reality system.

position error caused by thermal drift can be compensated usinga local scan mechanism, as shown in Fig. 2. Before starting themanipulation, a quick scan around the object can determine theactual position of the object. The operation immediately per-formed after the local scan can correct the position error dueto the thermal drift. Another factor that causes the position er-rors is that the bending of the cantilever in normal direction alsocauses a lateral displacement along the length direction of thecantilever [12]. Using a stiff cantilever and a position compen-sation algorithm can minimize this effect. The position com-pensation algorithm is discussed in Section II-C, and the relatedparameter calibration methods are discussed in Sections II-Dand E. The third factor causing the position errors is the van derWaals force and electrical static force between the nanoobjects.Since the directions (attractive or repulsive) and the magnitudesof these forces are unpredictable, the behavior of objects is verydifficult to predict when they are very close. Therefore, a smallposition error is inevitable.

Another issue for AFM-based nanomanipulation is the loss ofthe reference for the working space after changing the cantilever.Landmarks have to be used in order to solve this problem. Prac-tically, there are many landmarks available such as some bigfeatures within the working area. Using the inverted optical mi-croscope, the AFM can scan in a big area around the landmark.The working space can be easily recovered by zooming into theoriginal working space.

A. Measurement of 3-D Interactive Forceson the Cantilever Tip

In order to measure the forces applied on the AFM tip, amodel of the cantilever–tip interaction has been developed in[12], as shown in Fig. 3. The 3-D interaction force on tip in thecantilever frame can be measured from the normal and lateralsignals from the photo diode output by the following equations:

(1)

where and are the signal outputs in normal and lateraldirections, is the spring constant of the cantilever, is thetorsional constant of the cantilever, and are the gains

LI et al.: “VIDEOLIZED” AFM FOR INTERACTIVE NANOMANIPULATION AND NANOASSEMBLY 607

Fig. 3. Force analysis model of the cantilever: O is the origin of Cantileverframe. w and l are the cantilever width and length, respectively. h is tip height,which includes the thickness of the cantilever. l is the cantilever length. T is thetip apex.

in the normal and lateral direction, is the tip height, whichincludes the thickness of the cantilever, is the length of thecantilever, and is the tip motion direction in cantilever frame.

The relationship between the AFM frame and cantileverframe is determined by the following rotational matrix:

(2)

where is the scanning angle during imaging. The 3-D forcesrepresented in the AFM frame are then

(3)

From (1), it is clear that the spring and torsion constantsand and the signal gains and must be calibrated inorder to obtain the actual forces acting on the tip. The springconstant is usually provided by the manufacturer or can becalibrated experimentally [13], [14], and the normal gaincan also be calibrated experimentally [12] ( nm/V forthe specific cantilever used in this paper). Although andcannot be obtained separately, their product can beobtained by directly calibrating the relation between the lateralforce and normal force [15]. There are at least three calibrationtechniques now available [16]–[18]. A simpler method is usedin this paper to calibrated the product .

B. Calibration of

Although can be estimated as (see Sec-tion II-D), it is very difficult to obtain . Fortunately, it is pos-sible to calibrate the product . If there is a tilted hardsurface with slope angle , as shown in Fig. 4, the followingequation must be satisfied when the tip is sliding up or downalong the tilted surface:

(4)

(5)

(6)

(7)

Fig. 4. Force analysis when the tip is sliding up and down on a tilted surface.

where are the load force and lateral forces, respec-tively, measured by AFM (when scanning angle

), are the actual repulsive force from the titled surface,and and are the frictional forcesincluding the frictions and the adhesive shear force .Here, is the friction coefficient.

Equations (4) and (5) can be solved as

(8)

and (6) and (7) can be solved as

(9)

From (5) and (7), it can be found that

Therefore,

(10)

(11)

From (1), it can be seen that the measured normal force is

and the measured lateral force is

Noting that when , it is clear that

(12)

and

(13)

Therefore,

(14)

(15)


Fig. 5. S and S with respect to S . @S =@S = 0:07 and @S =@S =0:02, respectively.

Finally, and can be found by solving the fol-lowing two equations:

(16)

(17)

Equations (16) and (17) are nonlinear and they can be solvedusing the graphic method or other techniques.

