Carbon nanotube-clamped metal atomic chainDai-Ming Tanga,1, Li-Chang Yina,1, Feng Lia, Chang Liua,2, Wan-Jing Yua, Peng-Xiang Houa, Bo Wua, Young-Hee Leeb,Xiu-Liang Maa, and Hui-Ming Chenga,2

aShenyang National Laboratory for Materials Science, Institute of Metal Research, Chinese Academy of Sciences, 72 Wenhua Road, Shenyang 110016,China; and bDepartment of Physics, Department of Energy Science, Sungkyunkwan University, Suwon 440-746, Korea

Edited* byMildred Dresselhaus, Massachusetts Institute of Technology, Cambridge, MA, and approvedMarch 29, 2010 (received for review December 30, 2009)

Metal atomic chain (MAC) is an ultimate one-dimensional structurewith unique physical properties, such as quantized conductance,colossal magnetic anisotropy, and quantized magnetoresistance.Therefore, MACs show great potential as possible componentsof nanoscale electronic and spintronic devices. However, MACsare usually suspended between two macroscale metallic electro-des; hence obvious technical barriers exist in the interconnectionand integration of MACs. Here we report a carbon nanotube(CNT)-clamped MAC, where CNTs play the roles of both nanocon-nector and electrodes. This nanostructure is prepared by in situ ma-chining a metal-filled CNT, including peeling off carbon shells byspatially and elementally selective electron beam irradiation andfurther elongating the exposed metal nanorod. The microstructureand formation process of this CNT-clamped MAC are explored byboth transmission electron microscopy observations and theoreti-cal simulations. First-principles calculations indicate that strongcovalent bonds are formed between the CNT and MAC. The elec-trical transport property of the CNT-clamped MAC was experimen-tally measured, and quantized conductance was observed.

in situ microscope ∣ nanoconnector ∣ electrode ∣ quantized conductance ∣half-metallicity

Electron transport across a metal atomic chain (MAC) presentsultimate miniaturization and novel functionality of electronic

devices (1–4). Several techniques, including scanning tunnelingmicroscopy (STM) (5), mechanically controlled breaking junction(6), high resolution transmission electron microscopy (HRTEM)(7), self-terminated electrodeposition (8), and controlled electro-migration (9), have been developed for the preparation ofsuspended MACs. However, how to connect and integrate theextremely small MACs remains a challenge, because the fabri-cated MACs are usually held between two macroscale electrodes.Carbon nanotube (CNT) is a one-dimensional tubular structurewith intriguing electronic properties. Various CNT-based electro-nic devices, such as ballistic wire (10), field effect transistor (11),and logic circuit (12), have been demonstrated. It was recentlyreported by Rodríguez-Manzo et al. that CNTs can form strongcovalent heterojunctions and interconnect with particles of differ-ent metals, by theoretical calculations and in situ TEM observa-tion (13–15). Therefore, CNT can be a promising candidate forinterconnecting and integrating MACs. Nevertheless, no suchCNT-MAC structure is experimentally demonstrated yet.

In this study, we design and fabricate a unique CNT-MAChybrid structure: CNT-clamped MAC. Our fabrication strategyis to fill a metal nanorod into a CNT, and then peel off the carbonshell surrounding the filled metal nanorod and thin theexposed metal nanorod by strong electron irradiation within aTEM. Finally, a CNT-clamped MAC is formed through furtherelongation and thinning by the strain existing in the TEM sampleor by stretching using a TEM-STM holder.

