Metal coating on suspended carbon nanotubes and itsimplication to metal±tube interaction

Y. Zhang, Nathan W. Franklin, Robert J. Chen, Hongjie Dai *

Department of Chemistry, Stanford University, Stanford, CA 94305, USA

Received 5 June 2000; in ®nal form 29 September 2000

Abstract

Coating of various metals on suspended single-walled carbon nanotubes (SWNT) is carried out by electron-beam

evaporation. Transmission electron microscopy studies reveal that Ti, Ni and Pd coatings on the suspended tubes are

continuous and quasi-continuous, resulting in nanotube-supported metal nanowire structures. In strong contrast, Au,

Al, and Fe coatings on the suspended SWNTs only form isolated discrete particles on the nanotubes. These results shed

light into the nature of metal±tube interaction, an important topic to many fundamental and practical aspects of na-

notubes. Ó 2000 Elsevier Science B.V. All rights reserved.

1. Introduction

Carbon nanotubes [1,2] are quasi one-dimen-sional systems ideal for studying the physics inmolecular scale wires and exploring chemicallyderived nanostructures for future electronics [3±6].Enabling low resistance ohmic contacts to na-notubes are critical to elucidating their intrinsicelectrical properties and obtaining functionalelectronic devices with useful characteristics [5,6].This requires an in depth understanding of thenature of metal±nanotube interaction. Such un-derstanding is not at hand currently as systematicexperimental and theoretical work to address thisissue has been lacking.

We have found through a series of studies thatlow resistance metal±tube ohmic contacts can beconsistently achieved with Ti, Nb and Ni metals

[5±10]. The resistance of a Ti contacted metallicsingle-walled carbon nanotubes (SWNT) is as lowas 12 kilo-ohms for several microns tube length[10], and is the lowest resistance measured withindividual SWNTs among all reported results.Low resistance semiconducting samples are alsoobtained with Ti and Ni contacts and have led tohigh-transconductance nanotube ®eld-e�ect tran-sistors [9]. On the other hand, non-ohmic highresistance contacts are typically resulted with Auand Al in our sample fabrication approach [11].These electron transport results suggest that theelectrical characteristics of metal±tube systems issensitive to the metal type, and motivate us toinvestigate the interactions between nanotubes andvarious metals.

Metal±nanotube interaction is also important interms of forming nanowires on nanotube tem-plates [12,13] or `substrates' [14]. It could be agood approach to obtaining metallic or super-conducting nanowires by continuously coating thesidewalls of nanotubes with metals. An excellent

24 November 2000

Chemical Physics Letters 331 (2000) 35±41

www.elsevier.nl/locate/cplett

* Corresponding author. Fax: +1-650-725-0259.

E-mail address: hdai@chem.stanford.edu (H. Dai).

0009-2614/00/$ - see front matter Ó 2000 Elsevier Science B.V. All rights reserved.

PII: S 0 0 0 9 - 2 6 1 4 ( 0 0 ) 0 1 1 6 2 - 3

example was given recently by Tinkham and co-workers [14] in obtaining superconducting nano-wires by uniformly sputter-coating a SWNTbundle with Mo±Ge. Forming metal wires on ananotube by deposition methods is essentiallygrowing ®lms on a quasi one-dimensional sub-strate. Similar to the growth of ®lms on a two-dimensional surface [15], the metal±substrate(nanotube) interaction must play a central role inthe process of metal wire formation.

This Letter presents results of various metalstructures formed on suspended single-walled na-notubes by electron-beam deposition. Dependingon the metal type, formations of continuous metalnanowires or decorations of isolated discrete par-ticles are observed by transmission electron mi-croscopy (TEM). These nanostructures lead toinformation about the interactions between na-notubes and various metals, therefore sheddinglight into the metal±tube contact issue. It is alsofound that pre-treating nanotubes with surfactantmolecules improve the uniformity of metal depo-sition on nanotubes. Thus, as grown or pre-treatedcarbon nanotubes could be used as novel one-di-mensional substrates for obtaining various metalnanowires.