Here is an example on how to find for the cantileverwith a spring constant of 0.12 N/m.

• First, a titled glass surface was made and then scannedwith a 270 scanning angle (the cantilever frame and AFMframe are overlapped in this case and ) and 10- mscanning range.

• Second, using the height information from the image, it isvery easy to find the glass slope angle .

• Third, applying different load signals by changing thescanning setting point and recording both the appliedvoltage and detected friction voltage and from thefriction loop, the plots of and with respect to areobtained, as shown in Fig. 5.

• Fourth, the slopes of and with respect to are foundas and .

• Finally, and can be obtained by graphicallysolving (16) and (17), as shown in Fig. 6. The results are

nN/V and .

C. Position Control

By mapping the position information of the joystick to theAFM coordinates, the cantilever tip can be controlled to moveon or above the sample surface. However, the cantilever is a longand thin beam so any further deformation with respect to theoriginal position in the imaging mode will cause displacementof the tip both in the normal and lateral direction, as shown inFig. 7.

Fig. 6. Graphic solution of (k K =h) and �. The cross point is the solutionthat shows that (k K =h) = 37:8 nN/V and � = 0:086.

Fig. 7. Lateral displacement of the AFM tip due to the bending of thecantilever: �n moving of the piezotube end in normal direction causes �l tipmovement in lateral direction.

A detailed analysis in [12] shows that the tip displacementcorresponding to the normal and lateral output signals are

where and are constants, which need to be calibrated (seeSection II-D). Finally, the new tip position with respect to theorigin of the AFM frame can be found using

where is a scaling constant from the haptic joystick to AFMframe, is the scanning range, are the coordinatesof the joystick position, and is same transformation matrix asdefined in Section II-A.


D. Estimation of the Displacement Constant

Calibration of experimentally is still an open problem,but it can be estimated based on certain assumptions. Assumethe quad-photodiode detector has the same sensitivity both inthe normal and lateral direction, namely, the normal and lateralsignal outputs should be equal if the bending angle in normalplane is equal to the twisting angle. Usually the bending andtwisting angles are very small so they can be estimated as

Noting that , the bending angle can be estimated as

Considering the same sensitivity both in the normal and lateraldirection, it can be found that

The displacement in the -direction caused by the twistingangle can be estimated as

Finally, it can be found that . Since and areknown, and can be calibrated, then is obtainable. Forthe cantilever with a 0.12-N/m spring constant, the displacementconstant nm/V.

E. Calibration of the Displacement Constants

By inscribing several straight lines on a soft surface at dif-ferent cantilever bending levels, but with the same piezotubeend trajectory, the lateral displacement can be directly mea-sured. After setting the scanning angle to 0 , the AFM framerotates 180 corresponding to the cantilever frame. Therefore,the displacement along the -direction in the cantilever frameis equivalent to the displacement along the negative -direc-tion in the AFM frame. Fig. 8 shows four lines inscribed on apolycarbonate surface using the cantilever with spring constant0.12 N/m at different pushing depths. It can be seen that the dis-placement increases along the negative -direction in the AFMframe.

In order to minimize the measurement error, a large separa-tion distance between lines is preferred. This requires a largepushing depth , which may cause the saturation of the normalsignal output because the maximum normal signal output is only

10 V. Since nm/V, any pushing depth larger than890 nm will cause signal saturation. However, this problem canbe solved by considering a virtual output , whichis equivalent to in analysis. The displacement constant isfinally obtained as

where is the distance between two lines andis the virtual signal output difference between two lines. Forthe cantilever with 0.12-N/m spring constant, the displacement

Fig. 8. Lateral displacement caused by different pushing depths. The linescorresponding to the pushing depth are 2.0, 1.5, 1, and 0.5 �m from the left- toright-hand side. The distance between two neighboring lines is around 130 nm.

constant nm/V according to the measured data fromFig. 8.

