Design and control of electron transport properties

0 downloads 0 Views 2MB Size Report
Sep 8, 2009 - ical molecular switches induced by cis-trans isomerization of azobenzene (23–25), making the process potentially more rele-.
Design and control of electron transport properties of single molecules Shuan Pana,1, Qiang Fua,b,1, Tian Huanga, Aidi Zhaoa, Bing Wanga, Yi Luoa,b, Jinlong Yanga,2, and Jianguo Houa,2 aHefei

National Laboratory for Physical Sciences at Microscale, University of Science and Technology of China, Hefei, Anhui 230026, China; and bDepartment of Theoretical Chemistry, School of Biotechnology, Royal Institute of Technology, S-106 91 Stockholm, Sweden Edited by Ellen D. Williams, University of Maryland, College Park, MD, and approved July 20, 2009 (received for review March 20, 2009)

hydrogen tautomerization 兩 melamine molecules 兩 rectifying effect 兩 switching property

E

lectron transport is a fundamental process that controls physical properties and chemical activities of molecular and biological systems. Over the years, different electron transport behaviors of a variety of molecules have been observed, and much effort has been made to elucidate the underlying mechanisms (1–12). Despite of these recent advances, it remains a great challenge to actively control the electron transport in a molecule and to systematically change its behavior from one type to another since this requires not only precise control of molecular structure, but also accurate activation of different electron tunneling processes. Controlling electron transport at the molecular level has important consequences for many applications, such as molecular electronics (13–26), biosensors (27), and solar cells (28, 29). For certain molecules, change of electron transport properties could take place due to their specific response to the change of molecular conformation or orientation (19–25), chemical reactions (26), and tautomerization (18). The latter experiment (18) has attracted considerable attention owning to the facts that the switching process involved does not result in drastic molecular conformation changes as often occurring in mechanical molecular switches induced by cis-trans isomerization of azobenzene (23–25), making the process potentially more relevant to applications in memory devices. Although many studies have been conducted over the years, only a limited number of special molecules can be chosen for such experiments. In this joint experimental and theoretical study, we demonstrate the possibility of changing the electron transport behavior of an ordinary molecule, melamine, with the help of surface chemistry and scanning tunneling microscope (STM) in a controllable manner. It is shown that the involvement of a dehydrogenation process can make the molecule standing on a Cu (100) surface and behaving like a conducting molecule. By applying a high-voltage pulse, the energetically unfavorable tautomerization reaction can be realized to generate a stable tautomer that works as a rectifier. Meanwhile, the mechanical oscillation of a N-H bond in the tautomer activated by inelastic electron scattering leads to conductance switching with tunable www.pnas.org兾cgi兾doi兾10.1073兾pnas.0903131106

frequency depending on the external bias and injecting current. Collectively, a dual functionality, rectifying and switching, has been integrated in a single molecular device. The observed non-integer power law dependence of the switching rate has also been well-described by a dynamic model. Results and Discussion Single melamine molecules adsorbed on the terraces of Cu (100) surface exhibit a ‘‘dumbbell’’ shape protrusion in topographic STM images (Fig. 1A). The structure and orientation of the adsorbed molecule can be well determined by comparing experimental STM images with first principles calculations. There are three amino groups in a free melamine molecule (Fig. 1B). Upon adsorption, two hydrogen atoms in two amino groups are dissociated to allow the molecule chemically bonded to the Cu substrate and standing upright on the bridge site of the surface (Fig. 1C and D) (see SI Text and Figs. S1 and S2). Theoretically simulated STM image (Fig. 1E) resembles very well the experimental results and attributes the observed ‘‘dumbbell’’ shape to the ␲ bond of the molecule. There are three additional possible tautomers for a free melamine molecule (30, 31). However, upon adsorption, the dehydrogenated melamine has only one additional tautomer in which one of hydrogen atoms in the last amino group is transferred to a neighboring nitrogen atom to form an N-H bond. The dehydrogenated melamine and its tautomer have very different conjugations, which could in principle lead to different electron transport properties. We applied high-voltage pulse to activate the tautomerization process. After imaging the molecules at low bias voltage and low tunneling current (typically 兩 V 兩 ⬍ 1.0 V, 兩 I 兩 ⬍ 0.6 nA), we placed the tip on top of single molecule, temporarily suspended the feedback loop and then applied a positive high-voltage pulse (⫹2.3 V to ⫹ 2.8 V) to the molecule. Fig. 2A shows the STM images for three melamine molecules taken from the same area as that given in Fig. 1 A after applying a pulse of 2.4 V to the molecules 1 and 3. The topographic images of the activated molecules, 1⬘ and 3⬘, obtained at bias of 0.2 V show asymmetric ‘‘cashew nut’’ patterns which are significantly different from the symmetric ‘‘dumbbell’’ pattern of original molecules, indicating that the symmetric structure of these molecules was changed. If we apply a highervoltage pulse again on the modified molecule (typically 0.5 V ⬍ 兩 V 兩 ⬍ 1.0 V), the opening of the asymmetric ‘‘cashew nut’’ pattern of molecule 1⬘ may reverse, see molecule 1⬙ in Fig. 2B as also obtained at bias of 0.2 V. Hence the switching process is Author contributions: B.W., J.Y., and J.H. designed research; S.P., Q.F., T.H., A.Z., B.W., and J.H. performed research; Y.L. contributed new reagents/analytic tools; S.P., Q.F., T.H., A.Z., B.W., Y.L., J.Y., and J.H. analyzed data; and B.W., Y.L., J.Y., and J.H. wrote the paper. The authors declare no conflict of interest. This article is a PNAS Direct Submission. 1S.P. 2To

