LETTER pubs.acs.

org/NanoLett

Molecular Conduction through Adlayers: Cooperative Effects Can Help or Hamper Electron Transport
Matthew G. Reuter,*,, Tamar Seideman, and Mark A. Ratner

Department of Chemistry, Northwestern University, Evanston, Illinois 60208-3113, United States

S b Supporting Information

ABSTRACT: We use a one-electron, tight-binding model of a molecular adlayer sandwiched between two metal electrodes to explore how cooperative eects between molecular wires inuence electron transport through the adlayer. When compared to an isolated molecular wire, an adlayer exhibits cooperative eects that generally enhance conduction away from an isolated wires resonance and diminish conductance near such a resonance. We also nd that the interwire distance (related to the adlayer density) is a key quantity. Substrate-mediated coupling induces most of the cooperative eects in dense adlayers, whereas direct, interwire coupling (if present) dominates in sparser adlayers. In this manner, cooperative eects through dense adlayers cannot be removed, suggesting an optimal adlayer density for maximizing conduction. KEYWORDS: Electron transport, cooperative eects, adlayers, tight-binding models

he ow of electric current through molecules (electron transport) and the transfer of charge between molecules and substrates are interesting problems for both their inherent conceptual issues and their applications to solar cells, thermoelectrics, scanning probe microscopies, and molecular electronics.14 Even though electron transport through a single molecular wire is reasonably well understood,3,4 cooperative eects5 between molecular wires complicate the scaling of these results to multiple molecular wires in parallel.622 Using Ohms law,23 conventional electronics predicts that the conductance through N wires is simply N times an isolated wires conductance

GN NG1

however, previous studies have found that molecular wires obey a pseudo-Ohms law7 GN NGeff 1 2

where Ge is an eective single wire conductance that should be 1 independent of N. In many cases, cooperative eects cause Ge > G1,68,12,14,18,21,22 1 although cases where Ge < G1 have been reported.8,1113,18,22 1 Besides our fundamental interest in how cooperative eects change G1 into Ge, this phenomenon has practical implications. For 1 instance, analyses of certain experimental setups, including several mechanical break junctions and scanning probe microscope-based junctions, often assume that only a single molecule connects the electrodes when electron transport is measured.4 In actuality, there could be multiple molecules, and understanding when cooperative eects increase (decrease) the conductance per wire may aid in the interpretation of these experiments. Moreover, cooperative eects should be visible in conductance histograms (methods for
r 2011 American Chemical Society

statistically analyzing experimental data),24 and the ability to infer more about the underlying physics of conduction from existing data would be an exciting advance. Computational studies using ab initio methods (usually within the density functional theory) have been used to investigate such changes in per-wire conductances,7,8,14,15,17,21 nding that cooperative eects (i) arise from direct interwire coupling and substrate-mediated coupling8,14,15 and (ii) can cause signicant electrostatic eects at the molecule-electrode interfaces.8,17 Despite this eort, ab initio methods exhibit several problems that have confounded the development of a set of guidelines for interpreting the underlying physics of cooperative eects in electron transport. First, ab initio methods convolute many eects together (including any cooperative or electrostatic eects), making it dicult to attribute changes in conductance to a specic source. Second, numerous approximations are required to calculate electron transport with ab initio methods.2 Alternatively, simple model systems (e.g., tight-binding models) can provide qualitative, mechanistic insights and, in some cases, exact, analytical results at the expense of quantitative accuracy. We recently investigated this change in conductance per wire using tight-binding model systems with two molecular wires,22 generally nding that Ge > G1 when the Fermi level is far from an 1 isolated wires resonance and Ge < G1 when the Fermi level is 1 near such a resonance. The combination of direct, interwire coupling (e.g., interactions) and substrate-mediated coupling, each a source of cooperative eects, causes this behavior by energetically shifting the systems conduction channels away
Received: July 10, 2011 Revised: September 19, 2011 Published: October 18, 2011
4693
dx.doi.org/10.1021/nl202342a | Nano Lett. 2011, 11, 46934696

