phys208_trygve_slides.pdf

Upload: azhar-mahmood

Post on 14-Apr-2018

214 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    1/51

    1 Harmonic Crystals

    To describe a crystal, we start with balls placed in a regular way with

    springs holding them in place. A ball moved form its equilibrium posi-tion will experience a force. Since this force is the result of a deviation froman equilibrium position, the Taylor expansion of the potential cannot have alinear term. Also ignoring the non-interesting constant term the potential is

    V(qi) =1

    2

    ij(qj

    qj0)

    2 +

    O (q q0)3 , (1)

    where qj0 is the equilibrium position. Thus for q q0 sufficiently small we canapproximate the potential to be the harmonic potential. If we also assumethat all springs are the same, we can write the potential as

    V(q) =

    2(q q0)2. (2)

    1.1 Mono-atomic chain

    For simplicity, consider the one-dimensional case. Also we assume that ourchain is infinitely long, so that we wont have to care about boundary effects.

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    2/51

    We take all balls and all springs to be the same. The balls have mass m andthe equilibrium distance between two balls are a. We let ui denote ball isdeviation from its equilibrium position. The Lagrangian of the chain is,

    L = T V =j

    m2

    u2j 2 (uj+1 uj)2

    , (3)

    which gives the equations of motion:

    0 =d

    dt

    L

    ui L

    ui = mui + (ui ui1) (ui+1 ui)

    = mui + (2ui ui+1 ui1)To solve this set of coupled equations, we try to find a solution on the formun = ue

    i(kant). We find that this is a solution provided that and k are

    related by

    2 =

    m

    2 eika eika

    =2

    m(1 cos ka) = 4

    msin2

    ka

    2.

    (4)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    3/51

    The solution un = uei(kant) is a wave with wavenumber k and angular fre-

    quency .

    Equation (4) is called the dispersion relation of the wave. [...]

    1.2 Density of states

    To find the density of states, it is easiest to start with a finite chain. But westill do not want to be bothered by boundary effects. To avoid this, we impose

    periodic boundary conditions. If the chain has N balls the periodic boundaryconditions tell us that eikan and eika(n+N) describe the same mode. Thus theseN balls can sustain no more than N different modes. eikan = eika(n+N) implieskaN = 2, thus the k-values for these N modes is evenly distributed over ainterval of 2/a. We take this interval to be /a < k < /a. From this

    argument we see that independent of then size ofN the density of k-states is(k) =

    N a

    2. (5)

    To find the density of modes , we use

    () = 2(k)k. (6)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    4/51

    The factor 2 is because there are two branches of k. Rearranging (6) andtaking N we find

    () = 2(k)

    k1 N= 2(k)k

    1

    (7)

    On the branch 0 < k < /a the dispersion relation is = 0 sinka2 , where we

    have defined =

    4/m. By differentiating it we find

    k = 0a2 cos ka2 = a2

    20 20 sin2 ka2 = a2

    20 2. (8)

    Putting everything together, we find that the mode density is

    () =2N

    1

    20 2

    . (9)

    For 0 can expand the mode density in /0,

    () =2N

    0

    1

    1 0

    2=

    2N

    0

    1 + O (/0)2 (10)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    5/51

    which shows that for low frequencies the mode density is essentially constant.When 0 the density diverges.

    1.3 Di-atomic chain

    We now make a small variation of the case described above. We keep allsprings the same, but we replace every other ball with a ball with a differentmass.Coordinates ui and vi

    1 are associated with the balls with mass m1 and

    m2, respectively. The Lagrangian for this chain isL = T V =

    j

    m12

    u2j +m22

    v2j

    2

    (uj vj)2 + (uj vj1)2

    . (11)

    Note that the potential terms count each spring exactly once, as we of courseshould do. The equations of motion are

    0 = ddt

    Lui

    Lui

    = m1ui + (2ui vi vi1),

    0 =d

    dt

    L

    vi L

    vi= m1vi + (2vi ui ui+1).

    1We take ui and vi to be adjacent balls. For concreteness let ui be to the left ofvi

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    6/51

    In analogy with the mono-atomic string we look for solutions on the formun = ue

    i(kant), vn = vei(kant). Inserting this into the equations of motion

    we find:

    m12 =

    u(2u veika veika) = 21 v

    ucos ka

    ,

    m22 =

    v(2v ueika ueika) = 2

    1 u

    vcos ka

    .

    Eliminating u/v, we find that the dispersion relation for the di-atomic chain

    is 1 m1

    2

    2

    1 m2

    2

    2

    = cos2 ka, (12)

    or

    4 21

    m1+

    1

    m22 +

    42

    m1m2sin2 ka = 0. (13)

    Solving for 2 we find

    2 =

    1

    m1+

    1

    m2

    1

    m1+

    1

    m2

    2 4

    m1m2sin2 ka

    (14)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    7/51

    Before we analyse this dispersion relation any further, lets see what happensif we put m1 = m2 = m.

