1.

Introduction to Mesoscopic Physics I

1.1

1 INTRODUCTION AND BASIC CONCEPTS 1 1.1 Introduction and Basic Concepts Brief Overview

1.1

1 INTRODUCTION AND BASIC CONCEPTS

1.2

Mesoscopic Physics deals with the physcics of small condensed objects (hence with a collection of atoms in contrast to atomphysics). Since these objects are often in the nanometer size-regime, mesoscopic physics is generally thought as of a (sub-)discipline of Nanoscience. Nanoscience is a popular research discipline even though a clear-cut definition is not straightforward. This is mainly because of the interdisciplinary nature of this eld. Nanoscience involves physics, chemistry, and biochemistry and progress is often derived from technology, i.e. the methods allowing to shape objects on the nanometer scale. In view of this, it is natural that scientists which are trained in dierent elds will adopt dierent denitions. In a rst attempt we may dene nanoscience as the research concerned with physical objects for which the nanometer length-scale is essential. In biology, it may be the size of the molecule of interest. In solid-state physics, it is a piece reduced in size in one or more dimensions to below 1 m. This denition is extremely wide as it includes everything listed below: surface science (lattice, reconstructions, adhesion, growth), thin lms (also organic monolayers), interfaces (multilayers, solid-solid-, but also solid-liquid interfaces, wetting and growth phenomena), internal interfaces (domain boundaries, type II superconductors, phase segregation, grain boundaries, dislocations...), defects (atomic defects, line defects, compostional defects...), devices (state-of-the art transistors, research devices for transport studies), all scanning-probe microscopy (SPM) techniques (STM, SFM, SNOM, ...), metrology (nanometer as a measuring standard), macromolecules, and single atoms/molecules (atomic manipulation), and many more. As can be seen, nanoscience covers an enormous great deal of subjects. A certain distinction may be obtained by looking at the technology which is applied for the fabrication and analysis. We can distinguish four major fabrication methods:

1. Lithography has a long tradition. It has been developed to a high degree of perfection driven by the needs of large-scale IC manufacturing. For nanoscience, electron-beam lithography is usually applied to structure a solid into small electric circuits with an accuracy of 5 10 nm. Research on these nano-devices has stimulated a lot of theoretical and experimental eort during the last decade. It has had an enormous impact on the understanding of physics on the nanometer scale. The physics of such devices are now often termed mesoscopic physics, though this term has a much wider scope. 2. The Scanning-Tunnelling Microscope (STM) and all related scanningprobe microscopes (SPM) cover a length-scale from several m down to the scale of a single atom. Initially, research using SPM could be dened as being mostly surface science. However, in the meantime scientists have learned to perform much more than just viewing the arrangement of atoms on a surface. Apart from imaging, atoms can be manipulated, surfaces can be modied, and mesoscopic objects can be characterized. 3. Synthesis. This chemical route is quite dierent to the physical lithography approach where a piece is cut-down. In chemistry and biochemistry, nano-objects are build up from atoms (ions) and small molecules (synthesis). Also, one may use such structures as templates for other processes. Such templates, zeolites for example, have the advantage that they are monodisperse, all units are of equal size. Furthermore, the building up of more complicated structures can be realized by the application of principles, such as self-assembly and reproduction, derived from mother nature. 4. Finally, there are other methods, nucleation and growth, for example, which most often fall under the area of material science. Driven by the enormous pace of the integrated-circuit industry, which is using the traditional lithography approach, mesoscopic physics has developed rapidly. Initially, mesoscopic physics was dealing with static electrictransport phenomena in small devices. This is what we could now term classical mesoscopics. In the meantime, Mesoscopic Physics is progressively expanding into other areas as well, e.g. non-linear phenomena, dynamic eects, mesoscopic magnetism, mesoscopic superconductivity... Interest in devices of reduced dimensions (often in the nanometer) came up, because scientist were curious to observe quantum eects not present