Nano Letters

LETTER

Figure 1. (a) Cartoon depiction of an example adlayer. (b) System schematic of the adlayers modeled in this work. Each molecular wire (maroon circle) sits atop an electrode site (light orange circle), and is separated from its four nearest neighbors by (m2 + n2)1/2 electrode sites.

from those of an isolated wire. Separating the molecular wires reduces the impact of cooperative eects on conduction; however, an interference eect between the two couplings allows unconventional methods to control cooperative eects. Specically, balancing the direct and substrate-mediated couplings removes cooperative eects in systems with two wires. In this Letter, we extend our analysis to systems of uniformly dense adlayers, seeking to establish the relationship between G1, Ge, and the Fermi level as the number of parallel molecular wires 1 increases from two to innity. This generalization is important for several reasons. First, eq 2 is approximate and is more accurate when N is large.18 Second, adlayers are generally easier to characterize and have applications in sensing, logic, and computation.2427 As before,22 we use a one-electron, tight-binding model within the LandauerImry (coherent scattering) formalism,28 yielding exact, semianalytical (i.e., analytical up to numerical quadrature) results. After introducing our model, we calculate electron transport properties through adlayers of various densities. In addition to conrming that Ge is usually larger than G1 far o 1 resonance (and less than G1 near resonance), we nd that in certain cases there is an optimal adlayer density for maximizing electron transport through the adlayer. We model each electrode by a semi-innite cubic lattice of single-state sites, where each site has energy and couples to its nearest neighbors with element Ve.18,22,29,30 Likewise, each molecular wire is represented by a single state with energy ; this state is conceivably the highest occupied or lowest unoccupied molecular orbital. We construct a uniformly dense adlayer by placing wires in a regular grid, as depicted in Figure 1. Each wire is m electrode sites in one direction and n in the other from its four nearest neighbors; such an adlayer is denoted (m,n) and has density (m2 + n2)1. For simplicity, we limit direct, interwire coupling to nearest neighbors, and assume that this direct coupling, , is real. The wire Hamiltonian, Hwire, is explicitly stated in Section 1 of the Supporting Information. Next, each wire sits atop a single electrode site (as shown in Figure 1), and couples to this site with element V. The electrode-wire coupling operator, V, is also discussed in Section 1 of the Supporting Information. Finally, we appeal to the detailed discussion of these parameters in Section 5 of ref 18 for choosing values representative of realistic systems. Specically, Ve = 0.82 eV, = 0 eV (centering the metal band around E = 0), and V = 0.45 eV. Due to our focused interest in electron transport through the adlayer of molecular wires, we write an eective Hamiltonian for the wires, H f Hwire + 2(E), where (E) is the self-energy of coupling the wires to an electrode1 and the factor of 2 reects

Figure 2. Transmission (per wire) spectra for molecular wire adlayers of various densities. The density of adlayer (m,n) is (m2 + n2)1. In the absence of direct, interwire coupling ( = 0), cooperative eects dramatically change the transmission spectra of dense adlayers when compared to the transmission through an isolated wire (top rows). Since the substrate-mediated coupling decays as the density decreases, the transmission spectra approach that of an isolated wire as m and n increase, as expected. While the addition of direct coupling causes small changes in the transmission through dense adlayers, it seems unable to remove cooperative eects, unlike systems with two wires.22 Direct coupling tends to amplify cooperative eects in sparser adlayers.

identical coupling of the wires to both electrodes (for simplicity). Formally,1 (E) = VGelec(E)V, where Gelec(E) is the (retarded) Green function of an isolated electrode, which is semianalytical for these electrodes.30 Within a coherent scattering formalism, the zero-bias conductance per wire at zero temperature is G0 2e2 TEF h 3

where EF is the Fermi level, and after exploiting identical coupling to both electrodes the transmission (per wire) through the adlayer is TE TrEGEEG E 4 For use in eq 4, (E) is the spectral density, (E) = i[(E) (E)], and G(E) is the Green function of the adlayer connected to both electrodes, GE EI Hwire 2E1 5

where we assume the wire states are orthonormalized. While omitted for clarity, our model produces analytical or semianalytical forms for all of these quantities, which are derived and discussed in Sections 2 and 3 of the Supporting Information. We begin by examining the transmission through an isolated wire, that is, = 0 and m, n f , which provides a basis for comparison. This transmission, denoted Tiso(E), is depicted in Figure 2. As is typical of conduction through a single molecule,3 Tiso(E) is quasi-Lorentzian, where the sole maximum corresponds to the wires resonance. This resonance signies that the
4694
dx.doi.org/10.1021/nl202342a |Nano Lett. 2011, 11, 46934696