    2 =2

    m1 1 sin

    2 ka = m

    (1

    cos ka) . (15)

    Picking the minus-sign this is exactly the dispersion relation for the mono-atomic chain, (4). But what about the plus-sign? ch

    Now it is time to take a closer look at equation (14) in the case where themasses are different. To simplify the notation we write the dispersion relation

    in terms of 1/ = 1/m1 + 1/m2 and M = m1 + m2,

    2 =

    1

    1 4

    Msin2 ka

    . (16)

    Since < 4M the discriminant is always non-negative, and 1. This isreassuring since then will be real for all combinations of m1 and m2.

    If 0 < k /a equation (16) can be expanded to give2 =

    1

    1 2

    Msin2 ka + O sin4 ka

    =

    1 1 2

    M(ka)2 + O (ka)

    4

    .(17)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    8/51

    2 Heat capacity in a solid

    2.1 The Einstein model

    The Einstein model approximates the heat capacity in a solid with the assump-tion that each oscillator is independent and has a certain frequency . Sincewe work in three dimensions a solid consisting of N atoms will have 3N suchoscillators. Since all oscillators are identical, the ratio of number of oscillatorsin the (n + 1)th quantum state to number of oscillators in the nth quantumstate is ch

    Nn+1Nn

    = eh. (18)

    Thus Nn enh. Then the average excitation number of an oscillator is

    n = sesh esh . (19)The summation in the denominator is simple since

    xs =

    1

    1

    x

    . (20)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    9/51

    Using a simple trick, we can do the summation in the nominator almost justas simply,

    sxs = xd

    dx xs =x

    (1 x

    )

    2. (21)

    Thus the average occupation number is

    n = eh

    1 eh =1

    eh 1 . (22)

    The average thermal energy of an oscillator is then

    n

    h, thus for the whole

    sample, containing 3N such oscillators, the thermal energy is

    U = 3Nnh = 3N heh 1 (23)

    To find the heat capacity we simply differentiate the thermal energy withrespect to temperature,

    CV =

    U

    T

    V

    = 3NkB

    h

    kBT

    2 eh(eh 1)2

    = 3NkB ET

    2eE/T

    (eE/T

    1)2

    ,

    (24)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    10/51

    where we have defined the Einstein temperature E = h/kB. At low temper-atures, T E we find

    CV 3N kB ET 2

    e

    E/T (25)

    which shows that the heat capacity is exponentially decaying when T 0. Athigh temperatures, T E we expand the exponentials to find

    CV 3NkB ET 2

    1

    (1 + E/T 1)2 = 3N kB (26)

    which is the result of Dulong and Petit.

    2.2 The Debye model

    The Debye model is a simple extension of the Einstein model where the as-sumption that all oscillators have the same frequency is removed. Instead weassume that the oscillator frequencies can take any value up to a yet unknowncut-off, D, and that the modes have a quadratic distribution g() = F .

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    11/51

    For each mode the calculation of average occupation number we did for theEinstein model still applies. Thus

    n =1

    eh 1 .With these assumptions the thermal energy is

    U =

    D

    0

    h

    eh

    1

    F 2d. (27)

    To fix the value of F we integrate the density of states up to the cut-off andrequire this to be equal to the total number of oscillators,

    3N =

    D

    0

    F 2d =1

    3F 3D. (28)

    Thus the density of states is

    g() =9N 2

    3D. (29)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    12/51

    To calculate the heat capacity we differentiate the thermal energy with respectto temperature

    CV

    = UT

    V

    =

    T

    9Nh

    3DD0

    3

    eh 1d

    =9Nh2

    kBT23D

    D0

    4eh

    (eh 1)2 d

    = 9NkB T

    D3

    D/T

    0

    x4ex

    (ex

    1)2

    dx,

    (30)

    where we have defined the Debye temperature, D = hD/kB. At low temper-atures, T D we integrate all the way to infinity to. Then the integral istemperature independent and

    CV T3 (31)as T 0 in agreement with experiments. At high energies T D,

    CV 9N kB

    T

    D

    3 D/T0

    x2dx = 3N kB (32)

    just as for the Einstein model.

