You are on page 1of 22

Quantised conduction

TN2992

Joeri de Bruijckere, Thomas Horstink, Stevie-Ray Janssen and Martijn Schmeetz

TN2992: Experimenteel eindproject

February 14, 2013

Contents
1 Introduction 2 Theory 2.1 Fermions . . . . . . . . . . . . . 2.1.1 Fermi-Dirac statistics . 2.1.2 Free electron gas . . . . 2.2 Conduction . . . . . . . . . . . 2.2.1 Classical conduction . . 2.2.2 Quantum conduction . . 2.3 2-Dimensional-Electron-Gas . . 2.4 Ballistic transport . . . . . . . 2.5 Quantised conductance . . . . . 2.6 Low Resistance Measurements . 3 Experiment 3.1 Introduction . . . . . 3.2 Sample description . 3.3 Measurement results 3.4 Conclusion . . . . . Acknowledgements . . . . 1 2 2 2 3 4 4 5 6 7 9 11 14 14 14 16 18 19

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

1. Introduction
In summary we have reported the rst measurements of the conductance of single ballistic point contacts in a two-dimensional electron gas. A novel quantum eect is found: The conductance is quantised in units of e2 / [1] [2]. This quantum eect was discovered by accident at the TU Delft university in 1988. Unexpectedly, plateaus were observed in the conductance. The step gain was measured to be equal to integer multiples of e2 / . The purpose of this report is to provide the reader with a theoretical basis concerning quantised conductance and to provide the reader with experimental results of the performed reproduction of the experiment mentioned above. This report is the result of minor students (listed on the front page) and is part of the TN2992 course at the TU Delft. The report is structured as follows: The rst part governs descriptions of the theories behind the quantum conductance and some of the elements required for the experiment, including: fermion characteristics, a classical and quantum mechanical approach on conduction, an explanation of 2DEG, the requirement for ballistic transport, an overview of the quantised conductance eect, concluding with some theory on low resistance measuring. The second part governs the experimental result, including sections dedicated to the setup, processing of the obtained results and a summary of conclusions of the reproduced experiment.

2. Theory
2.1 Fermions
The charge carriers in the sample of our experiment are, like in almost every electrical circuit, electrons. The Standard Model of particle physics treats electrons as elementary particles, which means that they are among the fundamental building blocks of our universe and cannot be divided into smaller particles. This idea is generally accepted among physicists; as the Standard Model in its entirety is. Still, many theories have been proposed, advocating the existence of even smaller particles: preons. But so far none have been found in experiments [3] and since the Standard Model has been so successful, preon models gain little interest nowadays. The elementary particles of the Standard Model have the property of being completely identical to particles of the same kind. You could, for example, not distinguish between two electrons; their DNA is exactly the same. This holds consequences for describing systems that have multiple elementary particles of the same kind. It makes that the wave function of the system cannot simply be written as a product of multiple one-particle wave functions, because you could not tell which particle belongs to which wave function. For such a system of multiple identical particles (for simplicity, we consider two: one in state a and one in state b ), a wave function has to be constructed as: (r1 , r2 ) = a (r1 )b (r2 ) b (r1 )a (r2 ) (2.1)

where the vectors r1 and r2 represent position coordinates of the particles. The normalisation constant is left out, for its little importance in this explanation. It can be seen that there are two possible wave functions for this system of identical particles: one goes with a plus sign, in which case we call its particles bosons and the other goes with a minus sign, for which we call its particles fermions. Now, we suppose that the two identical particles occupy the same state, say a . For bosons, the two products in Equation (2.1) get simply added. But, for fermions something peculiar happens: (r1 , r2 ) = a (r1 )a (r2 ) a (r1 )a (r2 ) = 0. The wave function reduces to zero ! Therefore, the probability of nding two identical fermions in the same state, is zero. This remarkable conclusion is a fundamental principle in quantum mechanics and is known as the Pauli exclusion principle. It holds only for identical fermions. As we mentioned, the wave function for a system of identical bosons just adds the two products of wave functions, so these particles are not forbidden to occupy the same state and do not obey the exclusion principle. It happens to be that electrons are fermions, so they should also obey the Pauli exclusion principle. In atoms, this is what prevents the electrons from decaying all into their ground states, which would decrease the diversity in all our atoms and would make a very dull universe. By the exclusion principle, the electrons in our sample should all be in dierent states, having dierent combinations of spin and momenta. The energies are therefore varying from electron to electron. A particular branch of physics, called Fermi-Dirac statistics, describes these energies of single particles, for a system of multiple identical fermions. It can be used to determine the probability for fermions to be in a state with a certain energy.