and Q.F. contributed equally to this work.

whom correspondence [email protected].

may

be

addressed.

E-mail:

[email protected]

or

This article contains supporting information online at www.pnas.org/cgi/content/full/ 0903131106/DCSupplemental.

PNAS 兩 September 8, 2009 兩 vol. 106 兩 no. 36 兩 15259 –15263

CHEMISTRY

We demonstrate in this joint experimental and theoretical study how one can alter electron transport behavior of a single melamine molecule adsorbed on a Cu (100) surface by performing a sequence of elegantly devised and well-controlled single molecular chemical processes. It is found that with a dehydrogenation reaction, the melamine molecule becomes firmly bonded onto the Cu surface and acts as a normal conductor controlled by elastic electron tunneling. A current-induced hydrogen tautomerization process results in an asymmetric melamine tautomer, which in turn leads to a significant rectifying effect. Furthermore, by switching on inelastic multielectron scattering processes, mechanical oscillations of an N-H bond between two configurations of the asymmetric tautomer can be triggered with tuneable frequency. Collectively, this designed molecule exhibits rectifying and switching functions simultaneously over a wide range of external voltage.

Fig. 1. STM image and adsorption structure. (A) Topographic image of isolated melamine molecules adsorbed on Cu (100) (V ⫽ 0.2 V, I ⫽ 0.5 nA). Atomically resolved image of substrate is given in the insert (V ⫽ 5 mV and I ⫽ 20 nA). The scale for apparent high of the molecule is given. (B) Adsorption structure of single melamine molecule. Hydrogen atoms within the red dashed circle are dissociated when the molecule is adsorbed on Cu (100) at RT, resulting in the dehydrogenated melamine, whose top and side views, as well as simulated STM image, are shown in C–E, respectively.

completely reversible. When the scanning bias of ⫺0.8 V is used for imaging, a symmetric pattern appears, see molecule 1⵮ in the insert of Fig. 2B. This symmetric pattern of molecule 1⵮ arises from averaged image of two configurations because of fast switching rate under such a high voltage. It can be seen that the pattern for molecule 1⵮ is noticeable different from the image of intact molecule (see molecule 2 in the insert). The measured I-V curves for one of the molecules before and after modified by pulses are shown in Fig. 2C. It can be seen that the I-V curve of the modified molecule shows significant rectifying effect with a