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    13/51

    3 Electrons in a periodic potential

    3.1 Blochs theorem

    Blochs theorem states that the eigenstates of the one-electron Hamiltonian

    H = h2 2

    2m+ U(r), U(r + R) = U(r). (33)

    can be written as

    nk(r) = eikrunk(r), unk(r + R) = unk(r). (34)

    Consider a one-dimensional infinite atomic lattice. The potential set up bythis lattice is naturally periodic, and can be written as

    V(x) =

    (x xi) =

    n=

    U(x na), (35)

    where a is the lattice constant. U are the independent wells localised aroundeach atom. We now consider the solution on a subset of size L = Na. We

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    14/51

    start by expanding the wave function (x) and the potential U(x) in Fourierseries,

    (x) = (kn)eiknx, (36)U(x) =

    U(Kn)e

    iKnx. (37)

    Note that the allowed wave numbers in the two expansion are not the samesince the periodicity of requires (x) = (x + Na) giving

    kn =

    2n

    aN , (38)whereas U has period equal to the lattice spacing, U(x) = U(x + a) giving

    Kn =2n

    a. (39)

    Inserting the expansions into the one-particle Schrodinger equation we find

    0 =

    h22m

    x + U(x)

    (x)

    =n

    h2

    2mk2n

    (kn)e

    iknx +m,n

    U(Km)(kn)ei(kn+Km)x.

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    15/51

    Multiplying through by eikix and using orthogonality this simplifies to

    h2

    2mk2i (ki) +

    m

    U(Km)(ki Km) = 0. (40)

    Note that for a fixed ki equation (40) couples only the Fourier modes with wavenumber ki, ki + K, ki + 2K , . . ., where K is the (uniform) spacing of the modesin the expansion ofU(x). This means that the original eigenvalue problem hasbeen split up into N independent problem, one for each ki in the first Brillouin

    zone. For each of these problems ki(x) can be written on the form

    ki(x) =m

    (ki Km)ei(kiKm)x = eikixm

    (ki Km)eiKmx. (41)

    Thus we have proved Blochs theorem (in one dimension).

    3.2 Tight binding

    To study the energy bands arising from the periodic structure in a crystal weknow introduce the approximation of Linear Combination of Atomic Orbits

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    16/51

    (LCAO). The basic assumption is that the solutions (r) to the one atomproblems has small overlap and that the extra potential energy due to thecrystal is small compared to the atomic potential energy. If the Hamiltonianfor one single atom is

    Ha(r) = T + U(r), (42)

    the Hamiltonian for the crystal is

    Hc(r) = T + V(r) = T +

    RU(r R). (43)

    The solution can now be written as

    (r) =R

    c(R)(r R), (44)

    where (r) is the solution to the single-atom problem. The coefficients c(R)can found by diagonalising the Hamiltonian, but we will rather exploit the factthat the solutions must follow Blochs theorem. Assume therefore that c(R)can be written on the form

    c(R) = eikR. (45)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    17/51

    The energy of a state is

    E =

    k,E|H|k,E

    k,E

    |k,E

    . (46)

    The only problem now is that if there are N possible values for R the numera-tor will contain N3 terms and the denominator N terms. But here comes thetight binding approximation to our rescuesince we assume that the single-atom wave functions have small overlap only the interaction with the nearestneighbours needs to be considered. Before continuing we do yet a simplifica-

    tion: we assume that the solutions (r) are spherically symmetric (s-states),such that only the distance to the neighbours come in. The nearest neighbourterms then give

    = d3r(r)U(r d)(r) + d

    3r(r d)U(r)(r), (47)

    where |d| is the distance to the nearest neighbour. Using this the energysimplifies to

    E(k) =N E0 N

    d

    Nd

    eikd

    N E0

    d

    deikd, (48)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    18/51

    where E0 is the single-atom energy and N ( 3N 1)3. As a simple appli-

    cation, consider a cubic lattice where there are 6 nearest neighbours. Then

    E(k) = E062(cos kxa +cos kya +cos kza) E066+ a2k2, (49)where the last approximation is good only for small kx, ky, kz.

    3.3 Weak binding

    To learn more about the band structure it is also useful to consider the weakbinding approximation. If there is no potential at all the electrons can bedescribed by plane waves, eikr, with energy E = h2k2/2m. If we contain theelectrons to move in a box of linear size L each of the the momentum will be

    quantised,E =

    h2

    2m

    2

    L

    2(n2x + n

    2y + n

    2z). (50)

    Having periodic boundary conditions we should only consider wave vectorsover an interval 2/a, where a is the lattice spacing. We will take this interval

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    19/51

    to be /a < k/a. Wave vectors outside this interval should be moved in.In a weak potential the picture will be very similar, but close to the atomsin the lattice the potential cannot be completely neglected. The effect is thatfor k close to n/a there will be deviations from the no potential solutionproducing gaps in the energy spectrum.

