www.elsevier.com/locate/carbon

Carbon 44 (2006) 1130–1136

TEM image simulation study of small carbon nanotubesand carbon nanowire

Takuya Hayashi a,*, Hiroyuki Muramatsu a, Yoong Ahm Kim a, Hiroshi Kajitani a,Shinji Imai a, Hideyuki Kawakami a, Masamitsu Kobayashi a, Toshiharu Matoba a,

Morinobu Endo a, Mildred S. Dresselhaus b

a Department of Electrical and Electronic Engineering, Shinshu University, 4-17-1 Wakasato, Nagano 380-8553, Japanb Department of Physics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

Received 16 May 2005; accepted 12 November 2005Available online 5 January 2006

Abstract

Recent findings of extremely small diameter carbon nanotube and nanowire in the core of a multi-walled carbon nanotube (MWCNT)have attracted interests from broad range of researchers. Direct observation of carbon nanotube is usually done using a transmissionelectron microscope (TEM). When nanotubes become smaller, it becomes harder to correctly understand the TEM images, not onlybecause of the weak scattering, but also due to the artifact that starts to appear because of the interference effect and the inappropriatedefocus condition.

In this study, we have shown that the artifact such as ghost fringes due to inappropriate defocus conditions of the TEM appear in thecore of an MWCNT, and can be misinterpreted as either carbon nanowire or small carbon nanotube. It is also shown that, in the TEMimage, it is hard to distinguish a single-walled nanotube bundle from a double-walled carbon nanotube bundle. Finally, we propose thatthe cross-sectional observation is necessary for the correct characterization of single- and double-walled carbon nanotube bundles.� 2005 Elsevier Ltd. All rights reserved.

Keywords: Carbon nanotubes; Transmission electron microscopy; Microstructure

1. Introduction

The quest for a mass production process for carbonnanotube synthesis has recently advanced considerably.The catalytic chemical vapor deposition (CCVD) method[1] is now considered to be the most promising methodfor producing carbon nanotubes in quantity [3–6]. Thestudy of the CCVD process is now moving to develop away to control the quality, properties, and growth direc-tions of the nanotubes [7,8].

Among such trends, one of the most interesting studies isthe synthesis of small diameter carbon nanotubes, especiallysingle-walled nanotubes. Thanks to the improvement of the

0008-6223/$ - see front matter � 2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.carbon.2005.11.017

* Corresponding author. Fax: +81 26 269 5208.E-mail address: hayashi@endomoribu.shinshu-u.ac.jp (T. Hayashi).

CCVD method, we are now able to produce sub-nanometertubes such as those of only 0.43 nm diameter, as reportedearlier [9], and also larger diameter tubes, but still of sub-nanometer size (0.6–0.9 nm diameter) on a regular basis.More recently, double-walled carbon nanotube (DWCNT)production is becoming an exciting field owing to its high-stability and interesting physical properties [10,11].

Fabrication of a one-dimensional nanowire in a single-walled carbon nanotube (SWCNT) was reported by Meyerand coworkers, and the possibility of using the inner spacehas attracted much interest [12]. Recently, carbon nano-wires were reported to be synthesized in the core part ofa multi-walled carbon nanotubes, which is also a very inter-esting novel sub-nanometer material [13]. Another reporton the discovery of an even smaller carbon nanotube witha diameter of 0.3 nm [14] has attracted interest, as well.

Fig. 1. (a) Model of MWCNT with (left) and without (right) a smallnanotube in the core. (b) Model of carbon nanotube with and without acarbon nanowire in the central core. (c) Model of (10,10) SWCNT bundle.The coordinate system used in the image simulation is also indicated. Theincident beam direction was rotated 30� as indicated in the image. (d)Model of DWCNT bundle with (15,15) outer tube and (10,10) inner tube.The incident beam direction was rotated 30� as indicated in the image.

Fig. 2. Simulated images and corresponding line profiles of MWCNT: (a)without the 0.4 nm tube and (b) with 0.4 nm tube in the core.