Results and DiscussionThe metal can be Fe, Co, Ni and their alloys that serve as thegrowth catalyst of CNTs and are encapsulated in situ (16, 17),or a variety of metals that are electrochemically deposited intoCNTs by anodic aluminum oxide (AAO) template method(18). As an example, here we present the formation and structure

of a CNT-clamped Fe AC. Firstly, Fe-filled CNTs were preparedby floating catalyst chemical vapour deposition (FCCVD) usingferrocene as catalyst precursor and acetylene as carbon source(16). Fig. 1A shows the TEM image of a CNT filled with a Fenanorod. We can see that the outer and inner diameters ofthe CNTare about 25 nm and 10 nm, respectively. Selected areaelectron diffraction pattern (Fig. 1A, Inset) indicates that thefilled Fe nanorod is single crystalline and body-center cubic(bcc) structured. When electron beam (with a current densityof 100–300 A∕cm2) was focused on the carbon shells surroundingthe Fe nanorod, the CNTwalls were removed element- and site-selectively (Fig. 1B) (19). And then the exposed Fe nanorod wasfurther deformed and thinned, under strong electron irradiation.Because the energy of electron beam in our work (200 and300 keV) is lower than the threshold energy (higher than400 keV) for atom displacements from bulk Fe crystal lattices,the deformation and thinning are possibly due to electron irradia-tion induced surface sputtering effects and resulted surface diffu-sion (14, 20–22). During the thinning process, the Fe nanorodkept the single crystalline structure, with lattice corresponding

Fig. 1. Fabrication process of a CNT-clamped Fe nanorod. (A) TEM imageof an original Fe-filled CNT. (Inset) Selected area electron diffraction patternof the filled Fe nanorod. (B) TEM image showing that part of carbon shellssurrounding the Fe nanorod are peeled off. (C) HRTEM imageof the exposed Fe nanorod. (D) Schematic drawing of the CNT-clamped Fenanorod.

Author contributions: D.-M.T., L.-C.Y., F.L., C.L., and H.-M.C. designed research; D.-M.T. andL.-C.Y. performed research; W.-J.Y., P.-X.H., and B.W. contributed new reagents/analytictools; D.-M.T., L.-C.Y., F.L., C.L., Y.H.L., X.-L.M., and H.-M.C. analyzed data; and D.-M.T.,L.-C.Y., F.L., C.L., and H.-M.C. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.1D.M.T. and L.C.Y. contributed equally to this work.2To whom correspondence may be addressed. E-mail: cliu@imr.ac.cn or cheng@imr.ac.cn.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.0914970107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.0914970107 PNAS ∣ May 18, 2010 ∣ vol. 107 ∣ no. 20 ∣ 9055–9059

ENGINEE

RING

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

1

to (110) clearly resolved (Fig. 1C). Finally, a CNT-clamped Fenanorod was produced, as schematically shown in Fig. 1D.

When the diameter of the Fe nanorod was reduced to approxi-mately 6 nm (Fig. 2 A and B), the intensity of the electron beamwas decreased to a level normally used for TEM observations(10–30 A∕cm2). The Fe nanorod was elongated and thinnedspontaneously, possibly due to the strain existing in the TEMsample (23–25). Atomic steps are observed during the thinningprocess as marked by arrows in Fig. 2 C and D. Distortion isanother deformation manner for the Fe nanorod (marked bydashed lines in Fig. 2 E and F). The fact that the distortion doesnot cause rupture indicates that the bending energy can be ab-sorbed by the very fine nanorod (24). Fig. 2G shows that anFe nanorod contains only 3 layers of Fe atoms, and a single-atomwide Fe AC is finally formed in Fig. 2H. This single-atom wide FeAC is composed of 5 atoms with 3 atoms suspended, as schema-tically shown in Fig. 2I. The interatom distance at the center ofthe Fe AC is about 2.2 Å. The larger interatom distance of the FeAC (compared with the interplane distance of bulk Fe(110)plane) is consistent with previous investigations on elementary(Au) (24, 26) and alloy (Au-Ag) ACs (27).