2.1.1

Fermi-Dirac statistics

In a system that contains numerous electrons, like our sample, a lot of dierent possible states are occupied by the electrons. For large numbers of particles, we can use statistical mechanics to describe the system. Fermi-Dirac statistics are concerned with the energy states within a system of single 2

particles, that obey the Pauli exclusion principle. It describes how the electrons (or more generally: the identical fermions) occupy the dierent energy states. Fermi-Dirac statistics is only applicable to fermions. Bosons, particles that do not obey the Pauli exclusion principle, are described by a similar kind of statistics, called Bose-Einstein statistics. But, since we are dealing exclusively with electrons, we do not need Bose-Einstein statistics. According to Fermi-Dirac statistics the expected fraction of electrons n with a certain energy E is given by the equation: 1 n(E ) = (E )/k T (2.2) B e +1 where is the chemical potential, kB is the Boltzmann constant and T is the temperature. This equation is known as the Fermi-Dirac distribution. At zero temperature this distribution has a rectangular shape, as in Figure 2.1 (a), which means that all energy states are occupied up to a certain energy: the Fermi energy, indicated as EF in Figure 2.1. The Fermi energy is the energy corresponding to the highest occupied state at absolute zero. When the temperature increases, the shape of the distribution becomes smoother, as in Figure 2.1 (b). The Fermi energy is then no longer the highest energy in the system. The chemical potential is also called the Fermi level. This is the energy for which the occupation number is half the total number of particles. Close to absolute zero the Fermi level is equal to the Fermi energy and the terms get therefore often mixed up. When the temperature rises however, the Fermi level increases and the terms become less equal.

Figure 2.1: Fermi-Dirac distribution at absolute zero (a) and above absolute zero (b) Close to absolute zero temperature only the energy states around the Fermi energy are partially occupied. The states at lower energies are completely lled, so no electrical conduction can take place at these levels. That means that at very low temperatures, only the electrons having an energy close to the Fermi energy can be conducting. This is an important aspect in the understanding of quantised conductance and we will use this fact later on.

2.1.2

Free electron gas

The electrons in our sample can be treated by approximation as particles in a box, where they are not subjected to any forces. This idealised model is known as the free electron gas. Because of the heterostructure, the electrons in our sample are conned in one direction, so we actually have a two-dimensional free electron gas. If we suppose that the electrons are conned in the z -direction and the dimensions of the sample are of lengths lx and ly , the potential V (x, y ) for this system is: V (x, y ) = 0 if 0 < x < lx and 0 < y < ly otherwise (2.3)

Solving the Schr odinger equation for this system leads to wave functions in the following form: nx ny = sin nx x sin lx 3 ny y ly (2.4)

where nx = 1, 2, 3, ... and ny = 1, 2, 3, ... are positive integers, each combination representing a stationary state. That leaves us with a set of discrete stationary states. For simplicity the normalisation constant is omitted. Now, we introduce a vector that characterises the direction and magnitude of an electrons momentum: the wave vector, k (kx , ky ). Its magnitude is related to the wavelength by k 2/ and to the momentums magnitude by k p/ . One way of representing the stationary states of this system is by drawing them in a k -space as intersection points of the lines that satisfy kx = (/lx ), (2/lx ), ... and ky = (/ly ), (2/ly ), ... The intersection points representing states of equal energies lie on a circle in this k -space. The outermost circle, where the energies equal the Fermi energy EF , is known as the Fermi circle, drawn in Figure 2.2. For all points on the Fermi circle the magnitude of the wave vector k = kF .

Figure 2.2: Fermi circle in k -space with the possible stationary states indicated as dots. Actually, Figure 2 is a simplication of a k -space where we cannot draw a smooth circle through the dots (stationary states) with the highest energies. The resolution of the discrete points is too low and the drawn Fermi circle does not make a lot of sense in this case. In reality, the number of dots is a large number in the order of Avogadros number. Through such a large number of dots, it is possible to draw a very smooth Fermi circle, like the one in Figure 2.2. The circle represents the line between occupied and unoccupied states at absolute zero. The analogue for a three-dimensional system would be the Fermi surface, for the states at the Fermi energy would lie on a sphere.

2.2

Conduction

Electrical conductance is a quantity which denes a materials ability to conduct an electric current, often noted by the Greek letter . It is the reciprocal quantity of the better known quantity resistance A (). Conduction is expressed in the SI-unit siemens (S) and S = V . Conductance has in the past been described with classical mechanics and has correctly described certain eects like the Hall eect. But it also had its shortcomings, which lead to redescribing the phenomenon of electric conduction using quantum mechanics. We will use conduction in a metallic solid to describe the conductivity, which is essentially the movement of electrons in the solid. A metal in solid state arranges the atoms in a lattice where outer-shell electrons (valence electrons ) can freely move between the atoms. We say that a free electron gas is present in the solid, which enables the electrons to move within the solid. When a potential dierence is applied over the solid, the electric eld which is created by the potential dierence forces the electrons to move in the direction of the positive potential, creating a current.