T. Hayashi et al. / Carbon 44 (2006) 1130–1136 1131

Analytical techniques for characterizing such sub-nano-meter materials are limited to several techniques like trans-mission electron microscopy (TEM), Raman spectroscopy,and scanning probe microscopy. Among these methods,most of the studies related to nanotubes, nanowires andnanofillings use TEM as an indispensable tool for analyz-ing the structure, diameter, and crystallinity of the sub-nanometer carbon materials.

However, we always have to be careful when using TEMon such small size materials, because of the interferenceeffect caused by the highly coherent electron source andthe improper defocus conditions for small tubular struc-tures, resulting in a ghost image (artifact) that appears inthe TEM photographs. This problem applies to larger scalematerials as well, but the danger of misinterpreting theimage for larger scale materials is small compared to theproblem for sub-nanometer materials, that we are nowtreating. Another rather obvious issue requiring specialcare is the overlapping of the samples with an amorphouscarbon film on the specimen grid [15], or overlapping withother materials such as outer tube walls. Other issue lies ina bundle state of the tubes, which makes it difficult to iden-tify single- and double-walled carbon nanotubes. Whenthere is a TEM image of a bundle, it is often very hardto tell if they really are single-walled or not. This alsoapplies to the double-walled nanotube case. There is a lit-erature that briefly mentions the artifacts that appears inthe core of an MWCNT and some image simulations onSWCNT [16], but has not extensively treated the nanowiresand bundle simulations and does not mention about theeffect of the overlapping tube layers.

In this study, we used the TEM image simulation toinvestigate the artifacts that appears on the TEM imageof sub-nanometer size carbon materials, such as multi-walled carbon nanotube, single-walled carbon nanotubebundle, and double-walled carbon nanotube bundlemodels.

2. Experimental

TEM image simulation was performed with the multi-slice TEM simulation program integrated into Cerius [2].The multi-slice method is the most commonly used TEMimage simulation method. The conditions (accelerationvoltage was 200 keV, Cs = 1 mm) were set identical toour own TEM (JEOL JEM-2010FEF) and is one of thegeneral condition for the 200 keV TEM instruments usedin the world. The Scherzer defocus condition, which isthe condition for obtaining optimal resolution, is about�50 nm at 200 keV for our TEM.

We have performed the image simulations of the follow-ing models:

1. Multi-walled tube with/without small nanotube in thecore (Fig. 1a). The MWCNT used in the simulationwas a 5-layered nanotube with a core diameter of 1.1and 0.34 nm interlayer spacing. The small nanotube

was a 0.4 nm diameter tube that is considered to bethe smallest diameter tube ever observed [17]. The defo-cus value used in the image simulation ranged from �45to �65 nm. Similar simulation was performed for the

1132 T. Hayashi et al. / Carbon 44 (2006) 1130–1136

0.3 nm diameter tube inside the MWCNT at the acceler-ation voltage condition of 120 keV. The defocus rangehas been shifted accordingly.

2. Multi-walled tube with/without a carbon nanowire inthe core (Fig. 1b). The MWCNT used in this simulationwas a 4-layered nanotube with a core diameter of 0.7and 0.34 nm interlayer spacing. The carbon nanowirewas modeled according to the proposed model in Ref.[6]. The defocus conditions at the image simulationswere set to +10, 0, �10, and �50 nm.

Fig. 3. Simulated images of an MWCNT with and without 0.3 nm tube in the c(b) and (d) shows the simulated image with 0.3 nm tube. Image (e) was obtainthe ghost tube appearing at the core of an SWCNT (indicated by the arrows). Ifringe appearing at the core (indicated by the arrows).

