IOP PUBLISHING NANOTECHNOLOGY

Nanotechnology 18 (2007) 185503 (5pp) doi:10.1088/0957-4484/18/18/185503

Cutting and sharpening carbon nanotubesusing a carbon nanotube ‘nanoknife’XianLong Wei, Qing Chen1, Yang Liu and LianMao Peng

Key Laboratory for the Physics and Chemistry of Nanodevices and Department of Electronics,Peking University, Beijing 100871, People’s Republic of China

E-mail: qingchen@pku.edu.cn

Received 23 January 2007, in final form 1 March 2007Published 5 April 2007Online at stacks.iop.org/Nano/18/185503

AbstractA new method has been developed to precisely cut and to sharpen carbonnanotubes using a ‘nanoknife’, which is a short carbon nanotube adhered to ametal tip. The mechanism for the cutting and the sharpening was proposed tobe local vaporization of carbon caused by Joule heating. The ‘nanoknife’ wasalso found useful to cut other nanotubes and nanowires. The cutting processwas also found useful to construct complex carbon nanotube structures.

1. Introduction

Carbon nanotubes (CNTs) have been found in the last 15years to have many remarkable properties and can be usedin many fields. Among all the promising applications, themost probable one that can be realized in the near future isto use CNT as a tip. Due to small tip radii of curvatureand high aspect ratio, a CNT can be used as a field emissiontip [1, 2]. A CNT attached to a conventional atomic forcemicroscope (AFM) or scanning tunnelling microscope (STM)tip can be used for imaging and measuring [3, 4], as well asfor nanofabrication [5] and nanolithography [6]. Chemicalfunctionalized CNT tips can be used as probes for chemicalforce microscopes or for biology applications [7]. Nanotubeshaving a small magnetic particle on their tips can be used formagnetic force microscopes [8]. However, for all the aboveapplications, it is important to control the length and the endstructure of the CNT tips.

Several methods have been developed to cut CNT tips.Chemical etching [9] and mechanical milling [10] can cut alarge number of CNTs at the same time, but it is difficultto achieve precise cutting. A scanning probe microscope(SPM) is an effective tool to precisely cut CNTs dispersedon a substrate [11–13]. However, the structure of the CNTscut by the above methods is easily destroyed during thefurther process of mounting CNTs to the supporting tips. Itis desired to modify the length and the end structure of theCNT tips after they have been mounted to the final support.Applying a dc bias between a CNT tip and a conducting filmsurface [14] or a conducting tip [15, 16] can cut CNT with a

1 Author to whom any correspondence should be addressed.

relatively large diameter (>5 nm). About 50 nm of the tip wasremoved at a time [14]. The length of a single-walled CNT(SWCNT) rope [3] or an individual SWCNT [14] extendingout from a supporting tip can be adjusted by applying a voltagepulse between the tip and a conductive surface 10–50 nmaway. Reproducible etching of 2–20 nm per pulse may beachieved by adjusting the pulse amplitude [14]. Using micro-translators or nanomanipulators one can control the mountingand adjust the length of CNT extruded out during the mountingprocess [17, 18]. Using a piezo-driven nanomanipulator in ascanning electron microscope (SEM) a CNT can be connectedto a metal tip. When a voltage is applied between the tip andthe nanotube, the CNT may be broken by Joule heating. As thebreaking is normally at the defect position in the nanotube, thismethod can control the length of the CNT tip by adjusting themounting position of the CNT [17]. However, the accuracyof this method depends on how accurate one can judge theposition of the defect and the method is not very suitable forperfect CNTs. A more precise method has been developedrecently using the electron beam in a SEM [19, 20]. At theselected length a nanotube is cut by scanning a single lineperpendicular to the nanotube with the electron beam. After2 min, the electron beam starts to break the carbon bonds andcuts the CNT tip [19]. Water or O2 has been found to beable to increase the cutting speed dramatically [20]. Such amethod is not very suitable for thick CNTs as long exposuretime may be needed and sample drift and beam drift may affectthe accuracy. For example, approximately 25 nm long materialwas removed from a CNT during the cut due to beam driftand sample drift when the CNT (with a diameter being about16 nm) was cut by a beam with 3 nm spot size [20].

