sign in sign up
<<
Home , Mesoscopic physics

arXiv:cond-mat/0106256v1 [cond-mat.mes-hall] 13 Jun 2001 ag flweege h iesae aetesame the have eigenstates the For energies low directions. transverse of the multiplied range in wire a function the are localized along eigenstates running a poten- wave electronic by hard-wall plane The a a surface. by to regard given wire’s can the subjected one at being tial fact, as in Often, the the prevents that escaping. field from electric an charged electrons produce positively wire The the wire: in long ions a inside an ple, iesos h nwris answer The dimensions? interest. enormous of is of problem systems effects matter a combined condensed currently in the interactions understanding quantum and disorder fact, a In affect interaction system. exam- electron-electron for (impurities, and disorder below. ple) the how on discussed of light questions new be difficult shed have will systems the which low-dimensional as And such systems, research, mesoscopic of fields of new entirely to led nanotechnol- cur- and [2,3]. the nanoscience to dimensional. ogy central of low are fields as here fashionable covered regarded rently subjects be the essen- will of be too Many to these believed and are tial, dimensionality physics, reduced ef- in the of interest where fects superconductors, great high-temperature of In as systems such sciences. are life and there physical be- addition, other boundaries and the subject biologically blurred the even tween has and and materials of important use field but the the article, in involved developments this recent have of exciting in- most subject the not the of in some would such molecules entities one and low-dimensional Traditionally as occurring naturally [1]. at clude scale engineered atomic alloy the varying com- a positionally of or consists which structure, “nanostructure,” semiconductor semiconductor nanometer-scale the is tems ar device and electronic for, stream state-of-the-art potential chitectures. constant the new by a of, also by realization and fueled discoveries years, striking twenty dur- of tremen- past systems been the quantum has low-dimensional ing There in interest world. full dous our electrons, the of exploring as dimensions from restricted three such is photons, or degrees-of-freedom, , microscopic of o r lcrn,sy etitdfo oigi three in moving from restricted say, electrons, are How has phenomena quantum low-dimensional of study The sys- low-dimensional of workhorse and paradigm The motion the where one is system low-dimensional A ok r rvddwe possible. when students. provided undergraduate are physics books advanced to accessible be re umr ftepyiso o-iesoa unu sy quantum low-dimensional of physics the of summary brief A eateto hsc n srnm,Uiest fGeorgia of University Astronomy, and Physics of Department .INTRODUCTION I. unu hnmn nLwDmninlSystems Low-Dimensional in Phenomena Quantum confinement ae o exam- for Take, . ihe .Geller R. Michael Ags 3 2013) 13, (August - 1 aotutr ti fe eesr oflo rwhwith growth semiconductor follow a to necessary of often construction is the it sys- nanostructure complete two-dimensional To a best, tem. at in, resulting only, rection edn oacmoiinlygae lo fteform the of alloy graded compositionally Al a to leading n oetasta etiteeto oint ee than fewer to dimensions. motion electron three restrict that potentials ing reetia neto,se oiindpnetpoten- position-dependent varying a By sees tial. excitation, injection, optical electrical doping, or to through added band electron conduction an the Hence, varying minimum. spatially band a conduction produces variation structure band ing n h lmnmfraction aluminum the ing emeiay(B)mcie.Tk,freape the example, for Al Take, alloy machines. homogeneous (MBE) molecular- epitaxy high-vacuum beam in compositionally alloys “growing” semiconductor by graded is structures dimensional irsoi oeta rdcdb h lo Al alloy The the by structure. produced band distinct potential a microscopic with each AlAs, and GaAs and asadAA.I atclr h nrygpbetween with gap varies energy bands the of conduction and particular, that valence In between the pro- AlAs. interpolating as and structure regarded GaAs band be a may periodic, ducing strictly not although ytm,ognzdb hi “dimension.” the their of by quantum remainder organized low-dimensional particular The systems, to devoted discussed. is are article phenomena dimensions quantum low of features in general In some II. III Section Section of in fabrication introduced The is articles structures possible. low-dimensional review when provided recent are to books and References under- The physics students. advanced to graduate systems. accessible quantum be should low-dimensional material of physics the the along motion only leaving di- out,” transverse wire. those “frozen in is motion factor rections that wave means plane This the changes. only and eigenfunction transverse ra fasncaostgte ihafraction a 1 with and together minum atoms arsenic of array I AIGLWDMNINLQUANTUM LOW-DIMENSIONAL MAKING II. npatc,hwvr ti osbet vary to possible is it however, practice, In tutrsta ofieeetosaemd ychang- by made are electrons confine that Structures h otcmo ehdo arctn low- fabricating of method common most The hsatcewl rvd eybifitouto to introduction brief very a provide will article This c ( r c ) Ga orsodt h rsaln crystalline the to correspond 1 = eeecst eetrve rilsand articles review recent to References 1 − c tm sgvn h aeilshould material The given. is stems ( r tes eri 30602-2451 Georgia Athens, , ) − s where As, c c prpitl,oecnegne confin- engineer can one appropriately, fglim h pca ae of cases special The gallium. of STRUCTURES c Ga c 1 − aissaily h result- The spatially. varies c scnitn faperiodic a of consisting As c uigcytlgrowth, crystal during c c c noedi- one in Ga . c 1 falu- of − c c 0 = As, lithography, the selective etching of prepared surfaces. A. Effective Mass After coating a surface with a protective material, the “resist,” patterns are imprinted on the resist in a pho- Electrons in semiconductor nanostructures move in the tographic processes with focused light, electron beams, presence of as many as three fields; the periodic or nearly or even atoms. After the imprinted pattern is removed periodic potentials produced by the atoms in the crystal, chemically, the underlying semiconductor is etched away, fields applied externally, and the electron-electron inter- leaving an environment that confines electrons both in action potential. The atomic potentials, which vary at the MBE-growth direction and laterally (perpendicular the few Angstrom scale, are usually varying much more to the growth direction). Another common way to pro- rapidly than the others. In this case, assuming the elec- duce lateral confinement is to use lithographic techniques tronic states have energies near the bottom of the conduc- to pattern metallic electrodes, or “gates,” on the surface tion band or near the top of the valence band, there is an of an crystal (grown by MBE, for example) that has elec- extremely useful description whereby the original prob- trons confined in a buried two-dimensional layer parallel lem of an electron moving in the presence of the atomic to that surface. By applying voltages to these electrodes, as well as other potentials is replaced by the much sim- electrons in the layer can be depleted from or attracted pler problem of an electron, now with a different mass, to the regions below the electrodes. moving in the presence of the slowly varying fields only Finally, it is in many cases necessary to attach electri- [4]. For example, the effective mass of an electron in a cal contacts to the electron inside these nanostruc- GaAs conduction band is about 0.067 times the ordinary tures and to the gates on their surfaces. bare mass. In general, the effective mass depends on the The lower limit to the size of the structures one can electron’s propagation direction, and can even vary with make is usually determined by the size of the patterns position. one can make with lithography, which, in turn, is usually determined by the quality of the image formed during the exposure stage. At the time of this writing the semi- B. Density of States conductor industry can produce, in an integrated circuit, wires with a thickness of only 10 nm, more than 10 times smaller that the of visible light. An important distinction between systems with differ- To make even smaller structures naturally occurring ent dimensionality is their density of states N(ǫ), which nanometer-scale systems have used. Examples include is the number of states per unit “volume” LD in an en- self-assembled nanostructures and structures incorporat- ergy range ǫ to ǫ + dǫ, divided by dǫ. L is the linear size ing molecules or biological materials. A self-assembly of the system, and D can be either 1, 2, or 3. (The zero- technique to make arrays of few-nanometer quantum dots dimensional case has to be treated separately). A sim- (small “artificial atoms” where electrons are confined in ple general formula can be derived to determine the en- all three dimensions) by growing a thin layer of highly ergy dependence of N(ǫ): Given an excitation (electron, strained material on top of another crystal has been par- , photon) with disperion relation ǫ(k) ∝ |k|α, ticularly successful. The strained layer relaxes by break- the number of states contained within a D-dimensional ing into small islands, which form the quantum dots. An sphere of radius k in momentum space is proportional to excellent example of the use of molecules to make low- (L/2π)DkD and the number per unit volume is therefore dimensional systems is provided by large carbon sheets proportional to kD or ǫD/α. The density of states is evi- that can be rolled into stable hollow spheres, known as dently the derivative of the latter with respect to energy, D − Bucky balls, or into carbon nanotubes. It is even pos- so N ∝ ǫ α 1. Phonons and photons (α = 1) in three sible to put other molecules inside a Bucky ball, to ar- dimensions have N ∝ ǫ2, and in lower dimensions would range Bucky balls into a molecular crystal, and to elec- have N ∝ ǫD−1 as long as the dispersion remains linear. trically contact individual nanotubes. Electrical contacts Electrons (α = 2) have a density of states proportional 1 − 1 have also been attached to other molecules, turning them to ǫ 2 , ǫ0 (i.e., energy independent), and ǫ 2 in 3, 2, and into “.” and even DNA strands have 1 dimensions, respectively. The expressions for D = 2 been used to engineer quantum structures at the nanome- and D = 1 assume only one transverse mode or sub- ter scale. band present. When more are present, these additional branches simply add to N(ǫ).