First-principles calculation is effective in investigating the for-mation mechanism and physical properties of MACs (28–31). Weused this method to simulate the formation process of Fe ACsand to explore the electronic structure of CNT-clamped Fe

ACs. Our simulations start from a Fe nanorod oriented along[110] direction (Fig. 3AI), based on the results of the aboveHRTEM characterizations. This nanorod is elongated graduallyalong the axis direction with a step of 0.4 Å. Along with the elon-gation, the center atoms in the second and fourth layer extrudeout (Fig. 3 A II–III). Then, the nanorod shows obvious distortionperpendicular to the [110] direction (Fig. 3 A III–IV), consistentwith the experimentally observed slip of Fe (110) planes. The na-norod is elongated continuously and an AC emerges (Fig. 3 V–VI). The AC is stretched further as the elongation continues, andfinally a single-atom wide Fe AC composed of 4 atoms forms(Fig. 3AVII) before rupture (Fig. 3AVIII). The interatom dis-tances of the Fe AC were calculated to be 2.27, 2.24, and2.30 Å, which are larger than the interplane distance of bulkFe (110) plane, coinciding well with the TEM observations. Alongwith the structural evolution, we calculated the total bindingenergy of the Fe nanorod as a function of its length, and the re-sults are shown in Fig. 3B. There are two abrupt drops in the totalenergy, corresponding to the slip along (110) and the final rup-ture. By slipping, stress can be released and the total energy isdecreased. Therefore, both HRTEM observations and theoreticalsimulations indicate that the slip is an important deformationmode in the formation of a Fe AC.

To explore the electronic structure of CNT-clamped Fe ACs,we construct a (5, 5) single-walled carbon nanotube (SWCNT)-clamped Fe AC (the configuration of the Fe AC is determinedfrom the simulated elongation of Fe nanorod as shown inFig. 3AIII), and its fully relaxed structural model is shown inFig. 3CI. The calculated total densities of states (DOS) (Fig. 3CII)indicate that the SWCNT-clamped Fe AC is metallic with anasymmetric distribution for majority and minority states nearthe Fermi energy (EF) level. Fig. 3 CIII and 3CIV show thecalculated local DOS for an Fe AC and SWCNT, respectively.The Fe AC behaves as a metal for majority states whereas itshows a band-gap of approximately 0.50 eV for its minority states.The local DOS of SWCNTshows symmetrical distribution for themajority and minority spin near the EF. The charge densities oftwo slices along the x and y axes are plotted in Fig. 3D. Covalentbonding is clearly present at the interface between the SWCNTand Fe AC, which is consistent with the calculations and observa-tions of the electronic structure of the junction between CNTsand metal particles (13). The average Fe-C bond length was cal-culated to be 2.06 Å, which is comparable to that of bulk Fe3C(ranging from 1.90 to 2.06 Å) (32), indicating a strong connectionbetween the SWCNTand Fe AC. To investigate the structure andelectronic properties of an Fe AC clamped by larger CNTs with>1 shells, we constructed and studied a double-walled carbon na-notube (DWCNT)-clamped Fe AC (Fig. S1). Covalent bondingwas established at the interface of the Fe AC and the inner layerof DWCNT, with no obvious interaction between the Fe AC andthe outer shell of DWCNT. In addition, spin resolved local DOScalculations found that the Fe AC is also half metallic. Therefore,our density functional theory (DFT) studies show that, after thehybridization and connection, the Fe AC forms a strong bondingwith the inner layer of a CNT and retains its half-metallicityfeature.

Having realized the connection of CNTand Fe AC, we furtherinvestigated the electronic conductance of the obtainedCNT-clamped Fe AC in situ by using a TEM-STM holder.The experimental configurations of the fabrication and measure-ments are schematically shown in Fig. 4A. The detailed experi-mental procedure can be found in Materials and Methods.Briefly, an Fe-filled CNT is suspended between a gold STMtip and a gold wire. Then, strong electron beam is used to peeloff the carbon shells and to thin the exposed Fe nanorod of theFe-filled CNT. When the diameter of the Fe nanorod reachesseveral nanometers, the electron beam is reduced to a levelfor normal TEM observations and a tensile force is applied by

Fig. 2. Formation of a single-atom wide Fe AC. (A)–(H) High resolution TEMimages showing the formation process of the Fe AC. Atomic steps can beobserved at the surface [marked by arrows in (C) and (D)]. Distortion along(110) is marked with dotted lines in (E). The scale bar is 2 nm. (I) Schematicdrawing of the Fe AC shown in (H); the projected interatom distances aremarked in Å.