    3.4 Dynamics of electrons in a lattice

    We start out with Schrodinger equation, and assume a Bloch-type solution

    2

    ,,k(x) = eikxu,k,

    h2

    2m

    x+ ik

    2+ U(r)

    u,k(x) = (k)u,k(x) (51)

    2For simplicity we work in one dimension only

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    20/51

    Differentiating with respect to k, multiplying from the left by u,k and inte-grating over x we find

    h

    mu,kih

    x

    + hku,kdx + u,kHu,kk

    dx

    =

    u,k

    ku,k + u

    ,ku,k

    k

    dx,

    (52)

    where H is the operator on the left-hand side of equation (51). Since is aneigenvalue of H, u,k(H

    ) = 0 and the equation reduces to

    h

    m

    u,k

    ih

    x+ hk

    u,kdx =

    u,k

    ku,kdx =

    k, (53)

    or in terms of ,k(x)

    1

    m,kih

    x,kdx = 1

    h

    k

    . (54)

    But the left-hand side is just p/m, thus we have found that the expectationvalue of the velocity of an electron in the lattice is

    v = 1h

    (k)

    k. (55)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    21/51

    Now we go on to consider the time development of the velocity (for brevitywell denote the velocity v rather then v),

    dvi

    dt=

    d

    dt1

    h

    ki = 1

    h

    ki

    d

    dt=

    1

    h

    ki

    kj

    dkj

    dt = 1

    h2

    2

    kikj

    d

    dt(hkj) (56)

    Since hk is something like the momentum of the particle, its time derivativemust be like a force. But then equation (56) looks like Newtons law, but witha somewhat strange mass

    ai

    = 1m

    ij

    Fj

    ,1

    m=

    2

    kikj. (57)

    The tensor 1/m is the inverse effective mass. Note that the effective masscan be more or less anything, including a negative number!

    4 The Fermi gas model

    To describe an electron gas we have to take into account that electrons arefermions. Such particles respect the Pauli principle, i.e. no more than oneparticle can occupy a certain state.

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    22/51

    But there is also another aspect in which the electrons are substantially dif-ferent from classical particlesall electrons look exactly the same. Indeed, iftwo electrons are interchanged this must be considered the same configuration.This means that ifP(r1, r2) is the probability for finding electron 1 in positionr1 and electron 2 in position r2 (and all other quantum numbers identical) wemust have

    P(r1, r2) = P(r2, r1). (58)

    To write down a wave function that conforms with these two facts the wavefunction have to be antisymmetric under interchange of any two electrons.If we assume the electrons to be non-interaction (independent electron ap-proximation) such a wave function can be built from the single particle wavefunctions,

    (x1, x2, . . . , xn) = det

    1(x1) 1(x2) 1(xn)2(x1) 2(x2) 2(xn)

    ... . . .

    n(x1) n(x2) n(xn)

    . (59)This construction is called a Slater determinant. Interchanging two electronscorresponds to interchanging two rows (or equivalently, two columns), which

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    23/51

    produces the minus-sign we need. Also, if two electrons are in the same state,two rows will be identical and the determinant vanishes.

    4.1 Fermi gas at T = 0

    Considering a cube with side lengths L and imposing periodic boundary con-ditions the wave function of an electron must satisfy

    (x,y,z ) = (x + L,y,z )

    (x,y,z ) = (x, y + L, z)

    (x,y,z ) = (x,y,z + L)

    (60)

    Since we use the independent electron approximation, the wave functionsshould simply be plane waves. This means, for instance, that eikxx = eikx(x+L)

    which implies that the momentum is quantised, kxL = nx2. Since the mo-mentum is quantised, so is also the energy,

    Enxnynz =h

    2m

    2

    L

    n2x + n

    2y + n

    2z

    . (61)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    24/51

    At T = 0 all the lowest states will be occupied by one electron. Since electronscan have spin up or down, there may be two electrons in each momentumstate. Thus if there are N electrons in a solid of volume V, we can find themaximum momentum kF as

    N = 2V

    d3k

    (2)3(kF k) = V

    2

    kF0

    k2dk =V

    32k3F. (62)

    We could also do the same in terms of energy. Since = h2k2/2m, d =

    (h

    2

    /m)kdk. Let g() be the density of states. Then,

    N =

    F0

    g()d =h2

    m

    F0

    g()kdk. (63)

    Comparing with equation (62) we find that

    g() =V

    22

    2m

    h2

    3/21/2. (64)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    25/51

    4.2 Fermi gas at T > 0

    To the Fermi energy we can associate a temperature, the Fermi temperature

    TF = F

    kB= h

    2k2F2mkB

    = (32n)2/3h2mkB

    (65)

    where n is the electron density. Putting in typical numbers we find that TF isof the order of 103 104 K. Only at this high temperatures will a substantialamount of the electrons be in excited states. At more moderate temperatures

    only the electrons which are already close to the Fermi energy can be excited.Electrons with energy far below the Fermi energy cannot be excited since thePauli principle denies them to go to a state which is already filled. This explainswhy the electron contribution to the heat capacity is so smallonly those fewelectrons which are within kBT of the Fermi energy can contribute.To proceed, replace (63) with

    N =

    0

    g()f()d (66)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    26/51

    where f() is the Fermi distribution function,

    f() =1

    e() + 1. (67)

    Note that at T = 0 the exponential will be 0 or + depending on the sign of . Thus

    f(; T = 0) =

    1 < 0 >

    (68)

    To be consistent with equation (63) we must have = F at T = 0.