typical ratio of 20 to 25 at 2 V. The operation with high voltage pulse has thus successfully turned the nonfunctional melamine molecule into a rectifier (see Figs. S3 and S4). To determine the structure of the modified melamine molecule, we have carried out systematic calculations for several possible adsorption configurations (see SI Text and Fig. S5). Two stable structures for the tautomer of dehydrogenated melamine have been located which are 0.71 eV and 0.99 eV above the dehydrogenated melamine, respectively. The two configurations of the tautomer, labeled as melamine-C1 and melamine-C2, are shown in Fig. 2F and H, which differ only by the mutual position of the N-H bonds. Melamine-C1 has lower energy with two N-H bonds pointing to opposite directions, while melamine-C2 is slightly higher in energy with two N-H bonds directing to the same side. The simulated STM images for melamine-C1 and melamine-C2 are given in Fig. 2G and I, demonstrating asymmetry patterns that agree well with the experimental observations for the modified molecules, 1⬘, 1⬙, and 3⬘ shown in Fig. 2 A and B. The calculated I-V curves for dehydrogenated melamine, melamine-C1 and melamine-C2, are shown in Fig. 2 J. The differences in absolute current between the calculated and the experimental ones could be attributed to two possible sources: the simple theoretical approximation adopted in the calculations and the slight uncertainty of the exact tip position in experiment. The dehydrogenated melamine behaves like a normal conductor, while remarkable rectifying effect is revealed for melamine-C2 and less pronounced for melamine-C1. The different conduction behavior can be related to the symmetries of the molecules. It is found that the dehydrogenated melamine molecule, differing from its counterpart in gas phase, has a planar geometry and possesses C2v symmetry with a symmetric ␲ conjugation, whose local density of states (LDOS) with respect to the atomically sharp tip is evenly distributed below and above the Fermi level.

Fig. 2. Rectifying and switching behavior of a modified single melamine molecule. (A and B) STM topographic images of the same area in Fig. 1 A (V ⫽ 0.2 V, I ⫽ 0.5 nA) in which molecules 1 and 3 are activated by applying a ⫹ 2.4 V pulse on each of them. The black dots marked on the images give the STM tip position at where all conductance measurements are taken. Insert, STM images acquired with a sample bias of ⫺0.8 V. The scale for apparent high of the molecule is given. (C) I-V Curves measured for Cu (black), melamine (red), and melamine-C (blue). Fluctuation features are shown in the inserts. (D and E) Top and side views, respectively, of the computational model for the melamine adsorbed on Cu (100). The dashed line represents the unit cell. Calculated STM images of the optimized structures of melamine-C1 (F) and melamine-C2 (H) are given in G and I, respectively (at V ⫽ ⫺0.6 V and height of 7 Å from metal substrate). (J) The calculated I-V curves for dehydrogenated melamine, melamine-C1 and melamine-C2 on Cu (100). 15260 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0903131106

Pan et al.

The tautomerization reduced the symmetry of the molecule and its ␲ system, resulting in the increase of molecular LDOS below the Fermi level and consequently the rectification. It is noted from the calculated I-V curves that the current flow under a fixed STM tip site (black point marked in Fig. 2 A) is different for melamine-C1 structure and for melamine-C2 structure. The fact that both structures can be observed in the experiments implies that it might be possible that these two structures represent low and high conductance states in a switch (see Movie S1). By inspecting the experimental I-V curves of these two configurations, additional fluctuation features at higher bias can indeed be observed, as clearly illustrated in the inserts of Fig. 2C. To examine the nature of these features, we measured the current over a period continuously under fixed bias. The results for two biases, ⫺0.6 V and ⫺0.9 V, are shown in Fig. 3A and B, respectively. In both cases, the measured current switches back and forth between high and low conductance states with different rates. The measured voltage and current dependent switching rate is summarized in Fig. 3C (see SI Text and Fig. S4). A power-law relationship between the rate R and the current I, R ⬀ IN, is obtained at a given bias. Such power-law dependence has been observed in the study of rotational motion of O2 molecule on the platinum (111) surface (32). It has been shown that the value N should correspond to the number of electrons needed to over the rotational barrier. We plotted the relationship between the number of electrons (N) and the external biases in Fig. 3D. Several important conclusions can be drawn from these results. We noted that the low and high conductance states do not follow the same trends. The number of electrons needed for the transition from low conductance state to high conductance state is always larger than that needed for the reverse transition at a given bias. Moreover, two-electron process becomes dominant below the voltage of 0.7 eV for the former and 0.9 eV for the latter. It can thus been estimated that the energy barriers from the high to the low conductance states and vice versa could be around 0.7 eV and 0.9 eV, respectively. It is noted that the estimated energy difference between the two conductance states is around 0.2 eV, in good agreement with the calculated energy difference of 0.28 eV for these two stable configurations of the tautomer. To verify these experimental Pan et al.