3. Bundle of small-diameter SWCNTs (Fig. 1c) andDWCNTs (Fig. 1d). For this case SWCNT andDWCNT crystals consisting of 6 tubes were built fromSWCNTs with a diameter of 1.1 nm, and the DWCNTshave an outer diameter of 1.8 nm and an inner diameterof 1.1 nm. The models were rotated along the z-axisfrom 0� to 30� at 5� step, taking x-axis as the incidentbeam direction. Defocus conditions were set to vari-ous values, ranging from 0 to �105 nm with �15 nmstep.

ore. Image (a) and (c) shows the simulated image without 0.3 nm tube, anded from the simulation with 1200 keV acceleration voltage. (f) Example ofnset shows the simulated image of an SWCNT (df = �40 nm) with a ghost

Fig. 4. (a) Image simulation of a carbon nanotube with a nanowire. Apart without a nanowire is shown in the �50 nm image to helpunderstanding the difference between various images. (b) Image simulationof a carbon nanotube without a nanowire in the core. (c) Simulated imageof an MWCNT with a carbon nanowire contained over part of the length,as indicated in the image. We can see that the nanowire is resolved at1200 keV acceleration voltage.

T. Hayashi et al. / Carbon 44 (2006) 1130–1136 1133

4. Similar to 3, but an SWCNT bundle with a DWCNT inthe center of the bundle, and a DWCNT bundle with anSWCNT in the center of the bundle were simulated forthe defocus condition of �50 nm by rotating along thez-axis from 0� to 30� at 5� step, taking x-axis as the inci-dent beam direction.

3. Results and discussion

Fig. 2a and b shows the results of the image simulationwith and without a 0.4 nm SWCNT in the core of theMWCNT. Full results are shown in supplementary materi-als 1–5. With the SWCNT core (Fig. 2b), we can find thecontrast corresponding to the core through the entire defo-cus range. Interestingly, in the image simulation withoutthe core SWCNT (Fig. 2a) at the defocus range of �45to �50 nm, we can see the parallel contrast at the hollowpart that looks like the contrast of a tube, although nothingis supposed to exist in that location. This is the ghost imagedue to the interference fringes that arise from certain defo-cus conditions. Line profiles at the bottom of each imageshow the contrast across the tube (along the line indicatedin the images of Fig. 2a and b), respectively. The valley partcorresponds to the dark fringe of the simulated image,meaning that there is a tube wall in that position. The lineprofile also shows that at the defocus range of �45 to�50 nm, there is a ghost image appearing as a contrast atthe core part. We can note that the contrast level of theghost fringe is brighter compared to the contrast of the realSWCNT core. When there are more layers in the tube, thecontrast of the actual tube in the core becomes lower, whilethe ghost fringe remains the same. This fact makes it hardto determine from single TEM image whether the fringeappearing at the center of a tube image is coming from areal SWCNT or not. However, we can avoid this problemby obtaining a series of images changing the defocus condi-tion in a stepwise manner.

When there is a 0.3 nm nanotube in the core of anMWCNT, it becomes harder to interpret the image com-pared to the 0.4 nm tube case as shown in Fig. 3a–d. Theghost fringes can be found in some defocus conditions,making it hard to see if there really is something in the core.The difference from the 0.4 nm case is that the image nearthe Scherzer defocus does not resolve the 0.3 nm tube, andwe only see a single wide dark contrast. When there is no0.3 nm tube in the core, we do not see any contrast at thecore part. Therefore, in this condition, we are able tounderstand that there is something in the core, but it isnearly impossible to define what it is. It is only when weuse higher acceleration voltage TEM that we can see thecontrast of the innermost tube with 0.3 nm diameter(Fig. 3e). Fig. 3f shows an example of the observed imagewith possibly a ghost tube appearing at the central part ofan SWCNT. We see the ghost fringe only in one part of thetube because that part fulfilled the interference condition atthe specific defocus value. When we perform the image sim-

ulation for the isolated 0.3 nm tube, we could see thefringes in the simulated images. Therefore, the loss of con-trast was due to the overlapping by the outer MWCNTshells.