1 INTRODUCTION AND BASIC CONCEPTS

1.3

1 INTRODUCTION AND BASIC CONCEPTS

1.4

in macroscopically large solids (i.e. in the bulk). In general, we may say that there is a size below which a solid does no longer behave bulk-like. This length separates the macroscopic solid-state physics from mesoscopic physics. Importantly, it is assumed that the sample is large on the atomic scale, such that uctuations in the number of atoms contained in the sample are not important. That is why the term Meso has been used (in between a single atom and a bulk solids). Quite in contrast are STM-based techniques. Here a sharpened tip is brought into very close proximity to the sample so that wavefunction overlap between the foremost tip atom and the surface is appreciable. Hence, the interaction is always on the atomic scale (one atom). For a single local STM experiment (experiment over one atom on the surface without averaging) uctuations of the number of atoms matters. Hence, such an experiment does not fall into the classic denition for mesoscopic physics given above. The energy window investigated is large ranging from meV to 10 eV. Nonlinear transport is expected for these voltages. In addition, extreme conditions can be generated in the tip-substrate region: strong electric eld strength, large local forces, and extreme current densities. Such conditions allow the manipulation of atoms and small molecules and will bring about new physical insight in matter on the sub-nanometer scale. At present, however, many experiments are not fully understood because we are lacking theoretical background on this scale. Apart from atomicscale properties, (classical) mesoscopic eects are sometimes accessible by STM as well. This is the case if the measured quantity is averaged over a certain image window of size >> A, or if the quantity is not strongly dependent on the rapid varying atomic potentials. An example of the latter is the measurement of the spatially resolved BCS gap parameter in a type II superconductor. This is possible, because varies on the length scale of the coherence length which is typically much larger than interatomic distances (an notable exception are high-Tc superconductors). For STM-based experiments it therefore appears useful to distinguish mesoscopic experiments from atomic-scale ones. An other example may help to understand this distinction. A magnetic-force microscope (a SPM with a magnetic probing tip) can be used to measure local magnetic properties of a ferromagnet, for which the order parameter varies on a scale large compared to atoms. Since the probing tip averages the measured interac-

tion force over a length scale of order 10 50 nm, it is not sensitive to the local atomic potential landscape. Hence, the collective mesoscopic information can be obtained. In general however, we can expect that atomic potentials may obscure collective eects. This is the case because of the very dierent energy scales: looking at a single atom, the energy scale for STM experiment is given by the dierence of empty or occupied orbitals, which is in the eV range. In contrast, the exchange energy in a ferromagnet is at least an order of magnitude smaller. For a superconductor the energy scale is even much smaller which makes STM-measurement of this collective property very dicult. In view of the important developments using scanning-probe methods, which extended the size-regime of the objects of interest to a few or even a single atom it is fair to extend the denition of Mesoscopic Physics to also include the physics of these systems. This is justied by noting that the atom under the tip apex is not isolated, as it is in atomphysics, but it always resides on a substrate to which it electronically couples. How do we distinguish mesoscopic from macroscopic? A piece of solid can be considered bulk-like (macroscopic) if all properties are either scale independent or can be written as intensities which are then scale independent. For example, a piece of Au has a specic heat independent of its size. Or, a metallic wire has a specic resistivity per unit length. If a bulk solid is cut down into smaller pieces, there will be a critical size (in one, two or three dimension) below which some of the properties can no longer be described by intensities. This is the regime of mesoscopic physics. Furthermore, on two seemingly identical samples dierent amplitudes of one property may be measured. Hence, an average value for a particular property has to be obtained by averaging over an ensemble of representations (ensemble average). The sample-to-sample variations are commonly termed mesoscopic uctuations. Therefore, a solid is bulk-like if ensemble averaging is not necessary, dierent bulk-pieces are identical. It is then said that self averaging applies. If the size of a sample approaches the atomic scale mesoscopic uctuations can get very large. In STM, it may happen that the magnitude of the uctuation is of order of the average eect itself. Hence, the eect may be