0957-4484/07/185503+05$30.00 1 © 2007 IOP Publishing Ltd Printed in the UK

Nanotechnology 18 (2007) 185503 X L Wei et al

Figure 1. SEM images showing a CNT (the lower one) is cut by aCNT ‘nanoknife’ repeatedly and becomes shorter and shorter.

Besides the length, the end structure of CNT tip is alsoimportant for applications in field emission tips or AFM andSTM tips. Electrically driven vaporization has been usedto peel and sharpen multi-walled CNTs [21]. Control ofcontact angle and applied voltage between CNT AFM tips andconductive substrate when CNT tips intermittently contact thesubstrate can also sharpen CNT tips [22]. When CNTs are cutby an electron beam, applying a low voltage across the CNTscan sharpen the end of the CNTs to the size of a single-walledtube [19]. But so far, all these methods cannot control the finalshape of the CNT tips. Recently, in situ transmission electronmicroscope (TEM) manipulation and observation show that thetip structure of CNTs can be shaped controllably using the fieldevaporation phenomenon [23]. However, this method is noteasy for many users.

In the present work, we describe a new method to cut andto sharpen CNTs using a ‘nanoknife’, which is a short CNTadhered to a metal tip. Our method is easy to use, effective andhas better control on the length and the shape of the CNTs.

2. Experimental details

A MM3A-nanoprobe system (Kleindiek) installed in a FEI XL30F SEM or a TEM sample holder (Nanofactory) integratedwith a STM for a FEI Tecnai G20 TEM was used to mount andmodify CNT tips. Multi-walled CNTs were synthesized by thechemical vapour deposition (CVD) method using ferrocene ascatalyst and cyclohexane as carbon source. The CNTs weretransferred into SEM or TEM through a Pt tip, which hadbeen dipped in CNT powder to pick up some CNTs. W tipswere used for nanomanipulation and for supporting CNT tips.W tips were made by chemical etching in 5 mol l−1 NaOHsolution. To ensure good electrical contact between W tips andCNTs, W tips were first cleaned using an electrically inducedmelting process. While applying a 4 V bias between a W tip

Figure 2. TEM images taken before (a) and after (b) a CNT is cut bya ‘nanoknife’ (the lower one). (c) A high resolution TEM imageshowing the structure of the CNT and the ‘nanoknife’ after thecutting.

and another metal tip (W or Pt) inside an electron microscope,we manipulated the two tips to contact each other head tohead. At the moment the two tips touched, the W tip wasmelted, a thin contamination layer was peeled off and the tipwas cleaned.

The CNT ‘nanoknife’ was made as follows. Under theobservation of SEM, a clean W tip was manipulated intocontact with a single CNT attached to a Pt tip. A sweep voltage(from 0 to 5 V) was then applied between the Pt tip and the Wtip. The CNT would break usually at a point midway betweenthe two tips [24] leaving a segment of the CNT adhering to theW tip. The CNT adhered to a metal tip was able to be used as a‘nanoknife’ to cut or to sharpen other nanotubes or nanowires.

3. Results and discussion

Figure 1 shows SEM images taken when a CNT (the lower one)was cut by a CNT ‘nanoknife’ repeatedly becoming shorterand shorter. To cut a CNT, a ‘nanoknife’ was moved usinga nanomanipulator to contact with the nanotube to be cut ata selected position. A dc voltage of about 5 V was appliedbetween the ‘nanoknife’ and the nanotube during the process.When the contact was made with the voltage added, the CNTwas cut at the contact point and the ‘nanoknife’ remainedalmost unchanged except that the cutoff segment attached tothe ‘nanoknife’ with the same pose when it was cut off. Sucha segment of the CNT can be cleaned easily by just movinga tip or a CNT to nearly touch the segment. Therefore, a‘nanoknife’ can be used many times, as shown in figure 1. Tosee the structure change more clearly, the same experiment wasperformed in TEM. Figure 2 shows TEM images taken before(a) and after ((b) and (c)) a CNT was cut by a ‘nanoknife’ (thelower one). Careful observation shows the total length of thetwo segments after cutting is roughly the same as the original

2

Nanotechnology 18 (2007) 185503 X L Wei et al

Figure 3. (a)–(c) Schematic diagrams showing the cutting process.(d) Measured Lknife/LNT for every cut we performed using CNTswith roughly the same diameter. (e) The statistics of desired lengthand actual obtained length of CNTs after cutting.

length of the CNT before cutting, meaning that the cutting islocalized without much material loss. The tip end of the CNTafter cutting is thinner than its original size and is not flat.A high resolution image (figure 2(c)) shows the ‘nanoknife’is almost undamaged. Compared with TEM, SEM providesa larger space and can accommodate more nanomanipulators.Therefore, most of our cutting experiments were performedusing more than one nanomanipulator in SEM.