    At T = 0 f() will no longer be a step function, and will no longer be exactlyequal to F. But at T TF will still be close to F and f() will look muchlike a step function, only with a bit smoothening at the step.

    To find the value of at finite temperature we must calculate the integral (66).To calculate this integral, lets define () such that

    g() =

    ().

    Then

    N =

    0

    g()f() =

    0

    f() = ()f()|0

    0

    ()f

    (69)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    27/51

    Since () vanishes at = 0 and f() vanishes when the boundaryterm does not contribute. Changing the integration variable to = ( )we find

    N = ()

    f

    d =

    ()

    (e/2 + e/2)2 d. (70)

    To proceed, let T TF /kB. Then . We also expand around = 0,

    () = (0) +

    =0

    +2

    2

    2

    2 =0

    + . . .

    = (0) +2

    2

    g

    =0

    + . . .

    =V

    32

    2m

    h2

    3/21 +

    3

    8

    (kBT)2

    22 + . . .

    Inserting this into our integral, we find

    N V32

    2m

    h2

    3/2

    d

    1

    (e/2 + e/2)2 +

    3

    8

    (kBT)2

    22

    (e/2 + e/2)2

    (71)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    28/51

    Now that we have determined we can go on to compute other quantities.For instance, the energy of the Fermi gas is given by

    E =0 g()f()d, (72)

    and then the heat capacity can be found simply by differentiating with respectto temperature.

    5 p-n junction

    If a semiconductor is n-doped an extra energy level per doping atom i addedjust below the conduction band. Correspondingly for a p-doped semiconductoran extra energy level per doping atom is added just above the valence band.

    To make a p-n junction we p-dope part of the semiconductor and n-dope theother part. But since thermodynamics require the chemical potential to be thesame in all the sample at equilibrium something must happen close to wherethe p-doping and the n-doping meet. What happens is that in this regionthe electrons reorganise such that the potential on the n-side is lowered. The

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    29/51

    reason this happens is that on the p-side there is surplus of holes, and on then-side there is surplus of electrons. The electrons and holes annihilates. Butsince both the n-side and p-side originally were neutral this means that closeto the interface there will be a positively charged zone on the n-side and anegatively charged zone on the p-side.

    But as the positive charge builds up on the n-side and the negative chargebuilds up on the p-side there is created a electric field pushing the electronsback to the n-side and the holes back to the p-side. At some stage this fieldwill get strong enough to stop the electron and holes meeting to annihilate.

    To find the electrical potential at which the equilibrium is reached, we mustequate the different currents involved. The process bringing electrons to-wardsthe p-side and holes to-wards the n-side is diffusion. The diffusion current isgiven by Ficks law,

    J = Dn (73)where D is the diffusion constant and n is the number density of carriers. Toavoid double-counting we should consider either the electrons or the holes.Choosing the electrons we find that the charge current caused by diffusion is

    jdiff = eJ = eDn (74)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    30/51

    The current caused by the electric field which is build up is

    jdrift = E = enE, (75)

    where again n is the number density and is the mobility. At equilibrium

    0 = jdrift +jdiff = en(r)E + eDn(r) (76)

    Assuming that there only is change in the x-direction, the equation becomes

    n(x)Ex(x) = D dn(x)dx

    , (77)

    which gives us the electrical field in terms of the electron distribution:

    E(x) =

    D

    1

    n

    dn

    dx

    . (78)

    But we are really more interested in the potential,

    (x) = () x

    Ex(x)dx. (79)

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    31/51

    Since the interface between the n-doped and the p-doped semiconductor isvery depleted of charge carriers, the resistance there will be much larger thanin the rest of the semi conductor. Thus practically all of the potential fall willbe in this region. Thus the potential difference between the p-side and the

    n-side is

    Vc = () () =

    Exdx =D

    n()n()

    dn

    n=

    D

    log

    n()n() . (80)

    Far from the interface the number densities are that of p-doped and n-dopedsemiconductors, respectively. Since most of the electrons introduced by theimpurities are free at normal temperatures, n() = ND. In the p-region weuse the semiconductor equation to find n() NA = n2i (T). Here NA andND are the densities of acceptor and donor atoms, respectively. Thus

    n(

    )

    n() =ND

    NA

    n2i (T).