2.2.1

Classical conduction

The rst description of the behaviour of the free electrons in a metal lattice was described by the Drude Model, proposed in 1900 by Paul Drude. Drude assumed the free electrons to act as a gas 4

within the lattice, so electrons move fast in random directions. With this assumption it is logical that the electrons often collide with the positively charged atoms (ions) in the lattice and other electrons moving through the solid. Keep in mind that these electrons move this way, even if there is no potential dierence applied. It is part of Drudes assumption that the electrons behave as a gas. The speed the electrons move with within the lattice is the thermal velocity. In Drudes model the collisions are considered isotropic so the kinetic energy equals the thermal energy which is solely depended on temperature T . The thermal energy of an electron can be derived by assuming that it can move freely 3 in 3 dimensions, as Ut = 2 kb T , equating this with the kinetic energy of an electron we get: 3 1 me v 2 = kb T 2 2 v= 3kb T me (2.5) (2.6)

Where kb is the Boltzman constant, T is the temperature and me is the mass of an electron. But now suppose we do apply that potential dierence over the metal. The electrons will be aected by a force pointing to the positive potential due the electric eld induced by the potential dierence. This force accelerates the electrons in the opposite direction of the eld, while their free motion properties stay intact. Now every time an electron collides with something, the electric elds resulting force pushes the electron in the opposite direction of the electric eld. What we observe is a so called pinball eect where the electrons do move in the direction of the force but still interact with the ions and other electrons in the form of collisions. The net result of this pinball machine is movement of the electrons (along quite a bumpy road) in the opposite direction of the electric eld with a small velocity, called the drift velocity vd . The electric current can now be described in two ways; using Ohms law, but also using the newly acquired vd . Remembering that electrical current is nothing more than the amount of charge that crosses an area per unit of time, we can say that: I= V Q = = neAvd R t (2.7)

Where n is the amount of electrons, e is the elementary charge, A is the cross-sectional area. Another aspect that can be used to describe conduction is the mean free path, , which is the average length of the path an electron can cover before it collides with an ion in the lattice. Assuming the magnitude of the electrons velocity equals roughly the thermal velocity, we can say that = v where is the average time between two collisions. We should note that the probability of an electron having a collision in a time interval dt equals dt [4] and is independent of speed and momentum but does rely on r, the radius of the ions in the lattice. Drudes model was a nice beginning but it had some fundamental aws, varying from the description of heat conduction, to the actual electric conduction and its dependency on temperature. By the 1930s quantum mechanics was used to adjust Drudes model to the free electron model, sometimes referred to as the Drude-Sommerfeld model.

2.2.2

Quantum conduction

The Drude model left out certain aspects that were not known at the time; most importantly the wavecharacteristics that particles have according to quantum mechanics, but also the quantised energies that electrons have when roaming freely through the lattice. To describe the behaviour of electrons in a free electron model using quantum mechanics, we say that electrons are in a potential well with innite barriers; the electrons do not leave the metal and are free to roam inside[5]. Now note that the electrons possible energy levels are quantised and well dened. They are the integer multiples of the ground state E0 . Drudes model states that only the valence electrons participated in 5

conduction. This remains the same in Sommerfelds model, except that these free electrons do not have 3 the average thermal energy 2 kb T but have several distinctive energy levels. Some of these electrons are at the highest energy level, the fermi energy EF . The electrons with energy EF participate in conduction, because these electrons are the only ones capable of jumping to the next empty energy level. Suppose the kinetic energy of lattice ions equals kb T . The maximum energy an electron can gain from a collision with such lattice ion is kb T ! Only the electrons within kb T of the Fermi energy can move to the next state. Since not all the electrons can occupy the high energy levels due to the Pauli exclusion principle there are only a few electrons that contribute to the conduction. So only a few electrons, with a specic energy EF , participate in the electric conduction. Now instead of using the thermal velocity (a result of the thermal energy ) we use EF . Equating EF to the kinetic energy we gain the Fermi velocity: 2EF vF = (2.8) me This change is fundamentally dierent from Drudes assumptions. The electron energy and velocity are not as dependent on T as they were before! This gets rid of a lot of inconsistencies between the Drude model and experimental results regarding electron velocity and temperature. Another important aspect when viewing electrical conduction with quantum mechanics is that the collisions between an electron and lattice ion are no longer considered as if two balls bounce o each other. We see the electron as a travelling wave through the lattice. Now it is logical that if the electrons wavelength is longer then the space between two ions in the lattice, the electron can move freely without colliding with the ions. Now, if the lattice contains impurities this will interfere with the electron thus adjusting the mean free path, . But these impurities rise from the thermal vibration of the ion, so is also no longer dependent on the radius of the ion but rather on the radius of its thermal vibrations, which logically is depended of T . This corrects the relation problems between and T which were present in the Drude model.