Several factors may affect the cutting process. Fig-ures 3(a)–(c) show the process schematically. The directionof the applied voltage was found to have no effect. But wefound from many experiments that there is a threshold voltagebelow which no cutting can be made. Our experiments wereperformed by the following process: adding an applied volt-age, performing the cutting experiment; if the CNT was notcut, raising the applied voltage by 0.1 V and performing thecutting again. We found that the threshold voltage is 3–3.5 Vfor most of our experiments. To ensure the cuts, we used 5 Vdc voltage in our experiments. The effect of the length of theCNTs was studied using CNTs with roughly the same diam-eter. For convenience, we named two distances as indicatedin figure 3(b). The distance on CNT 1 (which would act as a‘nanoknife’) between contact points A and C is called Lknife.The distance between contact points B and C on CNT 2 (whichwould be cut) is called LNT. After every cut, Lknife and LNT

were measured and Lknife/LNT was calculated. Figure 3(d)shows the result. Among 109 cuts we performed using CNTswith roughly the same diameter, 85 cuts (about 78%) happenedwith Lknife/LNT < 1. We also performed 50 cuts on CNTs with

Figure 4. (a) and (b) SEM images showing a CNT tip sharpened by a‘nanoknife’. (c) and (d) SEM images taken before (c) and after (d) aCNT tip was thinned from the side.

different diameters (from 40 to 60 nm) and found similar re-sults that about 72% (36 cuts) happened with Lknife/LNT < 1.The diameter of CNTs was found to have little effect on thecutting. Therefore, it is the length ratio Lknife/LNT that decideswhich CNT would be cut. To perform a controllable cut, a CNTis selected and the position to be cut is decided, so that LNT isknown. One can then adjust the position of the ‘nanoknife’ tolet the contact position satisfy Lknife � LNT and make the cut.By using a nanomanipulator one can precisely control the con-tact position, which is also the cutting position, to nm scale.We performed a total of 159 cuts in SEM with CNTs rang-ing from several hundred nanometres to several micrometreslong. More than 80% of the cuts have an accuracy of at least50 nm, which is roughly the same as the diameter of the CNT‘nanoknife’. Figure 3(e) shows the statistics of desired lengthand actual obtained length of CNTs after cutting. The accu-racy for shorter desired lengths is higher. The reason will bediscussed later. Using a thinner CNT ‘nanoknife’ may furtherimprove the accuracy.

Besides cutting, the ‘nanoknife’ can also sharpen CNTs.Figures 4(a) and (b) are SEM images showing a CNT (thelower one) being sharpened. Similar to the cutting process, the‘nanoknife’ was moved into contact with the nanotube to besharpened. A lower voltage (4 V instead of 5 V) was appliedbetween the ‘nanoknife’ and the CNT. For sharpening a CNT,the tip end of the ‘nanoknife’ was moved in a direction havinga small angle with the axis direction of the CNT and to contactthe side wall of the CNT. When the ‘nanoknife’ touched theCNT, the CNT lost mass locally at the contact position andbecame thinner. Using this process, we can also sharpen thewhole CNT tip as shown in figures 4(c) and (d). Figure 5 showsTEM images taken before (a) and after ((b) and (c)) a tip wassharpened. After being sharpened, the tip end became very thinwith a diameter less than 5 nm. In a previous report, a multi-walled CNT has been sharpened by a larger CNT contacting

3

Nanotechnology 18 (2007) 185503 X L Wei et al

Figure 5. TEM images taken before (a) and after (b) a CNT tip(the upper one) was sharpened by a ‘nanoknife’. (c) High resolutionTEM image showing the structure of the sharpened tip.

it end-to-end under a bias [21]. In the case of the end-to-end contact, multi-walled CNTs can only be sharpened fromthe end. The process shown here can sharpen multi-walledCNTs at any position on the sidewall as shown in figure 4. Theshape of the end is also controllable by controlling the movingdirection of the ‘nanoknife’ during sharpening.