    But to complete this we need to know what D/ is in terms of known quan-tities. To establish the relationship between D and consider a constant Ex

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    32/51

    giving the equation

    D dndx

    = Exn(x),

    which has solution

    n(x) eExx/D. (81)But from Boltzmann statistics we have

    n(x) eeExx/kBT. (82)

    Since these to exponentials must be the same to ensure consistency, we findthat

    D

    =

    kBT

    e, (83)

    which gives

    Vc =

    kBT

    e log

    NDNAn2i (T) . (84)

    Now we go on to study the charge density distribution, (x). We assumethat (x) has a constant value in a region of thickness dp on the p-side of the

    junction and another constant value in a region of thickness dn on the n-side

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    33/51

    of the junction. Everywhere else we assume (x) = 0. We also assume that allelectrons from the region dn fill all the holes in the region dp. Together withcharge conservation this implies

    NAdp = NDdn.From Poissons equation we know that dEx/dx = 4(x). Thus for dp < x dn

    (87)

    Continuity at x = 0 gives

    2eNAd2p = Vc 2eNDd2n.

    fi d d d d d N d N

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    34/51

    Defining d = dp + dn, and using dnND + dpNA we can write

    dn =NA

    NA + NDd, dp =

    NDNA + ND

    d.

    Inserting this for dn and dp we find

    Vc = 2e

    NAN

    2D + NDN

    2A

    (NA + ND)2

    d2 = 2ed2

    NANDNA + ND

    ,

    or solving for d,d =

    Vc

    2e NA + ND

    NAND. (88)

    If an external voltage over the p-n junction, the resulting current is stronglydependent on the direction of the voltage. Since the voltage drop over the

    p-n junction happens over such a small distance the electric field here is verylarge. Therefore very high external fields must be applied to change the driftcurrent appreciably. But the diffusion current can influenced by changing thedensities. That is, if we connect the positive side of a battery to the p-sidewe remove electrons from the depletion zone. Thus we make room for new

    l diff h j i D i h h d dd

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    35/51

    electrons to diffuse across the junction. Doing the other way around we addelectrons where they are already abundant, causing less diffusion.

    At V = 0 the current over the p-n junction is

    j = jdiffjdrift = j0 j0 = 0.If V = 0 jdrift will not be much changed, but in the drift current there will bean extra Boltzmann factor eeV/kBT. Thus the current with an external field is

    j = (eeV/kBT 1)j0. (89)At V = 0 we retain j = 0 as we should. At V > 0 there is a current growingexponentially as we increase V, and at large negative values of V we are leftwith only the drift current which is very small.

    6 Magnetic properties of solids

    If a solid is exposed to a static magnetic field, there will be a response to theapplied field. What this response is, however, is strongly dependent on the ma-terial of which the solid is made. The basic types of responses are diamagnetic,

    ti d f ti A di ti t i l d b tti

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    36/51

    paramagnetic and ferromagnetic. A diamagnetic material responds by settingup a weak field opposing the applied field. A paramagnetic material sets up aweak field in the same direction as the applied field, whereas a ferromagneticmaterial sets up a strong field in the same direction as the applied field.

    6.1 Diamagnetism

    Diamagnetism is present in all materials, but it is so weak that if the material isparamagnetic the diamagnetism will not be seen. But if there are no competingeffects the solid will respond to an external magnetic field by setting up a weakfield in the opposite direction. Diamagnetism can be viewed as a manifestationof Lenzs lawwhen the material is exposed to an external magnetic field, theelectronic orbits adjust so as to set up a magnetic field opposing the appliedfield.

    6 1 1 Cl i l l ti

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    37/51

    6.1.1 Classical evaluation

    To find an expression for the magnetic susceptibility of a material, consider acharge revolving in a circle inside the solid. (This charge will be a classical

    electron). First we have to fix the sign conventions; Assume that the motionis in the xy plane, and let the positive z-direction define the positive revolutiondirection. The angular frequency of the revolution must be treated as a signedquantity according to this definition. In the radial direction we define to-wardsthe origin as the positive direction.

    If the particle revolving has charge q, the radius of the circle is a and theangular frequency is 0 we have the current

    I =q02

    ,

    which gives rise to a magnetic moment

    m =IA

    c=

    q0a2

    2c(90)

    If the charged particle has mass m, there must be a radial force Fr = m20a to

    keep it in its orbit. Now we apply an external magnetic field H in the positive

    z direction This gives rise to an extra radial force Assuming that the particle

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    38/51

    z-direction. This gives rise to an extra radial force. Assuming that the particleis confined to move only in its circle of radius a the magnetic force is

    Fm =

    qaH

    c

    , (91)

    where is the frequency with the external field present. Since the radial forcenow is modified, but the circle of motion remains the same there must havebeen a frequency shift. Thus = 0. To find the frequency shift we write theequation of motion for the system with the applied magnetic field

    m2a = Fr + Fm = m20a qaH

    c ,

    or

    2 20 = qH

    mc. (92)

    Assuming that the magnetic field is weak, such that 0 we find

    = qHmc

    + 0

    qH2mc

    . (93)

    is known as the Larmour frequency. The associated energy shift is

    E= h = BH.

    = eh2mc is known as the Bohr magneton

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    39/51

    B = eh2mc is known as the Bohr magneton.