2.3

2-Dimensional-Electron-Gas

In order to study quantum conductance, a 2-Dimensional-Electron-Gas (2DEG) was used. A 2DEG is a gas of electrons in which the electrons can only move in 2 directions. Movement in the third dimension is restricted to certain quantised energy levels, which allows motion in that direction to be ignored. This 2DEG can then be used to represent a wire of which the conductance can be investigated. [6] [7] [8] [9] [10] [11] To create a 2DEG, a heterostructure of semiconductors was used. The rst is an n-type AlGaAs and the other is intrinsic GaAs. An n-type semiconductor is a semiconductor that has been doped. Doping is the addition of impurities (e.g. atoms of a dierent material than the semiconductor is made of) to the semiconductor crystal structure. There are two types of doped semiconductors: The p-type, in which atoms with fewer electrons than atoms in the semiconductor material are added to the crystal structure. This creates holes which can move through the structure and potentially create electric current. The n-type, in which atoms with more electrons than atoms in the semiconductor material are added to the crystal structure. The semiconductor then contains excess electrons which can freely move within the material. An intrinsic semiconductor on the other hand, is a semiconductor made up of a pure element or compound. So it contains no impurities whatsoever. In metals, electrons can change energy levels quite easily, because of the low energy costs. They have very good conductivity, because there are numerous available energy states at the Fermi level. In semiconductors, the electrons can only have energy levels in so called energy bands. These energy

bands are separated by so called gaps. Such a gap is a measure for the energy needed to jump to a higher energy band. The energy bands are being lled from the bottom following the Pauliexclusion principle, which states that no two quantum-fermions can occupy the same quantum state simultaneously.

Figure 2.3: The dierence between the energy bands of an intrinsic semiconductor and an n-type semiconductor. The n-type semiconductor has an additional dopand band, containing the excess free electrons [6]. The bands are occupied up to the valence band. The band above the valence band is called the conduction band and these bands are separated by a gap. The conduction band only contains very few free electrons for semiconductors under normal conditions. The band gap determines how much energy it takes for an electron to move from the valence band to the conduction band. The two dierent semiconductors have dierent sized energy bands (see gure 2.3). The n-type semiconductor also has a dopand band, which contains the excess electrons. These electrons can move to the conduction band more easily. The semiconductor n-AlGaAs has a larger dierence between the valence energy and the conduction energy than i-GaAs (see gure 2.4). The Fermi energy of n-AlGaAs is higher than the conduction energy of i-GaAs. When they are brought together in a heterostructure, the excess electrons of nAlGaAs can move very easily from their dopand band to the conduction band of i-GaAs, since they have a higher energy. This charge transfer causes the conduction band and valence band to line up. This shifts the valence and conduction energy levels. All the free electrons from n-AlGaAs get trapped in a well at the junction between n-AlGaAs and i-GaAs, formed by the shift in energy levels. This well is limited by the new Fermi energy. The energy needed for these electrons needed to escape is too high. These electrons are now trapped in the well and can only move in the plain where n-AlGaAs and i-GaAs touch. This is called a 2-dimensional-electrons-gas.

2.4

Ballistic transport

If we want to explore the true nature of electrical conduction we have to look at the behaviour of electrons without interfering too much with a systems imperfections. As we now know, the mean free path of an electron in a crystal lattice is often (always ) very small and it becomes rather dicult to observe the electron solely on that path. This makes measuring the pure conductivity pretty clumsily because it is near impossible to correct the conductivity for all the systems imperfections. With ballistic conduction we describe the transport of electrons in a medium with negligible resistance due to scattering. Also, it is a very interesting question from an elementary point of view; what exactly 7