Nano Letters

LETTER

Figure 3. The conductance (per wire) through an adlayer relative to that through an isolated wire as a function of EF, each wires site energy (), and the direct, interwire coupling (). The solid black curves trace a ratio of one; cooperative eects do not aect conductance in these cases. At the isolated wires resonance (E1 , dashed black lines), cooperative R eects never increase conductance over the isolated wire. Moreover, only suciently sparse adlayers attain the isolated wire conductance near resonance. O resonance, cooperative eects tend to enhance conductance by increasing the density of wire states at these Fermi levels.18 Note that E1 6 , in general, due to molecular level shifting induced by R the electrodes; see Section 2 of the Supporting Information for details.

Figure 4. The eects of interwire distance, (m2 + n2)1/2, on transmission/conductance (per wire) for various energies. (a) The amount of direct coupling that maximizes transmission, max, for systems with two wires.22 (b) max for adlayers. (c) The corresponding maximum transmission (per wire) for adlayers. In both cases, max f 0 as the distance increases, as expected. Unlike systems with two wires, which always have maximum transmissions (per wire) of one22 (indicating the net cancellation of cooperative eects), cooperative eects cannot be negated in dense adlayers. This suggests an optimal adlayer density for maximizing conduction through adlayers.

conduction channel is completely open [Tiso(E1 ) = 1], and R occurs at the energy E1 , which is shifted from by the coupling R to the electrodes [(E1 ), see eq 5]. R Figure 2 also shows the transmission spectra through adlayers of various densities. Dense adlayers (those with smaller m and n, see Figure 1) exhibit substantially dierent transmission spectra from that of the isolated wire. Transmission decreases near the isolated wires resonance and increases away from this resonance. Moreover, substrate-mediated coupling induces most of the cooperative eects in dense adlayers, as evidenced by the fact that variations in the direct coupling produce only modest changes in the transmission. Sparser adlayers [here (0,2) and less dense] largely mimic isolated wires in the absence of direct coupling, conrming that substrate-mediated coupling decays as the distance increases.7,14,22,30 In these adlayers, direct coupling is the primary source of cooperative eects, when present (interwire coupling is also expected to decay with the interwire distance). In this manner, cooperative eects diminish as the interwire distance increases, yielding a true Ohms law relationship [eq 1]. Similar to their role in systems with two wires,22 cooperative eects in adlayers tend to decrease transmission near the isolated wires resonance (E1 ), while increasing it away from E1 . The R R isolated wires conduction channel is completely open at resonance, and cooperative eects cannot improve the transmission. Conversely, cooperative eects lead to a band of wire states (generalizing the energetic splitting seen in two wires22), and the availability of more wire states in the adlayer increases transmission o resonance.18 Translating to conductance with eq 3, these results conrm the generalization of our ndings from systems with two wires to adlayers. As displayed in Figure 3, Ge > G1 o 1 resonance and Ge < G1 near resonance, in most cases. 1 Unlike for systems with two wires, the adlayer density, which xes the interwire distance, becomes an important factor in electron transport through adlayers. As we previously showed,22

