nature nanotechnology | VOL 1 | DECEMBER 2006 | www.nature.com/naturenanotechnology 173 REVIEW ARTICLE Electron transport in molecular junctions Building an electronic device using individual molecules is one of the ultimate goals in nanotechnology. To achieve this it will be necessary to measure, control and understand electron transport through molecules attached to electrodes. Substantial progress has been made over the past decade and we present here an overview of some of the recent advances. Topics covered include molecular wires, two- terminal switches and diodes, three-terminal transistor-like devices and hybrid devices that use various different signals (light, magnetic fi elds, and chemical and mechanical signals) to control electron transport in molecules. We also discuss further issues, including molecule���electrode contacts, local heating- and current-induced instabilities, stochastic fl uctuations and the development of characterization tools. N. J. TAO Department of Electrical Engineering Arizona State University, Tempe, Arizona 85287 e-mail: nongjian.tao@asu.edu Although molecular electronics has been proposed as an alternative to silicon in post-CMOS devices, molecules with unique functions may have applications that are complementary to the silicon- based microelectronics. To date, many molecules with wonderful electronic properties have been identifi ed and more with desired properties are being synthesized in chemistry labs. In addition to electronic properties, many molecules possess rich optical, magnetic, thermoelectric, electromechanical and molecular recognition properties, which may lead to new devices that are not possible using conventional materials or approaches (Fig. 1). Despite there being many unsolved issues in molecular electronics, important and solid advances have been made over the past decade. Th ese advances include: (1) demonstration of simple molecular device functions (2) development of many diff erent experimental approaches for measuring electron transport in single molecules and theoretical methods for describing the electron transport properties (3) emergence of new characterization techniques that help to bridge theories and experiments (4) hybrid devices, such as molecular sensors, which have been actively pursued parallel to the eff orts in pure electronic devices. Th is paper will provide an overview of some of these advances and discuss remaining issues that must be solved in order to reach the dream of functional devices. Molecular electronics is a diverse and rapidly growing fi eld. Having limited space and references, we have inevitably overlooked important work in preparing this review. Fortunately, there are many excellent reviews covering various aspects of molecular electronics, which will amend these defi ciencies (see, for example, reviews published within the past three years1���3). MOLECULAR WIRES A wire-shaped molecule that can effi ciently transport charge has been actively pursued as it may provide interconnections for molecular devices. Th e interest can also be traced back to its close relevance to charge transfer in biological systems, such as photosynthesis and DNA, as well as to conducting polymers. Such a molecule is oft en referred to as a molecular wire, although the defi nition is still a subject of debate4. To date, many wire-shaped molecules have been studied, which can be loosely divided into two categories: saturated and conjugated chains. SATURATED CHAINS A well-studied molecular system is the alkane consisting of saturated C���C bonds terminated by linkers that can bind to electrodes5���11. Th ese molecules have large gaps between their highest occupied molecular orbitals (HOMO) and lowest unoccupied molecular orbitals (LUMO) and are considered to be poorly conducting wires. However, they serve as a model system for testing both experimental techniques and theoretical calculations. Experiments have found that the conductance (G) of alkanes decreases exponentially with molecular length (L), and can be described by G = Aexp(�����L), where A is a constant and �� is a decay constant varying between ~0.7���0.9 �����1 (Fig. 2a). Similar decay has also been found in peptide chains12. Th e exponential decay, together with characteristic current���voltage (I���V) curves and temperature independence7,13, suggests electron tunnelling as the conduction mechanism for these molecules. Although the exponential decay of the conductance with length is similar in most experiments, comparison of the absolute conductance values of diff erent experiments is not straightforward. Th is is partly because some experiments measure many molecules and others measure single molecules. Even in the case of single-molecule measurements, there may be two or more conductance values for the same molecule resulting from diff erent molecule���electrode contact geometries13���17. CONJUGATED MOLECULES Because of the rapid decrease in tunnelling rate with distance, in most experiments the conductance is too small to be measured for alkane chains longer than 2���3 nm. For long-distance charge transport, conjugated molecules with alternating double and single bonds or delocalized �� electrons are better candidates. Th ese molecules have much smaller HOMO���LUMO gaps compared to alkanes and should therefore transport charge more effi ciently. However, for short conjugated molecules, off -resonance tunnelling is still believed to be the conduction mechanism. With increasing Nature Publishing Group ��2006

REVIEW ARTICLE 174 nature nanotechnology | VOL 1 | DECEMBER 2006 | www.nature.com/naturenanotechnology length, it is thought that tunnelling is replaced by hopping in which charges hop from one site to the next along the molecule. Hopping is thermally activated and has weaker length-dependence than the tunnelling process4. Conjugated molecules have been studied by measuring the charge transfer rate through the molecules bridged between an acceptor and a donor group4, or between a redox group and a bulk electrode18. Th ese experiments have established that: (1) charge transport can occur over a much greater distance in conjugated molecules than in alkanes (2) a crossover from a tunnelling regime to a hopping regime occurs when the length of the molecule is increased. Direct conductance measurements of conjugated molecules have been reported using diff erent approaches, and interesting phenomena have been observed19���22, although systematic studies into the length dependence