Figure 2.4: A heterostructure of semiconductors with their valence and conduction bands before (a) and after (b) charge transfer. At the junction of the two semiconductors a 2DEG is formed (Ec is the conduction energy, Ef the Fermi energy, and Ev the valence energy) [7]. drives resistance if there is no scattering? Is it zero like in a superconductor? Is there some elementary value for conduction? In order to observe ballistic transport we must try to make the electrons mean free path long enough so that it is longer then the system through which the electron propagates. In this way, the electrons motion is constant and is not subjected to alternations due to impurities. Keep in mind that normally, especially at temperatures well above the absolute zero, scattering due to impurities dominates the motion behaviour of the electron. In order to achieve the long mean free path we must alter the factors that shorten , the scattering eects. For there are multiple sources for dierent scattering events, we must rst gure out which are applicable to the electrons in our 2DEG. Then we can use the Matthiesen rule for adding up the scattering events to get a proper total scattering value. Matthiesens rule says: 1/ = i 1/i where i indicates all the dierent scattering events. Note that this formula does not really represent the real relationship of the mean free path, because many individual scattering events eect each other. Scattering could change the velocity, which alters the impact of other scattering events. Matthiesens rule is a nice guideline though. The fact that the electrons are inclined in a two dimensional space alters the scattering possibilities [12]; The most dominant scattering sources are the so called ionised impurities of the AlGaAs/GaAs heterostructure, but also phonon scattering and interface impurities take their part into describing the mean free path. Many of these scattering sources have dierent ways to be treated; ionised and interface impurities decrease if the AlGaAs and GaAs donors are further removed from the 2DEG but this also decreases the electron density and therefore also the electron mobility [12]. The electron mobility characterises how quickly an electron can move when pulled by an electron eld. Interface impurities are the deviations from a perfect crystal lattice. These deviations account for scattering as well. This value is usually very small [12], because the crystal structures of AlGaAs and GaAs t quite neatly on each other, from a lattice point of view. Phonon scattering characterises the scattering which is induced by the vibrational motion of the lattice. There are dierent types of 8

phonon scattering, but for now we will only use the fact that they are all temperature dependent and more or less negatively aect . Some of these scattering phenomena can be changed by altering the systems geometry, some can be changed by using purer materials; for instance, in the AlGaAs certain Ga atoms are replaced with Al atoms. This is done in a random order which impuries the lattice, thus creates scattering events. The lattice vibrations, caused by thermal vibrations for instance, can be reduced by cooling the system down, resulting in less scattering. Applying techniques to decrease the amount of possible scattering events in a system increases the mean free path, giving a longer distance over which electrical conductance can take place without scattering. This is exactly what has been done with the 2DEG samples used for ballistic transport. Also remember that, by having ballistic transport, with no interaction between the systems medium and the electron, it is possible to account for quantum mechanic eects of conduction; normally scattering interaction with the system by colliding entangles the electrons system with the mediums system and this destroys the observable quantum eects! Now how long should this be in order to observe quantised conductance in our sample? The point contact created with the 2DEG and electron depletion creates a gate with a width, say W and a length of L. In order for the electron to propagate between the gates freely without scattering, the mean free path, should be much larger than L. If this is the case, the contact is known as a Sharvin point contact, which allows for quantised conduction.

2.5

Quantised conductance

In Section 2.1.1 we mentioned that conduction is mainly caused by electrons with the Fermi energy. At low temperatures the electrons with the Fermi energy are the electrons with the highest energies. High energies correspond to short wavelengths and thus do the electrons with the Fermi energy have the shortest wavelengths. For conduction to take place, the conducting material should be wider than this wavelength. Otherwise, you could say that the electrons would not t through the material. This becomes clearer when you imagine the electron to be a wave. Only when the connement is wide enough for a standing wave to t in (see Figure 2.5), electrons will go through. Such a connement could only take place at nano-scale, for the wavelengths of electrons are in this order of magnitude.

Figure 2.5: Connement of width W with the rst standing wave to t in. The standing wave in Figure 2.5 represents the lowest energy state to t through a connement of width W , having a wavelength of = 2W . So only when the width becomes comparable to W F /2 where F is the wavelength for electrons with the Fermi energy, conduction becomes possible. This would result in a step when we would measure the conduction as a function of W ; a step from zero conduction for W < F /2 to nonzero conduction for W > F /2. If we would increase the width even further, we would measure more steps. The conduction increases stepwise with increasing width. This is due to the fact that for every certain interval of W an extra stationary state becomes able to pass through the connement, which results in a contribution to the total conduction. 9

The stepwise behaviour can be made clear with the Fermi circle, Figure 2.2. The stationary states with the Fermi energy are discrete points on the circle, which means that there is no continuous transition between the various states on the circle. If the connement is wide enough for the rst electrons to pass, then these electrons would be in the stationary state with their momentum parallel directed along the passage. Obviously the electrons directed along the passage, pass most easily. For electrons that are angled away from the direction of the passage, a wider connement is required for them to pass and take part in the conduction. Each subsequent state on the Fermi circle requires a wider connement and because the states are discrete, the conduction with increasing width will increase in steps rather than linear. It happens to be that each state contributes to the conduction with an amount of e2 / where e represents the elementary charge and the reduced Planck constant. The quantised conduction for N propagating states can therefore be written as:
N

G=
n=1

e2

(2.9)

This formula is conrmed by experiments, which have resulted in patterns like Figure 2.6. Here, the conductance is not plotted as a function of the width, but as a function of gate voltage, which has basically the same eect. The gate voltage depletes electrons below it and because of its geometry a passage at nano-scale can be created; a quantum point contact. With decreasing negative gate voltage the electrons get less depleted and the passage becomes wider.

