TEAS Lecture 2
TEAS Lecture 2
Lecture 2
Contents
             2.1      Introduction . . . . . . . . . . . . .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   13
             2.2      Mathematical model . . . . . . . .       .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   15
             2.3      Modes . . . . . . . . . . . . . . . .    .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   18
             2.4      Transmission . . . . . . . . . . . .     .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   23
             2.5      Near-equilibrium (linear) transport      .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   26
             2.6      Transport in the bulk . . . . . . .      .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   27
             2.7      Summary . . . . . . . . . . . . . .      .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   31
             2.8      References . . . . . . . . . . . . . .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   31
2.1 Introduction
                                                               13
October 9, 2012   9:55                            7975: Near-Equilibrium Transport                        transport
             Fig. 2.1. Illustration of a model nanoscale electronic device. The voltage, V , lowers the
             Fermi level of contact 2 by an amount, qV .
             move the states up and down in energy (as in a transistor), but in these
             lectures we will assume a two-terminal device and set U = 0.
                 The channel of our device is connected to two ideal contacts, which
             are assumed to be large regions in which strong scattering maintains near-
             equilibrium conditions. Accordingly, each contact is described by an equi-
             librium Fermi function (or occupation number),
                                                             1
                                           f0 =                            ,                     (2.1)
                                                  1+   e(E−EF )/kB TL
                Although this model is very simple, we shall see that it is also very
             powerful. We shall be concerned with two questions:
             (1) How is the electron density in the device related to the Fermi levels in
                 the contacts, to the density of states, and to the characteristic times?
             (2) How is the electron current through the device related to the same
                 parameters?
             (1) The channel of the device is described by a band structure, E(k). This
                 assumption is not necessary; for the more general case, see Ref. [1].
             (2) The contacts are large with strong inelastic scattering that maintain
                 near-equilibrium conditions.
             (3) We assume that electrons feel a self-consistent (mean-field) potential,
                 U , due to the other electrons and the applied biases. (This assumption
                 breaks down for “strongly correlated transport”, such as single electron
                 charging.) In practice, we would find the self-consistent potential by
                 solving the Poisson equation. It is important for devices like transistors,
                 but in these notes, we restrict our attention to two-terminal devices and
                 set U = 0.
             (4) All inelastic scattering takes place in the contacts. Electrons flow from
                 left to right (or right to left) in independent energy channels.
             (5) The contacts are reflectionless (absorbing). Electrons that enter the
                 contact from the channel are equilibriated according to the Fermi level
                 of the contact.
             Although these assumptions may appear restrictive, we will find that they
             describe a large class of problems. Having specified the model device, we
             turn next to the mathematical analysis.
             To develop the mathematical model, consider first the case where only the
             first (left) contact is connected to the channel. Contact 1 will seek to fill
             up the states in the channel according to EF 1 . Eventually, contact 1 and
             the channel will be in equilibrium with number of electrons between E and
October 25, 2012   10:0                       7975: Near-Equilibrium Transport                 transport
E + dE given by
                                      ′
                                     N01 (E)dE = D(E)dEf1 (E) ,                       (2.2)
                                      ′
                                     N02 (E)dE = D(E)dEf2 (E) ,                       (2.4)
                 In practice, both contacts are connected at the same time and both
             inject or withdraw electrons from the channel. The total rate of change of
             the electron number in the device is
                                          D(E)dE          D(E)dE
                            N ′ (E)dE =          f1 (E) +        f2 (E) ,             (2.7)
                                            2               2
October 9, 2012   9:55                            7975: Near-Equilibrium Transport              transport
             where we have assumed that τ1 = τ2 and used eqns. (2.2) and (2.4). Finally,
             we obtain the total, steady-state number of electrons in the channel by
             integrating over all of the energy channels,
                                                Z                             
                                                      D(E)          D(E)
                              Z
                         N=       N ′ (E)dE =              f1 (E) +      f2 (E) dE .    (2.8)
                                                       2             2
                 Equation (2.8) is the answer to our first question. It gives the number
             of electrons in the channel of the device in terms of the density-of-states of
             the channel and the Fermi functions of the two contacts. Finally, a word
             about notation. The quantity, N ′ has units of number / energy; it is the
             differential carrier density, N ′ (E) = dN/dE|E .