The mechanism of the cutting and sharpening is proposedto be due to local vaporization of CNT caused by Joule heating.In the process of cutting or sharpening, two CNTs biasedwith a certain voltage are manipulated into contact. At themoment they touch each other, a current passes through thetwo CNTs, Joule heat is generated. Assuming the contactsbetween the CNTs and the metal tips are good, it is the CNT–CNT contact point that has higher electronic resistance andgenerates more heat. In the vacuum, most of the heat canonly be dissipated through the CNTs to the supporting metaltips. If the thermal conductivity is the same for both CNTs,the CNT with a longer L (which is the distance between theCNT–CNT contact and the CNT–metal tip contact) dissipatesJoule heat less effectively, the local temperature goes up untilthe carbon–carbon bonds are broken and the CNT is cut orsharpened, while the CNT with a shorter L can dissipatesJoule heat faster and remain almost undamaged. Such amechanism can explain why Lknife/LNT is the key factoraffecting the cutting process. In some cases, if one CNThas more defects than the other, which may cause its thermalconductivity to be lower than the other, it may dissipate heatless effectively and be cut even if it is shorter. That couldbe the reason that there is about 20% of cuts not satisfyingLknife/LNT < 1. The phenomenon that the cutting accuracyfor shorter desired length is higher can also be explained.The CNTs used in the present work were synthesized by theCVD method and have defects. Longer CNTs have moredefects which raise the resistance and generate more heatlocally with the same current, which also lowers the thermal

Figure 6. SEM images taken before (a) and after a ZnO nanowirewas cut by a CNT ‘nanoknife’ adhered to the right tungsten tip. Thevoltage applied between the two tips was 10 V.

conductivity of the CNTs and makes the heat dissipation lesseffective. Also, defect positions are easier to break thanperfect positions due to weak bonds. All these effects makesome long CNTs break at defect positions instead of thecontact position.

The cutting and sharpening process was also monitored bymeasuring the current going through the contact. We found thatcutting or sharpening CNTs happened in almost no time (lessthan 0.01 s). Most of the time spent in our experiments wasused to make the ‘nanoknife’ contact the CNT to be cut at theselected position through nanomanipulation. Our method canbe speeded up by improving the manipulation technique.

The present ‘nanoknife’ is also useful to cut othernanotubes or nanowires. Semiconducting nanowires, suchas ZnO and Bi2S3 nanowires, were cut using the CNT‘nanoknife’. The threshold voltage for cutting semiconductingnanowires was found to be much higher than that for cuttingCNTs. For example, we found that the threshold voltage forcutting ZnO nanowires is about 10 V. It was also found that thecutting position is not very localized and lots of material is lostduring cutting (shown in figure 6). The reason might be thatthe electronic conductivity of the semiconducting nanowiresis not as good as CNTs, so that a higher voltage is neededto get a high enough current. Also, due to the relativelyhigh resistivity, the whole nanowire is generating heat whena current is running through it, so the Joule heat is not focusedat the contact point.

The segment attached to the ‘nanoknife’ after the cuttingis also useful: together with the ‘nanoknife’, a T-junction isformed. Interconnecting carbon nanotubes may be importantelements of future electrical nanodevices. Y-CNTs have beenreported to have many interesting transport properties [25, 26].So far, most Y-CNTs are directly synthesized which isnot easy and not controllable. Until now, controllableinterconnections of CNTs have only been constructed withthe help of nanomanipulators in TEM and electron-beam-induced amorphous carbon deposition [27]. Using the cuttingprocess without cleaning the ‘nanoknife’, we can constructsome complex CNT nanostructures. The length of branchesand the connecting angles in the nanostructures can all becontrolled by adjusting the relative position of the CNTs inthe cutting process. Figure 7 shows SEM images taken whena comb-like CNT structure was constructed by the cuttingprocess. The junctions can be stabilized by amorphous carbondeposited using electron-beam-induced deposition. Afterstabilization, the constructed structure can be moved out forsome applications.

4

Nanotechnology 18 (2007) 185503 X L Wei et al

Figure 7. SEM images showing a comb-like CNT structure beingconstructed by the cutting process.