III. MANIPULATION OF NANOPARTICLES

In order to update the real-time “videolized” AFM image,the nanoparticle’s behavior has to be modeled. A detailed anal-ysis of tip–substrate–particle interaction was discussed in [12].Using the augmented reality system, manipulation of latex par-ticles has been performed to verify the effectiveness of the real-time visual display. Latex particles with diameters of 100 nm aredistributed on a glass surface. The real-time AFM image is dis-played in the augmented reality interface, as shown in Fig. 9(b).A new scanning image, as shown in Fig. 9(c), shows that thefinal result matches the display in the augmented reality inter-face. The little mismatch of the final position of the latex parti-cles between the augmented reality display and the final result isdue to the inevitable position errors, as analyzed in Section II.Under the assistance of the augmented reality system, manip-ulation of nanoparticles becomes very straightforward. The ex-perimental result shown in Fig. 10 is done by pushing more than100 nanoparticles to form complex patterns within a half hour.

IV. MANIPULATION OF NANOWIRES

In order to enable the nanoassembly of nanoobjects usingthe augmented reality system, the tip–wire–surface interactionmodel and the wire behavior model during nanomanipulationhave to be developed to update the AFM image in real time.

A. Modeling the Behavior of the Nanowire Under Pushing

The nanowire under pushing may have different kinds of be-havior, which depends on its own geometric properties. Definethe aspect ratio of a nanowire as

Practically, a wire with an aspect ratio of usually be-haves like a rope, which will deform or bend under pressure.Fig. 11 shows a nanowire with an aspect ratio of 25, which wasbent by pushing it at one end.


Fig. 9. Pushing latex particles (diameter of 100 nm) on a glass surface with anoperation range of 5 �m. (a) Image of latex particles on a glass surface beforemanipulation. (b) The real time display on the augmented reality during pushingoperation. (c) A new scanning image after manipulation.

Fig. 10. Pushing particles on a glass surface to form patterns. The nanoobjectsare latex particles in diameters of 100 nm. The work area is 10 �m� 10 �m.

Fig. 11. Silver nanowire with aspect ratio of � = 30 was bent due to theexternal pushing force. (a) The nanowire before pushing (b) The nanowire afterpushing, which was bent in the middle.

The rotation behavior was observed for nanowires with aspectratios of . The deformation of the nanowires with largeaspect ratio needs further investigation in order to model thebehavior of the nanowire. In this paper, only the nanowires withaspect ratio less than 15 will be discussed. The detailed externalforces applied on the nanowire in this case have been shown inFig. 12(a). The nanowire with dimension of is pushedat point and it rotates around point . The pushing forcefrom the tip causes the friction or shear force , which can becalculated as

Fig. 12. External forces applied on nanowire in surface plane. (a) Detailedforce model. Here, S is the static point, T is the pushing point and also the tipposition, L is the length of the rod, d is the width of the rod, l is the lengthbetween the pushing point and the reference end, s is the distance between thestatic point S and the reference end, F is the pushing force from the tip, Fis the friction and shear force from tip, f is the evenly distributed friction andshear force density on the rod, and f is the evenly distributed friction and shearforce density along the axis of the wire. (b) Simplified force model. Here, A isthe reference point, B is the other end of the wire, and the other parameters arethe same as in (a).

where and are the friction and shear coefficients betweenthe tip and nanoobject, which depend on the material propertiesand the environment, and is the adhesion force between thetip and the nanowire. The shear force is along the wire axisdirection when the pushing direction is not perpendicular to thewire axis. Fortunately, it is easy to prove that the force doesnot cause any significant wire movement along the wire axisdirection. Assuming that the shear forces between the wire andsubstrate surface are equal along any direction during moving,then

The shear force is usually proportional to the contact area andthe contact area between the wire and surface is much greaterthan that between the tip and wire. Therefore,

and also note that

Since is usually is very small, it is finally reasonable to assumethat

This means that the wire will have no motion along the wire axisdirection. Therefore, the static point must be on the axis of thewire.


Fig. 13. Possible position of the static point S. When l < L=2; S fallsbetween

p2L=2 to L, when l > L=2; S falls between 0 to (1 �

p2=2)L to

L, and when l = L=2, the solution is not unique.

Considering the above analysis, the wire can be simplified asa line segment. The external forces applied on the wire in thesurface plane can be modeled as shown in Fig. 12(b). The staticpoint can be either inside or outside of the wire. First, assumethat is inside the wire. In this case, all the torques aroundare self-balanced during smooth moving as follows:

namely,

(18)

There must be an , which can minimize , in which the wirebegins to rotate once the pushing force reaches this minimumforce . Therefore, the static point can be determined bythe following equation:

(19)

Since we have assumed that , a unique solution of(19) for any , except can be determined by

(20)

The results are shown in Fig. 13. It can be seen that the staticpoint will never fall between and .