                 We should note the similarity of eqn. (2.8) to the standard expression
             for the equilibrium electron number in a semiconductor [4],
                                                      Z
                                            N0 =          D(E)f0 (E)dE .                (2.9)
             The difference is that eqn. (2.9) refers to the number of electrons in equi-
             librium whereas eqn. (2.8) describes a device that may be in equilibrium
             (if EF 1 = EF 2 ) or very far from equilibrium if the Fermi levels are very
             different.
                 We should remember that N is the total number of electrons in the
             channel, and D(E) is the total density-of-states, the number of states per
             unit energy. In 1D, D ∝ L, the length of the channel. In 2D, D ∝ A, the
             area of the channel, and in 3D, D ∝ Ω, the volume of the channel. For
             device work we usually prefer to express the final answers in terms of the
             electron density (per unit length in 1D, per unit area in 2D, and per unit
             volume in 3D).
                 Having answered our first question, how the electron number is related
             to the properties of the channel and contacts, we now turn to the second
             question, the steady-state current. When a steady-state current flows, one
             contact tries to fill up states in the channel and the other tries to empty
             them. If EF 1 > EF 2 , contact 1 injects electrons and contact 2 removes
             them, and vice versa if EF 1 < EF 2 .
                 The rates at which electrons enter or leave contacts 1 and 2 are given
             by eqns. (2.3) and (2.5). In steady state,
                                                   F1 + F2 = 0 .                       (2.10)
October 9, 2012   9:55                          7975: Near-Equilibrium Transport                 transport
                                            D(E)
                                          Z
                                     N=          (f1 + f2 ) dE
                                             2
                                                                                        (2.15)
                                        2q         D(E)
                                           Z
                                     I=      γ(E)π         (f1 − f2 ) dE .
                                        h             2
2.3 Modes
             The fact that the current is proportional to (f1 − f2 ) makes sense, and 2q/h
             is a set of fundamental constants that we shall see is important, but what is
             the product, γπD/2? It is an important quantity. According to eqn. (2.13),
October 9, 2012   9:55                            7975: Near-Equilibrium Transport                          transport
             Fig. 2.2. A simple, 2D electronic device with channel width, W and length, L. For
             the calculation of the average x-directed velocity, ballistic transport is assumed, i.e. the
             channel is much shorter than a mean-free-path for scattering.
                                           qN ′ (E)dE   ~ (f1 + f2 )
                                             ′
                                                      =              .                           (2.17)
                                            I (E)dE     γ (f1 − f2 )
                                                      L
                                           τ (E) = 
 +    ,                          (2.19)
                                                    υx (E)
             which is just the average transit time of carriers across the channel.
                 To evaluate τ (E), we need hυx+ (E)i, the average velocity in the +x di-
             rection. From Fig. 2.2, we see that for ballistic transport, in which electrons
             travel across the device without changing direction,
                                       
 +    
                                        υx (E) = υ(E) hcos θi .                       (2.20)
                                                              D(E)
                                          M (E) ≡ γ(E)π                               (2.23)
                                                               2
             and using γ = ~/τ and D = D2D W L, we find
                                                             h
 +   
                            M (E) = W M2D (E) = W              υ (E) D2D (E) .        (2.24)
                                                             4 x
             Similar arguments in 1D and 3D yield
                                              h
 +    
                            M (E) = M1D (E) =   υx (E) D1D (E)
                                              4
                                                   h
 +    
                            M (E) = W M2D (E) = W    υ (E) D2D (E)                    (2.25)
                                                   4 x
                                                 h
                                          D(E)
                                         Z
                                   N=          (f1 + f2 ) dE
                                           2
                                                               ,                  (2.29)
                                      2q
                                         Z
                                   I=      M (E) (f1 − f2 ) dE
                                      h
             which shows that to compute the number of electrons and the current, we
             need two different quantities, D(E) and M (E). The density-of-states is a
             familiar quantity. For parabolic energy bands, we know that the 1D, 2D,
             and 3D densities-of-states are given by
                                                   s
                                          L              2m∗
                   1D : D(E) = D1D (E)L =                        H(E − Ec )
                                          π~           (E − Ec )
                                            m∗                                    (2.30)
                   2D : D(E) = D2D (E)A = A     H(E − Ec )
                                            π~2
                                              p
                                            m∗ 2m∗ (E − Ec )
                   3D : D(E) = D3D (E)Ω = Ω                  H(E − Ec ) ,
                                                  π 2 ~3
October 25, 2012   10:0                       7975: Near-Equilibrium Transport                     transport
             where W is the width of the 2D channnel and A is the cross sectional area
             of the 3D channel. Figure 2.3 compares the density-of-states and number
             of modes (conducting channels) in 1D, 2D, and 3D for the case of parabolic
             energy bands (E(k) = Ec + ~2 k 2 /2m∗ ).