2. Experimental

SWNTs were grown directly on gold micro-grids by chemical vapor deposition (CVD) ofmethane [16,17] using a ¯uid-phase catalystprecursor-material [18,19]. Preparation for thecatalyst precursor-material involved dissolving atriblock copolymer, aluminum, iron and molyb-denum chlorides in a mixed ethanol and butanolsolution [18,19]. Gold micro-grids used for TEMwere dipped into the precursor solution and im-mersed for a few seconds. The grid was then re-moved from the solution and then calcined at400°C for �12 hr in air. This led to a catalyst ®lmcoating on the gold grid-lines and extending intothe enclosed holes. CVD growth was carried outwith the catalyst coated gold grid in a tube furnaceat 900°C under a 1000 ml/min methane ¯ow for 15min. SWNTs grown and suspended over the holesin the TEM grid were used for metal coating study.

Coating of the nanotubes grown on the goldmicro-grid was carried out by electron-beam de-position of various thickness of Ti, Ni, Pd, Au, Al,and Fe. Several samples were pre-treated by sur-factant molecules. This was carried out by im-mersing the samples in a 0.1 mM Triton X-100aqueous solution for 30 min followed by rinsingwith pure water. Metal evaporation was moni-tored by a quartz oscillator to determine the rateand thickness of material deposited. The deposi-tion rates were �0.02, 0.2 and 0.2 nm/s for ®lmthickness of 0.5, 5 and 15 nm, respectively. Themetal coated nanotubes were examined by a Phil-ips CM20 TEM operated at 200 kV.

3. Results and discussion

TEM of the as grown samples reveals the syn-thesis of individual SWNTs (diameter 1±4 nm) andsmall bundles of SWNTs. Many nanotubes aresuspended over the holes of the gold-grid. Thesuspended nanotubes are nearly free of defects andamorphous carbon overcoating [17], and ideal forinvestigating the deposition of metal atoms onmolecularly clean nanotube substrates.

The results of various metal coatings on asgrown SWNTs for 15, 5 and 0.5 nm thickness areshown in Figs. 1±4. For 15 nm deposition, Tiforms continuous nanowires on suspendedSWNTs (Fig. 1a). Ni and Pd also form uniformcoating on SWNTs (Fig. 1b and c, respectively),but the wire structures are occasionally discon-nected at certain locations (Fig. 1c) and are moreso for Pd than Ni. The deposited Pd and Ni are inthe form of ®ne particles connecting together onthe nanotubes. On the other hand, Au, Al, and Fedeposited on the nanotubes form disconnectedcrystalline particles, leaving sections of the na-notubes free of metal coating (Fig. 1d±f, respec-tively). The Au particles decorating SWNTs are upto �60 nm wide, larger than the deposited ®lmthickness. This indicates that Au atoms depositedon the tubes have migrated and merged together toform large particles.

The striking di�erence in Ti and Au coatingscan also be seen in scanning electron microscopy(SEM) data shown in Fig. 2. Ti coating clearly

36 Y. Zhang et al. / Chemical Physics Letters 331 (2000) 35±41

leads to the formation of continuous and uniformnanowires on the nanotubes, whereas Au coatingleads to disconnected large particles decoratingnanotubes.

For 5 nm metal depositions, Ti coating onthe nanotubes is also continuous and uniform

(Fig. 3a). For Ni and Pd however, the disconti-nuity in the coating becomes apparent (Fig. 3b andc, respectively). Au, Al, and Fe still form discreteparticles but with reduced sizes compared to theparticles made by the 15 nm deposition. The Auand Al particles appear to be elongated along theaxes of the nanotube (Fig. 3d±f), which di�er fromthe particles formed under 15 nm depositions(Fig. 1d and e). The Fe particles are more or lessspherical in shape (Fig. 3f).