III. PHYSICS IN QUANTUM SYSTEMS OF REDUCED DIMENSIONS C. Mesoscopic Physics

Physics in low-dimensional systems is often different Mesoscopic physics is an exciting new field of science than in three dimensions. We now discuss some of these made possible by nanostructures [5]. A mesoscopic sys- differences and the theoretical tools used to understand tem is one that is in some sense between a microscopic them. and macroscopic system. It is typically much larger than

2 a few atoms or molecules, yet it is small enough that affected by the magnetic field even though they are never the degrees-of-freedom (usually electrons) have to be re- in contact with it, a consequence of the Aharonov-Bohm garded as fully quantum-mechanical. More precisely, a effect. mesoscopic system has a size L that is larger than a mi- Nondissipative current-carrying states can occur in a croscopic lengthscale a (for example the Bohr radius), yet superconducting ring too, but in the superconducting smaller than the -coherence length Lϕ, which is the case the current-carrying state is not, strictly speaking, characteristic length beyond which a particle loses phase an equilibrium state, but is instead an extremely long- coherence. Lϕ generally depends on the dimension and lived metastable nonequilibrium state. the temperature of the system, as well as on microscopic As as the ring becomes larger, the magnitude of the details. Thus, in simple terms, a mesoscopic system is decreases. One reason is that when one that is larger than microscopic and in which quan- L exceeds Lϕ the electrons no longer are moving coher- tum mechanics manifests itself fully. ently around the ring and the Aharonov-Bohm effect is In what follows we shall discuss a few examples and no longer operative. The second reason is that even if concepts from mesoscopic physics. is Lϕ was always larger than L, the magnitude of the maximum persistent current decreases as 1/L.

1. Aharonov-Bohm Effect 3. Phase-Coherent Transport In 1959 Yakir Aharonov and David Bohm predicted that a magnetic field can infleuence the quantum- Electrons in a mesoscopic conductor (a conductor mechanical phase of a charged particles, even if the par- whose dimensions are smaller than Lϕ) move as a wave, ticles are prevented from entering the region contain- not as a particle. The behavior is similar, in fact, to an ing the magnetic field. This proposal, although a di- electromagnetic wave propagating inside a wave guide. rect consequence of , violated every- This wave-like nature leads to many unusual physical one’s intuition and was extremely controversial at the properties [9]. time. The Aharonov-Bohm effect has been observed hun- Perhaps the most profound is the origin of resistance in dreds of times and shows up everywhere in the study of mesoscopic conductors. Ordinarily resistance is caused mesoscopic systems. Excellent accounts of it are given by inelastic collisions of the current-carrying electrons in quantum mechanics textbooks [6] and review articles with disorder (impurities and other imperfections), other [7,8]. electrons, and with lattice vibrations (phonons). In The basic idea is that a charged particle moving around a mesoscopic conductor, however, an electron typically a closed loop of circumference L accumulates, along with travels the entire length of the system without undergo- its kinematical phase 2πL/λ, where λ is the deBroglie ing an inelastic collision. Thus it might seem that there wavelength of the particle, an additonal Aharonov-Bohm would be no resistance at all. But to measure the resis- phase given by 2πΦ/Φ0, where Φ is the magnetic flux tance in a mesoscopic system one has to attach electrical enclosed by the ring and Φ0 ≡ hc/e is the quantum of contacts or leads to it, which are macroscopic. What magnetic flux. The Aharonov-Bohm phase changes the occurs, then, in a mesoscopic conductor, is that the re- energies of charged particles and also can shift quantum sistance is caused by a combination of elastic scattering interference patterns. But the effect can only be observed of electron waves from disorder and inelastic scattering if the particles are sufficiently phase coherent. in the macroscopic leads, although the magnitude of the resistance is determined exclusively by the former. This phenomena is succinctly described by a formula, 2. Persistent Currents originally proposed by Rolf Landauer in 1957, for the con- ductance G (reciprocal of the resistance) of a mesoscopic A simple example of a mesoscopic effect is as follows: system. Landauer’s formula is simply G = |t|2e2/h, Take a small metal or semiconductor ring of circumfer- where t is the quantum transmission amplitude for an ence L. The mesoscopic ring can be made using the electron to propagate through the system (|t|2 is the lithographic techniques described above. The ring is an transmission probability). The ratio e2/h of fundamen- ordinary conductor, not a superconductor. It seems rea- tal constants has dimensions of conductance and is about sonable and is usually the case that in the absence of a 1/26kΩ. The Landauer formula shows that the conduc- magnetic field the ground state of the ring is non-current- tance of a mesoscopic conductor is directly related to carrying. However, if the phase-coherence length Lϕ is the quantum mechanical probability that an electron can larger than L then by threading the ring with a magnetic propagate through the system without scattering elasti- flux Φ the ground state (or finite-temperature equilib- cally. It makes no reference to the strength of inelas- rium state) becomes a current-carrying state, and current tic scattering that actually causes the energy dissipation will flow around the ring without ever dissipating, even normally associated with a resistor. in the presence of disorder. The electrons in the ring are Another important consequence of the phase-coherent