cooperative eects between two wires could always be negated at an arbitrary energy, regardless of interwire distance, by balancing the direct and substrate-mediated couplings. This eect is depicted in Figure 4a, where we plot the value of direct coupling required to remove cooperative eects, max, as a function of the interwire distance for systems with two wires. Since substratemediated coupling decays with distance, max approaches 0 with increasing distance, and in this sense the convergence of max toward 0 estimates the persistence of cooperative eects. At the isolated wires resonance (E1 ), max can also be interR preted as the amount of direct coupling that maximizes transmission/ conductance (per wire) through the wires. In systems with two wires, this maximum transmission (per wire) is always one.22 A close examination of Figures 2 and 3, though, suggests that dense adlayers do not realize transmissions (per wire) of one at E1 , R partly resulting in our general conclusion that Ge < G1 near 1 resonance. Investigating further, Figure 4b plots max for adlayers and Figure 4c shows each adlayers corresponding maximum transmission. Akin to systems with two wires, adlayers exhibit larger |max| when the wires are close; |max| decays to zero as the wires separate. On the other hand, the maximum transmission (per wire) through an adlayer is only one after the wires are suciently separated. This result seems unintuitive at rst but is easily explained. In two-wire systems, the direct coupling (max) balances the substrate-mediated coupling, a quantity that strongly depends on the interwire distance. With only one pair of wires, the choice of max is optimal, and high transmissions are realized. In adlayers, conversely, we cannot nd a max that simultaneously balances the substrate-mediated couplings to all other wires unless the substrate-mediated couplings are uniform. This only occurs when the adlayer is suciently sparse, that is, all substratemediated couplings are zero. In the context of maximizing conductance through an adlayer, this suggests an optimal adlayer density. When EF E1 , we put as many molecules as possible in R
4695
dx.doi.org/10.1021/nl202342a |Nano Lett. 2011, 11, 46934696

Nano Letters the adlayer (maximizing the number of conduction channels), but require them to be suciently separated to minimize cooperative eects. Away from resonance (EF 6 E1 ), cooperative R eects tend to increase conductance, and the optimal adlayer density is instead limited by the adsorption chemistry (the number of molecules in the adlayer). In summary, we have presented a semianalytical formulation of electron transport through parallel molecular wires in uniformly dense adlayers, exploring the role of cooperative eects (i.e., direct, interwire coupling and substrate-mediated coupling) on transmission and conductance. Reminiscent of inelastic eects,31,32 cooperative eects generally increase conductance (Ge > G1) 1 when the Fermi level is far from an isolated wires resonance and decrease it (Ge < G1) near such a resonance. These results help 1 explain previous observations regarding changes in the conductance per wire as the number of wires varies.616,1821 Although our model is of single-electron type and thus neglects interelectronic interactions, it has the signicant advantage of providing exact answers, much like the Newns self-energy or the Ising model. The eects of electronic repulsion can, to some extent, be introduced into one-electron models by reparameterization; however, true multielectron eects, such as correlation, will be absent from our results. For several reasons, the interwire distance, or equivalently, the adlayer density, is a key property. First, substrate-mediated coupling dominates cooperative eects in dense adlayers, whereas direct coupling (if present) induces most cooperative eects in sparser adlayers. Second, only in the absence of substratemediated coupling (sparser adlayers) can wires in an adlayer behave like isolated wires. Such net negation of cooperative eects requires the simultaneous balance of direct and substratemediated couplings between all wire pairs, which is only possible when the substrate-mediated coupling is uniform (and thus zero). Taken collectively, these results suggest optimal adlayer densities for maximizing current through adlayers. Near an isolated wires resonance the wires need to be suciently separated to eliminate substrate-mediated coupling, but away from such a resonance cooperative eects usually increase conductance, favoring denser adlayers.

LETTER

for a fellowship. We thank the NSF (Grant CHE-1012207) and the MRSEC program of the NSF (DMR-0520513) for support.