As the metal deposition is further reduced toapproximately 0.5 nm, we ®nd that Ti coating onthe sidewalls of nanotubes remains homogeneousbut with ¯uctuations in the amount of metals de-posited along the nanotube lengths (Fig. 4a).Coating of 0.5 nm of Au (Fig. 4b) and Fe (Fig. 4c)leads to the formation of particles with small sizesand low densities. Other features of the particlesare similar to those formed by depositing 5 nm ofmetals.

It is well established by extensive research inthin ®lm deposition on planar substrates that, thestructure and morphology of the ®lm is largelydictated by the interactions between the deposited

Fig. 1. TEM images of: (a) Ti, (b) Ni, (c) Pd, (d) Au, (e) Al, (f)

Fe coating on carbon nanotubes with a thickness of 15 nm.

Fig. 2. SEM images of: (a) Ti (15 nm) and (b) Au (15 nm)

coated SWNTs.

Y. Zhang et al. / Chemical Physics Letters 331 (2000) 35±41 37

atoms and the substrate [15]. The various observedmetal nanostructures formed on suspendedSWNTs can lead to information about metal±tubeinteractions, since the nanotubes are used as quasione-dimensional substrates. According to a clas-sical nucleation theory [15], the condensation orsticking coe�cient for atoms impinging on a sub-strate from the vapor phase is proportional toexp�Eb=KBT �, where Eb is the binding energy of an

atom with the substrate. This is proportional tothe rate that atoms can stabilize on the substrateand reach thermal equilibrium without being re-evaporated. The adatoms di�use around on thesubstrate and join critical nucleation centers forcluster growth. The di�usion rate is proportionalto exp�ÿEdiff=KBT �, where Ediff is the di�usionactivation energy. Empirically Ediff � Eb=4 [15],suggesting that weak binding allows fast surfacedi�usion. The density of nucleation centersstrongly depends on the condensate±substrate in-teraction. Large condensate±substrate interactionleads to a higher nucleation density [15].

The Ti coating on nanotubes appear very uni-form and continuous along the nanotube side-walls. This points to high nucleation density,strong metal±substrate interaction and Ti±SWNTbinding. On the contrary, the Au coating appearshighly discontinuous with very low nucleationdensity. This suggests weak Au±SWNT interactionand a small binding energy. The weak interaction

Fig. 3. TEM images of: (a) Ti, (b) Ni, (c) Pd, (d) Au, (e) Al, (f)

Fe coating on carbon nanotubes with a thickness of 5 nm.

Fig. 4. TEM images of carbon nanotubes coated by (a) Ti, (b)

Au, (c) Fe with a nominal thickness of 0.5 nm.

38 Y. Zhang et al. / Chemical Physics Letters 331 (2000) 35±41

is correlated with a low activation barrier for dif-fusion, leading to rapid motion of Au atoms on thenanotube sidewalls. The low nucleation densityand high di�usion rate cause metal atoms or evensmall clusters to merge into isolated large particles[20]. A quantitative measure of the binding ener-gies between various metal atoms and SWNTs hasnot been carried out at the present time. Never-theless, based on the observed sticking behavior ofmetals on the nanotubes and the resulting metalnanostructures, we tentatively suggest thatEb�Ti� > Eb �Ni� > Eb�Pd� > Eb�Fe� > Eb�Al� >Eb�Au�.