3 nature of electron transport in a mesoscopic system is At low temperatures the dominant mechanism for de- a phenomenon called weak localization. To understand phasing is electron-electron scattering: A given electron weak localization it is first necessary to understand “or- feels a fluctuating electric field produced by the other dinary” Anderson localization (sometimes called strong electrons that scrambles its phase after some time τϕ. localization), named after Phillip Anderson. Anderson Theoretically it is predicted that the rate van- −1 β localization is the process whereby the eigenstates of an ishes at low temperature as τϕ ∝ T , where β is a electron in a disordered environment become spa- positive exponent, but recently Richard Webb and oth- tially localized around impurities, causing the system to ers have given experimental evidence for a saturation of −1 behave as an instead of a conductor. Weak lo- τϕ in the T → 0 limit. The physics of low-temperature calization is a very different process that also increases dephasing is currently a problem of great controversey the resistance of a disordered conductor (although it is and interest. not strong enough to turn a conductor into an insulator). It comes from a very special quantum interference ef- fect that occurs in systems with time-reversal symmetry: 5. Thouless Energy To find the probability P for an electron to propagate from point r to point r′, one has to add the quantum Mesoscopic physics research has also led to a profound amplitudes Ai for the electron to take all possible paths new discovery about quantum systems in general. In i, and then calculate the modulus squared, P = | P A |2. i i the 1970’s David Thouless and collaborators showed that (This expression is a consequence of Feynman’s path inte- any quantum system possess an important fundamental gral formulation of quantum mechanics, but one can also energy scale, now called the Thouless energy ET. The view it as a generalization of the double-slit interference Thouless energy is a measure of how sensitive the eigen- formula to an infinite number of “slits.”) The cross-terms states in a quantum system are to a change in bound- in this expression are responsible for interference. r r′ ary conditions. Specifically, ET is defined as the energy Usually when open paths with different from are change of a state at the Fermi energy caused by a change considered, the randomness in the Ai wash out any quan- from periodic to antiperiodic boundary conditions. ET tum interference effects. However, there is a special class r r′ would be zero in a system—for example an insulator— of paths, closed paths with = , where interference with localized eigenstates, because if the wave functions effects can be important. In systems with time-reversal don’t extend to the boundaries their energies will be inde- symmetry (which basically means that there can be no pendent of boundary conditions. Thouless s howed that applied magnetic field) there will always be pairs of closed the dimensionless ratio g ≡ ET/∆ǫ of ET with the en- paths and their time-reversed counterparts in the above ergy level spacing at the Fermi energy, ∆ǫ, determines summation that have the same amplitude. The result is r r whether the system is a conductor (g > 1) or an insula- that the probability to go from to , in other words tor (g < 1). In fact, g is equal to the conductance of the the probability to go nowhere is enhanced by quantum system in units of e2/h. interference effects, and this amounts to a measurable increase in resistance. D. Integrable Systems

4. Dephasing by Electron-Electron Interaction An extremely important theoretical aspect of low- dimensional quantum systems is that there is a large fam- As explained above, weak localization (and more gen- ily of one-dimensional models that are exactly solvable erally, the Aharonov-Bohm effect) occurs when the elec- or integrable, that is, the exact many-body wave func- tron motion is sufficiently phase-coherent. This fact can tions and energies can be obtained. This is especially be exploited to actually measure the phase-coherence fortunate because perturbation theory, which is needed length Lϕ or phase-coherence time τϕ, the characteris- to develop approximate descriptions of three-dimensional tic time beyond which the electron becomes decoherent, interacting systems, tends to break down and become in an electron system. Lϕ and τϕ are simply related to invalid in low dimensions. One of the first quantum each other and only one needs to be measured. The re- many-body problems to be solved exactly was the one- sistance increase due to weak localization depends on the dimensional spin-1/2 Heisenberg model with antiferro- number of closed paths that contribute to the summation magnetic nearest-neighbor interaction, solved by Hans 2 in P = | Pi Ai| , and a given closed path contributes as Bethe in 1931. Later other important quantum many- long as its length L is less than Lϕ (if L>Lϕ the elec- body problems were solved exactly, including (i) the one- tron would not have the phase coherence necessary to dimensional with delta-function interaction; (ii) exhibit interference). Thus, the magnitude of the weak the one-dimensional Hubbard model, which describes in- localization effect—which can be determined by “turning teracting electrons hopping on a lattice; (iii) the Kondo the effect off” by applying a magnetic filed—can be used model, which describes noninteracting electrons scatter- to infer the value of Lϕ or τϕ. ing from a localized magnetic impurity, and (iv) one-

4 dimensional spin-1/2 antiferromagnets with long-range their many phases as external parameters such as tem- (inverse square-law) exchange interaction. perature, pressure, and coverage are varied. Most bulk The formidable mathematical techniques required to phenomena, such as superconducitvity and magnetism, solve these problems exactly is more than compensated have surface counterparts that are interesting in their for by the rich phenomena exhibited by the solutions and own right. Surface science is blessed with an abundance by the new concepts they have introduced into mod- of surface-sensitive probes that have permitted extremely ern [10,11]. For example, the ex- detailed experimental studies. plicit solution of a one-dimensional model of interacting Fermions, known as the Tomonaga-Luttinger model, has excited-state solutions describing noninteracting bosons, B. Heterostructures and no “dressed” electrons or as predicted from the general picture of interacting Fermi systems known as Fermi theory. Understanding this solv- Semiconductor heterostructures, which are simply able model led in the 1980’s to a new general theory of atomically abrupt interfaces between two different semi- one-dimensional systems known as the- conductors, can be thought of as building blocks for more ory and to a powerful new mathematical technique called elaborate nanostructures [13]. The most important quan- bosonization. tities characterizing them are the conduction band and valence band offsets, which determine the height of the potential steps or discontinuities, seen by electrons and E. and Beyond holes. They can be classified into two categories: In type I heterostructures both electrons and holes are attracted One of the reoccuring themes in low dimensional elec- to the same side of the interface whereas in type II they tron systems is whether or not the system is a Fermi are attracted to opposite sides. liquid. As mentioned above, this term refers to a specific A single undoped cannot do much; it theory of interacting Fermi systems originally proposed has no free charge carriers and has a bulk-like DOS. by in 1957, for which (along this his pio- However, by adding donor impurities to the high-energy neering work on superfluidity) he was awarded the 1962 or barrier side of the heterostructure, electrons will es- Nobel prize. The modern definition of a Fermi liquid cape to the low-energy side of the interface, leaving the is a system of fermions (usually electrons) where the ef- then-ionized impurity centers behind. The dipolar charge fect of electron-electron interactions are sufficiently tame distribution formed from the electrons and the ionized that they can be accurately described with perturbation donors creates an electric field that confines the elec- theory. A Fermi liquid is in this sense “close” to a non- trons to the interface and bends the conduction band interacting electron gas. (and all bands) into a triangiular shape. If the den- In low dimensional systems, especially one-dimensional sity of electrons is low enough only the lowest transverse systems, Fermi liquid theory and many-body perturba- mode of the heterostructure will be occupied, resulting tion theory can break down, leading to what is called a in a two-dimensional electron gas (see below). This tech- non-Fermi liquid. The paradigm for non-Fermi is nique of modulation doping, a term that applies anytime the Lutinger liquid, which will be discussed further be- the dopant atoms are placed away from the region where low. the electrons or holes reside, minimizes the scattering of charge carriers from the ionized impurities, and leads to systems with exceptionally high mobility. IV. TWO-DIMENSIONAL QUANTUM SYSTEMS