9056 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0914970107 Tang et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

1

retracting the STM tip gently with a speed of approximately0.1 nm∕s to form an Fe AC. Electrical conductance is measuredat a constant bias of 12 mV simultaneously, along with the elon-gation, thinning and rupture of the Fe AC, with a recorded curveof conductance versus time shown in Fig. 4B. We can see that theconductance decreases in a stepwise way, and the conductanceplateaus are found to be near integral multiples of 0.5G0 (con-ductance quantum, which is 2e2∕h ∼ ð13 kΩÞ−1Þ. The last plateauis about 2G0, because the Fe nanorod ruptured before single-atom wide AC formed. Then, we reconnected the ruptured Fenanorod and stretched it more slowly. In this case, stepwise-decreasing plateaus were also recorded, and the last plateau isaround 0.5G0, as shown in Fig. 4C. Quantized conductance

for metal atomic contacts has been intensively studied (33).The conductance can be interpreted with the Landauer formula(G ¼ G0ΣTi, where G0 ¼ 2e2∕h, Ti is the transmission probabil-ity for the ith conductance channel) (34). For ferromagneticmetals, the conductance peaks are around integral multiples of0.5G0, rather than G0 for nonmagnetic metals such as gold(1, 2) and alkali metals (35), due to the lack of spin degeneracy(3, 8, 36, 37). Therefore, the conductance plateaus in our workcan be interpreted as quantized conductance of ferromagneticmetal (Fe), and the conductance steps are ascribed to structuralrearrangements during elongation. The plateaus are not always atthe integral multiples of 0.5G0, which was also found previouslyfor other metals, and this can be understood because the trans-

Fig. 3. DFTcalculations on the formation process of Fe AC and electronic properties of CNT-clamped Fe AC. (A) Simulation on the structural evolution of the FeAC. (I) The original [110] oriented Fe nanorod. (II) and (III) The center atoms in the second and fourth layer extruded out. (IV) Distortion along (110) occurs.(V)–(VII) Further elongation and formation of the single-atom wide Fe AC. (VIII) Rupture of the Fe AC. (B) The calculated total energy as a function of thenanorod length. The interatom distances and lengths of the nanorod are marked in Å. (C) Simulated structural model and the calculated density of states of theCNT-clamped Fe AC. (I) Fully relaxed ball-and-stick model of a CNT-clamped Fe AC. Carbon and iron atoms are marked in gray and blue, respectively. (II) Spinresolved total DOS for the CNT-clamped Fe AC. (III) and (IV) Spin resolved local DOS of the Fe AC and CNT. The red and blue lines denote the majority andminority states, respectively. And the Fermi energy (EF) is set to zero. (D) Charge density of two slices cut along the x and y axes. Strong covalent bonding isshown at the interface of the Fe AC and CNT.

Fig. 4. In situ electrical transport measurements of a CNT-clamped Fe AC. (A) Schematic diagram of the experimental process. (B) and (C) Conductance changesas a function of time for two formation and thinning processes of Fe ACs at a constant bias of 12 mV, where the contact ruptured comparatively abruptly in(B) and smoothly in (C).

Tang et al. PNAS ∣ May 18, 2010 ∣ vol. 107 ∣ no. 20 ∣ 9057

ENGINEE

RING

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

1

mission probability is not always one for one conductance chan-nel (33, 37). The plateaus last at a time scale of several seconds,therefore the current dependence on bias voltage (I–V character-istic) could be measured when suspending the elongation process,as shown in Fig. S2. When the contact is about 0.75 nm in width,corresponding to 3 to 4 Fe atomic layers, the I–V curve is almostlinear within the measurement voltage range (−0.5 to 0.5 V),with a calculated conductance of about 1.5G0. Previous studieshave shown that contamination of gas adsorption may lead tononlinear behavior for metal nanocontacts (38), so the linear be-havior observed in this study indicates that the conductance isOhmic (37) and the contact is clean (8). The above results showthat after the connection and assembly with CNT, the MACscould well retain their unique electrical transport properties, suchas quantized conductance.