It is known that in three-dimensional bulkmaterials, di�erent metals exhibit di�erent inter-actions with carbon. In general, the ability fortransition metals to bond with carbon atoms in-creases with the number of un®lled d-orbitals.Metals such as Al, Au and Pd have no d-vacanciesand negligible a�nity for carbon. Metals with fewd-vacancies such as Ni, Fe and Co exhibit ®nitesolubility for carbon in certain temperature ranges.3d and 4d metals with many d-vacancies such as Tiand Nb can form strong chemical bonds withcarbon and thus highly stable carbide compounds.For two-dimensional systems, experimental andtheoretical investigations of the interactions be-tween metals and the basal plane of graphite havebeen carried out previously. Although controversyexists, it is generally believed that the interactionsbetween deposited metals and a graphite basalplane are weak [20±28]. This includes Ti [21], Ni[22,23] and certainly the rest of metals [20,24±28]concerned by the current work. The interactionsare suggested to be through van der Waals forcesand do not involve chemical bond formations be-tween the metal and carbon atoms in the graphitebasal plane. This should be due to the highchemical inertness of the ordered sp2 carbon net-work of a graphene sheet. Several studies have alsobeen carried out to investigate the interactionsbetween metals and C60 (zero-dimension) [29±31].In has been shown that certain metals interactmuch more strongly with C60 than graphite in-cluding Ti [29] and Ni [31]. Deposition of Ti on C60

can lead to the formation of Ti±C carbide bondsrevealed by photoelectron spectroscopy using X-rays from a synchrotron source [29].

Interactions between metals and quasi one-di-mensional SWNTs could be signi®cantly di�erentthan those between metal and carbon in two- andzero-dimension. A nanotube di�ers from agraphene sheet in their drastically di�erent elec-tronic structures and that the cylindrical sidewallof a nanotube is curved and in a non-planar sp2

bonding con®guration. A nanotube di�ers fromC60 in dimensionality, radius and that C60 containspentagons in its structure, whereas the sidewall ofa nanotube contains exclusively hexagons. C60

should therefore be more reactive with metals thana nanotube.

A detailed theoretical account for the interac-tions between various metals and nanotubes iscurrently lacking and called for. Nevertheless, arecent theoretical study by Menon and co-workers[32] sheds some light into this important issue. Theauthors used a tight binding molecular dynamicsand ab-initio method to calculate bonding con-®gurations of Ni with the sidewall atoms on aSWNT, and compared the results with Ni bondingon a graphene sheet. Covalent bonding charac-teristics of Ni (on certain sites) with carbon atomson the nanotube was identi®ed from the calcula-tions. The interaction was found to be strongerthan ironic-like Ni bonding with a graphene sheet.The strong Ni±SWNT interaction was attributedto curvature-induced rehybridization of carbon sp2

orbitals with the Ni d-orbital [32]. This resultcould be consistent with our experimental obser-vation that Ni exhibits signi®cant condensation/sticking coe�cient on SWNTs, and the depositionleads to quasi-continuous Ni nanowires. There-fore, the Ni±SWNT interaction could have certaincovalent bonding characteristics, and is consistentwith the electrical transport result that low resis-tance ohmic contacts are obtainable in Ni con-tacted SWNTs.

Ti atoms deposited on nanotubes exhibit thehighest condensation/sticking coe�cient among Niand other metals. The Ti±SWNT interactionshould be stronger than that for Ni±SWNT andcould involve covalent bonding. This can be at-tributed to the high a�nity of Ti for carbide for-mation and the curvature induced rehybridizatione�ect. The intimate Ti±SWNT interaction is con-sistent with our result that low resistance ohmic

Y. Zhang et al. / Chemical Physics Letters 331 (2000) 35±41 39

electrical contacts to individual SWNTs can bereliably made by depositing Ti electrodes ontonanotubes. It is interesting to note that Ti has beenthe favorite metal for making ohmic contacts todoped diamonds. In this case, as deposited Tiforms carbide bonds with the very surface of dia-mond, and ohmic contacts are obtained by ther-mal annealing that leads to a layer of carbide atthe Ti±diamond interface [33].

Our coating results suggest that the interactionsbetween Au and Al with SWNTs are very weakand presumably van der Waals in nature. The lowcondensation coe�cient and di�usion barrier isevident for Au from the deposited metal struc-tures. This is also consistent with our electricalmeasurement result that Au and Al contactedSWNTs tend to be high in resistance and oftenexhibit non-ohmic behavior and with non-linearcurrent-voltage characteristics.