             Fig. 2.3. Comparison of the density-of-states, D(E), and number of channels, M (E),
             in 1D, 2D, and 3D. Parabolic energy bands are assumed in each case.
2.4 Transmission
             Figure 2.2 showed how electrons flow from contact 1 to contact 2 under
             ballistic conditions. Figure 2.4 shows the diffusive case.
             Fig. 2.4. A simple, 2D electronic device with channel width, W and length, L. In this
             case, diffusive transport is assumed — the channel is many mean-free-paths long.
             in the diffusive limit that we have considered (L ≫ λ), but it fails when L
             is short. The correct, general expression is
                                                        λ(E)
                                           T (E) =             ,                      (2.43)
                                                      λ(E) + L
             which reduces to eqn. (2.42) for the diffusive limit of L ≫ λ, but for the
             ballistic limit of L ≪ λ it approaches 1. This expression is reasonable,
             but we will see in Lecture 6 that it can be derived with relatively few
             assumptions and that it is valid not only in the ballistic and diffusive limits,
             but in between as well.
                 To summarize, we can write in general
                                              D(E)
                                      γ(E)π        = M (E)T (E) ,                     (2.44)
                                               2
             with M (E) being given by eqns. (2.25) and T (E) by eqn. (2.43). People
             speak of three different transport regimes:
Our simple transport model can be used to describe all three regions.
                                     2q         D(E)
                                        Z
                                I=        γ(E)π        (f1 − f2 ) dE
                                     h             2
                                                                           (2.46)
                                     2q
                                        Z
                                I=        T (E)M (E) (f1 − f2 ) dE .
                                     h
             There is no limitation to small applied biases yet, but if we apply a large
             bias, then there could be a lot of inelastic scattering that would invali-
             date our assumption that the current flows in independent energy channels.
             Since our interest is in near-equilibrium transport, we now simplify these
             equations for low applied bias.
October 9, 2012   9:55                           7975: Near-Equilibrium Transport             transport
                The two Fermi functions in eqn. (2.46) are different when there is an
             applied bias. Recall that an applied bias lowers the Fermi level by −qV . If
             the applied bias is small, we can write
                                                            ∂f0
                                       (f1 − f2 ) ≈ −           ∆EF .               (2.47)
                                                            ∂EF
             From the form of the equilibrium Fermi function,
                                                         1
                                       f0 =                         ,               (2.48)
                                                 1 + e(E−EF )/kB TL
             we see that
                                                 ∂f0    ∂f0
                                                     =−     .                       (2.49)
                                                 ∂EF    ∂E
             Equations (2.49) and (2.47) can be used in eqn. (2.46) along with ∆EF =
             −qV to obtain
                               2Z                       
                               2q                    ∂f0
                          I=          T (E)M (E) −         dE V = GV .          (2.50)
                                h                    ∂E
             The final result,
                                      2q 2
                                                                 
                                                              ∂f0
                                             Z
                                 G=              T (E)M (E) −       dE ,            (2.51)
                                       h                      ∂E
             is just the conductance in Ohm’s Law, but now we have an expression that
             relates the conductance to the properties of the material. It is important
             to remember that this expression is valid in 1D, 2D, or 3D, if we use the
             appropropiate expression for M (E).