Fig. 4. Transition path for tautomerization and bond rotation processes. (A) Energy profile for tautomerization and rotation of N-H bond of dehydrogenated melamine molecule adsorbed on Cu (100) surface. (B) Schematic representation of rotation of N-H bond between the two stable configurations. (C) Calculated double well potential connected by melamine-C1 and melanmine-C2 structures and the key parameters that control the inelastic tunneling processes. P is the probability of inelastic tunneling process at external bias Ep, ␶ is the lifetime of the intermediate state at where the N-H bond has tendency to relax back to its origin. T is the transmission probability.

observations, we searched the possible transition paths for tautomerization and bond rotation processes (see SI Text and Fig. S7). The potential map is summarized in Fig. 4. The first transition state (TS1) for tautomerization is found to be located at 1.90 eV above the dehydrogenated melamine molecule, which explains why a pulse of 2.4 V can produce the asymmetric tautomer in the experiment. The second transition state (TS2) has been located in between two configurations of the tautomer, which is about 1.06 eV and 0.78 eV above melamine-C1 and melamine-C2, respectively. In this transition state, the N-H bond is placed along the centre of the molecule and points out of the molecular plane. The N-H bond length in TS2 remains unchanged. The most favourable motion of the N-H bond is thus an out-of-plane rotation. It should be noted that the reversible tautomerization process is always competing with the mechanical motion of the N-H bond. From energetic point of view, the former is much less favorable than the latter. In other words, the reversible tautomerization process does not contribute to the molecular switches observed here, which is completely different from what was reported in (18). On the other hands, the motion of N-H bond is very local and does not induce drastic conformation changes of the molecular framework, avoiding the complications involved in the well-studied cis-trans isomerization of azobenzene (23–25). Our new mechanical switch has thus combined the best of these two approaches. The calculated potential surface along the motion of N-H bond is given in Fig. 4C. The direct proton transfer probability is found to be very small and the possible involvement of elastic tunneling process can thus be ruled out. We have developed a statistic model (see SI Text) to describe the inelastic tunneling process based on the assumption that if the total energy absorbed by the molecule is higher than the energy barrier, the N-H bond can rotate over to reach another configuration with certain probability T. Each inelastic tunneling process is considered as a random event and controlled by two other key parameters as PNAS 兩 September 8, 2009 兩 vol. 106 兩 no. 36 兩 15261

CHEMISTRY

Fig. 3. Switching behavior with tuneable rates. (A and B) Current traces continuously measured at bias of ⫺0.6 V and ⫺0.9 V. (C) Switching rate, R, in the high conductance state as a function of tunneling current, I, under various sample biases. The solid lines are least-square fits to the data, which follow a power-law, R ⬀ IN. (D) Number of possible inelastic electrons under different bias. L and H refer to the low and high conductance states. The calculated results are presented in dashed-lines. The observed large difference between the positive and negative biases is discussed in SI Text, Figs. S4 –S6, and Table S1.

highlighted in Fig. 4C, namely the probability of inelastic tunneling events (P) at injection energy Ep, and the lifetime (␶) that the excited N-H bond is needed to spontaneously relax back to its origin. It is quite clear that multielectrons event can take place only if the molecule is able to absorb enough energy within the lifetime ␶. Our Monte Carlo simulations have shown that experimental values for the number of electrons, N, are statistically averaged values, resulting from the mixture of various N-electron events (see Fig. S6 and Table S1). For instance, at the external bias of ⫺0.6 V, for high conductance state the inelastic tunneling process is mainly a mixture of 80% two-electron and 18% three-electron processes, while for low conductance state, it is dominated by a combination of 48% two-electron and 40% three-electron processes. It should be noted that the existing static models, like the one used by Stipe et al. (33) could only lead to integer value for N. The number of electrons, N, under different biases simulated by our statistic model are in very good agreement with the experiments as shown in Fig. 3D. To conclude, we have demonstrated in this study how one can change electron transport properties of a non-functional molecule through design of molecular structure and control of electron tunneling processes. By using a single melamine molecule adsorbed on a Cu surface as a model system, we have shown that with the help of surface chemistry and STM manipulation, one can tune the function of the melamine molecule from a normal conducting molecule to a rectifier, and eventually to a switch with controllable rate. Materials and Methods Melamine (99%) was from Aldrich. We performed the experiments with a low-temperature scanning tunneling microscope (STM) (Omicron) operating under a base pressure of 3 ⫻ 10⫺11 mbar. The Cu (100) sample was cleaned by Ar⫹ ion sputtering and annealing cycles. Melamine molecules were thermally evaporated (by heating to ⬇90 °C) onto a Cu (100) substrate held at room temperature to allow melamine molecules to be chemisorbed on the Cu