The quasi-continuous coating of Pd on SWNTspoints to a high condensation/sticking coe�cientfor Pd on nanotubes, which could be related to thesimilarity between Pd and Ni in the periodic table.The observed weak Fe±SWNT interaction issomewhat surprising as Fe shares certain commonchemical characteristics as Ni in terms of carbonsolubility or carbide formation. The di�erence inFe and Ni depositions on SWNTs is not under-stood at the present time.

It is important to note that metal coating onnanotubes could be in¯uenced by various deposi-tion conditions including temperature. We believethat heat generated in metal evaporations may beresponsible for the shape changes of Au and Alcrystallites (Fig. 1d and e, and 3d and e), due tofacilitated di�usions and re-arrangements of metalatoms or small clusters at high temperatures. Also,it is necessary to employ clean SWNTs for theinvestigation of metal±tube interactions. If exten-sive amorphous carbon overcoating exists on thesidewall of nanotubes, the outcome of metal de-position could be drastically di�erent from cleantubes. It is known from previous work that metalcoating on amorphous carbon can from continu-ous ®lms, as opposed to large clusters on the basalplane of graphite [34]. We have found that a thinlayer of Triton X-100 surfactant molecules pre-absorbed on SWNTs allow for nearly uniform and

continuous Au coating (Fig. 5), in strong contrastwith the isolated Au particles formed on cleannanotubes. This also suggests that rational pre-treatment of nanotubes could enable quasi-one-dimensional nanotube substrates with tunablesurface properties for the formation of variousinteresting nanostructures.

Finally, we note that metal±substrate interac-tions strongly depend on the type of substrate[15,35]. It is well known that Ti, Ni, Cr and Aladhere well on SiO2 and glass substrate. The in-teractions are strong in these cases and are be-lieved to involve metal±oxygen bonds formed atthe interface during metal deposition [35]. Thestrong interactions between Ti and Ni with oxidesubstrates and carbon nanotubes must have dif-ferent mechanisms due to di�erent chemical natureof the metal±substrate interfaces. For Al, the in-teraction with oxide substrates appears muchstronger than with carbon nanotubes.

4. Conclusion

We have carried out depositions of varioustransition metals on carbon nanotubes to gain anunderstanding of metal±tube interactions. De-pending on the metal type, continuous metal(Ti,Ni,Pd) nanowires or isolated particles(Au,Al,Fe) are formed on nanotube substrates. Ti,Ni interact strongly with the sidewall of nanotu-bes, and the interactions are likely to be related to

Fig. 5. A nearly continuous Au nanowire formed on a surfac-

tant pretreated carbon nanotube.

40 Y. Zhang et al. / Chemical Physics Letters 331 (2000) 35±41

partial covalent bonding between the metals andcarbon atoms. Au and Al interact weakly withSWNTs through van der Waals forces. These re-sults shed light into electrical coupling betweennanotubes and metal contacts, an issue importantto fundamental electrical properties of nanotubesand their applications as high performance devic-es. Clean or pre-treated carbon nanotubes can beused as novel one-dimensional substrates for ob-taining various nanowires.

Acknowledgements

This work was supported by NSF, SRC/Motorola, a David and Lucille Packard Fellow-ship, a Terman Fellowship, ABB Group, Labo-ratory for Advance Materials (LAM) at Stanford,the Camile Henry-Dreyfus Foundation and theAmerican Chemical Society.

References

[1] S. Iijima, Nature 354 (1991) 56.

[2] S. Iijima, T. Ichihashi, Nature 363 (1993) 603.

[3] M.S. Dresselhaus, G. Dresselhaus, P.C. Eklund, Science of

Fullerenes and Carbon Nanotubes, Academic Press, San

Diego, 1996.

[4] C. Dekker, Phys. Today 52 (1999) 22.