1 INTRODUCTION AND BASIC CONCEPTS

1.5

1 INTRODUCTION AND BASIC CONCEPTS

1.6

obscured by these uctuations. An instructive example is the atomic-scale point contact which is realized by gently touching a metallic surface with a metallic tip in an STM conguration. It is now understood that the striking jumps observed in the electric resistance of such a contact are caused by spontaneous rearrangement of atoms in the contact regions. The measured resistance values do not directly resemble quantized conductance. This is expected because of a contact diameter which is comparable to the Fermi wavelength of the electrons. However, if all observed resistance values are averaged over a large ensemble, the quantization of the conductance starts to become evident. From this exercise it is seen that sample specic uctuations can be very large on the atomic scale. We should therefore be careful with future perspectives of atom- or single-molecule based electronics. Their state would be susceptible and dicult to interface unequivocally. Finally, we mention that the mesoscopic length scale depends on many parameters, the temperature and purity of the sample, for example. The scale may range from a few nanometers up to 100 m or more. Hence, mesoscopic physics is not necessarily nanoscience. It may equally well be microscience. An example is the quantized Hall eect which can be observed in rather large samples. Here, the (two-terminal) resistance is independent of the size of the sample demonstrating that a specic resistivity cannot be dened. In mesoscopic physics, the emphasis has been on electric transport properties, i.e. the static linear conductivity. Nonlinear and dynamical eects have only recently been considered. Many of the mesoscopic eects have their origin in a reduced dimensionality with respect to a length scale characteristic for conducting electrons in the bulk solid. Some important length scales are for example the Fermi wavelength F , the electron mean-free path l, and the phase-breaking length l . In nanometer wide constrictions, lithographically dened into a twodimensional electron gas, the electric conductance was surprisingly found to be quantized in units of 2e2 /h (e is the unit of charge and h the Planck constant). In the quantum-ballistic point contacts, as these constrictions are now classied, the conductance increases in discrete steps as the width of

the constriction is widened. It is now understood, that this eect is caused by the quantization of the transversal momentum in the constriction resulting in a set of one-dimensional wavefunctions (modes) with dispersion along the constriction. Since, according to the Landauer-B uttiker formula, each occupied propagating mode contributes exactly e2 /h to the total conductance, the conductance of the point contact is just found by counting the discrete number of occupied modes. The conductance quantization is a quantum size-eect originating from the constriction width which is comparable in size with the electron Fermi wavelength (order 10 nm in semiconductors). If the characteristic size L of the conductor is made much larger than the elastic mean-free path l, electrons diuse along the conductor. This regime is classically described by the Drude equation. However, even in this regime quantum eects can show up at suciently low temperatures. This is because elastic scattering changes the phase of electrons in a predictable manner: the phase memory is thereby not lost. The distance the electron diuses after which the phase memory is destroyed is called the phase-breaking length l . If l << L < l the transport in the conductor is diusive and phase coherent. That phase-coherent transport can be observed in normal metal wires has been demonstrated by the observation of Aharonov-Bohm oscillations in nanometer-sized metallic ring structures. Another phenomena caused by phase coherent transport is weak localization. An electron that diuses through the conductor starting from an arbitrarily chosen point P has a nite propability to return to P . If this closed path (loop) is shorter than l , constructive interference with the time-reversed loop is possible. This increases the probability for an electron to return to its starting point, hence, results in a decreased conductivity, which has been termed weak localization. This eect is removed in a magnetic eld which destroys time-reversal symmetry. Apart from weak localization, which can be observed even in samples of macroscopic size (larger than l ), there are in addition direct interference contributions to the conductance active in phase-coherent samples. Since identically prepared samples are dierent in their impurity conguration, conducting electrons are scattered in a potential-landscape which is dif-