    In an atom with closed shells all electron spins are paired, so that they sum to0. Therefore there is no permanent magnetic moment. However, if a magneticfield is applied the orbital frequency will be changed by the Lamour frequency.Substituting the Larmour frequency into (90), we find that as a response tothe applied field a magnetic moment

    m = q2a2H

    4mc2(94)

    is induced. So far we have viewed the orbital motion as confined to a plane. Inreal life, however, the electrons move in three dimensions. This means that a2

    in reality should be the mean square perpendicular distance from the magneticfield axis through the nucleus to the electron. If H is taken to point in the zdirection, a2 = x2 + y2. This should be related to the mean square radialdistance r2 = x2 + y2 + z2. If the atom is spherical x2 = y2 = z2which gives r2 = 3a2/2. To compare the result with experiments we shouldcalculate the susceptibility = mN/V H, where N is the number of atoms inthe volume V. From (94 we find

    = pe2N

    6mc2Vr2 (95)

    where p is the number of electron in the outer shell 3 Note that is negative

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    40/51

    where p is the number of electron in the outer shell. Note that is negative,as Lenzs law told us it should be.

    6.1.2 Quantum evaluation

    We apply an external magnetic field B = (0, 0, Bz). Since B = A anrealisation of this field in terms of A is A = (Bzy/2, Bzx/2, 0). If we nowintroduce the minimal coupling,

    p p e

    c A (96)

    3Strictly speaking we should include all electrons, each with its own value for r2i. But

    since usually only the outer shell electrons have a sizable contribution to r2, this is a goodapproximation.

    in the Hamiltonian we get

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    41/51

    in the Hamiltonian we get

    H =(p e

    cA)2

    2m+ V(r)

    = p22m

    e2mc

    p A + A p + e2

    2mc2A2 + V(r)

    = H0 e2mc

    L B + e2

    8mc2(x2 + y2)B2z

    = H0 m B + .

    (97)

    The energy of the magnetic field can also be written

    H =1

    8

    B HdV = 1

    8

    B (B4M)dV = 1

    8

    B2(14)dV. (98)

    Since diamagnetism is only important when all spins pair to zero, we consider

    m = 0. Thus the contribution containing must come from the -term.Thus

    = N p2V

    e2

    8mc2x2 + y2 = pe

    2N

    24mc2Vr2. (99)

    6.2 Paramagnetism

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    42/51

    6.2 Paramagnetism

    If an atom has angular momentum (including spin) not equal to zero, it willhave a permanent magnetic moment. An orbital angular momentum L con-

    tributes L = BL to the magnetic moment, whereas an unpaired electronspin contributes S = geB/2. ge is called the the gyromagentic ratio of theelectron, and is approximately equal to 2. Coupling the orbital angular mo-mentum to the spin, the total angular momentum J is found. The magneticmoment of the atom is then

    m = gJBJ = BL + 2S (100)gJ is called the Lande g-factor.

    6.2.1 Classical evaluation

    If a magnetic field H is applied the magnetic moment m will contribute Em =m H to the energy. Thus by aligning the magnetic moments to the appliedfield, the magnetic energy will be minimised. However, thermal excitationwill partially destroy the alignment. Neglecting any other effect which might

    hinder alignment, the resulting magnetisation of the material is

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    43/51

    hinder alignment, the resulting magnetisation of the material is

    M =mN

    Vcos

    =m

    N

    Vcos eEmd

    eEmd

    =mN

    V

    11

    cos emHcos d(cos )11

    emHcos d(cos )

    =N

    HVmH

    mH

    xexdxmHmH

    exdx

    =mN

    V

    emH + emH

    emH emH 1

    mH

    =N

    V 1

    tanh mH 1

    mH

    (101)

    If mH kBT the susceptibility is

    =2mN

    3kBT V=

    C

    T, (102)

    where we have defined the Curie constant C = 2mN/2kBT V.

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    44/51

    m / B

    6.2.2 Quantum evaluation

    Quantum mechanically the magnetic moment of a atom cannot take an arbi-trary value. The projection of J on H can only take 2J + 1 different values.Then the magnetic contribution to the energy is Em = mH = mJgJBH,J < mJ < J, and the magnetisation is

    M = NV

    JmJ=JgJBmJemJgJBHJmJ=J

    emJgJBH= N

    VgJBJBJ(y), (103)

    where

    BJ(y) =

    2J + 1

    2J tanh

    12J + 12J y

    1

    2J tanh

    1 y2J

    , (104)

    with y = gJBJH. BJ(y) is called the Brillouin function. For large J theBrillouin function asymptotically approaches the classic result.