When , there is no unique solution. A detailed anal-ysis will show that can be anywhere outside the wire whenthe force is applied in the exact middle of the wire, in which

becomes a bifurcation point. Now, assume the static pointis outside of wire, but on the left-hand side. Noting thatnow, the self-balanced torque equation becomes

Fig. 14. Minimum force required for the pushing wire at a different location.The pushing force is only (2 �

p2)fL when pushing the wire from the end.

The pushing force reaches the maximum value fL when pushing the wire inthe middle.

Namely,

(21)

By minimizing , it can be seen that only when canenable

Similarly, if is on the right-hand side, the results should be thesame. Practically, it is hard to keep at this bifurcation point

. Any disturbance will move the static point to oneend of the wire. Therefore, during manipulation, it is better toavoid pushing the exact middle of the wire because it is hard topredict the behavior of the wire in this case. The manipulationscheme of a nanowire has to go through a zigzag strategy inorder to position the nanowire to a specified orientation.

B. Modeling the Tip–Substrate–Wire Interaction

After the behavior of the wire has been modeled, the motionof the wire should be predictable once the minimum force isachieved by pushing the tip down to the surface. The minimumforce required for pushing the wire at different locations can besolved by the self-balanced force equations as follows:

(22)

The relationship between and are shown in Fig. 14. Itcan be shown that a maximum force is neededin order to push the wire in the middle and a minimum force

to push the wire at the end. This is anotherreason that we prefer pushing the wire from the end. Obviously,if the tip can generate a pushing force , then the wire is


movable on the surface. Otherwise, the wire will remain staticon the surface. It is obvious that

where is the repulsive force between the object and the sub-strate surface, is the sliding friction coefficient between thewire and the substrate surface, is the adhesion force betweenthe object and the substrate surface, and is the shear coeffi-cient. Now (22) becomes

(23)

Let

Then

Equation (23) then becomes

(24)

The interaction among tip, substrate, and wire can be modeledas shown in Fig. 15. Three main types of forces are shown inFig. 15, i.e., adhesive, repulsive, and frictional force. The la-beling of these forces is chosen in such way as to distinguishthem easily. The superscript of force could be one of “a,” “f,”or “r” representing “adhesive,” “frictional,” and “repulsive,” re-spectively. The subscripts could be a combination of “t,” “o,”and “s” representing “tip,” “object,” and “substrate,” respec-tively. For example, represents the adhesive force appliedto the tip by the object.

As illustrated in Fig. 15, the directions of the three main basictypes of forces are known, although the accurate force value isnot available. By assuming that the pushing direction is perpen-dicular to the body axis of the nanowire, the equilibrium condi-tion of the nanowire both in the horizontal and vertical directionscan be obtained as

(25)

(26)

The equilibrium condition of the tip in the normal and lateraldirections is

(27)

(28)

Define as the minimum value of to maintain equilib-rium under the condition that when the tip apex isstill contacting the substrate surface. Noting that since

and can be solved by (25), (26),and (27) based on whether the nanowire is sliding or rolling.

Fig. 15. Model of tip–wire interaction, where � is the twisting of thecantilever and is the open half angle of the tip apex. (a) Forces applied to thewire. (b) Forces applied to the tip.

Whether the nanowire is sliding or rolling is determined by theenergy cost during pushing. It has been observed that the energycost for rolling is larger than the sliding cases when pushinga carbon nanotube sideways. Therefore, the nanotubes shouldprefer sliding instead of rolling. Although the rolling behaviorof a carbon nanotube on a graphite substrate surface has beenobserved in [19], in most cases, the nanoobjects under manipu-lation should slide on the surface due to their irregular shape, de-formation, and strong adhesive force. During sliding, the rollingpotential due to the friction has been balanced by the irreg-ular distribution of the adhesive force . The object has norelative movement with respect to the tip, but slides on the sub-strate surface. Therefore, the friction can be assumed to bezero. The friction between the object and substrate surfaceis the force needed to be overcome in order to move the wire onthe surface. Therefore, , which can be calculated from(24). Using (24) to solve (25) and (26), the repulsive force be-tween the tip and wire can be obtained as