Two-dimensional systems are usually the easiest to C. Quantum Wells and Superlattices make in the laboratory, and, as discussed above, are the starting point for most lower-dimensional systems. Putting two heterostructures together makes a quan- tum well, which, when undoped, looks like the one- dimensional square-well potential familiar from elemen- A. Surface Science tary quantum mechanics. The depth of the well in the conduction band is equal to the conduction band offset, The study of atomically clean surfaces has emerged as and the same holds for the valence band. The main dif- a major area of modern and ference is that the motion is unrestricted in the two direc- chemistry [12]. Topics of current interest include the tions perpendicular to the growth direction. Thus, single- growth of surfaces (for example, by molecular-beam epi- particle eigenstates are labelled by a transverse mode taxy) the physical and chemical properties of atoms ad- quantum number for the motion in the growth direc- sorbed on surfaces, the physical and electronic structure tion, an in-plane wave vector, and a spin projection. The of the surfaces themselves, and the transitions between transverse modes form overlapping energy bands called

5 subbands, because they are bands within the conduc- 1. The Electron tion band. As electrons are added to the quantum well its shape changes because of the potential produced by Electrons in a metal or doped semiconductor are usu- the carriers themselves, but the properties remain qual- ally described as being in a liquid or gas-like phase, re- itatively unchanged. At low enough densities, only the flecting the absence of long-range positional order. How- lowest transverse mode (or subband) is occupied, and the ever, in 1934 Eugene Wigner predicted in that an electron motion again becomes two-dimensional, the motion along gas of low enough density, moving in a uniform back- the growth direction being completely quenched. ground of positive charge, should crystallize. The reason Quantum wells don’t have to be square in shape; for this is that at low densities the electron-electron inter- there is nothing that prevents one from varying the action energy becomes larger than the kinetic energy, and semiconductor’s composition during growth to produce by forming a crystal the interaction energy is minimized. other confining potentials. For example, parabolic quan- In three dimensions (and zero magnetic field) the Wigner tum wells have been made that mimic the fimiliar one- crystal is predicted to form, at zero temperature, at elec- 4 ∗ 3 −1 dimensional harmonic oscillator. tron number densities less than n3D = [ 3 π(rsaB) ] , ∗ 2 ∗ 2 Putting many quantum wells together by alternating where rs ≈ 67. Here aB ≡ κ¯h /m e is the effective Bohr radius in a solid with constant κ and effec- layers of well and barrier materials leads to a periodic ∗ square-well potential known as a superlattice, which is tive mass m . The density in ordinary three-dimensional analogous to a common textbook example of a one- is much too high to form a Wigner crystal. In dimensional “crystal” called the Kronig-Penny model two-dimensions (again zero magnetic field) the critical ∗ 2 −1 [14]. The term superlattice refers to the fact that there density is n2D = [π(rsaB) ] , with rs ≈ 37. In semicon- are two different lattices present, the microscopic one ductors the effective Bohr radius can be much larger than formed by the atoms and the larger one consisting of the that in a vacuum, making the required densities easier to periodic array of quantum wells. The periodic square- achieve. well potential leads to the formation of energy bands in- A strong magnetic field suppresses the kinetic energy of side each original band, called minibands, for the same the electrons, enabling a crystalline state at higher den- reason that the periodic potential in a crystalline solid sity. The physics of Wigner crystals in strong fields is a forms the original bands in the first place. The energy rich and interesting subject with a lot of current activity gaps between the minibands are called minigaps. [15].

2. The Metal-Insulator Transition

D. The Two-Dimensional Electron Gas As mentioned above, experiments have observed metallic states and transitions between metallic and insulating states in low-density high-mobility two- dimensional electron systems [16]. This was a great sur- The two-dimensional electron gas has made a tremen- prise, and the first reports, made in 1994, were initially dous impact on modern condensed matter physics, mate- treated with some skepticism. In 1958 Phillip Ander- rial science, and, to a lesser extent, son showed that a three-dimensional electron gas could and quantum field theory. It lies at the boundary be- undergo a “metal-insulator” transition to an insulating tween three-dimensional electron systems, which are gen- state in the presence of sufficiently strong disorder. The erally Fermi liquids, and one-dimensional systems, which insulating state is said to be localized because the single- are not. In fact, where the two-dimensional electron gas particle wave functions become localized in space rather stands in this regard is not known at present, although than being extended throughout the solid. Then David most indications are that is it a Fermi liquid (or possibly Thouless and Franz Wegner introduced the idea that the what is known as a marginal Fermi liquid). It also lies metal-insulator transition could be regarded as a quan- at another boundary, that between metals and insula- tum and they studied the problem us- tors. The three-dimensional electron gas in the presence ing scaling methods developed for the study of critical of disorder (impurities and defects) can be an electrical phenomena. In 1979, Elihu Abrahams and collaborators conductor or insulator, but in the presence of any disor- developed what is now called the scaling theory of lo- der a one-dimensional system is always an insulator. Pre- calization, which applies to noninteracting electrons and viously it was thought that the two-dimensional electron which predicts no metal-insulator transition in two di- gas was insulating, but recent experiments have shown mensions. Experiments in the early 1980 were consistent that at low enough densities it can in fact be conducting. with that prediction. This and a few of the many other exciting areas of two- There is no consensus yet for what the correct theory dimensional electron gas research are briefly described of two-dimesnsional disordered systems is. However, be- below. cause electron-electron interaction was not incorporated