1. Joachim C, Gimzewski JK, Aviram A (2000) Electronics using hybrid-molecular and mono-molecular devices. Nature 408:541–548. 2. Park J, et al. (2002) Coulomb blockade and the Kondo effect in single-atom transistors. Nature 417:722–725. 3. Liang WJ, Shores MP, Bockrath M, Long JR, Park H (2002) Kondo resonance in a single-molecule transistor. Nature 417:725–729. 4. Zhao AD, et al. (2005) Controlling the Kondo effect of an adsorbed magnetic ion through its chemical bonding. Science 309:1542–1544. 5. Tao NJ (2006) Electron transport in molecular junctions. Nature Nanotech 1:173–181. 6. Gaudioso J, Lauhon LJ, Ho W (2000) Vibrationally mediated negative differential resistance in a single molecule. Phys Rev Lett 85:1918 –1921. 7. Ashwell GJ, Tyrrell WD, Whittam AJ (2004) Molecular rectification: Self-assembled monolayers in which donor-(␲-bridge)-acceptor moieties are centrally located and symmetrically coupled to both gold electrodes. J Am Chem Soc 126:7102–7110. 8. Donhauser ZJ, et al. (2001) Conductance switching in single molecules through conformational changes. Science 292:2303–2307. 9. Moresco F, et al. (2001) Conformational changes of single molecules induced by scanning tunneling microscopy manipulation: A route to molecular switching. Phys Rev Lett 86:672– 675. 10. Blum AS, et al. (2005) Molecularly inherent voltage-controlled conductance switching. Nat Mater 4:167–172. 11. Lastapis M, et al. (2005) Picometer-scale electronic control of molecular dynamics inside a single molecule. Science 308:1000 –1003. 12. Sainoo Y, et al. (2005) Excitation of molecular vibrational modes with inelastic scanning tunneling microscopy processes: Examination through action spectra of cis-2-butene on Pd(110). Phys Rev Lett 95:246102. 13. Chen J, Reed MA, Rawlett AM, Tour JM (1999) Large on-off ratios and negative differential resistance in a molecular electronic device. Science 286:1550 –1552. 14. Zeng CG, Wang HQ, Wang B, Yang JL, Hou JG (2000) Negative differential-resistance device involving two C60 molecules. Appl Phy Lett 77:3595–3597. 15. Grobis M, Wachowiak A, Yamachika R, Crommie MF (2005) Tuning negative differential resistance in a molecular film. Appl Phy Lett 86:204102. 16. Zhao J, et al. (2005) Single C59N molecule as a molecular rectifier. Phys Rev Lett 95:045502. 17. Eigler DM, Lutz CP, Rudge WE (1991) An atomic switch realized with the scanning tunneling microscope. Nature 352:600 – 603. 18. Liljeroth P, Repp J, Meyer G (2007) Current-induced hydrogen tautomerization and conductance switching of naphthalocyanine molecules. Science 317:1203–1206.

15262 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0903131106

surface via dehydrogenation. The sample was then transferred into the STM cryostat which was precooled to 5 K for measurements. All simulations were performed with density functional theory (DFT) using Vienna Ab initio Simulation Package code (34, 35). The projector augmented wave (PAW) potentials were used to represent the interaction between ions and electrons (36). Exchange correlation was described by the generalized gradient approximation of Perdew, Burke, and Ernzerhof (PBE-GGA) (37). The surface was modeled by periodically repeated slabs consisting of four Cu layers and the adsorbed molecule. Slabs were separated by a 14 Å vacuum in between. The wave functions were expanded in the plane-wave basis up to a kinetic energy of 400 eV. The experimentally determined Cu lattice constant of 3.61 Å with a p(3 ⫻ 6) surface periodicity was used. Gamma-centered grids of (4 ⫻ 2 ⫻ 1) and (6 ⫻ 3 ⫻ 1) k points were used for geometry optimizations and electronic structure calculations, respectively. During the optimizations, the uppermost two copper layers as well as the adsorbed molecule were allowed to relax till atom forces below 0.02 eV/Å. Vibration frequencies were calculated from the Hessian matrix determined by means of finite differences for which all copper atoms were fixed. The nudged elastic band method was used to calculate energy barrier along the minimal energy pathway, (38) and the transition states were identified by vibration frequency analysis. Theoretical STM images were simulated by Tersoff-Hamann formula (39), that is, integrating spatially resolved density of states (DOS) in energy from a bias potential to the Fermi level. The I-V curves were also calculated with Tersoff-Hamann approximation. The conductance (current) for a given sample bias was obtained by integrating the DOS projected at the tip site in the energy range from the corresponding bias to the Fermi level. Within the energy range from ⫺2.0 V to 2.0 V, 44 sample biases were selected in the calculations of current and the I-V curve was then fitted according to these 44 values for the current. In the current simulations, the tip site derivates 1.07 Å from the center of the adsorbed molecule, as marked by black dots in Fig. 2 A and B. The actual physical size of the STM tip is considered by including local DOS within a 0.38 Å radius round the tip site. ACKNOWLEDGMENTS. This research was supported by National Basic Research Program of China (2006CB922000), National Natural Science Foundation of China (50721091, 50532040, 20533030, 10825415), Supercomputer Centers of USTC-HP, Chinese Academy of Sciences, and Shanghai. Y.L. thanks the support from Swedish Research Council.