[5] H. Dai, J. Kong, C. Zhou, N. Franklin, T. Tombler, A.

Cassell, S. Fan, M. Chapline, J. Phys. Chem. 103 (1999)

11246.

[6] H. Dai, Phys. World 13 (2000) 43.

[7] H. Soh, C. Quate, A. Morpurgo, C. Marcus, J. Kong, H.

Dai, Appl. Phys. Lett. 75 (1999) 627.

[8] J. Kong, C. Zhou, A. Morpurgo, T. Soh, C. Marcus, C.

Quate, H. Dai, Appl. Phys. A 69 (1999) 305.

[9] C. Zhou, J. Kong, H. Dai, Appl. Phys. Lett. 76 (1999) 1597.

[10] C. Zhou, J. Kong, H. Dai, Phys. Rev. Lett. 84 (2000) 5604.

[11] H. Dai, J. Kong, unpublished results.

[12] H. Dai, E.W. Wong, Y.Z. Lu, S. Fan, C.M. Lieber, Nature

375 (1995) 769.

[13] W.Q. Han, S.S. Fan, Q.Q. Li, Y.D. Hu, Science 277 (1997)

1287.

[14] A. Bezryadin, C.N. Lau, M. Tinkham, Nature 404 (2000)

971.

[15] K.L. Chopra, Thin Film Phenomena, McGraw-Hill, New

York, 1969.

[16] J. Kong, H. Soh, A. Cassell, C.F. Quate, H. Dai, Nature

395 (1998) 878.

[17] A. Cassell, J. Raymakers, J. Kong, H. Dai, J. Phys. Chem.

103 (1999) 6484.

[18] A. Cassell, N. Franklin, T. Tombler, E. Chan, J. Han, H.

Dai, J. Am. Chem. Soc. 121 (1999) 7975.

[19] N. Franklin, H. Dai, Adv. Mater. 12 (2000) 890.

[20] L. Barfotti, P. Jensen, A. Hoareau, M. Treilleux, B.

Cabaud, A. Perez, F.C.S. Aires, Surf. Sci. 367 (1996) 276.

[21] Q. Ma, A. Rosenberg, Phys. Rev. B 60 (1999) 2827.

[22] M. Baumer, J. Libuda, H. Freund, Surf. Sci. 327 (1995)

321.

[23] P. Kruger, A. Rakotomahevitra, J. Parlebas, C. Demang-

eat, Phys. Rev. B (1998) 5276.

[24] S.S. Peng, B.R. Cooper, Y.G. Hao, Philos. Mag. 73 (1996)

611.

[25] D. Tomanek, W. Zhong, Phys. Rev. B 43 (1991) 12623.

[26] Q. Ma, R. Rosenberg, Surf. Sci. 391 (1997) L1224.

[27] I. Moulett, Surf. Sci. 333 (1995) 697.

[28] J. Arthur, A. Cho, Surf. Sci. 36 (1973) 641.

[29] T. Ohno, Y. Chen, S. Harvey, G.H. Kroll, P. Benning, J.

Weaver, L. Chibante, R. Smalley, Phys. Rev. B 47 (1993)

2389.

[30] V. Vijayakrishnan, A. Santra, R. Seshadri, R. Nagarajan,

T. Pradeep, C.N.R. Rao, Surf. Sci. 262 (1992) L87.

[31] E. Parks, K. Kerns, S.R.B. Winter, Phys. Rev. B (1999).

[32] M. Menon, A. Andriotis, G. Froudakis, Chem. Phys. Lett.

320 (2000) 425.

[33] T. Tachibana, B. Williams, J. Glass, Phys. Rev. B (1992)

11975.

[34] W. Egelho�, G. Tibbetts, Phys. Rev. B 19 (1979) 5028.

[35] M. Ohring, The Materials Science of Thin Films, Academic

Press, San diego, 1992.

Y. Zhang et al. / Chemical Physics Letters 331 (2000) 35±41 41