(29)

In order to prevent the tip from squeezing the nanowire into thesubstrate surface, the angle should be less than a certain value,which can be found by letting the denominator of the secondterm in (29) equal zero, i.e., goes to infinity. Therefore,

Substituting (29) into (27), and letting and, the minimum value of to maintain the equilibrium

condition is

(30)

In order to maintain the pushing condition, has to be largerthan . The additional force is balanced by . If

, it means that the adhesive force itself can keep the tip incontact with the substrate surface. During manipulation, a small

is preferred since a large will induce a large friction forcebefore the tip contacts the object, which wears out the tip


Fig. 16. Minimum normal force �F to keep the tip on the surface in slidingmode as a function of � and �. (Here, the unit of �F is F . Assume F =

20F and � = 0:2.)

apex quickly. The simulation results of the minimum forcecorresponding to and have been shown in Figs. 16 and 17.From the simulation results, it can be seen that is proportionalto . Therefore, a stiff cantilever and a tip with a small apex openangle is also preferred for manipulation of nanowires.

C. Experiments on Assembly of the Nanowires

In order to make nanodevices, some of the nanostructureshave to be settled in place. For example, a carbon nanotubehas to be fixed between two electrodes in order to make it ananosensor. Fortunately, under the assistance of the augmentedreality system, it is possible to first create a fixture that can holda nanowire in place and then push the nanowire into it. Fig. 18(a)shows a silver nanowire with length of 1.3 m and diameter of100 nm (aspect ratio of ).

First, a trench, as a holder for the nanowire, can be createdusing the AFM tip by inscribing it into the surface. Fig. 18(b)shows the real-time result displayed in the augmented realityinterface. After a new scan, the image is shown in Fig. 18(c).A trench with length of 2.8 m has been created as a fixture. Itcan be shown that the new AFM image is almost the same as theimage captured from the augmented reality interface.

Finally, the nanowire can be pushed into the trench. In orderto push the nanowire, a zigzag strategy has to be used accordingto the theoretical analysis in Section IV-A. Under the assistanceof the augmented reality system, the nanowire can be easily ma-nipulated on the surface. As shown in Fig. 18(d), the shadowsleft by the nanowire indicate that the nanowire moves in a zigzagfashion. The real result from a new AFM image, as shown inFig. 18(f), has the same result as in the augmented reality inter-face, as shown in Fig. 18(e).

From this experimental study, it can be seen that assembly ofnanostructures using the AFM-based nanomanipulation systembecomes very easy with the assistance of the augmented realityinterface. The experimental results show the efficiency and ef-fectiveness of the newly developed system.

Fig. 17. Minimum normal force �F to keep the tip on the surface in slidingmode as a function of � and l. (Here, the unit of �F is F . Assume F =

50F , and � = 0:2.)

Fig. 18. (a) Silver nanowire with diameter of 100 nm and length of 1.3 �m ona smooth polycarbonate surface in scanning range of 5 �m� 5 �m. (b) Imagedisplayed in the augmented reality interface. A trench as a fixture has beencreated on the surface by inscribing the tip into the surface. (c) AFM imageby a new scan after the trench has been created. (d) and (e) Image displayedin the augmented reality interface. The pushing operation on nanowire drivesthe wire into the fixture. (f) AFM image by a new scan after the final pushingoperation.

V. CONCLUSION

It is well known that the main difficulty of nanomanipulationusing AFM is the lack of real-time visual feedback. Fortunately,the newly developed augmented reality system has solved thisproblem through a “videolized” AFM. The “videolized” AFMis achieved by locally updating the AFM image in real timebased on a system model and measured force informationduring nanomanipulation. It has been shown that nanoparticlesand nanowires can be easily manipulated using the AFM-basednanomanipulation system assisted by an augmented realityinterface.


Unlike manipulation in the macroworld, a lot of informa-tion during nanomanipulation is unavailable. The system has tocount on many experiments to develop a database, which canguarantee the effectiveness of the system. Fortunately, once thedatabase has been built up, it becomes very convenient to use it.In the current stage, several models have been developed such asthe nanolithography model on polycarbonate surface, the inter-action model of latex particles on a glass surface and polycar-bonate surface, and the interaction model of silver nanocubesand nanowires on a polycarbonate surface. Other nanoobjects,which are similar to these entities, can take the same models,but use different values of the parameters that only need a cali-bration.