6 in the 1979 scaling theory of localization, it is reasonable nonnteracting case and leads to a quantized transverse to believe that the metal-insulator transition is driven by and vanishing longitudinal conductance. Laughlin also interactions, and there are good experimental indications predicted that at filling factors close to ν = 1/q, with q of this as well. an odd integer, the ground states would consist of his correlated state plus quasiparticles or quasiholes hav- ing a fractional charge of magnitude e/q. Not only 3. The Quantum Hall Effect are these particles fractionally charged, they obey “frac- tional” or anyon statistics, intermeditate between Bose- Einstein and Fermi-Dirac statistics. Soon after Laugh- The quantum Hall effect is an extremely rich and ac- lin’s work, Duncan Haldane and Bertrand Halperin were tive area of research that can only be touched on here able to explain most of the other observed fractions by [17]. There are two effects, called the integral and frac- assuming that the quasiparticles themselves form corre- tional quantum Hall effects. They occur when the two- lated Laughlin states when their densities reach certain dimensional electron gas is placed in a strong perpendicu- values. lar magnetic field. The eigenstates of noninteracting elec- Laughlin’s fractionally charged quasiparticles are be- trons in such a field forms highly degenerate bands known lieved to have been observed experimentally, first by as Landau levels. The number of degenerate states in Vladimir Goldman and Bo Su in a 1995 tunneling exper- each Landau level is equal to the number of flux quanta iment, and more recently by two groups measuring the in the system BA/Φ0, where B is the field strength and shot noise (current fluctuations) in the fractional quan- A is the system area. The Landau levels are separated in ∗ tum Hall effect regime. The proposed fractional statistics energy by the cyclotron energyhω ¯ c, where ωc ≡ eB/m c. has not been directly observed yet. Robert Laughlin was If the magnetic field is strong enough all the electrons in awarded the 1998 for his profound the two-dimensional electron gas can be accomodated in work in this area. the lowest Landau level. Is is convenient to measure the There will not be room to do justice to the incredible electron density n in units of the density that fills ex- amount of important work that has been done on this actly one Landau level; this is called the filling factor ν, subject. But it is worth pointing out the physical reason which can be written as nhc/eB. The filling factor can be why the properties of the two-dimensional electron gas in controlled experimentally either by changing the electron a strong magnetic field are so unusual and often exotic. density or by changing the magnetic field. The reason is that electron-electron interactions are, in a When ν is equal to an integer it is observed (at low certain sense, much stronger than in an ordinary metal. temperatures) that the transverse or Hall conductance A dimensionless measure of the strength of interactions of the electron system is, to an extremely high precision, 2 2 is the ratio of the typical interaction energy, say e /r0, e equal to ν h , independent of any microscopic details such where r0 is the average interparticle distance, to the typ- as the nature and strength of the disorder. (The trans- ical kinetic energy of electrons at the Fermi energy. In verse conductance is the ratio of the current through the an ordinary metal the latter would simply be the Fermi sample divided by the voltage drop across the sample in energy, but in quantum Hall effect systems where the the direction perpendicular to the current flow). This in- electrons are in the lowest Landau level (ν ≤ 1) there is tegral quantum Hall effect was discovered experimentally no real kinetic energy, because all electrons in the low- in 1980 by and coworkers, for which est Landau level have the same energy, which, without he was awarded the 1985 Nobel prize in physics. At the loss of generality, can be taken to be zero. In this case same time that the transverse conductance is quantized, the dimensionless ratio is infinite! Thus, the quantum the ordinary longitudinal resistance vanishes. This in- Hall effect system is one where many-body effects are tegral quantum Hall effect can be understood to be a extremely important. consequence of the Landau-level structure of the nonin- teracting . More surprising was the experimental discovery start- 4. Composite Fermions ing in 1982 of states with quantized Hall conduc- tance at certain fractional values of ν, such as 1 2 3 3 2 1 2 3 4 5 4 3 2 7 4 5 , 9 , 13 , 11 , 7 , 3 , 5 , 7 , 9 , 9 , 7 , 5 , 3 , 9 , 5 , and so on. In a In 1989 Jainendra Jain proposed a theory unifying brilliant 1983 paper Robert Laughlin explained the sim- the integral and fractional quantum Hall effects [18,19]. plest fractional states (ν = 1/3 and 1/5) by essentially The idea is to make an exact transformation whereby guessing the exact many-body for the an even number of fictitious magnetic flux tubes, each ground and lowest extited states. His work showed that of strength hc/e, are attached to each electron. The re- at these special densities the two-dimensional electron sulting composite object (electron plus flux tubes) obeys system forms a highly correlated liquid-like ground state Fermi statistics and is called a composite Fermion. The with an energy gap of order 0.01 κe2/ℓ, where ℓ ≡ ¯hc/eB Fermi statistics in this case is a consequence of the un- is the magnetic length. This energy gap has the same derlying Fermi statistics of the bare electron combined effect as the energy gap between Landau levels in the with the Aharonov-Bohm phase shift produced by the

7 flux tubes; had an odd number of flux tubes been mathe- of this article. However, a few introductory statements matically attached the resulting object would instead be can be made. a composite boson. First, the superconducting state itself seems to be The flux-attanchment transformation is an exact of the same type as in conventional superconductors, transformation of the two-dimensional electron gas prob- where electrons (or holes) acquire an attractive inter- lem, and the many-body Hamiltonian is now much more action causing them to bind into bosonic pairs which complicated that it was originally. However, when then Bose condense. This description of superconduc- a mean field approximation is applied to this trans- tivity, now called BCS theory, was developed by John formed Hamiltonian an integral quantum Hall effect of Bardeen, , and Robert Schrieffer, for which the composite fermions results. This technique provides they won the 1972 Nobel prize in physics. Beautiful ex- a method to reliably calculate properties of fractional periments measuring the spontaneous magnetization of quantum Hall effect systems. The physics underlying the YBa2Cu3O7 rings have determined that the pair wave fractional quantum Hall effect is sufficiently complicated function has d-wave symmetry, the same symmetry as a that it cannot be described by a mean field theory for d orbital. the original Hamiltonian. The magic of the composite When undoped, the CuO layers are antiferromagnetic Fermion transformation is that it makes an exact trans- insulators. Electrons removed from them allow in-plane formation of the many-body Hamiltonian to a form where conduction by holes. As a hole moves around the sur- simple analysis become adequate. rounding spins must constantly readjust themselves to Even more interesting things happen when the trans- lower their antiferromegnetic exchange energy. It turns formation is applied to non-quantized Hall states, for out that the exchange energy can be kept lower if the 1 example when ν = 2 . In this case there are two real holes move together. Thus, the background antiferro- flux tubes per electron in the system. Now two fake flux magnetism is at least partly responsible to the pairing of tubes are attached, oriented opposite to the real ones, the holes. to each particle. On average, then, the fictitious mag- The non-superconducting state of these novel materials netic field cancels the real field, leaving no magnetic field has turned out to be at least as unusual as the supercon- left. Composite fermion theory then predicts that the duting state. For example, it is not even clear whether two-dimensional electron gas, although in a very strong the normal states are Fermi liquids or non-Fermi liquids. perpendicular magnetic field, should behave as though Certain common features in these materials, for example there were no field at all. This astonishing effect has the temperature dependence of the resistivity, have no been observed in a series of beautiful experiments. explanation to date.