The simulation of the tube with and without the nano-wire in the core of the MWCNT is shown in Fig. 4a and

1134 T. Hayashi et al. / Carbon 44 (2006) 1130–1136

b, respectively. From the image with the wire at the optimaldefocus condition, instead of a clear lattice fringe that iscoming from the nanowire, we can only see that the corebecomes dark compared to the tube without a nanowire.It is only when we compare (i) the image with a nanowire,with (ii) the image without a nanowire that we can find thedifference caused by the presence of a nanowire. At someother defocus condition, such as 0 nm or �10 nm, we canfind the fringe appearing at the core of the tube thatappears to correspond to the nanowire. The problem is thata similar fringe is appearing at the core of a tube without ananowire, which makes it hard to distinguish the actual

Fig. 5. (a) Simulated images of SWCNT and DWCNT bundle under selectDWCNTs with the simulation. Image (i) is similar to the 20� rotated (df = �45x + 30� (df = 0 nm), x + 20� (df = 0, �15 nm) or x + 15� (df = �15 nm).

nanowire from the ghost fringe. When we have changedthe acceleration voltage in the simulation to 1.2 MeV,which corresponds to the voltage of an ultrahigh-voltageTEM, we could see the fringe of the nanowire in the coreof a multi-shell nanotube near the Scherzer defocus condi-tion, and we could not find the ghost fringe in the tubewithout the nanowire (Fig. 4c). According to the simula-tion, although the resolution of a 200 keV TEM is highenough to see the graphite interlayer spacing and the iso-lated nanowire, it was not high enough to resolve the car-bon nanowire in the core of an MWCNT due to theinterference effect.

ed defocus conditions are shown. (b) Matching the observed images ofnm) bundle, and the lower part of (ii) is close to the bundle from either the

T. Hayashi et al. / Carbon 44 (2006) 1130–1136 1135

The image simulation results for an SWCNT bundle andDWCNT bundle are shown in Fig. 5a. For the SWCNTcase, near the optimal defocus condition from the x-direc-tion, the simulated image clearly shows the SWCNT, andthere is nothing confusing about interpreting the imageand in measuring the diameter of the tube. From thex + 30� direction images of an SWCNT bundle, we canfind a fringe that looks like the walls of DWCNTs evennear the Scherzer defocus condition. Most confusing defo-cus condition is from 0 to �15 nm, and we cannot actuallyrecognize what kind of tube is producing this image. In theactual image, where large bundles exist, it is hard to deter-mine from the image itself whether the bundles consist ofSWCNTs or of DWCNTs. This is because the directionof the bundle with respect to the incident beam is hard todefine, and the defocus condition makes it even harder tounderstand what kind of tubes are forming the bundle(see supplementary data obtained from other rotationangles).

From the cross-section image simulation (z-direction) atsome defocus condition, there appears to be a nanowire inthe core, which is, of course, an artifact. At df = �45 nm,the cross-section reproduces the actual tube very well.When shifting to df = 0 nm, we can see the hexagonalfringe as an extra layer appearing between the tubes. Inthe actual observation, it is very hard to find the bundleshaving an ideal direction for measuring the diameter, oranalyzing the structure. Therefore, to correctly characterizethe nanotube bundles from a TEM image, we need a cross-section of the tube with a correct defocus condition. Fromthe bundle lying on the grid, it is usually difficult to tell

Fig. 6. Simulated images of SWCNT bundle with a DWCNT (DWCNT/SWCNT), and DWCNT bundle with an SWCNT (SWCNT/DWCNT)from different incident beam angles.

which pairs of the fringes constitute a single tube, makingit harder to measure the correct diameter of the tube.