of the conductance of single conjugated molecules bridged between two electrodes are relatively rare. One such measurement has been made in carotenoid polyenes with diff erent lengths using a combined break junction and statistical analysis approach (Fig. 2b)23. Th e measured �� is 0.22 �� 0.04 �����1, which is in close agreement with the value obtained from fi rst- principles simulations (0.22 �� 0.01 �����1) (Fig. 2a). Th e small value of �� demonstrates that conductance drops off slowly with chain length, confi rming that carotenoid conjugated chains are relatively good molecular ���wires���. Charge transport in oligothiophenes with three and four repeating units have been measured, and the longer molecule was found to be more conductive than the shorter one24. Th is unusual length-dependent conductance was attributed to a smaller HOMO���LUMO gap and a closer position of the HOMO to the Fermi levels of the probing electrodes for the longer molecule. Th is conclusion is supported by the ultraviolet���visible absorption, electrochemical measurements, and also by the dependence of the conductance on the electrochemical gate, which shift s the HOMO relative to the Fermi levels. Compared to charge transfer rate measurements, direct conductance measurements are still at an early stage. For example, charge transfer rate experiments have established the important role of the electron coupling to the nuclear coordinates of the molecules and the surrounding solvent molecules25. Th e rate measurements have also revealed the crossover from tunnelling to hopping26. Th ese eff ects have yet to be studied and confi rmed in the direct single- molecule conductance measurements. Finally it is important to note that charge transfer rate and direct conductance methods measure relevant but diff erent quantities. Direct conductance measurement of a molecule requires the molecule to bridge two electrodes that are connected to a power source, forming a continuous circuit. Although the two electrodes act like a donor and acceptor, the energy levels of the electrodes are continuous bands, whereas the energy levels of the donor and acceptors in the charge transfer measurements are discrete. A relationship between charge transfer rate and conductance has been worked out by Niztan27, which requires certain parameters such as the coupling strength between the molecules and the two electrodes to be known. TWO-TERMINAL DEVICES Device applications demand molecules to work not only like a wire but also an active element that can perform one or a set of controllable functions. Silicon-based electronics relies critically on three-terminal devices, such as transistors. However, owing to the diffi culty of placing a third gate electrode close to a molecule, most studies in molecular electronics to date have focused on two-terminal devices. Nevertheless, two-terminal devices can exhibit interesting and important behaviour that is useful for applications. Examples include rectifi cation28, negative diff erential resistance29 and conductance switching eff ects30. MOLECULAR DIODES A diode or rectifi er is an important component in electronics that allows an electric current to fl ow in one direction, but blocks it in the opposite direction. Inspired by the visionary work by Aviram and Ratner28, building diodes using single molecules has been pursued by many groups31. Th e basic structure of early molecular diodes consisted of a donor and an acceptor separated by a �� bridge, with �� being some saturated covalent bond linking the donor and acceptor and providing a tunnelling barrier between them. In this donor��������acceptor structure, the diode behaviour was expected to occur as a consequence of diff erent thresholds at positive and negative bias voltages. Such Aviram���Ratner molecular diodes have been demonstrated using Langmuir���Blodgett fi lms32 and block copolymers33 sandwiched between two planar electrodes. When using self-assembly to build a diode, the orientation of the molecules must be controlled. One way to achieve this is to synthesize molecules terminating in a thiol at each end, both of which are protected with diff erent groups34. One protection group is removed fi rst, allowing the thiol at that end to bind to one electrode, followed by removal of the second protection group allowing the other end to link to another electrode. Th e experiments described above involve a large number of molecules. Th e single-molecule diode has been demonstrated in a molecule consisting of two weakly coupled conjugated units, using a mechanically controlled break-junction method35,36 (Fig. 3). Th e two conjugated units are diff erent ��� one is fl uorized and the other one is not ��� and this breaks the symmetry of the molecular junction. When sweeping the bias voltage, the energy levels of both units are shift ed relative to each other. Whenever an unoccupied level passes by an occupied one, an additional transport channel opens up and the current increases by a certain amount ��� one step. Th e height of each step depends on the bias polarity, which gives rise to the diode-like I���V curve results as shown in Fig. 3. Th e presence of donor and acceptor groups introduces non- symmetry to the molecular junctions. More generally, one may expect rectifi cation to arise from non-symmetric molecular junctions that do not necessarily contain a pair of donor and acceptor groups37. Indeed, diode-like I���Vs have been observed by using diff erent molecule���electrode contacts on the two ends38,39, non-symmetric molecules34,40,41, or both42. However, non-symmetry alone is not suffi cient to achieve a larger diode eff ect if the bias voltage drops entirely at the molecule���electrode interfaces43. Another type of molecular diode was proposed based Gate Electric Magnetic Mechanical Optical Chemical Electrochemical Drain Source Figure 1 Illustration of a single molecule attached to two electrodes as a basic component in molecular electronics. Electron transport through the molecule may be controlled electrically, magnetically, optically, mechanically, chemically and electrochemically, leading to various potential device applications. To reach the ultimate goal in device applications, experimental techniques to fabricate such an electrode���molecule���electrode junction, and theoretical methods to describe the electron transport properties must be developed. Both are challenging tasks, but rapid advances have been made in recent years. Nature Publishing Group ��2006