1 INTRODUCTION AND BASIC CONCEPTS

1.7

1 INTRODUCTION AND BASIC CONCEPTS

1.8

ferent in between samples. Therefore, the interference term contribute dierent for dierent samples. The ensemble averaged variance of sampleto-sample conductance dierences G has found to be universal and of order e2 /h. For this macroscopic quantum phenomenon the term universal conductance uctuations (UCF) is used. The phenomena described above are a consequence of the wave nature of electrons. However, we also know that electrons can manifest themselves as particles carrying the charge quantum e. This has been demonstrated in the classical experiment by Millikan where the charge of an oil droplet was found to be quantized. In electric conductors the particle nature does not usually show up because of the electrons forming an extended uid. However, if high resistive tunnelling junctions are used, the electric transport characteristic changes dramatically: now, electrons traverse the junction in a discrete fashion as packets carrying the unit charge e. A certain number of electrons can be stored on a capacitor connected to the tunnel junction. The number of electrons can precisely be controlled if the eective capacitances are suciently small such that the single-electron charging energy dominates over thermal energies. Here again, we enter the mesoscopic size-regime, but now from the high impedance side. Electron transport occurring in tunnel junctions dominated by the charge of single electrons are said to be in the single-electron tunnelling regime. Most of the new physical phenomena discovered in the mesoscopic sizeregime are eects observed in the static electric conductance of normal and superconducting nanostructures in the so-called linear response regime, i.e. for small applied voltages. Nonlinear, dynamic and magnetic eects have remained widely unexplored so far. This is a new opportunity for basic research in nanophysics. Electric noise phenomena is one of these opportunities. Electric noise can be caused by dynamical resistance uctuations with an approximately 1/f spectral density (1/f noise, icker noise) or it can be due to thermal and shot-noise with a frequency independent spectral density. Despite many careful studies, 1/f noise is still not well understood. There is however increasing evidence that part of the resistance uctuations are caused by the kinetics of lattice defects. Defect induced uctuations can

become very pronounced in nanometer-sized structures, e.g. in point contacts. In contrast to 1/f noise, shot-noise has only recently been considered in the mesoscopic size regime. Shot-noise originates from the discreteness of the electric charge. A train of particle which is uncorrelated in time (Poisson distribution) results in the classical shot-noise spectral density PP oisson = 2eI . The suppression of the current noise below PP oisson originates from correlations imposed by the Pauli principle or from Coulomb interactions. The former is important for metallic conductors and the latter for devices in the single-electron tunnelling regime. Noise measurements provide us with new information not obtainable from the static electric conductance measurement. References
1

STM related The Nanometer Age: Callenge and Chance by H. Rohrer in Micro-and Nano Engineering 27 3 (1995). Mesoscopic Physics, electron-wave related Y. Imry in Directions in Condensed Matter Physics, G. Grinstein and G. Mazento eds. (World Scientic, Singapore, 1986), p. 101; Nanostructured Systems, Mark Reed ed., Semiconductors and Semimetals, Volume 35 (Academic Press 1992). C. W. J. Beenakker and H. van Houten in Solid State Physics, H. Ehrenreich and D. Turnbull eds., Vol. 44, p. 73-174 (1993); B. L. Altshuler and A. G. Aronov in Elelctron-Electron Interactions in Disordered Systems, M. Pollak and A. I. Efros eds. (North Holland, Amsterdam, 1985); S. A. Washburn and R. A. Webb, Advance Physics 35, 375 (1986); R. A. Webb et al. in Localization, Interaction, and Transport Phenomena in Impure Metals, G. Bergmann, Y. Bruynseraede, and B. Kramer eds. (Springer, Heidelberg 1985); Quantum Transport in Semiconductor Submicron Structures, B. Kramer ed., NATO ASI Series E: Applied Sciences - Vol. 326, (Kluwer 1995).

1 INTRODUCTION AND BASIC CONCEPTS

1.9

1 INTRODUCTION AND BASIC CONCEPTS 1.2 Nanostructures

1.10

Mesoscopic Physics, particle related, i.e. single-elctron eects D. V. Averin and K. K. Likharev in Mesoscopic Phenomena in Solids, B. L. Altshuler, P. A. Lee, and R. A. Webb eds., (North-Holland, Amsterdam, 1991); H. van Houten, C. W. J. Beenakker, and A. A. M. Staring in Single Charge Tunnelling, H. Grabert and M. H. Devoret eds., NATO ASI Series B, Vol. 294 (Plenum, New York, 1992). Mesoscopic Physics, Textbook S. Datta Electronic Transport in Mesoscopic Systems, Cambridge Studies in Semiconductor Physics and Microelectronic Engineering, A. Ahmed, M. Pepper, and A. Broers eds. (Cambridge University Press 1995).