In addition to the CNT-clamped Fe AC, we also applied theabove fabrication philosophy to other metals and alloys. For ex-ample, Pt-filled CNTs and Fe-Ni alloy-filled CNTs were producedby an AAO template and chlorine-promoted FCCVD method,respectively. CNT-clamped Pt and Fe-Ni alloy ACs were success-fully fabricated by an approach similar to the fabrication of the FeAC as described above (Fig. S3, Fig. S4, Fig. S5, and Fig. S6).DFT calculations also confirmed that both the Pt and Co ACscan form strong connection with CNTs (Fig. S7 and Fig. S8).Therefore, a variety of metals are applicable for the fabricationof CNT-clamped MACs.

To sum up, we design and fabricate CNT-clamped MACs,which can provide a general approach for the interconnectionand integration of MACs with CNTs. The formation process ofCNT-clamped MACs was explored by both in situ TEM observa-tions and first-principles calculations. Electronic structure calcu-lations indicate that strong covalent bonds are formed at theCNT/MAC interface. The intriguing properties of Fe ACs, suchas half-metallicity and quantized conductance, are retained afterthe combination of Fe ACs and CNTs. The strategy we proposedis effective in fabricating a variety of MACs and in situ connectingthese MACs with CNTs, thus it may find potential applications inthe assembly of nano/subnano devices.

Materials and Methods1 Synthesis of Metal-Filled CNTs. Fe and Fe alloy-filled CNTs were produced byFCCVDmethod (16, 17). Pt-filled CNTs were produced by AAO template meth-od. More details can be found in the SI Text.

2 Fabrication of CNT-Clamped MACs. Two methods were used for the fabrica-tion of CNT-clamped MACs. The first one is HRTEM method developed byKondo et al. (7). It was done under Tecnai G2 F30 or F20 TEMs, using metalnanorod-filled CNTs as starting materials. Focused electron beamwas used topeel off the carbon shells by knocking-on effects (19) and to deform and thinthe exposed metal nanorod by surface sputtering and diffusion effects(14, 20–22) in the initial stage. When the diameter of the nanorod was re-duced to approximately 6 nm, the electron beam density was decreasedto normal level for HRTEM observations. The nanorod was elongated andthinned gradually by the strain in the TEM sample grid (23–25) to formCNT-clamped MACs. The images were taken by using CCD camera (GatanUltrascan 894) with an exposure time of 0.5 s.

The second method is in situ TEM-STM developed by Ohnishi et al. (1)using a Nanofactory holder. Metal-filled CNTs were adhered to a gold wireof 0.25 mm in diameter by simply touching the samples with a cleanly cutedge. To avoid organic contaminations, no conductive clue was used. A goldSTM tip, prepared by electrochemical etching, was manipulated by a piezo-electric tube to touch the metal-filled CNTs. Then a bias was applied betweenthe STM tip and the CNT to anneal the sample. The purpose of the annealingis to desorb possible adsorbed gases and get Ohmic contact between the CNTand the STM tip. The current was controlled to be around 10 μA, by tuningthe bias. Current voltage characteristics of the system at different stages areshown in Fig. S9. When the STM tip just touches the sample, only noisecurrent <2 nA can be measured, due to a thin inert layer on the tip. Duringthe annealing process, a nonlinear behavior is revealed and can be ascribedto the Schottky barrier between the tip and the CNT. After annealing, analmost linear behavior is demonstrated, and the conductance is around

1G0 after annealing, which is used as the resistance for correction, whenthe conductance of the CNT-clamped MAC is calculated. Simultaneously, itis found that the CNT is welded to the STM tip. The bias is also calibratedafter the annealing process. It was found that for our electronics, the zeropotential corresponds to about 18 mV. Therefore, when calculating conduc-tance, the bias should be corrected, accordingly. After that, similar to theHRTEM method, strong electron irradiation was used to peel off the carbonshells, by focusing the electron beam probe to be around 100 nm and swipingaround the targeted area. To avoid high temperature reactions and un-wanted phase transitions, the bias was set to be zero. Electron beamwas alsoused to thin the exposed Fe nanorod, similar to the first fabrication method.When the Fe nanorod was thinned to be around 6 nm, the electron beamintensity was reduced to normal level for HRTEM observations. The Fe nanor-od got further elongated and thinned until breaking, by retracting the STMtip gently at a speed of about 0.1 nm∕s. After the first rupture, the brokennanorod could be reconnected and subjected to the second elongation andthinning. Along with the fabrication of the Fe ACs by retraction, the conduc-tance was recorded with a constant bias.