E. High-Temperature Superconductors F. Two-Dimensional Magnetism

The phenomena of , and the related The discovery of high-temperature superconductors phenomena of superfluidity, are extremely interesting with antiferromagnetic CuO planes has also stimulated subjects that are covered in a number of excellent intro- new interest in two-dimensional magnetism, especially in ductory texts [20,21]. Superconductivity was discovered two-dimensional Heisenberg antiferromagnets. in 1911 by Kamerlingh Onnes, who observed an abrupt The Mermin-Wagner theorem, which states that a con- transition to a resistance-free state in mercury near 4 K, tinuous symmetry (in this case, rotational symmetry) and, since then, there has been a search for materials that cannot be broken in a two-dimensional system at fi- become superconducting at higher temperatures. In 1986 nite temperature, requires that magnetism in the two- a new class of oxides were discovered by George Bendorz dimensional Heisenberg model exist, if at all, at zero tem- and Karl M¨uller that have superconducting transition perature only. temperatures as high as 130 K. Bendorz and M¨uller were Whether or not there is long-range order at T = 0 de- awarded the 1987 Nobel prize in physics for this discov- pends on the parameters in the Hamiltonian, and this ery. leads to an interesting situation: By varying parame- These high-temperature superconductors have weakly ters in the Hamiltonian, the quantum ground state can connected two-dimensional sheets composed of Cu and change, say, from paramagnetic to antiferromagnetic. O, and most of these materials assert that the This type of transition is called a quantum phase tran- physics responsible for the superconductivity is a conse- sition because it is analogous to thermodynamic phase quence of these CuO4 planes. Thus, in this sense, high- transitons between phases that occur as the temperature temperature superconductors are also examples of low is varied, but here the temperature is always zero. dimensional quantum systems. In contrast with the Heisenberg model, a finite- There is no consensus yet as to the microscopic origin temperature magnetic state of the two-dimensional Ising of high-temperature superconductivity, and a discussion ferromagnet, which does not have continuous symmetry, of the competing proposals is beyond the intended level does exist.

8 The study of low-dimensional quantum magnetism tubes, isolated multi-wall tubes, or in bundles or ropes continues to be an exciting and active area of re- of parallel tubes. search. Sophisticated mathematical and computational Both their structural and electronic properties are un- techniques have been developed to extract their physical usual. They are predicted to be the strongest fibers properties [22]. known, and do not break when bent. In fact, the tube wall appears to simply straighten out after the bending forces are released. Great progress has been made in V. ONE-DIMENSIONAL QUANTUM SYSTEMS obtaining a microscopic understanding of the structural properties of carbon nanotubes. One-dimensional systems possess some of the most Much is also known about the transport properties exotic phases of condensed matter. In these systems, of nanotubes; however, there remains many unanswered the properties of even weakly interacting particles dif- questions. The sheets of carbon making up the tubes, fer dramatically from that of noninteracting ones, and when undoped, are insulators. When they are rolled into the generic conducting state of one-dimensional bosons, tubes they can be metals or insulators, depending on the fermions, or even spins is a Luttinger liquid instead of a tube diameter and the amount of twist, if any, introduced Landau Fermi-liquid. during their creation. Actually, it is surprising that a nanotube, if they are regarded as a one-dimensional crys- tal, can be conducting at all, because one-dimensional A. Quantum Wires metallic crystals are usually unstable to the formation of a periodic lattice distortion or strain that lowers their en- ergy but makes them insulating. Their tubular geometry, Quantum wires are extremely narrow wires where elec- however, is believed to make them sufficiently resistant tron motion is allowed in one direction, along the wire, to longitudinal strain. but confined in the other two directions. Most often Then there is the question of electron-electron inter- they are created by putting metallic gates on top of a action effects, which are usually very important in one- two-dimensional electron gas and applying voltages to dimensional systems. Luttinger liquids (see below) have deplete the electron gas underneath. This can produce been predicted to occur in nanotubes, complicated by the quantum wires of varying length whose width can be con- fact that the conducting channels (for each spin orienta- trolled during an experiment. In fact, varying the width tion) come in pairs. Electron-phonon interaction is also changes the number of transverse modes that have ener- believed to be strong in nanotubes, and there is even the gies below the Fermi energy and that contribute to the possibility of turning them into superconductors. conductance of the wire. The quantum wires come com- plete with electrical contacts, made from the underlying two-dimensional electron gas, which allow transport mea- surements to be made. C. The Luttinger Liquid State The Landauer formula, described above, predicts that 2 the conductance of such a wire is equal to 2e /h times In the 1950’s and 1960’s Sin-itiro Tomonaga and the number of transverse modes below the Fermi energy. Jaoquin Luttinger studied theoretical models of one- Thus, as the width of the wire is varied, the conductance dimensional metals (similar to quantum wires) and found 2 is expected to be quantized in units of 2e /h. By “quan- that they possessed properties entirely different than tized conductance” one means that the conductance is three-dimensional metals and different than that ex- 2 equal to 2e /h times an integer, and that integer in- pected by Landau’s Fermi liquid theory. In particular, creases as the wire is made wider. the low-energy eigenstates described elementary excita- This has been observed in many pioneering exper- tions that are bosons, not fermions. These bosons are iments, but recent experiments (on wires made using basically phonons (or quanta of the sound waves) of the the cleaved-edge-overgrowth technique) show quantiza- electron gas. tion in multiples of approxiately 0.7 times that which is The breakdown of Fermi liquid theory in one dimen- expected. These intriguing experiments are not under- sion is now known to be the norm, and the generic non- stood at this time. Fermi-liquid in one dimension is called the Luttinger liquid. The simplest type of Luttinger liq- uid is characterized by a single dimensionless parameter B. Carbon Nanotubes g, which, roughly speaking, characterizes the degree to which the system deviates from a Fermi liquid, defined Nature has provided beautiful molecular quantum as g = 1. The value of g is determined by the parame- wires called carbon nanotubes, formed by rolling car- ters characterizing the electron gas, like its density and bon sheets of hexagonal symmetry into tubes of varying electron-electron interaction strength. diameter [23]. They can have diameters as little as a Luttinger liquids are interesting because of their exotic nanometer, and can be produced as isolated single-wall behavior, including an unusual conductivity as a func-

9 tion of temperature, voltage, and frequency, and their Luttinger liquid behavior has been observed in quan- extreme sensitivity to disorder. Experimental efforts to tum Hall effect systems, but the values of g measured do observe them in the laboratory have been unsuccessful not generally agree with the theoretical predictions. This until only very recently. is another interesting unsolved problem in the physics of low-dimensional quantum systems.