Figure 2.6: Conductance as a function of gate voltage in a 2DEG at 0.6 K [1]. Quantised conductance only occurs at very low temperatures. The reason for this is that ballistic transport can take place at these temperatures. By this, the electrons in the propagating modes can pass unaected through the connement, i.e. with zero resistance. Despite this, the conduction in the system does not become innite. This is due to the fact that only a small number of modes is capable of passing through the connement and thus only a small fraction of electrons can be conducting. All the rest is blocked by the connement, which causes the resistance. When the connement gets widened so that an extra mode can pass, all electrons in that mode will become conducting at the same time, causing a step in the conductance. At higher temperatures the electrons get aected by impurities, which results in a nonzero resistance for even the propagating modes. If we again widen the connement for an extra propagating mode, the electrons in that mode not immediately start conducting altogether. Instead, this mode now has nonzero resistance and its electrons start conducting in a more continuous way. Moreover, when the temperature increases, the Fermi-Dirac distribution tends to a smooth curve like in Figure 2.1 (b), which means that more states become available for taking part in the conduction. The states that are conducting lie no longer on a thin Fermi circle, but rather on a circle with 10

a much thicker line. When the connement gets widened the propagating modes are added more continuously, because of their arbitrarily distributed energies. All this spoils the discrete behaviour of the conductance. It smears out the steps, giving a more linear correlation between conductance and width. When the temperature increases, the stepwise behaviour of Equation (2.9) reduces to the classical description of conductance for a quantum point contact. In two dimensions this is given by [1]: e2 kF W

G=

(2.10)

In this equation the conductance is obviously linearly proportional to the width and the discrete steps have vanished. Before quantised conductance was discovered, this equation was thought to be generally correct. No discrete behaviour was expected.

2.6

Low Resistance Measurements

The resistance of the device can be measured by either varying the current, measuring the voltage or varying the voltage and measuring the current using a digital multimeter (DMM). The resistance can then be calculated using Ohms Law: V = IR (2.11)

Consider a two-wire resistance measurement setup as in Figure 2.7. In this setup a current source and voltmeter are used to measure the desired resistance. The voltage reading however will not only measure the voltage across device VR , but it will also take into account lead and contact resistance Vlead , oset voltages and noise eects.

Figure 2.7: Two-Wire Resistance Measurement Setup; where VM is the measured voltage and VR the voltage over the test resistance. Consider rst the leading resistance contribution only, where the measured resistance is given by: VM = R + (2 Rlead ) I (2.12)

The expected resistance will have a very low magnitude in comparison to the lead resistance. The two-wire resistance measurement setup will thus not suce for the desired measurement. A four-wire resistance measurement setup can be used to overcome this problem caused by the lead resistances (Figure 2.8). The sense current will be very small due to the high resistance of the voltmeter in 11

comparison to the very small resistance of the devices. Because the sense current is negligible; VM VR , and the measured resistance can be determined with much greater accuracy: VM VR = I I (2.13)

Figure 2.8: Four-Wire Measurement Setup The circuit measures a nonzero voltage oset even when there is no current source. This oset is caused my multiple error sources including thermoelectric EMFs, osets generated by rectication of radio frequency interference and osets in the voltmeter input circuit. There are many ways to counteract these dierent error sources which are not mentioned in this report. The oset can be measured in a dry setting, such that it can be subtracted from the measured value to obtain a more accurate result. The noise eects can be reduced using a Lock-In amplier. The Lock-In amplier basically consists of two systems: a synchronous switch and a amplier/lter, see Figure 2.9. The switch is designed such that negative signals are multiplied with -1 such that the output is fully positive. If a reference signal with the same frequency as the input signal, is applied to the input signal: V (t) = V0 sin(2fm t + ) (2.14)

The phase of the reference signal can be modied such that both the signals are in phase. If they are, the result is then passed to the amplier/lter stage, where the signal is modied such that the output signal will be a constant DC signal. The result is a strong low signal-to-noise ratio signal. (If 1 d is then the output signal equals to zero.) 2

12

Figure 2.9: Schematic Lock-In Amplier

13

3. Experiment
3.1 Introduction
The phenomenon quantised conduction was rst observed by the Delft-Philips group in 1988. As part of the course TN2992 experimenteel eindproject we will try to replicate the phenomenon using equipment supplied by the Quantum Transport group at the Department of Applied Physics, Delft University. At 4.2 K we use a high mobility 2DEG of a GaAs-AlGaAs hetereostructure to create ballistic transport and use the electron depletion technique to create a variable gate width. With this set-up we expect to measure quantised conduction with steps of e2 / . The experiment has been conducted on 12 December 2012 at the Quantum Transport group at the Department of Applied Physics. Measurement results suggest one single step of conductance has been observed with size 8K1 which is within range of the expected value e2 / 1 . We propose the samples limited geometry and used doping to account as explanation for the observation of only one step instead of multiple steps.