The experimental studies show that the newly developed aug-mented reality system is capable of manipulating both nanopar-ticles and nanowires. As a result, it has made the nanoassembly,such as constructing nanostructures, making nanodevices, andbuilding nanosensors feasible.

REFERENCES

[1] H. McNally, M. Pingle, S. W. Lee, D. Guo, D. E. Bergstorm, and R.Bashir, “Self-assembly of micro- and nano-scale particles using bio-in-spired events,” Appl. Surf. Sci., vol. 214, pp. 109–119, 2003.

[2] G. Binning, C. F. Quate, and C. Gerber, “Atomic force microscope,”Phys. Rev. Lett., vol. 56, no. 9, pp. 930–933, 1986.

[3] Y. Kim and C. M. Lieber, “Machining oxide thin films with an atomicforce microscope: Pattern and objective formation on the nanometerscale,” Science, vol. 257, no. 5068, pp. 375–377, 1992.

[4] R. Luthi, E. Meyer, H. Haefke, L. Howald, W. Gutmannsbauer, and H.-J.Guntherodt, “Sled-type motion on the nanometer scale: Determinationof dissipation and cohesive energies of c ,” Science, vol. 266, no. 5193,pp. 1979–1981, 1994.

[5] D. M. Schaefer, R. Reifenberger, A. Patil, and R. P. Andres, “Fabricationof two-dimensional arrays of nanometer-size clusters with the atomicforce microscope,” Appl. Phys. Lett., vol. 66, pp. 1012–1014, Feb. 1995.

[6] T. Junno, K. Deppert, L. Montelius, and L. Samuelson, “Controlledmanipulation of nanoparticles with an atomic force microscope,” Appl.Phys. Lett., vol. 66, no. 26, pp. 3627–3629, Jun. 1995.

[7] R. Resch, C. Baur, A. Bugacov, B. E. Koel, A. Madhukar, A. A. G. Re-quicha, and P. Will, “Building and manipulating three-dimensional andlinked two-dimensional structures of nanoparticles using scanning forcemicroscopy,” Langmuir, vol. 14, no. 23, pp. 6613–6616, Nov. 1998.

[8] L. T. Hansen, A. Kuhle, A. H. Sorensen, J. Bohr, and P. E. Lindelof,“A technique for positioning nanoparticles using an atomic force micro-scope,” Nanotechnology, vol. 9, pp. 337–342, 1998.

[9] M. Sitti and H. Hashimoto, “Tele-nanorobotics using atomic force mi-croscope,” in Proc. IEEE Int. Intelligent Robots Systems Conf., Victoria,BC, Canada, Oct. 1998, pp. 1739–1746.

[10] M. Guthold, M. R. Falvo, W. G. Matthews, S. Washburn, S. Paulson,and D. A. Erie, “Controlled manipulation of molecular samples with thenanomanipulator,” IEEE/ASME Trans. Mechatronics, vol. 5, no. 2, pp.189–198, Jun. 2000.

[11] M. Sitti and H. Hashimoto, “Controlled pushing of nanoparticles: Mod-eling and experiments,” IEEE/ASME Trans. Mechatronics, vol. 5, no. 2,pp. 199–211, Jun. 2000.

[12] G. Y. Li, N. Xi, and M. Yu, “Development of augmented reality systemfor AFM based nanomanipulation,” IEEE/ASME Trans. Mechatronics,vol. 9, no. 2, pp. 199–211, Jun. 2004.

[13] J. P. Cleveland, S. Manne, D. Bocek, and P. K. Hansma, “A nondestruc-tive method for determining the spring constant of cantilevers for scan-ning force microscopy,” Rev. Sci. Instrum., vol. 64, no. 2, pp. 3967–3969,1993.

[14] J. E. Sader, J. W. M. Chon, and P. Mulvaney, “Calibration of rectangularatomic force microscope cantilever,” Rev. Sci. Instrum., vol. 70, no. 10,pp. 403–405, 1999.

[15] R. G. Cain, M. G. Reitsma, S. Biggs, and N. W. Page, “Quantitative com-parison of three calibration techniques for the lateral force microscope,”Rev. Sci. Instrum., vol. 72, pp. 3304–3312, 2001.