4. Conclusions

We demonstrated that a short CNT adhered to a metal tip canbe used as a ‘nanoknife’ to cut or to sharpen other nanotubesor nanowires. The effects of the applied voltage, the lengthand the diameter of the ‘nanoknife’ have been studied. Themechanism for the cutting and sharpening has been proposedto be local vaporization of carbon caused by Joule heating. Thecutting or sharpening process has been found to happen in lessthan 0.01 s and the method can be speeded up by improvingnanomanipulation techniques. The cutting process was alsoused to construct complex CNT structures.

Acknowledgments

We thank Professor J Zhang for the CNT samples. This workwas supported by NSF of China (60440420450, 90206048),the MOST (2006CBON0400), the Chinese Ministry ofEducation (20050001055 and NCET) and NCNNC.

References

[1] Doytcheva M, Kaiser M, Verheiijen M A, Reyes-Reyes M,Terrones M and de Jonge N 2004 Chem. Phys. Lett. 396 126

[2] de Jonge N, Allioux M, Oostveen J T, Teo K B K andMilne W I 2005 Phys. Rev. Lett. 94 186807

[3] Dai H J, Hafner J H, Rinzler A G, Colbert D T and Smalley R E1996 Nature 384 147

[4] Nguyen C V, Chao K J, Stevens R M D, Delzeit L, Cassell A,Han J and Meyyappan M 2001 Nanotechnology 12 363

[5] Aunstin A J, Nguyen C V and Ngo Q 2006 J. Appl. Phys.99 114304

[6] Dai H J, Franklin N and Han J 1998 Appl. Phys. Lett. 73 1508[7] Wong S S, Joselevich E, Woolley A T, Cheung C L and

Lieber C M 1998 Nature 394 52[8] Arie T, Nishijima H, Akita S and Nakayama Y 2000 J. Vac. Sci.

Technol. B 18 104[9] Ziegler K J, Gu Z N, Shaver J, Chen Z Y, Flor E L,

Schmidt D J, Chan C, Hauge R H and Smalley R E 2005Nanotechnology 16 S539

[10] Chen L, Pang X J, Zhang Q T and Yu Z L 2006 Mater. Lett.60 241

[11] Vanama L C, Wildoer J W G, Temminck Tuinstra H L J andDekker C 1997 Appl. Phys. Lett. 71 2629

[12] Park J Y, Yaish Y, Brink M, Rosenblatt S and McEuen P L2002 Appl. Phys. Lett. 80 4446

[13] Kim D H, Koo J Y and Kim J J 2003 Phys. Rev. B 68 113406[14] Li C and Cheung C L 2001 PhD Thesis Harvard University, pp

83–84 and 101[15] Kuzumaki T, Sawada H, Ichinose H, Horiike Y and

Kizuka T 2001 Appl. Phys. Lett. 79 4580[16] Akita S and Nakayama Y 2002 Japan. J. Appl. Phys. 41 4887[17] De Jonge N, Lamy Y and Kaiser M 2003 Nano Lett. 3 1621[18] Stevens R, Nguyen C, Cassell A, Delzeit L, Meyyappan M and

Han J 2000 Appl. Phys. Lett. 77 3453[19] Martinez J, Yuzvinsky T D, Fennimore A M, Zettl A,

Garcı́a R and Bustamante C 2005 Nanotechnology 16 2493[20] Yuzvinsky T D, Fennimore A M, Mickelson W,

Esquivias C and Zettl A 2005 Appl. Phys. Lett. 86 053109[21] Cumings J, Collins P G and Zettl A 2000 Nature 406 586[22] Nguyen C V, So C, Stevens R M, Li Y, Delziet L,

Sarrazin P and Meyyappan M 2004 J. Phys. Chem. B108 2816

[23] Wang M S, Wang J Y and Peng L M 2006 Appl. Phys. Lett.88 243108

[24] Collins P G, Hersam M, Arnold M, Martel R andAvouris Ph 2001 Phys. Rev. Lett. 86 3128

[25] Bandaru P R, Daraio C, Jin S and Rao A M 2005 Nat. Mater.4 663

[26] Park J, Daraio C, Jin S, Bandarua P R, Gaillard J and Rao A M2006 Appl. Phys. Lett. 88 243113

[27] Wang M S, Wang J Y, Chen Q and Peng L M 2005 Adv. Funct.Mater. 15 1825

5