3.2

Sample description

The sample used in our experiment was actually not designed for measuring quantised conductance. It serves as a chip to create and manipulate quantum dots on. Nevertheless, it has the required characteristics for making quantised conductance apparent. Just like in the original experiment, our sample has a heterostructure of GaAs and AlGaAs in which a 2DEG is conned. Furthermore, it has a gate with electrodes separated by distances (180-280 nm) in the same order of magnitude as the Fermi wavelength. A schematic overview of the sample is given in Figure 3.1.

14

Figure 3.1: Schematic overview of the sample used in this experiment. In the gure both gates (red) and ohmics (green/blue) are indicated. The gure gives a view perpendicular to the plane in which the 2DEG is conned. To make quantum point contacts, negative voltages have to be applied to opposing gates. The electrons beneath them get depleted and depending on the voltage, the electrons can pass or cannot pass between two electrodes. Because the sample contains multiple gates, there are several possibilities of making quantum point contacts. The two most suitable combinations have been used during the experiment, as will be discussed later on. The rst is the combination of QPC1 and LS. The shortest distance between these two electrodes is 280 nm. To measure the resistance for the point contact created by these gates, ohmics 10 and 9 have to be used. The other combination involves the gates T and D, which are separated by a distance of 180 nm. This point contact can be measured by a combination of the ohmic contacts 10 (or 9) and 3 (or 4). The ohmic contacts can be connected to a measurement device in order to make resistance measurements of the paths between them. The connections of all gates and ohmics are directed to a single bus. A schematic overview of the bus is given in Figure 3.2. The connections are indicated by the same labels as in Figure 3.1. With this connector, the sample can be connected to a measurement device and a voltage supply.

Figure 3.2: Schematic overview of the connector of the sample.

15

3.3

Measurement results

The experimental set-up has now been discussed. The following step is to discuss what measurements have actually been performed in order to observe quantum conductance. The goal was to use two ohmics and two gates to achieve an electric current for which the resistance could be measured. Before the actual measurements could be performed, the resistance of the ohmics of the sample was measured. This resistance will be subtracted from the actual measured resistance later on. The measurement for the ohmic resistance was performed at 4 K. A voltage source and a amp` ere meter were connected in series to the circuit. Next, these were connected over two of the ohmics, that were used later on for the actual sample measurements. First ohmics O9 and O10 were connected. The results of this measurement are displayed in table 3.1. Ohmics Input voltage (V) Oset voltage (V) Output voltage (V) S3b voltage source amplication (V/V) M1b amp` ere meter amplication (V/A) Actual voltage (V) Actual current (A) Resistance () O9 - O10 0.05 0.051 0.262 1.00E-03 1.00E+08 0.00005 2.11E-09 23696.68 O9 - O4 0.05 0.01 0.168 1.00E-03 1.00E+08 0.00005 1.58E-09 31645.57

Table 3.1: The measured resistances over the two ohmic pairs used in the sample measurements. The expected values of the resistance over the ohmics was in the k range. By performing simple calculations (see equation 3.1), the settings for the amplication of the amp` ere meter could be deducted. The resistance could best be measured with a voltage in the order of V. Since the input voltage was 50 mV, the amplication of the voltage source was set on 1.00 103 V/V, to convert the voltage to V. U I U 50 106 = = 5.0 108 A R 10 103

R=

I=

(3.1)

The expected value of the resistance was in the order of 10 k, and the output voltage should be around 1 V, meaning that the amplication of the amp` ere source could be calculated. It turned out to be 1.00 108 V/A. Using these settings, the measurement could be performed, and the output voltage and oset voltage were measured. The oset voltage was subtracted from the output voltage, the values were converted using the amplication factors and the actual voltage and current could be calculated. These values gave in turn the actual resistance over the ohmics (see table 3.1). As expected, the values were in the k range. In attempt to reproduce the experimental step-wise conductance plot, a voltage sweep was executed over the gate QPC1-LS. The result is shown in Figure 3.3. Instead of the expected multi-step plot, the result shows only a single step. Rather than a single jump, the plot shows a transition phase, after which a nearly constant value of L = 3.8464 105 1 . This value has to be corrected for the lead resistance caused by the ohmics leading to eq3.2. L = Lmeas + 1 = 3.8464 105 + 1/23696.68 = 8.0664 105 1 Rohmic (3.2)