D. Edge States of the Quantum Hall Fluid E. Spin Chains The most obvious place to look for a Luttinger liquid is in a . This has been attempted for years Spin chains are one-dimensional lattices of spins. without success because it has not been possible to make Although they occur in nature, embedded in three- clean enough quantum wires and disorder ruins the Lut- dimensional materials such as Mn(HCOO)2 · 2H2O, tinger liquid state, making it insulating. The quantum CsNiCl3, and RbNiCl3, they are also of great theoreti- Hall state of a two-dimensional electron gas, however, has cal interest because they exhibit dramatic disorder and provided an alternative. quantum fluctuation effects that occur in higher dimen- The electrical current present in an ordinary metal sions as well. One-dimensional spin systems also have a (when connected to a battery, for example) is just the large class of exactly solvable versions and can be effi- sum of the current carried by each electron. At the same ciently studied numerically, giving them certain advan- time, when the metal is in its ground or equilibrium state tages over higher-dimensional magnets. there is no net current flowing (assuming a macroscopic One of the most interesting quantum-fluctuation ef- conductor so there is no persistent current). This means fects in spin chains was discovered by Duncan Haldane that in thermodynamic equilibrium the currents from all in 1983. In higher-dimensional antiferromagnets it is be- the electrons present must sum to zero. Accordingly, to lieved that the ground states can be antiferromagneti- describe the current-carrying state of a metal it is suf- cally ordered and support a gapless spin-wave spectrum. ficient to keep track of changes of the state occupation Although true long-range order is prohibited in one di- numbers from their equilibrium values. mension, quasi-long-range order, where the spin-spin cor- In the linear-response regime, where the metal is only relation functions decay at large distances as a power- slightly out of equilibrium, only the occupation numbers law, is allowed, and it was assumed that antiferromag- near the Fermi energy change. This is why transport netic spin chains also had gapless spin-wave excitations. properties like resistance are referred to as “Fermi sur- What Haldane predicted, and what was later established face” properties. experimentally and by numerical studies, was that this Now, in the quantum Hall effect system, when the picture was correct for spin chains with half-integer spins 1 3 5 conditions are such that a quantized Hall conductance ( 2 , 2 , 2 , · · ·) but that integer spin system (1, 2, 3, · · ·) have is observed, the states at the Fermi energy are physi- finite energy gaps. cally located at the edges of the two-dimensional elec- tron gas. (In the bulk of the system, away from the edges, the Fermi energy is in an energy gap—either be- VI. ZERO-DIMENSIONAL QUANTUM SYSTEMS tween Landau levels or in the energy gap of many-body origin described by Laughlin.) As in an ordinary metal, Zero-dimensional systems are confined in all three di- to describe the current-carrying state it is sufficient to rections, and, as such, have discrete energy spectra (iso- keep track of the states at the Fermi energy, which are lated energy levels seperated by gaps). called “edge states.” The edge state concept has been extremely helpful in understanding strong-field magneto- transport in the quantum Hall effect regime and in other A. Quantum Dots low-dimensional electron systems [24]. Edge states corresponding to fractional quantum Hall effect states are especially interesting, because they are Quantum dots are nanometer-scale structures, often predicted to be Luttinger liquids. However, there some made from gated and compositionally graded semicon- important distinctions between these Luttinger liquids ductors, that can hold a few to a few thousand elec- (called chiral Luttinger liquids) and the Luttinger liquids trons [25,26]. In many respects they are like artificial described above: First, the value of g in a chiral Luttinger atoms, the inverse-square-law Coulomb force of the nu- liquid is predicted to be simply related to the filling fac- cleus being replaced by a linear, harmonic-oscillator-like tor ν of the electron gas, so it can be determined rather force provided by the confinement. However, unlike nat- easily. Second, g can be made to differ appreciably from ural atoms, the number of electrons in a quantum dot 1, making the associated non-Fermi-liquid effects larger. can be precisely controlled and varied during an exper- And third, edge states can be made to be entirely insen- iment, allowing, in some sense, one to “sweep” through sitive to disorder [24]. the periodic table and beyond.

10 Like other nanostructures, quantum dots are stud- the conduction electrons interact very strongly with the ied with both transport measurements and optical spec- spins of the electrons in the dot, enhancing the conduc- troscopy, with no magnetic field and in large fields. They tance of the dot when there is an odd number of electrons are interesting both from a fundamental physics point-of- in it. Because quantum dots are somewhat like atoms, view and for their potential applications. Applications several classic phenomena have also been include transistors that control currents at the single- observed in them. electron level, highly efficient low-power lasers, and even circuit elements for quantum computers (see below). 2. Quantum Computers

1. and the Single Electron Quantum dots in the near future may even become circuit elements for a revolutionary type of computer. A simple yet fascinating effect occurs when one tries Quantum computing is a new multidisciplinary sub- to pass current through a quantum dot: From a classical ject of great current interest to university researchers, point-of-view, transport through the dot has to occur via and also of great importance to government agencies and a discrete change of charge of the dot, because the total the information and technology industries [27]. The term charge of N electrons in the dot is −Ne, where −e is the “quantum computing” refers to the possibility of build- charge of one electron. But this means that the electro- ing a computer out of elements, called quantum logic static energy (Ne)2/2C, where C is the capacitance of gates, whose operation exploits the laws of quantum me- the quantum dot (approximately equal to its diameter), chanics to perform operations prohibited by conventional changes by a finite amount, which, for small enough dots, or classical logic gates. There is also a related subject can be larger than the thermal energy kBT and the en- called quantum information science in which the quan- ergy supplied by the battery eV . Here kB is Boltzmann’s tum properties of matter are to quantify, store, encode, constant, T is the temperature, and V is the applied and communicate information. Recent theoretical and voltage. In this situation no current can flow, a phenom- experimental breakthroughs in these subjects have at- ena called Coulomb blockade because it originates from tracted computer scientists, mathematicians, physicists, Coulomb interaction between electrons. and chemists [28]. This simple classical picture is not entirely correct, be- Whereas the basic unit of information in a classical cause quantum mechanically an electron can tunnel from computer is a binary digit or bit, a “0” or a “1” say, a one lead to the other, through the quantum dot, violat- quantum computer processes information in the form of ing energy conservation on the dot for a short period of a coherent superposition |ψi = α|0i+β|1i, called a quan- time. However, this is a small effect (although interest- tum bit or qubit. Thus, the elementary component in a ing in itself) for a quantum dot very weakly connected quantum computer is the two-level system familiar from to leads. quantum mechanics. A novel transistor can be made by adding a third lead There are two major efforts in quantum computing re- to the quantum dot. Unlike the first two leads, how- search. One emphasis is to assume that one has a number ever, the third lead is not connected to the dot itself, but of qubits that can be manipulated, and ask what they rather, to a metallic gate near the dot. Varying the volt- can be used for. Are there calculations that a quantum age VG on this gate changes the electrostatic energy of computer can do much faster than a classical computer? the electrons in the dots, and can be used to lower the The answer is yes: If a quantum computer could be built Coulomb blockade barrier. it could factor large numbers that would be essentially The transistor works as follows: In the presence of imposible to factor with a classical computer using the the metallic gate the classical electrostatic energy (when best known algorithms. And this assumed impossibility there are N electrons in the quantum dot) is E = of factoring large integers is the basis for current encrip- 2 2 N e /2C − eNVG. For a given value of VG the opti- tion methods used by the military and by banks. mum number of electrons on the dot (the number that The other emphasis is try to design and build a quan- would minimize E) is VGC/e. However, VGC/e is gener- tum computer. The qubit referred to above is of course ally not an integer, so N takes the value of the integer an abstract two-level system. What is the best physical closest to VGC/e. As the gate voltage is continuously realization of the qubit? Certainly one wants a qubit with varied, then, the number of electrons in the dot and the a long phase-coherence time τφ (see above), because this number passing through the leads changes one at a time. is the time during which the system remains quantum We therefore have a transisitor operating at the level of mechanical and during which a quantum computation single electrons! can be performed. Second, one has to figure out how to A number of other interesting phenomena have been manipulate qubits and have them interact to perform an observed in quantum dot systems. A beautiful example actual calculation. Electron spin states in quantum dots is the recent observation of the Kondo effect, which is have emerged as excellent candidate qubits. a low-temperature many-body effect where the spins of Quantum computing is very new and it is hard to pre-