3 In situ electronic conductance measurement. The conductance changeagainst the time was measured along with the retraction of the STM tip.The bias was set to be a constant of 12 mV (after calibration), and thesampling rate was 2 kHz. The conductance dependence on bias voltagewas measured when suspending the elongation process. The conductanceof the CNT-clamped MAC was calculated with correction using the resistanceafter the structure annealing, using the following formula

G ¼ ½ðV − V 0Þ∕I − R0�−1∕ð13kΩÞ−1;

whereG is the conductance with a unit ofG0 (conductance quantum), V and Iare bias applied and current measured, V0 (18 mV for our system) is thecalibrated zero bias point, R0 is the resistance of the system after annealing,and ð13 kΩÞ−1 corresponding to one conductance quantum.

4 First-principles calculations and simulation methods. We used the accuratefrozen-core full-potential projector augmented-wave method (39), imple-mented in the Vienna ab initio simulation package (40). The calculationswere based on DFT with the electronic exchange and correlation effectsdescribed by the generalized gradient approximation (41). We adoptedperiodic boundary conditions and placed a bcc Fe [110] oriented nanorodconsisting of 5 atomic layers involving 22 atoms in a 12 Å × 11 Å × 16 Åsupercell. To improve convergence in the case of partially occupied eigen-states at the Fermi level, a modest Gaussian smearing of σ ¼ 0.05 eV wasused. We used an energy cutoff of 350 eV for the plane-wave basis setand a cutoff of 1000 eV for the augmentation charges. The binding energieswere computed by taking an isolated spherical atom as a reference with atotal magnetic moment of 4 μB∕atom according to Hund’s rule. For bulkbcc Fe, the lattice constant and total magnetic moment were calculatedto be 2.83 Å and 2.19 μB, respectively, close to the corresponding experimen-tal values of 2.87 Å and 2.22 μB.

To simulate the formation process of the Fe AC, the nanorod was elon-gated along [110] direction by a step of 0.4 Å, meanwhile, the super cellwas prolonged by the same step in [110] direction to keep the separatedistance between the periodic images. At each step, symmetry-unrestrictedgeometry and spin optimizations were performed using conjugated gradientand quasi-Newtonian methods, except for the 8 Fe atoms located at the twoend layers of the nanorod, of which the coordinates along the [110] axis arefixed, until all the force components except for those constrained ones wereless than a threshold value of 0.01 eV∕Å.

A periodic geometry, which was constructed by connecting a (5, 5) SWCNTcomposed of 160 carbon atoms with a fully optimized Fe AC obtained duringabove elongation process in a 15 Å × 15 Å × 29.6 Å supercell, was adopted toinvestigate the atomic and electronic structure of CNT-clamped Fe AC. A fullygeometry relaxation was performed at the Γ-point only by minimizing theHellmann–Feynman forces on the atoms and stresses on the supercell, untilall the forces on each atom were smaller than 0.05 eV∕Å.

Pt or Co AC was obtained by using the same elongation process as that forFe AC with the initial Pt (Co) nanorod along [111] ([001]). The initial Pt (Co)nanorod consists of 123 (96) atoms. Locally stable Pt (28.4 Å) and Co (24.2 Å)ACs obtained during the elongation were selected to connect with a (8, 8)and (6, 6) SWCNT, respectively. CNT-clamped Pt (Co) AC was obtained aftergeometry optimization. The charge density plots of these two CNT-clampedMACs presented covalent bonding at the interface between CNT and Pt (Co)AC with an average bonding length of 2.00 (1.96) Å.