[16] E. Liu, B. Blanpain, and J. P. Celis, “Calibration procedures for fric-tional measurements with a lateral force microscope,” Wear, vol. 192,pp. 141–150, 1996.

[17] R. G. Cain, S. Biggs, and N. W. Page, “Force calibration in lateral forcemicroscope,” J. Colloid Interface Sci., vol. 227, pp. 55–65, 2000.

[18] D. F. Ogletree, R. W. Carpick, and M. Salmeron, “Calibration of fric-tional forces in atomic force microscopy,” Rev. Sci. Instrum., vol. 67,pp. 3298–3306, 1996.

[19] M. R. Falvo, R. M. Taylor, A. Helser, V. Chi, F. P. Brooks, S. Washburn,and R. Superfine, “Nanometer-scale rolling and sliding of carbon nan-otubes,” Nature, vol. 397, pp. 236–238, Jan. 1999.

Guangyong Li (S’01) received the B.S. degree inmechanical engineering from the Nanjing Universityof Aeronautics and Astronautics, Nanjing, China,in 1992, the M.S. degree in control and automationfrom the Beijing Institute of Control Engineering,China Academy of Space Technology, Beijing,China, in 1999, and is currently working toward thePh.D. degree in electrical and computer engineeringat Michigan State University, East Lansing.

From 2000 to 2001, he was a Research Scholarwith the Department of Electrical and Computer En-

gineering, National University of Singapore, where he was involved with mobilerobot control and neural-network control. His research interests include mobilerobot control, neural-network control, microrobotics/nanorobotics and systems,nanoelectronics, nanofabrication, and nanobioengineering.

Ning Xi (S’89–M’95) received the B.S. degree inelectrical engineering from the Beijing Universityof Aeronautics and Astronautics, Beijing, China, in1982, the M.S. degree from Northeastern University,Boston, MA, in 1989, and the D.Sc. degree insystems science and mathematics from WashingtonUniversity, St. Louis, MO, in 1993.

He is currently a John D. Ryder Professor of Elec-trical and Computer Engineering with Michigan StateUniversity, East Lansing. His research interests in-clude robotics, manufacturing automation, microsys-

tems/nanosystems, and intelligent control and systems.Dr. Xi was the recipient of the 1995 Best Paper Award presented at the

IEEE/The Robotics Society of Japan (RSJ) International Conference onIntelligent Robots and Systems, the 1998 Best Paper Award presented at theJapan–USA Symposium on Flexible Automation, the first Early AcademicCareer Award presented by the IEEE Robotics and Automation Society in May1999, and the National Science Foundation (NSF) CAREER Award.

Heping Chen (S’00) received the B.S. degree incontrol system engineering from the Harbin Instituteof Technology, Harbin, China, in 1989, the M.Eng.degree in electrical and electronic engineering fromNanyang Technological University, Singapore, in1999, and the Ph.D. degree in electrical and com-puter engineering from Michigan State University,East Lansing, in 2003.

He is currently a Post-Doctoral Research As-sociate with the Department of Electrical andComputer Engineering, Michigan State University.

His research interests include micromanufacturing/nanomanufacturing, micro-robotics/nanorobotics, manufacturing automation and control, system design,and implementation.

Dr. Chen was the recipient of the 2002 Highly Commended Award presentedby the Journal of Industrial Robots.


Craig Pomeroy (S’04) is currently working towardthe B.S. degree in computer and electrical engi-neering at Michigan State University, East Lansing.

He is currently a Research Assistant with theRobotics and Automation Laboratory, Departmentof Electrical and Computer Engineering, MichiganState University. His research interests includeembedded systems, robotics, real-time systems, andnanoscale manipulation.

Mathew Prokos is currently working toward the B.S.degree in computer science at Michigan State Univer-sity, East Lansing.

He is currently a Research Assistant with theRobotics and Automation Laboratory, Departmentof Electrical and Computer Engineering, MichiganState University. His research interests includenanodevices, distributed decision making in sensornetworks, and data visualization in human machineinterfaces.

ieee transactions on nanotechnology, vol. 4, no. 5 ... · the speciﬁc cantilever used in this...

Documents