    To take a really simple example, consider an atom with one unpaired electron

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    45/51

    y p p , pand no orbital angular momentum. This atom has J = 1/2, thus mJ can onlytake the values 1/2. Then the magnetisation is

    M = NV

    12gJB(e

    gJBH

    eg

    JBH

    )egJBH + egJBH

    =NgJB

    2V tanh(gJBH)

    . (105)

    For gJBH

    kBT, we expand the hyperbolic tangent and find the suscepti-

    bility to be

    =g2J

    2BN

    2kBT V. (106)

    6.3 Ferromagnetism

    Some materials can have a net magnetic moment even without an appliedmagnetic field. Furthermore, if an external field is applied, the material in-creases the field strength enormously. For these materials the independentatom approximation we did for the paramagnetic case does not hold. The

    auto-magnetisation of the material must be due to interactions between the

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    46/51

    gmagnetic moments of the atoms. In 1928 Heisenberg modelled this by addingat term

    Vm =

    2J(rij)Si

    Sj, (107)

    where Si is the spin of atom i and J(rij) describes the interaction strengthbetween atoms i and j. If J > 0 it is favourable for the spins to align, andwe get what we expect for ferromagnetism. If J < 0 we get what is knownas antiferromagnetism. The origin of this interaction is actually not magnetic,but electrostatic. The interaction arise because the charge distribution, as con-

    trolled by the electron wave function) must obey the Pauli principle. Assumingthat J is independent of temperature, we see that at some temperature themagnetisation must be destroyed by thermal agitation. This is indeed observedin experiments. The magnetisation vanishes when the temperature passes acritical temperature, Tc. At higher temperature the material is paramagnetic.

    Before proceeding with the Heisenberg theory, lets have look at one the Weissmolecular field theory. This is a phenomenological theory which explains manyaspects of ferromagnetism. Assume that there is an effective field BE = EMwhere M is the magnetisation of the solid, acting on the individual magneticmoments. In the paramagnetic phase, T > Tc, the total field is the sum

    of the applied field and E. Denoting the paramagnetic susceptibility 0 the

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    47/51

    magnetisation is

    M = 0(H + BE) = 0(H + EM) (108)

    Thus the normal susceptibility is

    =M

    H=

    01 0E (109)

    The temperature dependence of the paramagnetic susceptibility is given byCuries law, 0 = C/T. Thus

    =C

    T CE =C

    T Tc . (110)

    Note that at T T+

    c . This is typical for phase transitions.Recall that the quantum theory of paramagnetism showed that

    M =gJBJN

    VBJ(gJBJH). (111)

    To describe the ferromagnetic phase, assume that the same equation can be

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    48/51

    used to express the spontaneous magnetisation, Ms in the absence of an appliedfield if we replace H by BE = EMs. Thus

    Ms =gJ

    B

    JN

    V BJ(gJBJEMs). (112)

    This equation cannot be solved analytically, but using a numeric method wecan solve for Ms.

    Returning to the Heisenberg model we can now find approximate values for

    the exchange constant by using the results from the Weiss theory. In terms ofthe Heisenberg model the effective field BE must be understood as set up bythe effect of all neighbouring spins. Thus

    2j

    J(rij)Sj Si = i BE = gbSi BE (113)

    Using M gBSjN/V and Tc = EC = 2effN/3kBV we obtainj

    J(rij) 3kBTc2S(S+ 1)

    . (114)

    6.4 The Hall effect

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    49/51

    At the end of this section we will also discuss a completely different aspect ofmagnetism in solids. Consider a slab of a conducting material. Let there be

    a static electric field in the positive x-direction and a static magnetic field inthe positive z direction. The charge carriers is then subject to the force

    F = q(E + v B). (115)

    The applied electric field will give rise to a current in the x-direction. The

    charges contributing to this current will be affected by a magnetic force trans-verse to their motion. Since F = qv B, and the sign of v is opposite forpositive and negative charge carriers, the carriers will be drawn in the nega-tive y direction regardless of the sign of their charge. Since we are consideringa finite slab of material the carriers will have to accumulate at the boundary

    giving rise to a electric field in either the positive y-direction (positive carriers)or the negative y-direction (negative carriers).

    E = Exex + Eyey,

    B = Bzez,(116)

    and the corresponding forces are

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    50/51

    FE = q(Exex + Eyey)

    FB = qvxBzey(117)

    To find Ey we equate the forces in the y direction. But we do not want to keepthe velocity of the carriers in our final answer since this is hard to measure.Therefore we expresses the current in terms of the velocity,

    j = qnv

    and substitute this into the equation. Thus

    Ey = vxBz =jxBz

    qn. (118)

    We also define the Hall coefficient

    RH =Ey

    jxBz= 1

    qn. (119)

    Note that RH is only dependent on the sign of the charge carriers and thecarrier density. The former indicates that by measuring the sign of RH we

    can decide the sign of the charge carriers. However, measurements show thatl h H ll ffi d h

  • 7/29/2019 phys208_Trygve_SLIDES.pdf

    51/51

    some metals have a positive Hall coefficient, indicating positive charge carriers.Since we know that the charge carriers in metals are electrons, this shows thatwe need to refine the model.