9058 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0914970107 Tang et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

1

ACKNOWLEDGMENTS. We acknowledge helpful discussions with Prof. ZhongLin Wang, Prof. R. Saito, Dr. Oleg Lourie, Dr. Kui Du, Dr. Cheng-Hua Sun, andProf. Wei Liu. We thank Dr. Rui-Tao Lv, Dr. Qing-Feng Liu, andMs. Chun-XiangShi for providing Fe- and alloy-filled CNT samples. We thank Dr. Chuan-BinJiang for TEM support. H.M.C. and C.L. acknowledge financial support fromMinistry of Science and Technology of China Grants 2008DFA51400,2006CB932701, and 2006CB932703, and National Natural Science Foundation

of China Grants 90606008, 50672103, and 50921004. F.L. and L.C.Y. acknowl-edge support from the Supercomputing Center, Computer Network Informa-tion Center, Chinese Academy of Sciences. Y.H.L. acknowledges support fromthe STAR faculty project and World Class University program through the Na-tional Research Foundation funded by the Ministry of Education, Science,and Technology of Korea (R31-2008-000-10029-0).

1. Ohnishi H, Kondo Y, Takayanagi K (1998) Quantized conductance through individualrows of suspended gold atoms. Nature 395:780–783.

2. Yanson AI, Bollinger GR, van den Brom HE, Agrait N, van Ruitenbeek JM (1998) For-mation and manipulation of a metallic wire of single gold atoms. Nature 395:783–785.

3. Sokolov A, Zhang CJ, Tsymbal EY, Redepenning J, Doudin B (2007) Quantizedmagnetoresistance in atomic-size contacts. Nat Nanotechnol 2:171–175.

4. Smogunov A, Dal Corso A, Delin A, Weht R, Tosatti E (2008) Colossal magnetic aniso-tropy of monatomic free and deposited platinum nanowires. Nat Nanotechnol3:22–25.

5. Agrait N, Rodrigo JG, Vieira S (1993) Conductance steps and quantization inatomic-size contacts. Phys Rev B 47:12345–12348.

6. Muller CJ, Vanruitenbeek JM, Dejongh LJ (1992) Conductance and supercurrentdiscontinuities in atomic-scale metallic constrictions of variable width. Phys Rev Lett69:140–143.

7. Kondo Y, Takayanagi K (1997) Gold nanobridge stabilized by surface structure.Phys Rev Lett 79:3455–3458.

8. Chopra HD, Sullivan MR, Armstrong JN, Hua SZ (2005) The quantum spin-valve incobalt atomic point contacts. Nat Mater 4:832–837.

9. Bolotin KI, Kuemmeth F, Ralph DC (2006) Anisotropic magnetoresistance and aniso-tropic tunneling magnetoresistance due to quantum interference in ferromagneticmetal break junctions. Phys Rev Lett 97:127202.

10. White CT, Todorov TN (1998) Carbon nanotubes as long ballistic conductors. Nature393:240–242.

11. Tans SJ, Verschueren ARM, Dekker C (1998) Room-temperature transistor based on asingle carbon nanotube. Nature 393:49–52.

12. Bachtold A, Hadley P, Nakanishi T, Dekker C (2001) Logic circuits with carbon nanotubetransistors. Science 294:1317–1320.

13. Rodríguez-Manzo JA, et al. (2009) Heterojunctions between metals and carbonnanotubes as ultimate nanocontacts. Proc Natl Acad Sci USA 106:4591–4595.

14. Rodriguez-Manzo JA, et al. (2009) Growth of single-walled carbon nanotubes fromsharp metal tips. Small 5:2710–2715.

15. Rodriguez-Manzo JA, Wang MS, Banhart F, Bando Y, Golberg D (2009) Multibranchedjunctions of carbon nanotubes via cobalt particles. Adv Mater 21:4477–4482.

16. Cheng HM, et al. (1998) Large-scale and low-cost synthesis of single-walled carbonnanotubes by the catalytic pyrolysis of hydrocarbons. Appl Phys Lett 72:3282–3284.

17. Lv RT, et al. (2007) Effect of using chlorine-containing precursors in the synthesis offeni-filled carbon nanotubes. Carbon 45:1433–1438.