19. Moore AM, et al. (2006) Molecular engineering and measurements to test hypothesized mechanisms in single molecule conductance switching. J Am Chem Soc 128:1959 – 1967. 20. Lewis PA, et al. (2005) Molecular engineering of the polarity and interactions of molecular electronic switches. J Am Chem Soc 127:17421–17426. 21. Moore AM, et al. (2007) Real-time measurements of conductance switching and motion of single oligo(phenylene ethynylene) molecules. J Am Chem Soc 129:10352– 10353. 22. Kumar AS, et al. (2008) Reversible photo-switching of single azobenzene molecules in controlled nanoscale environments. Nano Lett 8:1644 –1648. 23. Alemani M, et al. (2006) Electric field-induced isomerization of azobenzene by STM. J Am Chem Soc 128:14446 –14447. 24. Choi B-Y, et al. (2006) Conformational molecular switch of the azobenzene molecule: A scanning tunneling microscopy study. Phys Rev Lett 96:156106 –156104. 25. Alemani M, et al. (2008) Adsorption and switching properties of azobenzene derivatives on different noble metal surfaces: Au(111), Cu(111), and Au(100). J Phys Chem C 112:10509 –10514. 26. Katano S, et al. (2007) Reversible control of hydrogenation of a single molecule. Science 316:1883–1886. 27. Willner I, Willner B, (2001) Biomaterials integrated with electronic elements: En route to bioelectronics. Trends Biotechnol 19:222–230. 28. Granstrom M, et al. (1998) Laminated fabrication of polymeric photovoltaic diodes. Nature 395:257–260. 29. Kim Y, et al. (2006) A strong regioregularity effect in self-organizing conjugated polymer films and high-efficiency polythiophene: Fullerene solar cells. Nature mater 5:197–203. 30. Ju SS, Han CC, Wu CJ, Mebel AM, Chen YT (1999) The fragmentation of melamine: A study via electron-impact ionization, laser-desorption ionization, collision-induced dissociation, and density functional calculations of potential energy surface. J Phys Chem B 103:582–596. 31. Jurgens B, et al. (2003) Melem (2,5,8-triamino-tri-s-triazine), an important intermediate during condensation of melamine rings to graphitic carbon nitride: Synthesis, structure determination by X-ray powder diffractometry, solid-state NMR, and theoretical studies. J Am Chem Soc 125:10288 –10300. 32. Stipe BC, Rezaei MA, Ho W (1998) Inducing and viewing the rotational motion of a single molecule. Science 279:1907–1909. 33. Stipe BC, et al. (1997) Single-molecule dissociation by tunneling electrons. Phys Rev Lett 78:4410 – 4413.

Pan et al.

37. Perdew JP, Burke K, Ernzerhof M (1996) Generalized gradient approximation made simple. Phys Rev Lett 77:3865–3868. 38. Henkelman G, Uberuaga BP, Jonsson H (2000) A climbing image nudged elastic band method for finding saddle points and minimum energy paths. J Chem Phys 113:9901–9904. 39. Tersoff J, Hamann DR (1985) Theory of the scanning tunneling microscope. Phys Rev B 31:805– 813.

CHEMISTRY

34. Kresse G, Furthmuller J (1996) Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comput Mater Sci 6:15–50. 35. Kresse G, Furthmuller J (1996) Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys Rev B 54:11169 –11186. 36. Kresse G, Joubert D (1999) From ultrasoft pseudopotentials to the projector augmented-wave method. Phys Rev B 59:1758 –1775.

Pan et al.

PNAS 兩 September 8, 2009 兩 vol. 106 兩 no. 36 兩 15263