             Equations (2.56) and (2.57) are standard results that are conventionally
             obtained from irreversible thermodynamics or by solving the Boltzmann
             Transport Equation [7]. We have obtained the standard expressions for
             bulk materials by assuming that the channel of our model device is much
             longer that a mean-free-path.
                 Real resistors can be linear even when quite large voltages are applied.
             How does this occur? It occurs because when the resistor is long, electrons
             do not drop down the total potential drop in one step. Instead, they con-
             tinually gain a little energy and then dissipate it by emitting phonons. If
             the resistor is long and the voltage drop not too large, then the electrons
October 9, 2012   9:55                          7975: Near-Equilibrium Transport                       transport
                                                                     dns
                                        Jnx = ns qµn Ex + qDn            ,                   (2.59)
                                                                     dx
             where
                                                 Dn   kB TL
                                                    =                                        (2.60)
                                                 µn     q
October 9, 2012   9:55                         7975: Near-Equilibrium Transport                transport
                                            d(Fn /q)
                               Jnx = σn
                                              dx
                                                                  
                                                               ∂f0
                                       Z
                               σn =        q 2 Dn (E)D2D (E) −       dE               (2.61)
                                                               ∂EF
                                                  1
                               f0 =                             ,
                                       1+   e(E−Fn (x))/kB TL
             and for electrons in the valence band, we have
                                            d(Fp /q)
                                Jpx = σp
                                              dx
                                                                  
                                                               ∂f0
                                       Z
                                σp =       q 2 Dp (E)D2D (E) −       dE               (2.62)
                                                               ∂EF
                                                1
                                f0 =                         .
                                       1 + e(E−Fp (x))/kB TL
             The total current is the sum of the contributions from each band. It is
             important to note that these equations refer to electrons in both the con-
             duction and valence bands. The occupation factor, f0 , describes the prob-
             ability that an electron state is occupied. It is often useful to visualize the
             resulting current flow in the valence band in terms of holes, but the expres-
             sions that we used were derived for electrons, and we did not inquire as to
             whether they were in the conduction or valence bands because it does not
             matter.
October 9, 2012   9:55                      7975: Near-Equilibrium Transport                   transport
2.7 Summary
             This has been a long lecture, but the final result is a simple one that we
             shall see is very powerful. Equation (2.51) describes the conductance of
             a linear resistor very generally. The conductance is proportional to some
             fundamental constants, (2q 2 /h), which we will see in the next lecture is
             the “quantum of conductance”, that is associated with the contacts. The
             conductance is related to the number of conducting channels at energy,
             E, M (E), and to the transmission, T (E), which is the probability that
             an electron with energy, E, injected from one contact exits to the other
             contact. We find the total conductance by integrating the contributions
             of all of the energy channels. Equation (2.51) is valid in 1D, 2D, or 3D
             — we simply need to use the correct expressions for M (E). It is valid for
             very short (ballistic) resistors or very long (diffusive) resistors and for the
             region in between. The next lecture will begin with eqn. (2.51). Finally, it
             should be mentioned, that we have assumed isothermal conditions — the
             two contacts are at the same temperature. The implications of temperature
             gradients will be discussed in Lectures 4 and 5.
2.8 References
             [5] Changwook Jeong, Raseong Kim, Mathieu Luisier, Supriyo Datta, and
                 Mark Lundstrom, “On Landauer vs. Boltzmann and Full Band vs. Ef-
                 fective Mass Evaluation of Thermoelectric Transport Coefficients”, J.
                 Appl. Phys., 107, 023707, 2010.
             [6] J.M. Ziman, Principles of the Theory of Solids, Cambridge Univ. Press,
                 Cambridge, U.K., 1964.
             [7] N.W. Ashcroft and N.D. Mermin, Solid–State Physics, Saunders Col-
                 lege, Philadelphia, PA, 1976.
             [8] M. Lundstrom, S. Datta, and M.A. Alam, “Electronics from the Bottom
                 Up”, http://nanohub.org/topics/ElectronicsFromTheBottomUp, 2011.