Fig. 6 shows the image simulation results of theDWCNT in an SWCNT bundle and SWCNT in aDWCNT bundle. By comparing them with the pureSWCNT and DWCNT bundles, we can see that there isalmost no distinctive difference between the mixed bundleand pure bundle. In the DWCNT in an SWCNT bundlecase, when we rotate 5�, we can find darker contrast thanin the pure SWCNT bundle case. However, in other anglesand bundle, it is hard to find pronounced differences. Thismeans that once in the actual TEM observation, it can bevery hard to tell from the TEM image of laid bundle if thebundle only consists of SWCNT/DWCNT or there aresome DWCNT/SWCNT mixed in it. It is therefore moreaccurate to observe the cross-section of the bundle to checkif it is made from same kind of tubes or not.

4. Conclusions

We have shown from the TEM image simulation andthe actual observation that the interference effect canindeed be a source of misinterpreting the artifact as anexisting material. Using higher acceleration voltage TEMcan be effective theoretically for the observation of extra-small carbon nanotubes and nanowires in the core of anMWCNT, although actual high acceleration voltage obser-vation might severely damage the specimen.

TEM image simulation of the pure and mixed DWCNTand SWCNT bundles from various directions showed thatacquiring the cross-section image of the tube or bundlecould be the only way to the correct characterization ofcarbon nanotube bundles. In the actual observation, it willbe virtually impossible to tell from the images of the laidbundle whether they are pure bundles or not.

To avoid the difficulties in interpreting the complicatedTEM images, we propose to observe the cross-section ofthe tubes and bundles, and when possible, combine withimage simulations and other characterization methods suchas Raman spectroscopy.

Acknowledgement

Present work was supported by The Cluster Projectfrom JSPS and ATI fund for young scientists.

Appendix A. Supplementary data

Supplementary data associated with this article can befound, in the online version, at doi:10.1016/j.carbon.2005.11.017.

References

[1] Oberlin A, Endo M, Koyama T. J Cryst Growth 1976;32:335.[2] Cheng HM, Li F, Su G, Pan HY, He LL, et al. Appl Phys Lett

1998;72:3282.

1136 T. Hayashi et al. / Carbon 44 (2006) 1130–1136

[3] Sen R, Govindaraj A, Rao CNR. Chem Mater 1997;9:2078.[4] Flahaut E, Govindaraj A, Peigney A, Laurent C, Rousset A, Rao

CNR. Chem Phys Lett 1999;300:236.[5] Su M, Zheng B, Liu J. Chem Phys Lett 2000;322:321.[6] Nikolaev P, Bronikowski MJ, Bradley RK, Rhommund F, Colbert

DT, et al. Chem Phys Lett 1999;313:91.[7] Fan SS, Chapline MG, Frankin NR, Tombler TW, Cassell AM, Dai

HJ. Science 1999;283:512.[8] Zhang ZJ, Wei BQ, Ramanath G, Ajayan PM. Appl Phys Lett

2000;77:3764.[9] Hayashi T, Kim YA, Matoba T, Esaka M, Nishimura K, Tsukada T,

et al. Nanoletters 2003;3:887.[10] Endo M, Muramatsu H, Hayashi T, Kim YA, Terrones M,

Dresselhaus MS. Nature 2005;433:476.

[11] Endo M, Hayashi T, Muramatsu H, Kim YA, Terrones H, TerronesM, et al. Nanoletters 2004;4:1451–4.

[12] Meyer RR, Sloan J, Dunin-Borkowski RE, Kirkland AI, NovotnyMC, Bailey SR, et al. Science 2000;289:1324.

[13] Zhao X, Ando Y, Liu Y, Jinno M, Suzuki T. Phys Rev Lett2003;90:187401.

[14] Zhao X, Liu Y, Inoue S, Suzuki T, Jones RO, Ando Y. Phys Rev Lett2004;92:125502.

[15] Qin L-C, Zhao X, Hirahara K, Ando Y, Iijima S. Chem Phys Lett2001;349:389–93.

[16] Wang ZL, Hui C, editors. Electron microscopy of nanotubes. Dordr-echt: Kluwer; 2003.

[17] Qin L-C, Zhao X, Hirahara K, Miyamoto Y, Ando Y, Iijima S.Nature 2000;408:50.