18. Kyotani T, Pradhan BK, Tomita A (1999) Synthesis of carbon nanotube composites innanochannels of an anodic aluminum oxide film. Bull Chem Soc Jpn 72:1957–1970.

19. Banhart F (1999) Irradiation effects in carbon nanostructures. Rep Prog Phys62:1181–1221.

20. Egerton RF, Li P, Malac M (2004) Radiation damage in the TEM and SEM. Micron35:399–409.

21. Egerton RF, Wang F, Crozier PA (2006) Beam-induced damage to thin specimens in anintense electron probe. Microsc Microanal 12:65–71.

22. Williams DB, Carter CB (2009) Transmission Electron Microscopy a Textbook forMaterials Science (Springer, US), 2 Ed.

23. Rodrigues V, Fuhrer T, Ugarte D (2000) Signature of atomic structure in the quantumconductance of gold nanowires. Phys Rev Lett 85:4124–4127.

24. Rodrigues V, Ugarte D (2001) Real-time imaging of atomistic process in one-atom-thickmetal junctions. Phys Rev B 63:073405.

25. Rego LGC, Rocha AR, Rodrigues V, Ugarte D (2003) Role of structural evolution in thequantum conductance behavior of gold nanowires during stretching. Phys Rev B67:045412.

26. Bahn SR, Jacobsen KW (2001) Chain formation of metal atoms. Phys Rev Lett87:266101.

27. Bettini J, et al. (2006) Experimental realization of suspended atomic chains composedof different atomic species. Nat Nanotechnol 1:182–185.

28. Delin A, Tosatti E (2004) The electronic structure of 4d transition-metal monatomicwires. J Phys-Condens Mat 16:8061–8074.

29. Delin A, Tosatti E (2003) Magnetic phenomena in 5d transition metal nanowires.Phys Rev B 68:144434.

30. Tung JC, Guo GY (2007) Systematic ab initio study of themagnetic and electronic prop-erties of all 3d transition metal linear and zigzag nanowires. Phys Rev B 76:094413.

31. Lucignano P, Mazzarello R, Smogunov A, Fabrizio M, Tosatti E (2009) Kondo conduc-tance in an atomic nanocontact from first principles. Nat Mater 8:563–567.

32. Wood IG, et al. (2004) Thermal expansion and crystal structure of cementite, fe3c,between 4 and 600 k determined by time-of-flight neutron powder diffraction. J ApplCrystallogr 37:82–90.

33. Agraït N, Yeyati AL, van Ruitenbeek JM (2003) Quantum properties of atomic-sizedconductors. Phys Rep 377:81–279.

34. Landauer R (1970) Electrical resistance of disordered one-dimensional lattices. PhilosMag 21:863–867.

35. Krans JM, Vanruitenbeek JM, Fisun VV, Yanson IK, Dejongh LJ (1995) The signature ofconductance quantization in metallic point contacts. Nature 375:767–769.

36. Rodrigues V, Bettini J, Silva PC, Ugarte D (2003) Evidence for spontaneous spin-polarized transport in magnetic nanowires. Phys Rev Lett 91:096801.

37. Komori F, Nakatsuji K (1999) Quantized conductance through atomic-sized ironcontacts at 4.2 k. J Phys Soc Jpn 68:3786–3789.

38. Mehrez H, et al. (2002) I-V characteristics and differential conductance fluctuations ofau nanowires. Phys Rev B 65:195419.

39. Blöchl PE (1994) Projector augmented-wave method. Phys Rev B 50:17953.40. Kresse G, Furthmüllerr J (1996) Efficient iterative schemes for ab initio total-energy

calculations using a plane-wave basis set. Phys Rev B 54:11169.41. Wang Y, Perdew JP (1991) Correlation hole of the spin-polarized electron gas, with

exact small-wave-vector and high-density scaling. Phys Rev B 44:13298.

Tang et al. PNAS ∣ May 18, 2010 ∣ vol. 107 ∣ no. 20 ∣ 9059

ENGINEE

RING

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

1