16

Figure 3.3: Result of Sweep voltage from -320mV to -200mV (201 data points) over gate QPC1-LS

Figure 3.4: Result of Sweep voltage from -500mV to 0mV (201 data points) over gate T-D

17

The theoretical conductance gain per step is given by equation 3.3. The obtained result diers only by 3% from the theoretical conductance gain, such that it is allowed to conclude that a single step was obtained. e2 = 7.818 105 1

G=

(3.3)

It was tried to use a dierent connection to reproduce the step-wise conductance plot. In this attempt a sweep ranging from -500mV to 0mV was used. The result is shown in g3.4. Again only a single step is observed. The nal conductance including the ohmic resistance correction is eq3.4, which deviates 11% from the theoretical step gain. L = Lmeas + 1 Rohmic = 3.7773 105 + 1/31645.57 = 6.9372 105 1 (3.4)

3.4

Conclusion

One step in conduction of e2 / has been observed where several steps were expected. In vain of multiple attempts with dierent setup congurations, no more steps were unveiled. The conduction did limit to a plateau values of 8.0664 105 1 and 6.9372 105 1 which are within acceptable range of the theoretical expected value 7.818 105 1 . An explanation for the single step can be that the width of the gates is not suciently large, even though the gate sizes in the sample were comparable to previous experiments [1]. It was assumed that the electron depletion has improved and that the Wmax is more constant over a longer distance. Moreover, the doping of the GaAs-AlGaAs heterostructure in the used sample is optimised for other applications then measuring quantum conductance. In other words, the sample used is not designed for measuring quantum conductance. The deviation in step height in comparison to the theoretical result is caused by a multitude of uncertainties, including: the uncertainty in ohmic resistance measurements and the uncertainty in sample resistance measurements. To minimise the measurement uncertainties, one could use noise reduction methods such as the AC lock-in technique and reduce the oset error.

18

Acknowledgements
To get a grasp of understanding certain quantum phenomena such as quantum conductance, one needs an understanding of quantum mechanics in general. In particular for bachelor-students from dierent faculties such as aerospace and mechanical engineering this knowledge is not readily present. This project enabled us to learn a bit about quantum mechanics and in particular quantum conductance. We would like to thank Dr. Ad Verbruggen for his supervision and support during this project. Dr. Pierre Barthelemey for helping us perform the experiments and the necessary preparations. Mohammad Shaei for supplying the samples and Pasquale Scarlino for enthusiastically giving us a tour through the lab. The project has been a great learning experience and it was an enjoying project to end our physics minor with.

19

References
[1] B. J. van Wees, H. van Houten, C. W. J. Beenakker, J. G. Williamson, L. P. Kouwenhoven, D. van der Marel, and C. T. Foxon. Quantized conductance of point contacts in a two-dimensional electron gas. Phys. Rev. Lett., 60:848850, Feb 1988. [2] H. Van Houten and C. Beenakker. Quantum point contacts. Physics Today, pages 2227, Jul 1996. [3] D. Lincoln. The inner life of quarks. Scientic American, 307:3643, Nov 2012. [4] Classical and quantum conductivity. quantum_conductivity, 2013. http://en.wikipedia.org/wiki/Classical_and_

[5] Igor Kuskovsky. Classical and quantum free electron models of electrical conductivity. Department of Physics of the Division of Mathematics and Natural Sciences, Queens College of the City University of New York, 2007. [6] Stanford University Rsasaki. Chapter 7. two dimensional electron gas system (2deg). http: //www.stanford.edu/~rsasaki/AP388/slide7, 2013. [7] 2deg. http://en.wikipedia.org/wiki/2DEG, 2013. [8] Semiconductor. http://en.wikipedia.org/wiki/Semiconductor, 2013. [9] Warwick University D. R. Leadley. Reduced dimensional structures. warwick.ac.uk/~phsbm/2deg.htm, 2013. http://homepages.

[10] Rice University Rui-Rui Du. 2deg materials and basic characterization. http://wls.iphy.ac. cn/Chinese/1219/2/rrdu.pdf, 2013. [11] Department of Physics Branislav K. Nikolic and University of Delaware Astronomy. Heterojunctions, interfacial band bending, and 2deg formation. http://www.physics.udel.edu/ ~bnikolic/teaching/phys824/lectures/band_bending_2deg.pdf, 2013. [12] Dominik Zumbhl. Gaas heterostructures and 2d electron gas. Quantum Coherence Lab at the Department of Physics, University of Basel, 2010.

20

You might also like