Mesoscopic physics is ultimately linked with nanotechnology, since it is the later that determines the type and size of structures we can fabricate, hence investigate. As model systems, simple electric curcuits are ususally considered. These are for example metallic wires straight ones or sometimes bent into a ring structure, single tunnel junctions between similar and different materials, highly transparent contacts between dierent materials etc. As materials, the focus is often on as simple as possible systems. For example, III-V semiconductor heterostructure, in which a nearly perfect two-dimensional electron gas can be realized, or metals like Au and Ag that have a nearly spherical Fermi surface derived from s-electrons. Today (1997), structures of width 50 nm can be made on a regular basis and state-of-the art (with tricks) reaches 10 nm. The set of images shown below give a few examples of nanostructures.

resist wire

wire

Fig 1.1 Conventional lift-o scheme for the fabrication of nanostructures. Metal is
evaporated from above onto a resist layer which is prestructured by lithography (exposed and developed). Lithography leaves an opening in the resist into which the metal is evaporated. Finally, the metal lm, which resides on the resist layer, is removed (lifted o) when the resist is stripped.

1 INTRODUCTION AND BASIC CONCEPTS

1.11

1 INTRODUCTION AND BASIC CONCEPTS

1.12

Tunnel Junction

Fig 1.3 This example shows a structure containing in its center a tunnel junction. Such junctions are realized by two subsequent evaporation runs in which the sample normal is tilted with respect to the evaporation source. In this way it is possible to overlay to wires over a short distance in a controlled manner.

Fig 1.2 Example of a 200 nm wide Au wire connected to large reservoirs to the left and right.

Fig 1.4 Metallic ring structures with four contacts attached. This structure served to investigate the Aharonov-Bohm eect in normal metal rings.

1 INTRODUCTION AND BASIC CONCEPTS

1.13

1 INTRODUCTION AND BASIC CONCEPTS

1.14

Fig 1.5 Extremely small metallic ring structure fabricated using high-energy electronbeam lithography.

Fig 1.6 This image shows an alternative fabrication method to obtain a large collection of nanowires. A wafer is prestructured by wet anisotropic etching to yield a regular triangular grooved surface. Then, metal is evaporated under a angle . Because of shadowing, small wires can be otained.

1 INTRODUCTION AND BASIC CONCEPTS

1.15

1 INTRODUCTION AND BASIC CONCEPTS

1.16

Fig 1.7 This gure shows the smallest structures that can lithographically be fabricated at present (1997). Here X-ray lithography has been used to structure an organic layer (resist). Fig 1.9 Dierent schemes that allow to laterally conne the electrons of a near-surface two-dimensional electron gas using either etching or a gating techniques.

Fig 1.8 Band-edge prole for a AlGaAs-GaAs modulation-doped heterostruture used to realize a two-diemensional electron gas.

Fig 1.10 Left, a quasi one-dimensional wire fabricated according to scheme c) of Fig. 1.9. Right, the gates of two point contacts fabricated ontop of a 2-dimensional electron gas using scheme d) of Fig. 1.9.

1 INTRODUCTION AND BASIC CONCEPTS

1.17

1 INTRODUCTION AND BASIC CONCEPTS

1.18

Fig 1.11 This is a demonstration of lithography with scanning-probe methods. Visible are metallic wires fabricated using STM.

Fig 1.12 This image shows four dierent cages built up from single Fe atoms. This structure has been fabricated with a aid of a scanning-tunnelling microscope. The wave pattern seen within the cages can be interpreted as standing waves originating from electrons in the Cu substrate that scatter o the Fe-atoms (border of cage).

1 INTRODUCTION AND BASIC CONCEPTS

1.19

1 INTRODUCTION AND BASIC CONCEPTS

1.20

mesopores: 4nm (selfassembling)

Fig 1.13 The visible nanowires (mushrooms) were fabricated electrochemically by galavnic deposition of metal into nanopores. With this method wires as small as < 5 nm can be fabricated.

Y-zeolite: 0.8 nm

Carbon-nanopores: 3nm

Fig 1.14 Dierent types of nanochannels that can be used as templates for the fabrication of extremely samll wires.