REFERENCES
(1) Datta, S. Electronic Transport in Mesoscopic Systems; Cambridge University Press: Cambridge, 1995. (2) Cuevas, J. C.; Scheer, E. Molecular Electronics; World Scientic: Hackensack, NJ, 2010. (3) Nitzan, A. Annu. Rev. Phys. Chem. 2001, 52, 681. (4) Chen, F.; Hihath, J.; Huang, Z.; Li, X.; Tao, N. J. Annu. Rev. Phys. Chem. 2007, 58, 535. (5) Naaman, R.; Vager, Z. MRS Bull. 2010, 35, 429. (6) Yaliraki, S. N.; Ratner, M. A. J. Chem. Phys. 1998, 109, 5036. (7) Magoga, M.; Joachim, C. Phys. Rev. B 1999, 59, 16011. (8) Lang, N. D.; Avouris, P. Phys. Rev. B 2000, 62, 7325. (9) Salomon, A.; Cahen, D.; Lindsay, S.; Tomfohr, J.; Engelkes, V. B.; Frisbie, C. D. Adv. Mater. 2003, 15, 1881. (10) Kushmerick, J. G.; Naciri, J.; Yang, J. C.; Shashidhar, R. Nano Lett. 2003, 3, 897. (11) Tomfohr, J.; Sankey, O. F. J. Chem. Phys. 2004, 120, 1542. (12) Lagerqvist, J.; Chen, Y.-C.; Di Ventra, M. Nanotechnology 2004, 15, S459. (13) Selzer, Y.; Cai, L.; Cabassi, M. A.; Yao, Y.; Tour, J. M.; Mayer, T. S.; Allara, D. L. Nano Lett. 2005, 5, 61. (14) Liu, R.; Ke, S.-H.; Baranger, H. U.; Yang, W. J. Chem. Phys. 2005, 122, 044703. (15) Geng, H.; Yin, S.; Chen, K.-Q.; Shuai, Z. J. Phys. Chem. B 2005, 109, 12304. (16) Long, M.-Q.; Wang, L.; Chen, K.-Q.; Li, X.-F.; Zou, B. S.; Shuai, Z. Phys. Lett. A 2007, 365, 489. (17) Wang, J.; Prodan, E.; Car, R.; Selloni, A. Phys. Rev. B 2008, 77, 245443. (18) Landau, A.; Kronik, L.; Nitzan, A. J. Comput. Theor. Nanosci. 2008, 5, 535. (19) Landau, A.; Nitzan, A.; Kronik, L. J. Phys. Chem. A 2009, 113, 7451. (20) Lin, L.-L.; Leng, J.-C.; Song, X.-N.; Li, Z.-L.; Luo, Y.; Wang, C.-K. J. Phys. Chem. C 2009, 113, 14474. (21) Wang, L.; Guo, Y.; Zhu, C.; Tian, C.; Song, X.; Ding, B. Phys. Lett. A 2010, 374, 778. (22) Reuter, M. G.; Solomon, G. C.; Hansen, T.; Seideman, T.; Ratner, M. A. J. Phys. Chem. Lett. 2011, 2, 16671671. (23) Halliday, D.; Resnick, R.; Walker, J. Fundamentals of Physics, 7th ed.; John Wiley: Hoboken, NJ, 2005. (24) Song, H.; Lee, T.; Choi, N.-J.; Lee, H. Appl. Phys. Lett. 2007, 91, 253116. (25) Chaki, N. K.; Vijayamohanan, K. Biosens. Bioelect. 2002, 17, 1. (26) Mativetsky, J. M.; Pace, G.; Elbing, M.; Rampi, M. A.; Mayor, M.; Samor P. J. Am. Chem. Soc. 2008, 130, 9192. i, (27) Browne, W. R.; Feringa, B. L. Annu. Rev. Phys. Chem. 2009, 60, 407. (28) Imry, Y.; Landauer, R. Rev. Mod. Phys. 1999, 71, S306. (29) Kalkstein, D.; Soven, P. Surf. Sci. 1971, 26, 85. (30) Reuter, M. G. J. Chem. Phys. 2010, 133, 034703. (31) Galperin, M.; Ratner, M. A.; Nitzan, A. J. Chem. Phys. 2004, 121, 11965. (32) Paulsson, M.; Frederiksen, T.; Ueba, H.; Lorente, N.; Brandbyge, M. Phys. Rev. Lett. 2008, 100, 226604.

ASSOCIATED CONTENT
S b

Supporting Information. A thorough description of our model, including derivations for all of the analytical and semianalytical results, is presented in the Supporting Information. This material is available free of charge via the Internet at http:// pubs.acs.org.

AUTHOR INFORMATION
Corresponding Author

*E-mail: mgreuter@u.northwestern.edu.
Present Addresses

Computer Science and Mathematics Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, U.S.A.

ACKNOWLEDGMENT We are grateful to Abraham Nitzan, Gemma C. Solomon, Thorsten Hansen, and Lisa A. Fredin for helpful conversations. M.G.R. thanks the DoE CSGF (Grant DE-FG02-97ER25308)
4696
dx.doi.org/10.1021/nl202342a |Nano Lett. 2011, 11, 46934696