11 dict if it will be successful in the near term. But if a Metallic nanocrystals have also been actively investi- quantum computer is successfully built, it will be revo- gated. They can be made out of superconductors, al- lutionary. lowing for the study of mesoscopic superconductivity. Making nanocrystals out of magnetic materials leads to fascinating new magnetic properties that may have use- 3. Exact Diagonalizaton ful data-storage applications. Metallic also have unusual optical properties because they are easily polarizable, and the propagation of an electromagnetic A theoretical technique that is particularly well suited wave though a cluster of metallic nanoparticles probes to study electrons in quantum dots is the direct numerical the electromagnetic modes of that cluster. solution of the many-body Schr¨odinger equation. This is usually done by expanding in a basis of noninteracting eigenstates, so the computational problem becomes that of finding the eigenvalues and eigenvectors of huge ma- trices, often called exact diagonalization. It is not tru- ely exact because it involves a truncation of the original infinite-dimensional Hilbert space, but this can often be [1] Physics of Low-Dimensional Semiconductor Structures, done in a controlled manner, yielding results that are edited by P. Butcher, N. H. March, and M. P. Tosi insensitive to that truncation. (Plenum Press, New York, 1993). Exact numerical studies of quantum many-body sys- [2] , edited by G. Timp (Springer-Verlag, tems are usually impossible because it is only possible to New York, 1999). handle a dozen or so particles. But small quantum dots [3] G. J. Milburn, Schr¨odinger’s Machines: The Quantum can be studied experimentally, making exact diagonaliza- Technology Reshaping Everyday Life (W. H. Freeman and tion methods extremely useful. Company, New York, 1997). [4] J. M. Ziman, Theory of , 2nd ed. (Cambridge Uni- versity Press, Cambridge, 1972). [5] Y. Imry, Introduction to Mesoscopic Physics (Oxford B. Artifical Molecules University Press, Oxford, 1997). [6] J. J. Sakurai, Modern Quantum Mechanics (The Ben- Another new area of research is the study of pairs of jamin/Cummings Publishing Company, Menlo Park, quantum dots, which then can be thought of as artifi- 1985). cial molecules. Just as in natural molecules, bonds can [7] S. Olariu and I. I. Popescu, Rev. Mod. Phys. 57, 339 be formed between the atoms by electrons shared be- (1985). tween dots. These systems permit controlled studies of [8] S. Washburn, in Mesoscopic Phonomena in Solids, edited (at least some) basic molecular phenomena. One problem by B. L. Altshuler et al. (Elsevier Science Publishers, New of current interest is the magnetic properties of artifical York, 1991). molecules. [9] C. W. J. Beenakker and H. van Houten in Solid State Physics: Advances in Research and Applications, vol. 44, edited by H. Ehrenreich and D. Turnbull (Academic Press, San Diego, 1991). C. Nanocrystals and Nanoparticles [10] D. C. Mattis, The Many-Body Problem: An Encyclope- dia of Exactly Solved Models in One Dimension (World There is considerable interest in nanometer-scale crys- Scientific, Singapore, 1993). tals containing, say, from 103 to 106 atoms [29]. These [11] Z. N. C. Ha, Quantum Many-Body Systems in One Di- can have structural, electronic, and optical properties mension (World Scientific, Singapore, 1996). very different than bulk solids. Spherical or nearly spher- [12] A. Zangwill, Physics at Surfaces (Cambridge University ical nanocrystals are sometines called nanoparticles. Press, Cambridge, 1988). [13] G. Bastard, Wave Mechanics Applied to Semiconduc- Semiconducting and insulating nanorystals have many tor Heterostructures (Les Editions de Physique, Cedex, similarities to quantum dots. In fact, the electronic prop- 1988). erties are often the same, because in both cases the elec- [14] C. Kittel, Introduction to Solid State Physics, 7th ed. tron is confined in all three directions. However the vi- (John Wiley and Sons, New York, 1996). brational properties can be quite different: A quantum [15] H. A. Fertig, in Perspectives in Quantum Hall Effects, dot is not usually mechanically isolated from a bulk solid, edited by S. Das Sarma and A. Pinczuk (John Wiley and so it supports a continuum of low-frequency vibrational Sons, New York, 1997). modes. But an isolated nanocrystal would have a dis- [16] See Search and Discovery article by G. B. Lubkin in crete vibrational spectrum, much like that of a drum Physics Today, July 1997, page 21. head. This means that any physical property involving [17] The Quantum Hall Effect, edited by R. E. Prange and S. phonons will be very different than that in a macroscopic M. Girvin (Springer-Verlag, New York, 1987). solid. [18] J. K. Jain, Physics Today, April 2000, page 39.

12 [19] Composite Fermions: A Unified View of the Quantum edited by M. Read (Academic Press, San Diego, 1992). Hall Regime, edited by O. Heinonen (World Scientific, [25] L. Jacak, P. Hawrylak, and A. W´ojs, Quantum Dots Singapore, 1998). (Springer-Verlag, Berlin, 1998). [20] D. R. Tilley and J. Tilley, Superfluidity and Superconduc- [26] U. Woggon, Optical Properties of Semiconductor Quan- tivity, 3rd ed. (Institute of Physics Publishing, London, tum Dots (Springer-Verlag, Berlin, 1997). 1990). [27] G. J. Milburn, The Feynman Processor: Quantum [21] M. Tinkham, Introduction to Superconductivity, 2nd ed. Entanglement and the Computing Revolution (Perseus (McGraw-Hill, New York, 1996). Books, Reading, 1998). [22] A. Auerbach, Interacting Electrons and Quantum Mag- [28] Introduction to Quantum Computation and Information, netism (Springer-Verlag, Berlin, 1994). edited by H. K. Lo, S. Popescu, and T. Spiller (World [23] R. Saito, G. Dresselhaus, and M. S. Dresselhaus, Physical Scientific, Singapore, 1998). Properties of Carbon Nanotubes (Imperial College Press, [29] S. V. Gaponenko, Optical Properties of Semiconductor London, 1998). Nanocrystals (Cambridge University Press, Cambidge, [24] M. B¨uttiker, in Semiconductors and , vol. 35, 1998).

13